Nitride Semiconductor Light-Emitting Diodes (LEDs), Second Edition: Materials, Technologies, and Applications

Nitride Semiconductor Light-EmittingDiodes (LEDs), Second Edition

Related titles

Buckley, Organic Light-Emitting Diodes (OLEDs)

(ISBN: 978-0-85709-425-4)

Dhoble, Swart and Kalyani, Principles and Applications of Organic Light EmittingDiodes (OLEDs)

(ISBN: 978-0-08-101213-0)

Jagadish and Mi, III-Nitride Semiconductor Optoelectronics: Semiconductors andSemimetals Series

(ISBN: 978-0-12-809723-6)

Woodhead Publishing Series in Electronic and

Optical Materials

Nitride SemiconductorLight-Emitting Diodes(LEDs)Materials, Technologies, andApplications

Second Edition

Edited by

JianJang HuangNational Taiwan University, Taipei, Taiwan

Hao-Chung KuoNational Chiao-Tung University, Hsinchu, Taiwan

Shyh-Chiang ShenGeorgia Institute of Technology, Atlanta, GA,United States

Woodhead Publishing is an imprint of ElsevierThe Officers’Mess Business Centre, Royston Road, Duxford, CB22 4QH, United Kingdom50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United StatesThe Boulevard, Langford Lane, Kidlington, OX5 1GB, United Kingdom

Copyright © 2018 Elsevier Ltd. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means,electronic or mechanical, including photocopying, recording, or any information storage andretrieval system, without permission in writing from the publisher. Details on how to seekpermission, further information about the Publisher’s permissions policies and ourarrangements with organizations such as the Copyright Clearance Center and the CopyrightLicensing Agency, can be found at our website: www.elsevier.com/permissions.

This book and the individual contributions contained in it are protected under copyright bythe Publisher (other than as may be noted herein).

NoticesKnowledge and best practice in this field are constantly changing. As new research andexperience broaden our understanding, changes in research methods, professional practices,or medical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge inevaluating and using any information, methods, compounds, or experiments describedherein. In using such information or methods they should be mindful of their own safety andthe safety of others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,assume any liability for any injury and/or damage to persons or property as a matter ofproducts liability, negligence or otherwise, or from any use or operation of any methods,products, instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication DataA catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication DataA catalogue record for this book is available from the British Library

ISBN: 978-0-08-101942-9 (print)ISBN: 978-0-08-101943-6 (online)

For information on all Woodhead Publishing publications visitour website at https://www.elsevier.com/books-and-journals

Publisher: Mathew DeansAcquisition Editor: Kayla Dos SantosEditorial Project Manager: Ana Claudia GarciaProduction Project Manager: Surya Narayanan JayachandranDesigner: Mark Rogers

Typeset by TNQ Books and Journals

http://www.elsevier.com/permissions

https://www.elsevier.com/books-and-journals

Contents

List of contributors xiPreface xv

Part One Materials and fabrication 1

1 Molecular beam epitaxy (MBE) growth of nitride semiconductors 3Qiandong Zhuang1.1 Introduction 31.2 Molecular beam epitaxial (MBE) growth techniques 31.3 Plasma-assisted MBE (PAMBE) growth of nitride epilayers

and quantum structures 41.4 Nitride nanocolumn (NC) materials 111.5 Nitride nanostructures based on NCs 161.6 Conclusion 19

References 19

2 MOCVD growth of nitride semiconductors 25Koh Matsumoto, Yoshiki Yano, Hiroki Tokunaga, Akinori Ubukata,Guanxi Piao, Akira Mishima, Tadakazu Ikenaga, Yuji Tomita,Toshiya Tabuchi2.1 Introduction 252.2 Growth mechanism 282.3 Carbon incorporation and Mg doping of GaN 312.4 Blue and green MQW 322.5 UV materials growth 35

References 40

3 GaN on sapphire substrates for visible light-emitting diodes 43Jae-Hyun Ryou, Wonseok Lee3.1 Importance and historical backgrounds of GaN epitaxial growth

and sapphire substrates 433.2 Sapphire substrates 463.3 Strained heteroepitaxial growth on sapphire substrates 523.4 Epitaxial overgrowth of GaN on sapphire substrates 59

3.5 GaN growth on nonpolar and semipolar directions 653.6 Outlook of LEDs on sapphire substrates 67

References 67

4 Gallium nitride (GaN) on silicon substrates for LEDs 79Matthew H. Kane, Nazmul Arefin4.1 Introduction 794.2 An overview of gallium nitride (GaN) on silicon substrates 794.3 Silicon overview 804.4 Challenges for the growth of GaN on silicon substrates 834.5 Buffer-layer strategies 864.6 Device technologies 934.7 Conclusion 117

References 118

5 Phosphors for white LEDs 123Zhanchao Wu, Zhiguo Xia5.1 Introduction 1235.2 Requirements for phosphors used in wLEDs 1245.3 The state-of-the-art phosphors for wLEDs 1275.4 New advances of future phosphors for wLEDs 1335.5 Future development of wLEDs phosphors 192

Acknowledgments 193References 193

6 Recent development of fabrication technologies of nitride LEDsfor performance improvement 209Ray-Hua Horng, Dong-Sing Wuu, Chia-Feng Lin, Chun-Feng Lai6.1 Introduction 2096.2 GaN-based flip-chip LEDs and flip-chip technology 2106.3 GaN FCLEDs with textured micro-pillar arrays 2136.4 GaN FCLEDs with a geometric sapphire shaping structure 2186.5 GaN thin-film photonic crystal (PC) LEDs 2256.6 PC nanostructures and PC LEDs 2276.7 Light emission characteristics of GaN PC TFLEDs 2316.8 Conclusion 237

References 238

7 Nanostructured LED 243Chien-Chung Lin, Ching-Hsueh Chiu, Da-Wei Lin, Zhen-Yu Li,Yu-Pin Lan, JianJang Huang, Hao-Chung Kuo7.1 Introduction 2437.2 Top-down technique for nanostructured LED 244

vi Contents

7.3 Bottom-up technique for GaN nanopillar substrates preparedby molecular beam epitaxy 263

7.4 Other nanostructures of interest for LEDs 2697.5 Conclusion 269

References 269

8 Nonpolar and semipolar LEDs 273Yuh-Renn Wu, C.-Y. Huang, Yuji Zhao, James Speck8.1 Motivation: limitations of conventional c-plane LEDs 2738.2 Introduction to selected nonpolar and semipolar planes 2778.3 Challenges in nonpolar and semipolar epitaxial growth 2858.4 Light extraction for nonpolar and semipolar LEDs 288

References 291Further reading 295

Part Two Performance of nitride LEDs 297

9 Efficiency droop in GaInN/GaN LEDs 299Houqiang Fu, Yuji Zhao9.1 Introduction 2999.2 Physical mechanisms of current droop in GaInN/GaN LEDs 3029.3 Progress of low-droop GaInN/GaN LEDs 3119.4 Thermal droop in GaInN/GaN LEDs 320

References 323

10 Photonic crystal nitride LEDs 327Martin D.B. Charlton10.1 Introduction 32710.2 Photonic crystal technology 33510.3 Improving LED extraction efficiency through PC surface

patterning 34110.4 PC-enhanced light extraction in P-side up LEDs 34710.5 Modelling PC-LEDs 35010.6 PC-enhanced light extraction in N-side up LEDs 36510.7 Summary 37310.8 Conclusions 374

References 376

11 Nitride LEDs based on quantum wells and quantum dots 377J. Verma, S.M. Islam, A. Verma, V. Protasenko, D. Jena11.1 Light emitting diodes 37711.2 Polarization effects in III-nitride LEDs 387

Contents vii

11.3 Current status of III-nitride LEDs 39711.4 Modern LED designs and enhancements 404

References 405Further reading 413

12 Colour tuneable LEDs and pixelated micro-LED arrays 415Yuk Fai Cheung, Zetao Ma, Hoi Wai Choi12.1 Introduction: motivation for color tuning and review of existing

technologies 41512.2 Stacked LEDs 41612.3 Group-addressable pixelated micro-LED arrays 43012.4 Conclusions 436

Acknowledgments 438References 438

13 Reliability of nitride LEDs 441Tzung-Te Chen, Chun-Fan Dai, Chien-Ping Wang, Han-Kuei Fu,Pei-Ting Chou, Wen-Yung Yeh13.1 Introduction 44113.2 Reliability testing of nitride LEDs 44113.3 Evaluation of LED degradation 44413.4 Degradation mechanisms 44713.5 Conclusion 452

References 452

14 Physical mechanisms limiting the performance and the reliabilityof GaN-based LEDs 455Carlo De Santi, Matteo Meneghini, Alberto Tibaldi, Marco Vallone,Michele Goano, Francesco Bertazzi, Giovanni Verzellesi,Gaudenzio Meneghesso, Enrico ZanoniIntroduction 45514.1 Modeling the performance-limiting effects in GaN-based

LEDs 45614.2 Degradation of LEDs under electrical and thermal stress 46714.3 Conclusions 481

References 481

15 Chip packaging: encapsulation of nitride LEDs 491Xiaobing Luo, Run Hu15.1 Functions of LED chip packaging 49115.2 Basic structure of LED packaging modules 49515.3 Processes used in LED packaging 49815.4 Optical effects of gold wire bonding 50215.5 Optical effects of phosphor coating 50515.6 Optical effects of freeform lenses 511

viii Contents

15.7 Thermal design and processing of LED packaging 51515.8 Conclusion 524

References 524

Part Three Applications of nitride LEDs 529

16 White LEDs for lighting applications 531Richard Kotschenreuther16.1 White LEDsddefinition of area 53116.2 Why “white LEDs”? 53116.3 The three-side-approach for lighting applications 53116.4 Fields of application 54216.5 LED light sources in the connected world 54616.6 Outlook 547

Abbreviations and Acronyms 549References 549Further reading 550Annex 1 551

17 Ultraviolet LEDs 553Hideki Hirayama17.1 Research background of deep ultraviolet light-emitting diodes 55317.2 Growth of low TDD AlN layers on sapphire 55717.3 Marked increases in IQE 56117.4 Aluminum gallium nitride-based DUV-LEDs fabricated

on high-quality aluminum nitride 56817.5 Increase in EIE and LEE 57617.6 Conclusions and future trends 583

References 584

18 Infrared emitters using III-nitride semiconductors 587Akhil Ajay, Yulia Kotsar, Eva Monroy18.1 Introduction 58718.2 High-indium-content alloys for IR emitters 58718.3 RE-doped GaN emitters 59018.4 III-nitride materials for ISB optoelectronics 59118.5 ISB devices 60118.6 Conclusions 605

References 606

19 LEDs for liquid crystal display (LCD) backlighting 619Chi-Feng Chen19.1 Introduction 61919.2 Types of LED LCD backlighting units 619

Contents ix

19.3 Technical considerations for optical films and plates 62419.4 Requirements for LCD BLUs 62519.5 Advantages and history of LED BLUs 62619.6 Market trends and technological developments 62919.7 Optical design 634

References 644

20 LEDs and automotive lighting applications 647John D. Bullough20.1 Introduction 64720.2 Forward lighting 64720.3 Signal lighting 65120.4 Human factor issues with LEDs 65220.5 Energy and environmental issues 65520.6 Future outlook 65620.7 For further information 656

References 656

21 LEDs for large displays 659Linas Svilainis21.1 Introduction 65921.2 LED display types 66021.3 Display parameters 66521.4 Technology in detail 68221.5 Summary 727

References 728

22 LEDs for projectors 737Linas Svilainis22.1 Introduction 73722.2 Projector technologies 73822.3 Applications 75422.4 Summary 757

References 757

Index 761

x Contents

List of contributors

Akhil Ajay CEA-Grenoble, INAC-PHELIQS, Grenoble, France

Nazmul Arefin University of Oklahoma, Norman, OK, United States

Francesco Bertazzi Istituto di Elettronica e di Ingegneria dell’Informazione e delleTelecomunicazioni, Consiglio Nazionale delle Ricerche, Torino, Italy; Politecnico diTorino, Torino, Italy

John D. Bullough Rensselaer Polytechnic Institute, Troy, NY, United States

Martin D.B. Charlton University of Southampton, Southampton, United Kingdom

Chi-Feng Chen National Central University, Taoyuan City, Taiwan

Tzung-Te Chen Industrial Technology Research Institute (ITRI), Hsinchu, Taiwan

Yuk Fai Cheung The University of Hong Kong, Hong Kong

Ching-Hsueh Chiu National Chiao Tung University, Hsinchu, Taiwan

Hoi Wai Choi The University of Hong Kong, Hong Kong

Pei-Ting Chou Industrial Technology Research Institute (ITRI), Hsinchu, Taiwan

Chun-Fan Dai Industrial Technology Research Institute (ITRI), Hsinchu, Taiwan

Carlo De Santi University of Padova, Padova, Italy

Houqiang Fu Arizona State University, Tempe, AZ, United States

Han-Kuei Fu Industrial Technology Research Institute (ITRI), Hsinchu, Taiwan

Michele Goano Istituto di Elettronica e di Ingegneria dell’Informazione e delleTelecomunicazioni, Consiglio Nazionale delle Ricerche, Torino, Italy; Politecnico diTorino, Torino, Italy

Hideki Hirayama Riken, Saitama, Japan

Ray-Hua Horng National Chung Hsing University, Taichung, Taiwan

Run Hu Huazhong University of Science and Technology, Hubei, China

C.-Y. Huang TSMC Solid State Lighting, Ltd, Hsinchu, Taiwan

JianJang Huang National Taiwan University, Taipei, Taiwan

Tadakazu Ikenaga TAIYO NIPPON SANSO Corp., Minato-ku, Japan

S.M. Islam Cornell University, Ithaca, NY, United States

D. Jena Cornell University, Ithaca, NY, United States

Matthew H. Kane Texas A & M University at Galveston, Galveston, TX, UnitedStates

Yulia Kotsar CEA-Grenoble, INAC-PHELIQS, Grenoble, France

Richard Kotschenreuther OSRAM GmbH, Munich, Germany

Hao-Chung Kuo National Chiao Tung University, Hsinchu, Taiwan

Chun-Feng Lai Feng-Chia University, Taichung, Taiwan

Yu-Pin Lan National Chiao Tung University, Hsinchu, Taiwan

Wonseok Lee LED Business Unit, LG Innotek, Paju-si, Korea

Zhen-Yu Li National Chiao Tung University, Hsinchu, Taiwan

Chien-Chung Lin National Chiao Tung University, Tainan, Taiwan

Chia-Feng Lin National Chung Hsing University, Taichung, Taiwan

Da-Wei Lin National Chiao Tung University, Hsinchu, Taiwan

Xiaobing Luo Huazhong University of Science and Technology, Hubei, China

Zetao Ma The University of Hong Kong, Hong Kong

Koh Matsumoto TAIYO NIPPON SANSO Corp., Minato-ku, Japan

Gaudenzio Meneghesso University of Padova, Padova, Italy

Matteo Meneghini University of Padova, Padova, Italy

Akira Mishima TAIYO NIPPON SANSO Corp., Minato-ku, Japan

Eva Monroy CEA-Grenoble, INAC-PHELIQS, Grenoble, France

Guanxi Piao TAIYO NIPPON SANSO Corp., Minato-ku, Japan

V. Protasenko Cornell University, Ithaca, NY, United States

Jae-Hyun Ryou University of Houston, Houston, TX, United States

James Speck University of California, Santa Barbara, CA, United States

Linas Svilainis Kaunas University of Technology, Kaunas, Lithuania

Toshiya Tabuchi TAIYO NIPPON SANSO Corp., Minato-ku, Japan

Alberto Tibaldi Istituto di Elettronica e di Ingegneria dell’Informazione e delleTelecomunicazioni, Consiglio Nazionale delle Ricerche, Torino, Italy

Hiroki Tokunaga TAIYO NIPPON SANSO Corp., Minato-ku, Japan

xii List of contributors

Yuji Tomita TAIYO NIPPON SANSO Corp., Minato-ku, Japan

Akinori Ubukata TAIYO NIPPON SANSO Corp., Minato-ku, Japan

Marco Vallone Politecnico di Torino, Torino, Italy

A. Verma Cornell University, Ithaca, NY, United States; Indian Institute ofTechnology, Kanpur, India

J. Verma University of Notre Dame, Notre Dame, IN, United States; IntelCorporation, Hillsboro, OR, United States

Giovanni Verzellesi Universit�a di Modena e Reggio Emilia, Reggio Emilia, Italy

Chien-Ping Wang Chung Yuan Christian University, Chung-Li, Taiwan

Yuh-Renn Wu National Taiwan University, Taipei, Taiwan

Zhanchao Wu Qingdao University of Science and Technology, Qingdao, People’sRepublic of China; University of Science and Technology Beijing, Beijing, China

Dong-Sing Wuu National Chung Hsing University, Taichung, Taiwan

Zhiguo Xia University of Science and Technology Beijing, Beijing, China

Yoshiki Yano TAIYO NIPPON SANSO Corp., Minato-ku, Japan

Wen-Yung Yeh Industrial Technology Research Institute (ITRI), Hsinchu, Taiwan

Enrico Zanoni University of Padova, Padova, Italy

Yuji Zhao Arizona State University, Tempe, AZ, United States

Qiandong Zhuang Lancaster University, Lancaster, United Kingdom

List of contributors xiii

This page intentionally left blank

Preface

Light-emitting diodes (LEDs) have extended their presence from being dim indicatorson instrument panels and children’s toys to highly efficient solid-state lighting (SSL)of daily life. This beautifully engineered technology, pioneered by Professor Holonyakin 1962 and enabled by numerous bright scientists and engineers with 50-plus years ofactive research and development, is transforming the way electric energy is utilized inthe creation of artificial lighting. Today, LEDs in SSL have reached a peak efficiencyof greater than 250 lm/W with a lifetime of greater than 60,000 h (approximately threeto five times longer lifetime than today’s fluorescent lamps!) and these “ultimatelamps” are currently produced using III-nitride (III-N) semiconductors.

Constant improvements in SSL device technology and a wide acceptance of SSLaround the world have led to the economies of scale for III-N LED technologies.According to a recent report, SSL sales reached $US 26 billion in 2016 and will reacha projected market size of $US 54 billion in 2022 with a compound annual growth rateof 13% between 2017 and 2022 (https://www.zionmarketresearch.com/market-analysis/led-lighting-market). Nitride-based LEDs are poised to replace the incandes-cent light bulbs that were brilliantly invented more than a century ago. Undoubtedly,better energy utilization in these ultimate lamps promises a significant reduction in thecarbon footprint, a crucial issue for scientists and all of the human kind who under-stand and are concerned about the human-activity-induced climate change.

The ubiquitous presence of LED technology is evident in all aspects of today’s con-sumer electronics and infrastructure, which require efficient and environmentallyfriendly photon emission to safeguard and enrich human life: traffic signal lights,pedestrian signage, and backlight sources for displays, just to name a few. Forexample, the replacement of compact cathode fluorescent lamps with eco-friendlymercury-free LEDs has enabled new generations of liquid crystal displays with lowerpower consumption, richer color reproduction, and improved response time in a highlycompact form. There are, however, several technological roadblocks and market chal-lenges to be overcome before LED technology can substantially impact the futureworld.

III-N material technology is a relatively new scientific research field. The quality ofepitaxial materials still leaves much room for improvement. The lack of native sub-strates for III-N materials may become a fundamental impediment to improvingLED efficiency and ultimately affecting manufacturing costs. The design and optimi-zation of the quantum mechanical structures needed for the manipulation and controlof electronephoton interactions significantly affect the performance of III-N LEDs.

https://www.zionmarketresearch.com/market-analysis/led-lighting-market

https://www.zionmarketresearch.com/market-analysis/led-lighting-market

These designs also need to work around the pronounced polarization charges in thesesemiconductors. In addition, the inevitable thermal effect and “mysterious” efficiencydroop phenomena have to be dealt with theoretically and experimentally. Packagingand thermal management issues for III-N LEDs are important for light extraction,and the human factor is an intricate but interesting topic in SSL businesses. The re-quirements for color rendering and lighting fixture retrofit have led to many importantLED developments that utilize phosphor-based wavelength conversion techniques andinclude integrated voltage regulators. In the end, the ultimate success of SSL is notonly through technological advancements but also by economic factors. Today, anLED lamp is more expensive than a fluorescent lamp per lumen. Although governmentsubsidies could boost the initial adoption of SSL LED technology, a sustainable SSLindustry also depends on the successful development of low-cost manufacturing anddevice innovation.

These intertwined scientific and technological issues in III-N LED developmenthave sparked tremendous research, development, and commercialization effortsaround the world. It should be noted that SSL technology is, in essence, a subset ofIII-N device technology. Engineering the bandgap energy of III-N materials providesnew opportunities for light emission in the ultraviolet (UV) and the infrared (IR) wave-lengths. UV-LEDs, for example, have helped the realization of compact UV light sour-ces for efficient sanitation and in the bacteria detections. These specialty LEDs willalso offer new opportunities for energy-efficient applications in different businesssectors.

This book aims to capture key development topics in contemporary III-N LED tech-nology and to provide its readers with an overview of the state of current technology.There are three parts to this book: materials and fabrication, performance consider-ations, and applications of nitride LEDs. Part I of the book concerns the basic technol-ogies that are currently employed in the physical device and chip fabrication. Althoughmetal-organic chemical vapor deposition (Chapter 2) is the major growth technologyfor contemporary commercial III-N LED manufacturing, we also include a discussionon molecular beam epitaxy (Chapter 1) to provide interested readers with the necessarybackground for other aspects of the advanced LED research. The choice of substrateshas a direct impact on the epitaxial quality and the cost of III-N LEDs. Currently, theavailable substrates for III-N LEDs range from (patterned) sapphire, silicon carbide,silicon, and free-standing or bulk GaN substrates. We will discuss III-N LEDs producedusing sapphire substrates in Chapter 3 and silicon substrates in Chapter 4. The phosphortechnology has evolved as an essential part of white-light LED manufacturing today,and it will be discussed in Chapter 5. Chapter 6 covers manufacturing technologiesfor current III-N LED manufacturing, and Chapter 7 presents new research develop-ment in nanostructured LEDs. Recent research of LEDs fabricated on less commoncrystalline planes such as nonpolar and semipolar directions is discussed in Chapter 8.

For readers who are interested in the performance matters of III-N LEDs, PART IIof the book provides a range of discussions from physics-based perspectives of III-NLED technologies. The efficiency droop of InGaN-based LEDs will be presented inChapter 9. A structural engineering using the concept of photonic crystal in LEDs ispresented in Chapter 10. Chapter 11 covers basic theory and design aspects of the

xvi Preface

active layers in nitride LEDs. New development in color tunable and the emergingmicro-LEDs is presented in Chapter 12. Chapter 13 discusses critical operational as-pects of the LED reliability employed in today’s SSL products, followed by a discus-sion on associated physical performance limiting and failure mechanisms of nitrideLEDs in Chapter 14. This part of the book is then concluded with a discussion onthe packaging nitride LEDs.

The applications for nitride LEDs are diverging and branching out in different elec-tronic systems these days. PART III of the book covers several application spaces ofnitride LEDs, including white LEDs for lighting (Chapter 16), UV-LEDs (Chapter 17),IR nitride LEDs (Chapter 18), LED backlighting in liquid crystal displays (Chapter19), LEDs in automotive lighting (Chapter 20), large-panel LED display (Chapter21), and LED projectors (Chapter 22). The LED technologies that are discussed inthis part of the book are not intended to provide an exhausted list of LED applicationsbut to serve a pedagogical purpose to help readers explore possible new ways to usethese uniquely engineered LEDs.

The editors of this book would like to express their gratitude for the contributingauthors of each chapter for their dedications and efforts in making the second editionof the book possible with timely updates for this fast-pacing technology. We are alsothankful for the tremendous editorial support by Ana Claudia Garcia, Kayla DosSantos, and the publishing team at the Elsevier. We hope this book may facilitate awider knowledge dissemination of III-N LED technology among students who studyoptoelectronic devices and professional engineers who are keen on new technologiesfor energy-efficient systems.

Last but not the least, the editors would like to express their immense gratitude toProfessor Emeritus N. Holonyak, Jr. at the University of Illinois at UrbanaeChampaign (UIUC), the late Professor Gregory Stillman at UIUC, Professor RussellD. Dupuis at Georgia Institute of Technology, and Professor Emeritus Milton Fengat UIUC for their pioneering work in compound semiconductor materials and devices,for their constant encouragement to younger generation of engineers, and for being ouracademic role models. Their relentless pursuit of engineering perfection with integrity,hard work, and unsurpassed perseverance in insisting on doing the right and importantthings have helped in making the world brighter for the future.

JianJang Huang, National Taiwan University, Taipei, TaiwanHao-Chung Kuo, National Chiao-Tung University, Hsinchu, Taiwan, and

Shyh-Chiang Shen, Georgia Institute of Technology, Atlanta, Georgia, USA

Preface xvii


Part One

Materials and fabrication


Molecular beam epitaxy (MBE)growth of nitride semiconductors 1Qiandong ZhuangLancaster University, Lancaster, United Kingdom

1.1 Introduction

It is well known that the only successful growth technique for the production ofcommercial nitrides for visible light sources is metal-organic chemical vapor deposition(MOCVD). Although molecular beam epitaxy (MBE) has many advantages with theepitaxial growth of various compound semiconductors and quantum structures, includingthe production of abrupt interfaces and sharp doping profiles and superior in situ growthmonitoring, it was not considered a promising alternative to MOCVD for producingnitride devices until the demonstration of the first pulsed laser diodes (LDs) with a400 nm emission wavelength grown by MBE.1 Since then, there has been significantprogress in MBE-grown nitride materials, nanostructures and related devices. Thisemerging growth technique has been used to create light sources operating at a widerspectral range and has generated new advanced devices. TwoMBE growth technologieshave been developed for nitride epitaxy: ammonia MBE and plasma-assisted MBE(PAMBE). The former MBE growth technique uses ammonia as the nitrogen precursor,while the latter uses plasma to atomize nitrogen gas. Significant progress in MBE-grownnitride devices has been demonstrated, but many challenges exist in the growth ofnitrides using MBE for high-performance devices, including high-quality bufferlayers, high-quality indium-rich InGaN alloys and their p-doping as well as n-dopingaluminum-rich AlGaN alloys. This chapter will review recent progress in the growth ofnitride materials and nanostructures usingMBE and the potential solutions to circumventthe challenges.

1.2 Molecular beam epitaxial (MBE) growth techniques

Ammonia MBE can be used to grow GaN at a high growth rate (overw1 mm/h) with afull width at half maximum (FWHM) of 540 arcsec as measured by X-ray diffraction.2

The GaN epilayers were initiated on a buffer layer grown by plasma-assisted MBE ata growth temperature of 500�C. Typical ammonia MBE growth requires a highammonia flow of w200 sccm, a high V/III flux ratio (of up to 103) that is NH3-richand a high growth temperature of 800e900�C, which is close to that of MOCVD.Hooper et al.1 produced the first room-temperature pulsed InGaN LD using ammonia

Nitride Semiconductor Light-Emitting Diodes (LEDs). https://doi.org/10.1016/B978-0-08-101942-9.00001-0© Woodhead Publishing Limited, 2014.

https://doi.org/10.1016/B978-0-08-101942-9.00001-0

MBE with an emission wavelength of 400 nm and a threshold current of 30 kA/cm2.Since then this group has continued to develop these lasers, producing the bestlasers using ammonia MBE, which emit at 405 nm and have a continuous-wave(CW) operation at room temperature with a threshold current of 3.6 kA/cm2, with amaximum CW output power of 45 mW per facet and a lifetime of 42 h.3 Althoughammonia MBE was used to produce the first nitride LD, it is not used by the majorityof the nitride MBE community due to the large consumption of ammonia, the highgrowth temperature, the corrosive nature of ammonia and the high hydrogen back-ground during the epitaxial process, which limit its extensive use for nitride growth.Sharp Laboratories of Europe Ltd is, perhaps, the only research group that is activein using ammonia MBE. PAMBE has become the technique that is used by themajority of the MBE community in nitride epitaxy.

Two plasma sources are used for PAMBE: electron-cyclotron resonance (ECR)4

and radio-frequency (RF) plasma.5 Molecular nitrogen is inert in MBE, but it canbe effectively cracked into reactive nitrogen species, that is neutral and chargedmolecular nitrogen (N2, N2

þ) and neutral and ionic atomic nitrogen (N, Nþ), andfree electrons. RF plasma sources are generally preferred to ECR plasma sources sincethey produce more neutral atomic nitrogen, which is favorable to the incorporation ofnitrogen during the epitaxial process. In addition, it has been shown that high qualitynitride materials are obtained from the metal-rich condition, with much lower growthtemperatures compared to ammonia MBE. This growth behavior has been extensivelyinvestigated theoretically and experimentally to understand the growth mechanisms.Theoretical work based on density-functional theory revealed the existence of anefficient lateral diffusion channel for adatoms on a semiconductor surface just belowthe thin metallic film.6 In particular, the activation energy of this so-called adlayer-enhanced lateral diffusion (AELD) is small and hence enables high-quality step-flowepitaxy at temperatures much lower than estimates based on the melting point of thematerial. Because of the low growth temperature, PAMBE has been identified as themain MBE growth technique for nitride materials.

1.3 Plasma-assisted MBE (PAMBE) growth of nitrideepilayers and quantum structures

There has been a worldwide effort to develop nitride materials, including InGaN andAlGaN alloys, and their quantum structures to produce LEDs and LDs operating atdifferent wavelengths, especially deep ultraviolet (d-UV), UV, visible light and whitelight. Such devices require aluminum-rich AlGaN or indium-rich InGaN alloys,relevant quantum structures and doping for the contacts. However, there are a numberof challenges including the degradation of materials with increasing indium content inInGaN alloys and the difficulty in obtaining p-doping, and the difficulty in obtainingn-doping in aluminum-rich AlGaN alloys as well as the high strain for specific wave-lengths. Enormous efforts have been made in growing nitrides using PAMBE tounderstand the growth mechanisms, to optimize the growth conditions and to createnew structures, which have resulted in various high-performance devices.

4 Nitride Semiconductor Light-Emitting Diodes (LEDs)

1.3.1 Gallium nitride (GaN) epilayers

MBE growth diagrams, also referred to as surface phase diagrams, are instrumental inproducing device-quality thin films. The GaN epilayers are generally grown on1e2 mm thick GaN templates grown by MOCVD on c-plane sapphire, which have anestimated dislocation density of 5 � 108 to 5 � 109 cm�2. The growth of GaNepilayers using PAMBE for different growth conditions has been studied using varioustechnologies, such as reflective high-energy electron diffraction (RHEED).7 RHEEDpatterns have been produced from the hexagonal GaN (0001) surfaces under differentgallium fluxes, where a high gallium flux is considered to be a gallium-rich condition,and a low one a nitrogen-rich condition. Fig. 1.1 shows typical RHEED patterns forGaN grown on c-plane sapphire under different conditions. The gallium-rich conditionproduced a streaky pattern suggesting two-dimensional (2D) growth, while thenitrogen-rich condition resulted in a spotty pattern indicating three-dimensional (3D)growth. This difference is explained by the different migration of excess species on thesurface between the nitrogen-rich and gallium-rich conditions. For the growth with aspotty RHEED pattern, the rough morphology can be recovered after the gallium shutteris closed. If the gallium flux is too high, then gallium droplets form on the surface, and thesurface recovery takes a longer time compared to nitrogen-rich growth, mainly due to theslow re-evaporation of excess gallium. Due to the similar atomic arrangement betweencubic (111) andhexagonal (0001) crystals, it is possible togrowcubic crystalline epilayerson hexagonal substrates. This was demonstrated by Okumura and coworkers at a lowgrowth temperature, which was far from the equilibrium growth conditions.7 The growthtemperature is dependent on the galliumflux; at a typical galliumfluxof 5 � 10�7 mbar, agrowth temperature less than640�Cwill lead to the growthof cubicGaN.A lower galliumflux requires a lower growth temperature.

Surface reconstruction was also studied to identify the optimal PAMBE growthconditions for GaN materials.8 The transition of 1 � 1 and 2 � 2 at different growthtemperatures and fluxes follows a curve as plotted in Fig. 1.2. It shows that a highergallium flux or lower growth temperature (upper left of a curve) yields the 1 � 1pattern, whereas a lower gallium flux or higher growth temperature gives rise to the2 � 2 pattern. This transition curve can be as a reference to determine the optimal

(a) (b)

Figure 1.1 RHEED patterns for GaN epilayers under: (a) a nitrogen-rich condition (galliumflux w4 � 10�7 Torr) and (b) a gallium-rich condition (gallium flux w5 � 10�7 Torr). Thesubstrate temperature, N2 flow and RF power were 800�C, 1.5 sccm and 400 W respectively.7

Molecular beam epitaxy (MBE) growth of nitride semiconductors 5

growth conditions. GaN epilayers grown at conditions close to the transition curve (onthe 2 � 2 side) have good carrier mobility, while epilayers grown in the regime of1 � 1 pattern yield gallium droplets.

Furthermore, atomic force microscopy (AFM)9e11 has been used to fi gure out thedependence of the morphology of the GaN epilayers on growth parameters. It wasconcluded that the gallium-rich condition is necessary for a smooth surface and resultsin strong photoluminescence (PL).9 Heying and co-workers12 established a detailedgrowth diagram for GaN using AFM and cross-sectional transmission electron micro-scopy (TEM). Three growth regimes are shown in Figs. 1.3 and 1.4 shows AFM

(1 × 1)

(2 × 2)

250

200

150

100

50

500 550 600 650 700 7500

350

300Fl

ux (1

012 a

tom

s/cm

2 .s

)

TS (°C)

N2 flow (ccm)6

4

21

φφGa

α N

Figure 1.2 Plot of Ga flux FGa versus growth temperature Ts showing the transition curves fordiffering N2 flow rates.8

20

18

16

14

12

Ga

flux

(nm

/min

)

550 600 650 700Substrate temperature (°C)

Ga dropletsGa stable

N-stable

IntermediateGa stable

C

B

A

Figure 1.3 Growth diagram showing gallium flux versus substrate temperature for thegallium-droplet, intermediate and nitrogen-stable growth regimes at a constant nitrogenflux of 2.8 nm/min.12


images of the surface morphology of the resulting GaN epilayers. They concluded thatgrowth within the nitrogen-stable regime results in a surface composed entirely ofinverted, pyramid-shaped pits initiated at threading dislocations (TDs). Epilayersgrown within the intermediate regime have flat surfaces between large pit featuresstabilized by low-angle grain boundaries. Films grown within the gallium dropletregime have atomically flat surfaces with no pit features. The reduction of the pitdensity was attributed to the change in the growth kinetics due to the increasingcoverage of the surface by metallic gallium.11

Due to the shortage of suitable substrates for nitride growth, nitride devices arenormally grown on free-standing GaN wafers, or sapphire or silicon substrates usinga thin layer of AlN as a nucleation layer to relax the strain (caused by lattice mismatch).A two-step growth procedure produces high-quality AlN epilayers with excellentcrystalline quality and a smooth surface.13 This procedure starts with a nitrogen-richgrowth condition then switches to aluminum-rich growth conditions. Faleev and

200 nm

200 nm

15 nm

7.5 nm

100 nm

100 nm

0 nm

0 nm

0 nm

1 μm

(a)

(b)

(c)

Figure 1.4 AFM images of GaN grown within (a) the nitrogen-stable regime (gallium flux:14.5 nm/min), (b) the intermediate regime (gallium flux: 15.8 nm/min) and (c) the gallium-droplet regime (gallium flux: 18.2 nm/min).12


co-workers developed the procedure for the growth of this buffer layer, which hasa low threading dislocation density of 1.75 � 105 cm�2, on sapphire.14 Lee and co-workers developed a novel buffer technique to produce free-standing GaN using athin MBE-grown zinc-polar ZnO layer on sapphire.15 This buffer layer was used toproduce a strain-free GaN epilayer with extensive photoluminescence indicating thepromising potential of nitrides grown on sapphire by MBE for advanced devices.

1.3.2 Aluminum nitride (AlN) epilayers

High-quality AlN epilayers grown by PAMBE have been obtained after acomprehensive study of growth.16 It was found that the III/V flux ratio and the growthtemperature are the critical parameters to achieve high-quality AlN layers in terms ofmorphology and crystalline quality. A III/V ratio close to stoichiometry and highgrowth temperatures (�900�C) lead to optimal AlN epilayers, which exhibit anFWHM of 10 arcmin in X-ray diffraction and an average surface roughness of48 Å. High-quality AlGaN alloys with an aluminum content ranging from 10% to76% were also obtained at an optimal growth temperature of 770�C.17 The aluminummole fraction has a linear dependence on the aluminum flux, indicating the ease ofcontrolling of the AlGaN alloy compositions with PAMBE.

1.3.3 Indium gallium nitride (InGaN) and indium nitride (InN)epilayers

InGaN alloys have attracted increasing interest due to their large tunability of bandgapenergy, high carrier mobility, superior light absorption and radiation resistance.However, it is still challenging to obtain high-quality InGaN alloys due to the largedifferences in bond energies and bond lengths, the large lattice mismatch as well asthe different thermal dissociation temperatures of GaN and InN. Since InGaN alloyshave a low thermal dissociation temperature, PAMBE is a good technique for growingindium-rich InGaN alloys. Kraus and co-workers investigated the incorporation ofindium during the PAMBE growth of InGaN.18 They found that the incorporationof indium is linearly dependent on the indium flux but inversely dependent on thegrowth temperature. It was thought that these behaviors were due to the differentsticking coefficients of gallium and indium, thermal desorption and the segregationof indium. They also found that indium incorporation is dependent on the growthregime. For the near stoichiometric regime, indium desorption is the dominant process.For the metal-rich regime, the effect of indium segregation is severe and produces ahigher indium content phase. This was observed in the growth of InGaN with anindium content of 25%. Extensive PL was observed for an InGaN epilayer grown at520�C with an indium content of 18.6%. InGaN epilayers with a higher indium content(>70%) were also obtained on InN templates using PAMBE at an optimal growthtemperature of 550�C. It was found that the quality of the materials is dependent onthe InN template, so a specific process to nitridate sapphire substrates was developedto produce an optimal InN buffer layer.19


1.3.4 Nitride-based InGaN/GaN multi-quantum wells (MQWs)

The growth of InGaN/GaN MQWs using MBE has been extensively studied forapplications for LEDs and LDs. Exceptional efficiency was achieved in InGaN/GaNMQW-based light-emitting devices despite a very high density of threading dis-locations. This was attributed to the carriers’ localization at the in-plane potentialfluctuation due to compositional inhomogeneities in the MQWs.20 The width of thequantum well, indium composition and growth temperature can affect the distributionof the indium composition in the wells, which modifies the optical properties of theInGaN/GaN QWs.21e23 In addition, the intrinsic electric field in InGaN/GaN MQWsalso plays an important role in emission efficiency and energy and was optimizedfor the emitting devices. However, realizing highly efficient long-wavelength visibleLEDs and LDs remains very challenging due to the significant drop of the internalquantum efficiency if the indium content is increased in the InGaN QWs.24 Forinstance, the performance of green LEDs is much lower than that of blue LEDs. Thiswas attributed to several factors, including indium segregation in the QWs, a piezo-electric field owing to the quantum confined Stark effect, the generation of dislocationsand the decomposition of the QWs in the doping layer during growth or post-growthannealing. Due to these challenges, the longest visible wavelength obtained for LDsis around 485 nm. The LD used was grown by MOCVD with an indium content of30% in the QWs.25 To overcome these problems, an attempt was made to growInGaN/InGaN MQWs using PAMBE. Siekacz et al. produced MQWs exhibiting PLat a wavelength of 510 nm.83 Furthermore, InN/InGaN MQWs have been obtainedusing PAMBE, and the PL was successfully extended into the near infrared, which isvery important for telecoms. For instance, PL at wavelengths of 1.55 mm26 and1.75 mm27 have been demonstrated for these MQWs. On the other hand, GaN/AlGaNMQWs have been developed and LEDs operating at a deep ultraviolet wavelength of273 nm have been obtained.28

1.3.5 Doping in nitride materials

When fabricating devices, it is essential to be able to introduce p-type and n-typematerials. Silicon and magnesium are typical dopants used widely for n-type andp-type GaN, respectively, and GaN-based devices, including LEDs and lasers, havebeen produced. However, doping is challenging for structures that require aluminum-rich AlGaN and indium-rich InGaN alloys. It is difficult to obtain p-type doping InGaNwith an indium content above 30% using magnesium.29 Recently carbon has receivedconsiderable interest as an acceptor because it has a similar atomic radius and electro-negativity as nitrogen. CBr4 has been shown to be an effective carbon source for GaNgrown by PAMBE30,31 and a doping level up to 1019 cm�3 was reported in PAMBE-grown GaN.32 The calculation of the ionization energy of carbon as an acceptor(substituting for nitrogen) and as a donor (substituting for gallium or indium) in InGaNalloys across the entire range of indium compositions revealed that carbon incorporationis more favorable when it acts as an acceptor (substituting for nitrogen) leading to p-typedoping.33 This theoretical study combined with experimental work indicates the great


potential of carbon as a dopant for realizing p-doping indium-rich InGaN alloys. ForAlGaN alloys, silicon has difficulty in producing n-type doping if the aluminum contentis above 49%, which is mainly due to the increased ionization energy as shown inFig. 1.5.34 A higher electron concentration can be obtained using MBE compared withMOCVD, and a doping concentration of 8 � 1018 cm�3 in AlGaN with 50% aluminumhas been demonstrated35; however, AlN doping is still a challenge, which impedesprogress in producing deep ultraviolet AlN LEDs.

1.3.6 Light emitters based on nitride MQWs

These impressive research efforts in the development of MBE-grown III-nitridematerials have resulted in LEDs and LDs with a wide spectral range from 273 to480 nm (see Table 1.1). Grandjean et al. reported UV LEDs grown by ammonia

700

600

500

400

300

200

100

0

AlGaN

Si dopedUndoped

Al mole fraction0.0 0.60.4 0.8 1.00.2

Act

ivat

ion

ener

gy (m

eV)

Figure 1.5 Activation energiesof silicon-doped (full circles)and nominally undoped (opencircles) AlxGa1 � xN. The linesare drawn as guides.34

Table 1.1 Some of the key developments and milestones achievedduring the development of MBE-grown LEDs and LDs

Device Group, Year

390 nm LEDs Grandjean et al., 199836

480 nm LEDs Waltereit et al., 200437

408 nm LDs Skierbiszewski et al., 200438

411 nm LDs Skierbiszewski et al., 200639

White light LEDs Damilano et al., 200840

273 nm LEDs Liao et al., 201128


MBE operating at a wavelength of 390 nm in 1998.36 In 2004, Waltereit et al. producedLEDs operating at 480 nm with an external efficiency >1.5%, which is close to that forLEDs grown by MOCVD.37 Skierbiszewski and co-workers have made a considerablecontribution to nitride LDs. In 2004, they produced blue-violet InGaN/GaN MQWsLDs operating at 408 nm, room temperature and pulsed operation with a thresholdcurrent of 12 kA/cm2 and a high output power of 0.83 W38; in 2006, they achievedroom temperature InGaN/GaN MQW LDs operating at a wavelength of 411 nm witha threshold current of 4.2 kA/cm2 and a high output power of 60 mW.39 Liao and co-workers produced LEDs operating at sub-300 nm wavelengths (275 nm) in 2011.28 Inaddition, white light LEDs have been proposed by monolithically stacking blue andyellow nitride MQWs; they have a reduced cost and increased efficiency comparedwith phosphor-based LEDs. Although such LEDs have been produced by MOCVD,the drawback of this structure is that the chromaticity coordinates strongly depend onthe injection current. Damilano et al. produced PAMBE-grown white LEDs using ayellow converter of five-period In0.2Ga0.8N (4 nm)/GaN (7.5 nm) MQWs followed bymonolithic-grown blue light MQWs.40 The InGaN and GaNwere grown at temperaturesof 550�C and 800�C, respectively. However, optimal light mixing conditions are stillnot obtainable in such conventional MQW structures.

1.4 Nitride nanocolumn (NC) materials

One-dimensional nanocolumns (NCs) are newly emerging materials and have attractedincreasing attention in the last few years due to a number of advantages. NC materialshave large aspect ratios and 3D stress relief mechanisms leading to dislocation-freestructures.41 Their small footprint helps to release strain and thermal expansion. Inlight-emitting device applications, these structures have a high light extraction efficiency.The light emission colors can be tuned from blue to red bymodifying the NC diameter.42

Furthermore, core-shell NC structures are obtainable, which supress the strong surfacerecombination and improve the efficiency of the light-emitting devices.43 Consequently,nitride NC materials are a promising candidate for a breakthrough development inlight-emitting devices.

1.4.1 Self-catalyst growth of GaN NCs using MBE

Nitride NCs can be grown by MBE using catalytic or self-catalytic (catalyst-free)methods. The growth method using a catalyst is called the vapor-liquid-solid (VLS)mechanism and is generally adopted in chemical vapor deposition (CVD). Metalssuch as gold,44 nickel45 and molybdenum46 have been used as the catalyst. In MBE,nitride NCs are more generally grown without a metal catalyst. The earliest reportof the growth of GaN NCs using PAMBE was from Sophia University, Tokyo,47

and then a group from Ciudad University, Madrid, also produced GaN NCs grownby PAMBE.48 Both groups reported that the nitrogen-rich growth condition leads tothe formation of GaN NCs in PAMBE. Since then the best growth conditions forGaN NCs have been explored and optimized to achieve high-quality NCs withcontrollable geometry, aspect ratio and area density. The V/III flux ratio is the crucial


factor for producing NCs of GaN,48 AlN49 as well as InN.50 Fig. 1.6 shows the typicalmorphology of GaN NC materials grown under nitrogen-rich condition (a) and a GaNcompact epilayer grown under gallium-rich condition (b). These observations wereinterpreted as being due to a mechanism where the high V/III flux ratio reduces thediffusion distance of gallium adatoms and supresses the coalescence of nucleationsites.51 It was consequently concluded that VLS is not the growth mechanism inPAMBE-grown NCs, instead, the growth of NCs is driven by a process that involvesnucleation and the diffusion of adatoms on the surface:

1. GaN precursor islands nucleate on the surface. Such islands are plastically relaxed and resultfrom different growth modes. For GaN NCs grown on an AlN buffer layer, the nucleation ofGaN islands is driven by the StranskieKrastanow (SK) growth mode; for GaN NCs grownon a different substrate, such as bare oxidized52 or nitridated51 silicon substrates, the islandsform due to the VolmereWeber (VW) growth mode.

2. GaN islands develop with further deposition and then initiate the growth of NCs when theyreach a critical size.

3. The growth of NCs strongly depends on two contributions: one is growth due to the directimpinging of atoms on an NC apex. The other contribution occurs when adatoms arriveon the surface. They diffuse to the base of an NC then climb up along the lateral sidewallsof the NC to the apex and become incorporated into the crystal.

Fig. 1.7 shows these growth processes. Furthermore, Debnath et al.53 suggested thatadsorption at the tip of an NC would have a signifi cant effect on an NC with a thickdiameter, and this was used to interpret the observation that an ensemble of NCs isnormally a mixture of short and thick NCs with long and thin ones.

High-quality GaN NCs have been successfully grown on a few different substratesincluding silicon (111),54 sapphire (0001)55 and silicon (001).56 Although there is a biglattice mismatch, the PAMBE-grown NCs on these substrates demonstrate fullyrelaxed epitaxial growth resulting in strain-free, dislocation-free and extendeddefect-free high-quality single crystals, which have excellent optical properties.57e59

For the NCs grown on silicon substrates, there is a standard procedure to desorb thenative oxide54: the silicon substrate is first etched by diluted HF, then it is transferredinto the MBE system followed by thermal treatment in a vacuum at 800�C, which

(a) (b)

Figure 1.6 Scanning electron micrographs (SEMs) of GaN NCs grown directly on Si (111)substrates under (a) nitrogen-rich and (b) gallium-rich conditions.48


ensures an excellent surface condition and a 7 � 7 RHEED pattern is visible. Threedifferent procedures have been used to start the growth of NCs: on a buffer layer ofGaN grown at low temperatures of 500e600�C,54 on a buffer layer of AlN17 andthe direct growth of NCs on a bare silicon substrate.57,60 The growth of the GaNNCs follows at a typical temperature of 720�C. The influence of the growth conditionson NC geometry, such as lateral dimension, height and number density, has beencomprehensively investigated. It was reported that the tuneable range can befrom <20 nm to w800 nm for the lateral dimension, from 50 nm to 3 mm for theheight, and of the order of w106e107 cm�2 for the area number density.54 UniformGaN NCs with a narrow lateral dimension variation of 20e40 nm on silicon (001)have been produced56; these NCs are strain-free and exhibit intense narrow excitonicPL indicating the high quality.

Besides these hexagonal GaN (h-GaN)NCs, cubic GaN (c-GaN)NCs have also beenobtained by PAMBE on a silicon (111) substrate.61 After the thermal de-oxidization ofsilicon, a thin AlN layer was grown to improve the orientation of the GaN NCs. h-GaNNCswere then grown in a nitrogen-rich condition at a growth temperature of 850�C.Thesubstrate temperature was then reduced to 580�C to grow a c-GaNNC section using thesame nitrogen and gallium fluxes. A PL emission peak at 3.27 eV was observed andattributed to the band-edge transition from the c-GaN NC section. Due to the highhomogeneity of c-GaN, this PL linewidth is narrow and makes the donor-bound

Desorbedatoms

ImpingingGa beam

(c) VW nucleus

(b)

Criticalnucleus

(a)

Sub-criticalnucleus

Adsorbedatoms

Figure 1.7 Growth of NCs from stable nuclei including direct incorporation from the impinginggallium flux (j(L)), and growth where gallium diffuses on the substrate (j(D)) to the base of anNC and up to the apex.51


exciton visible. The realization of a high quality c-GaN section on h-GaN NCs couldlead to novel devices.

1.4.2 Aluminum gallium nitride (AlGaN) NCs

In addition to the success in growing GaN NC materials using PAMBE, wide bandgapAlGaN NCs have been attempted. Landré et al. grew AlN NCs using catalyst-freePAMBE and described the structural and optical properties.62 The AlN NCs were grownon a substrate of silicon (100) covered with a thin layer of a few monolayers of SiO2.The aluminum/nitrogen flux ratio was fixed to about 1/6, to ensure nitrogen-richconditions. The growth temperature was in the range 900e950�C, which is higherthan that for GaN NCs to compensate for the low diffusion rate of aluminum comparedto gallium. The resulting AlN NCs were assessed by high-resolution TEM, Ramanspectroscopy and PL measurements. It was observed that the NCs were completelystrain relaxed, which is thought to be because they match the SiO2/Si. PL with an energyof 6.04 eV at 10K was observed, dominated by near-band edge emission. It was alsofound that there was no wetting layer at the base of the NCs, suggesting that the growthobeys a VW mode. AlGaN NCs with an aluminum content up to 30% have beenproduced.63,64 It was found that the incorporation of aluminum increases the diameterof the resulting NCs. In addition, it was observed that beryllium-doping led to largevertical GaN ribbons or GaN NCs with a slightly conical shape.63 These observationsindicated that surface diffusion has a strong effect on the growth of NCs.

1.4.3 InN and InGaN NCs

Due to the narrow bandgap energy of InN (0.6 eV) and the high carrier mobility, thegrowth of InN NCs by PAMBE has been extensively studied and optimized.65,66 Thegrowth temperature and the V/III flux ratio are the major factors that determine theresulting InN structures. Hsiao et al. grew InN structures on an InN buffer layer andan AlN buffer layer. A fixed indium flux of 2.5 � 10�7 Torr was used with variousnitrogen/indium flux ratios at various growth temperatures.65 They found that mate-rials grown at a temperature above 530�C and nitrogen/indium flux ratio of 40 weregrain-like InN, but a lower temperature (450e500�C) led to InN NCs, while InNgrown at temperatures of 450�C with a nitrogen/indium flux ratio of 20 producedan epilayer. Chang et al.67 developed a new PAMBE technique for growing high-quality InN NCs on a silicon (111) substrate using an in situ indium seeding layer.This growth technique produced well-separated and uniform InN NCs, which werenearly defect free and not tapered, as shown in Fig. 1.8. The PL has a narrow spectrallinewidth of 13 meV. In addition, a low residual carrier concentration ofw1 � 1016 cm�3 was achieved in such InN NCs. These achievements indicate thegreat potential of this technique for device-quality InN NCs. The growth temperatureis also a major factor for modifying the morphology of InN materials grown with thistechnique. Fig. 1.9 shows InN materials grown in a nitrogen-rich condition at differenttemperatures. X-ray diffraction measurements indicate that the resulting materials aremono-crystalline.


1.4.4 Overgrowth of nitride NCs

The overgrowth of high-quality nitride epilayers on GaN NCs has been achieved. Inparticular, these overgrown epilayers were reported to be strain free and dislocationfree as discovered for a 2.7-mm-thick overgrown GaN epilayer on GaN NCs.68 Asapphire substrate was etched by a mixture of H2SO4 and H3PO4 at 110�C for30 min followed by the deposition of titanium onto the back, then it was loaded intothe MBE system and thermally cleaned at 950�C for 30 min. Growth started afternitridation at 750�C for 20 min. An AlN buffer layer was grown at 850�C with anitrogen flow of 5.3 sccm and RF power of 450 W. After this, the GaN NCs startedto grow at a growth rate of 2.3 um/h under a nitrogen-rich condition. Then the growth

1 µm

1 µm

McGill 8.0 kV 13.0 mm x 30.0 k SE(U)

Figure 1.8 SEM image of a single non-tapered InN NC and an InN NC ensemble grown on anSi (111) substrate (inset).67

(a) (b)

Figure 1.9 SEM images of InN grown at different temperatures: (a) InN grains grown at 550�Cand (b) short InN NCs grown at 450�C.


condition was changed to a gallium-rich regime, leading to the direct growth of a GaNepilayer on the NCs. Fig. 1.10 shows a scanning electron micrograph of the resultingfree-standing GaN epilayer grown on the GaNNCs on a sapphire substrate. This methodis a new route for fabricating high-quality dislocation-free GaN epilayers on foreignsubstrates, and is very important to the nitride community.

1.5 Nitride nanostructures based on NCs

1.5.1 Quantum disks embedded in NCs

GaN quantum disks embedded in AlGaN NCs and AlN/GaN Bragg reflectors wereproduced by Risti�c et al.69 Such structures are extremely attractive for single photonsources. The nanocavities formed in these NCs are crack free and defect free, whichwas attributed to relaxation at the silicon interface and the high aspect ratio. The PLwas tuned by modifying the thickness of the GaN quantum disks and the aluminiumcontent of the AlGaN. PL emissions at peak energy varied from 3.4 to 4.0 eV.

Recently, InGaN/GaN dots-in-NCs have attracted increasing attention due to anumber of unique properties including significantly reduced threading dislocationdensities, suppressed polarization fields, enhanced light extraction efficiency, thecapability of accommodating InGaN dots with a larger indium content, as well asthe capacity for monolithic integration with large area and low-cost silicon substrates.Consequently such nanostructures have great potential in producing advanced LEDscompared with conventional thin-film MQWs. It has been a big challenge to realizeGaN-based LEDs operating in the red spectral range due to the huge lattice mismatch.Mi et al. explored the growth of such nanostructures using PAMBE. Red LEDs basedon such nanostructures operating at room temperature were obtained with an internalefficiency of up to 30%.70 They are very important for realizing phosphor-free solid-state lighting and full color displays. Full-color white light InGaN/GaN dots-in-NCsLEDs with an internal efficiency of 56.8% have been produced.71e73 Fig. 1.11 is aschematic of a core-shell NC (CSNC) white light LED. The emission of white lightis realized by a stack of InGaN nanodisks (NDs) embedded in a GaN nanorod p-njunction. To optimize the light mixing effect, different thicknesses (10e25 nm) and

1 µm

Free-standingGaN

Nano columns

Al2O3

Figure 1.10 A cross-sectionalSEM image of overgrown GaNon NCs on a sapphire substrate.68


indium content in the InGaN nanodisks were obtained through varying the growthtemperature and indium/gallium flux ratio.73 The number and positions of the nano-disks are important for obtaining the appropriate light mixing conditions. Additionally,in such LED devices, an indium content up to 50% has been obtained in the InGaNdots. These results clearly indicate the advantages of 1D NC materials for realizingoptimal light mixing conditions.

1.5.2 Core-shell NCs

It is well known that NC materials have a large number of surface states leading tosignificant surface recombination, which reduces carrier mobility and radiativerecombination efficiency in NCs. Consequently, reducing the number of surface statesis needed before the advantages of NCs can be exploited to produce highly efficientNC devices. Core-shell NCs (CSNCs) are a promising class of semiconductor nano-structures. The effect of surface states is suppressed resulting in a significant potentialfor optoelectronic device applications. Hestroffer et al. produced GaN/AlN CSNCsusing PAMBE.74 Growth started from GaN NCs under a nitrogen-rich conditionand a growth temperature of w850�C followed by deposition of AlN onto the top,which resulted in the formation of AlN shell layers around the GaN core NCs. Growthwas attributed to the significant lateral growth of AlN due to the limited diffusion ofaluminium on the NW facets (see references in Nguyen et al.74). GaN/AlGaN CSNCshave been produced with a wide tuneable emission spectral range of 280e400 nm.75

To obtain such CSNCs, 200 nm-high GaN core NCs were first grown at a temperatureof 750�C and a growth rate of 3 nm/min, then AlGaN shells were grown at atemperature of 800�C and a growth rate of 1.5 nm/min. The aluminium content inthe AlGaN shell layers was controlled by varying the aluminium/gallium flux ratio.

Ni/Au

Ti/Au

Si(111)

T3

T2

T1

p-GaN

5 InGaN NDs

n-GaN

Figure 1.11 InGaN/GaN CSNC whitelight LED: the active regions containmultiple InGaN nanodisks with variousthicknesses and indium content.73


The nitrogen plasma was maintained in a nitrogen-rich condition: an N2 flow of 1 sccmand an RF power of 350 W. A high internal efficiency up to 58% from such CSNCshas been realized. In addition, InN/InGaN CSNCs have also been produced.76 InNcore NCs were first grown at a growth temperature of 470�C and a growth rate of3.3 nm/min (a 0.5 nm-thick indium seeding layer was used to initiate NCs growth),then the InGaN shell layers were produced at a growth temperature of 500�C andgallium and indium beam equivalent pressure fluxes of 6.4 � 10�8 and 1 � 10�8 Torr,respectively. A high internal quantum efficiency of 62% for PL at room temperaturewas obtained and attributed to strong carrier confinement and the nearly intrinsicInN core NCs. These studies reveal the advantages for LED and other optoelectronicdevice applications.

1.5.3 Selective area growth of NCs

Although there has been significant progress in PAMBE-grown nitride NCs in the lastfew years, controlling the size, position and geometry of NCs as well as the compositionof the alloys still remains a significant challenge. This causes difficulties in the control ofemission wavelengths and material processing of NC devices. Consequently, producingcontrollable NCs with uniform height is highly demanding. Selective area growth onpatterned substrates has been proposed for circumventing the difficulties to produce or-dered and uniform NCs. In this approach, NCs only grow on regions of the patternedsubstrate with pores. Various mask materials have been considered including SiO2/GaN,77 Ti/GaN78e80 and SiNx/AlN.

81 The first selective area growth of GaN NCswas demonstrated by Kishino and co-workers on Ti/GaN/sapphire templates.78,80 Theeffect of the growth parameters on the morphologies of the resulting materials hasbeen studied. Growth temperature is a critical factor for NCs: a growth temperature inthe range 880e900�C produced NCs. There was no selective area growth at lowertemperatures because there is insufficient diffusion; however, higher temperaturesincreased the inhomogeneity of the geometry of the NCs and decreased the diameterof the NCs due to enhanced gallium desorption and diffusion.80 They also reportedthat the V/III flux ratio dramatically affected the geometry of the resulting NCs. Ahigh flux ratio, i.e., a nitrogen-rich condition, increased the height and diameter of theNCs.78 The effect of the aperture diameter of the pattern on the geometry of theresulting NCs was systematically investigated on NCs grown on SiNx/AlN templates.81

Single NCs were obtained for a pattern with an aperture diameter �500 nm. A largeropening produced coalesced nanostructures with rough and faceted tops.

Regularly aligned NCs in a 2D array leads to the photonic crystal effect, which canbe used to enhance and tune the light emission from an NC ensemble. Kouno et al.82

produced a rectangular array of GaN NCs consisting of eight-period InGaN/GaNMQWs. The 2D array had a horizontal lattice constant of 230 nm and a vertical latticeconstant of 245 nm. The emission of a specific wavelength was enhanced in such 2DNCs array as shown in Fig. 1.12. The figure clearly shows that the experimentalspecific peak emission at a wavelength of 471 nm coincides with the calculatedspectrum based on a 2D finite-difference time domain (2D-FDTD) method using theassumption of a refractive index dispersion of GaN. This observation opens a newroute for tuning the light emitted from NC arrays.


1.6 Conclusion

In summary, MBE has demonstrated its capacity for producing high-quality nitridematerials. A variety of devices have been realized with performances comparable tothose grown by MOCVD. This growth technique has also exhibited its advantages inproducing ultraviolet and near infrared LEDs. These achievements indicate the greatpotential ofMBE for producing commercial devices.More importantly, the NCmaterialsgrown by MBE have seen significant progress in the last few years. It is proposed thatMBE is a very promising route for producing high-performance solid-state light-emitting devices.

References

1. Hooper SE, Kauer M, Bousquet V, Johnson K, Barnes JM, et al. InGaN multi quantum welllaser diodes grown by molecular beam epitaxy. Electron Lett 2004;40:33e4.

2. Yang Z, Li LK, WangW. GaN growth by molecular beam epitaxy at high growth temperaturerates using ammonia as the nitrogen source. Appl Phys Lett 1995;67:1686e8.

3. Tan WS, Kauer M, Hooper SE, Barnes JM, Rossetti M, et al. High-power and long-lifetimeInGaN blueeviolet laser diodes grown by molecular beam epitaxy. Electron Lett 2008;44:351e2.

4. Ohtani A, Stevens KS, Beresford R. Microstructure and photoluminescence of GaN grownon Si(111) by plasma-assisted molecular beam epitaxy. Appl Phys Lett 1994;65:61e3.

5. Yang Z, Guarin F, Tao IW, Wang WI, Iyer SS. Approach to obtain high quality GaN on Siand SiC-on-silicon-on-insulator compliant substrate by molecular-beam epitaxy. J Vac SciTechnol B 1995;13:789e91.

6. Neugebauer J, Zywietz TK, Scheffler M, Northrup JE, Chen H, et al. Adatom kinetics onand below the surface: the existence of a new diffusion channel. Phys Rev Lett 2003;90:056101.

(a) (b)Room-temperatureHe-Cd laser

526.8 nm

325 nm, 1.0 mW

471.3 nm

RT-

PL

inte

nsity

(a.u

.)

400 450 500 550 600 400 450 500 550 600Wavelength (nm) Wavelength (nm)

Opt

ical

resp

onse

(a.u

.)

465.2 nm

Calculationby 2D-FDTD methodTE-mode response

Figure 1.12 (a) Low-excitation room-temperature PL (RT-PL) spectra of eight-period InGaN/GaNMQWs embedded on top of GaN rectangular-lattice NCs, and (b) light response spectrumcalculated by a 2D-FDTD method.82


7. Okumura H, Balakrishnan K, Hamaguchi H, Koizumi T, Chichibu S, et al. Analysis of MBEgrowth mode for GaN epilayers by RHEED. J Cryst Growth 1998;189/190:364e9.

8. Hacke P, Feuillet G, Okumura H, Yoshida S. Monitoring surface stoichiometry with the(2 � 2) reconstruction during growth of hexagonal-phase GaN by molecular beam epitaxy.Appl Phys Lett 1996;69:2507.

9. Yu Z, Buczkowski SK, Giles NC, Myers TH, Richards-Babb MR. The effect of atomichydrogen on the growth of gallium nitride by molecular beam epitaxy,. Appl Phys Lett 1996;69:2731.

10. Tarsa EJ, Heying B, Wu XH, Fini P, DenBaars SP, et al. Homoepitaxial growth of GaNunder Ga-stable and N-stable conditions by plasma-assisted molecular beam epitaxy. J ApplPhys 1997;82:5472.

11. Heying B, Tarsa EJ, Elsass CR, Fini P, DenBaars SP, et al. Dislocation mediated surfacemorphology of GaN. J Appl Phys 1999;85:6470.

12. Heying B, Averbeck R, Chen LF, Haus E, Riechert H, et al. Control of GaN surfacemorphologies using plasma-assisted molecular beam epitaxy. J Appl Phys 2000;88:1855.

13. Falth JF, Davidsson SK, Liu XY, Anderson TG. Influence of Al/N flux ratio duringnucleation layer growth on the structural properties of AlN grown on sapphire by molecularbeam epitaxy. Appl Phys Lett 2005;87:161901.

14. Faleev N, Li H, Schaff WJ. Low density of threading dislocations in AlN grown onsapphire. J Appl Phys 2007;101:093516.

15. Lee SW, Minegishi T, Lee WH, Goto H, Lee HJ, et al. Strain-free GaN thick films grown onsingle crystalline ZnO buffer layer with in situ lift-off technique. Appl Phys Lett 2007;90:061907.

16. Calleja E, Sanchez-Garcia MA, Monroy E, Sanchez FJ, Munoz E, et al. Growth kinetics andmorphology of high quality AlN grown on Si(111) by plasma-assisted molecular beamepitaxy. J Appl Phys 1997;82:4681.

17. Calleja E, et al. Growth of III-nitrides on Si(111) by molecular bema epitaxy doping,optical, and electrical properties. J Cryst Growth 1999;201/202:296.

18. Kraus A, Hammadi S, Hisek J, Bub R, Jonen H, et al. Growth and characterization of InGaNby RF-MBE. J Cryst Growth 2011;323:72.

19. Naoi H, Kurouchi M, Muto D, Araki T, Miyajima T, et al. Growth of high quality In-richInGaN alloys by RF-MBE for the fabrication of InN-based quantum well structures. J CrystGrowth 2006;288:283.

20. Chichibu S, Azuhata T, Sota T, Nakamura S. Spontaneous emission of localized excitons inInGaN single and multiquantum well structures. Appl Phys Lett 1996;69:4188.

21. Stevens M, Bell A, McCartney MR, Ponce FA, Marui H, et al. Effect of layer thickness onthe electrostatic potential in InGaN quantum wells. Appl Phys Lett 2004;85:4651.

22. Lin Y-S, Ma K-J, Hsu C, Feng S-W, Cheng Y-C, et al. Dependence of compositionfluctuation on indium content in InGaN/GaN multiple quantum wells. Appl Phys Lett 2000;77:2988.

23. Schulze F, Blasing J, Dadgar A, Krost A. Time-delayed indium incorporation in ultrathin(InxGa1 � xN/GaN) multiple quantum wells grown by metaorganic vapor phase epitaxy.Appl Phys Lett 2003;82:4558.

24. Wetzel C, Detchprom T. Development of high power green light emitting diode chips.MRSInternet J Nitride Semicond Res 2005;10:2.

25. Kim KS, Son JK, Lee SN, Sung YJ, Paek HS, et al. Characteristics of long wavelengthInGaN quantum well laser diodes. Appl Phys Lett 2008;92:101103.


26. Grandal J, Pereiro J, Bengoechea-Encabo A, Fern�andez-Garrido S, S�anchez-García MA,et al. InN/InGaN multiple quantum wells emitting at 1.5 mm grown by molecular beamepitaxy. Appl Phys Lett 2011;98:061901.

27. Che S-B, Ishitani WTY, Yoshikawa A, Matsuda T, et al. Fine-structure N-polarity InN/InGaN multiple quantum wells grown on GaN underlayer by molecular beam epitaxy. ApplPhys Lett 2005;86:261903.

28. Liao Y, Thomidis C, Kao C, Moustakas TD. AlGaN based deep ultraviolet light emittingdiodes with high internal quantum efficiency grown by molecular beam epitaxy. Appl PhysLett 2011;98:081110.

29. Dahal R, Pantha B, Li J, Lin JY, Jiang HX. InGaN/GaN multiple quantum well solar cellswith long operating wavelengths. Appl Phys Lett 2009;94:063505.

30. Green DS, Mishra UK, Speck JS. Carbon doping of GaN with CBr4 in radio-frequencyplasma-assisted molecular beam epitaxy. J Appl Phys 2004;95:8456.

31. Poblenz C, Waltereit P, Rajan S, Heikman S, Mishra UK, et al. Effect of carbon doping onbuffer leakage in AlGaN/GaN high electron mobility transistors. J Vac Sci Technol B 2004;22(3):1145.

32. Zado A, Tschumak E, Gerlach JW, Lischka K, As DJ. Carbon as an acceptor in cubic GaN/3C-SiC. J Cryst Growth 2011;323:88.

33. Tablero C. Ionization energy levels in C-doped InxGa1 � xN alloys. Appl Phys Lett 2010;97:192102.

34. Zeisel R, Bayerl MW, Goennenwein STB, Dimitrov R, Ambacher O, et al. DX-behavior ofSi in AlN. Phys Rev B 2000;61:R16283.

35. Ahoujja M, McFall JL, Yeo YK, Hengehold RL, Van Nostrand JE. Electrical and opticalinvestigation of MBE grown Si-doped AlxGa1 � xN as a function of Al mole fraction up to0.5. Mater Sci Eng 2002;B91e92:285.

36. Grandjean N, Massies J, Leroux M, Lorenzini P. Ultraviolet GaN light-emitting diodesgrown by molecular beam epitaxy using NH3. Appl Phys Lett 1998;72:82.

37. Waltereit P, Sato H, Poblenz C, Green DS, Brown JS, et al. Blue GaN-based light-emittingdiodes grown by molecular-beam epitaxy with external quantum efficiency greater than1.5%. Appl Phys Lett 2004;84:2748.

38. Skierbiszewski C, Wasilewski ZR, Siekacz M, Feduniewicz A, Perlin P, et al. Blue-violetInGaN laser diodes grown on bulk GaN substrates by plasma-assisted molecular-beamepitaxy. Appl Phys Lett 2004;86:011114e6.

39. Skierbiszewski C, Wi�sniewski P, Siekacz M, Perlin P, Feduniewicz-Zmuda A, et al. 60 mWcontinuous-wave operation of InGaN laser diodes made by plasma-assisted molecular-beamepitaxy. Appl Phys Lett 2006;88:221108.

40. Damilano B, Dussaigne A, Brault J, Huault T, Natali F, et al. Monolithic white lightemitting diodes using a (Ga,In)N/GaN multiple quantum well light converter. Appl PhysLett 2008;93:101117.

41. Zubia D, Hersee SD. Nanoheteroepitaxy: the application of nanostructuring and substratecompliance to the heteroepitaxy of mismatched semiconductor materials. J Appl Phys 1999;85:6492.

42. Sekiguchi I, Kishino K, Kikuchi A. Emission color control from blue to red with nanocolumndiameter of InGaN/GaN nanocolumn arrays grown on same substrate. Appl Phys Lett 2010;96:231104.

43. Waag A, Wang X, F€undling S, Ledig J, Erenburg M, et al. The nanorod approach: GaNNanoLEDs for solid state lighting,. Phys Status Solidi C 2011;8:2296.

44. Liu B, Bando Y, Tang C, Xu F, Golberg D. Quasi-aligned single-crystalline GaN nanowirearrays. Appl Phys Lett 2005;87:073106.


45. Li Q, Wang GT. Improvement in aligned GaN nanowires growth using submonolayer Nicatalyst films. Appl Phys Lett 2008;93:043119.

46. Shi F, Li H, Xue C. GaN nanorods catalyzed with Mo: effect of ammoniating time onmicrostructure, morphology, and optical properties. J Mater Sci Mater Electron 2010;21:1249.

47. Yoshizawa M, Kikuchi A, Mori M, Fujita N, Kishino K. Growth of self-organized GaNnanostructures on Al2O3(0001) by RF-radical source molecular beam epitaxy. Jpn J ApplPhys 1997;36(Part 2):L459.

48. Sanchez-Garcia M, Calleja E, Monroy E, Sanchez FJ, Calle F, et al. The effect of the III/Vratio and substrate temperature on the morphology and properties of GaN-and AlN-layersgrown by molecular beam epitaxy on Si(111),. J Cryst Growth 1998;183:23.

49. Karmann S, Schenk HPD, Kaiser U, Fissel A, Richter W. Growth of columnar aluminumnitride layers on Si(111) by molecular beam epitaxy. Mater Sci Eng B 1997;50:228.

50. Stoica T, Meijers R, Calarco R, Richter T, L€uth H. MBE growth optimization of InNnanowires. J Cryst Growth 2006;290:241.

51. Ristic J, Calleja E, Fern�andez-Garrido S, Cerutti L, Trampert A, et al. On the mechanisms ofspontaneous growth of III-nitride nanocolumns by plasma-assisted molecular beam epitaxy.J Cryst Growth 2008;310:4035.

52. Stoica T, Sutter E, Meijers RJ, Debnath RK, Calarco R, et al. Interface and wetting layereffect on the catalyst-free nucleation and growth of GaN nanowires. Small 2008;4:751.

53. Debnath RK, Meijers R, Richter T, Stoica T, Calarco R, et al. Mechanism of molecularbeam epitaxy growth of GaN nanowires on Si(111). Appl Phys Lett 2007;90:123117.

54. Tu LW, Hsiao CL, Chi TW, Lo I, Hsieh KY. Self-assembled vertical GaN nanorods grownby molecular beam epitaxy. Appl Phys Lett 2003;82:1601.

55. Calleja E, Sanchez-Garcia MA, Sanchez FJ, Calle F, Naranjo FB, et al. Luminescenceproperties and defects in GaN nanocolumns grown by molecular beam epitaxy. Phys Rev B2000;62:16826.

56. Cerutti L, Risti J, Fern�andez-Garrido S, Calleja E, Trampert A, et al. Wurtzite GaNnanocolumns grown on Si(001) by molecular beam epitaxy. Appl Phys Lett 2006;88:213114.

57. Park YS, Park CM, Fu DJ, Kang TW, Oh JE. Photoluminescence studies of GaN nanorodson Si (111) substrates grown by molecular-beam epitaxy. Appl Phys Lett 2004;85:5718.

58. Yi SN, Na JH, Lee KH, Jarjour AF, Taylor RA, et al. Photoluminescence properties of asingle GaN nanorod with GaN/AlGaN multilayer quantum disk. Appl Phys Lett 2007;90:101901.

59. Chen H-Y, Lin H-W, Shen C-H, Gwo S. Structure and photoluminescence properties ofepitaxially oriented GaN nanorods grown on Si(111) by plasma-assisted molecular-beamepitaxy. Appl Phys Lett 2006;89:243105.

60. Calarco R, Meijers RJ, Debnath RK, Stoica T, Sutter E, et al. Nucleation and growth of GaNnanowires on Si(111) performed by molecular beam epitaxy. Nano Lett 2007;7:2248.

61. Renard J, Tourbot G, Sam-Giao D, Bougerol C, Daudin B, et al. Optical spectroscopy ofcubic GaN in nanowires. Appl Phys Lett 2010;97:081910.

62. Landré O, Fellmann V, Jaffrennou P, Bougerol C, Renevier H, et al. Molecular beamepitaxy growth and optical properties of AlN nanowires. Appl Phys Lett 2010;96:061912.

63. Bertness KA, Roshko A, Sanford NA, Barker JM, Davydov AV. Spontaneously grownGaN and AlGaN nanowires. J Cryst Growth 2006;287:522.

64. Risti�c J, S�anchez-García MA, Calleja E, Sanchez-P�aramo J, Calleja JM, et al. AlGaNNanocolumns grown by molecular beam epitaxy: optical and structural characterization.Physica Status Solidi A 2002;192:60.


65. Hsiao C-L, Tu L-W, Chen M, Jiang Z-W, Fan N-W, et al. Polycrystalline to single-crystalline InN grown on Si(111) substrates by plasma-assisted molecular-beam epitaxy.Jpn J Appl Phys 2005;44(Part 2):L1076.

66. Chang Y-L, Li F, Mi Z. Optimization of the structural and optical quality of InN nanowireson Si(111) by molecular beam epitaxy. J Vac Sci Technol B 2010;28:C3B7.

67. Chang Y-L, Mi Z, Li F. Photoluminescence properties of a nearly intrinsic single InNnanowire. Adv Funct Mater 2010;20:4146e51.

68. Kusakabel K, Kikuchi A, Kishino K. Overgrowth of GaN layer on GaN nano-columns byRF-molecular beam epitaxy. J Cryst Growth 2002;237e239:988.

69. Risti�c J, Calleja E, Trampert A, Fern�andez-Garrido S, Rivera C, et al. Columnar AlGaN/GaN nanocavities with AlN/GaN Bragg reflectors grown by molecular beam epitaxy onSi(111). Phys Rev Lett 2005;94:146102.

70. Nguyen HPT, Zhang S, Cui K, Korinek A, Botton GA, et al. High efficiency InGaN/GaNdot-in-a-wire red light emitting diodes. IEEE Photon Tech Lett 2012;24:321.

71. Nguyen HPT, Cui K, Zhang S, Fathololoumi S, Mi Z. Full-color InGaN/GaN dot-in-a-wirelight emitting diodes on silicon. Nanotechnol 2011;22:445202.

72. Nguyen HPT, Zhang S, Cui K, Han X, Fathololoumi S, et al. p-type modulation dopedInGaN/GaN dot-in-a-wire white light emitting diodes monolithically grown on Si(111).Nano Lett 2011;11:1919.

73. Lin H-W, Lu Y-J, Chen H-Y, Lee H-M, Gwo S. InGaN/GaN nanorod array white light-emitting diode. Appl Phys Lett 2010;97:073101.

74. Hestroffer K, Mata R, Camacho D, Leclere C, Tourbot G, et al. The structural properties ofGaN/AlN coreeshell nanocolumn heterostructures,. Nanotechnol 2010;21:415702.

75. Wang Q, Nguyen HPT, Cui K, Mi Z. High efficiency ultraviolet emission from AlxGa1 � xNcore-shell nanowire heterostructures grown on Si(111) by molecular beam epitaxy. ApplPhys Lett 2012;101:043115.

76. Cui K, Fathololoumi S, Golam Kibria M, Botton GA, Mi Z. Molecular beam epitaxialgrowth and characterization of InN/InxGa1 � xN core/shell nanowire heterostructures onSi(111) substrates. Nanotechnol 2012;23:085205.

77. Hersee SD, Sun XY,Wang X. The controlled growth of GaN nanowires,. Nano Lett 2006;6:1808.

78. Kishino K, Sekiguchi S, Kikuchi A. Improved Ti-mask selective-area growth (SAG) byrf-plasma assisted molecular beam epitaxy demonstrating extremely uniform GaN nano-column arrays. J Cryst Growth 2009;311:2063.

79. Bengoechea-Encabo A, Barbagini F, Fernandez-Garrido S, Grandal J, Ristic J, et al. Under-standing the selective area growth of GaN nanocolumns byMBE using Ti nanomasks. J CrystGrowth 2011;325:89.

80. Sekiguchi H, Kishino K, Kikuchi A. Ti-mask selective-area growth of GaN by RF-plasma-assisted molecular-beam epitaxy for fabricating regularly arranged InGaN/GaNnanocolumns. Appl Phys Express 2008;1:124002.

81. Bertness KA, Sanders AW, Rourke DM, Harvey TE, Roshko A, et al. Controlled nucleationof GaN nanowires grown with molecular beam epitaxy. Adv Funct Mater 2010;20:2911.

82. Kouno T, Kishino K, Yamano K, Kikuchi A. Two-dimensional light confinement inperiodic InGaN/GaN nanocolumn arrays and optically pumped blue stimulated emission.Opt Express 2009;17:20440.

83. Siekacz M, Sawicka M, Turski H, Cywinski G, Khachapuridze A, et al. Optically pumped500 nm InGaN green lasers grown by plasma-assisted molecular beam epitaxy. J Appl Phys2011;110:063110.



MOCVD growth of nitridesemiconductors 2Koh Matsumoto, Yoshiki Yano, Hiroki Tokunaga, Akinori Ubukata, Guanxi Piao,Akira Mishima, Tadakazu Ikenaga, Yuji Tomita, Toshiya TabuchiTAIYO NIPPON SANSO Corp., Minato-ku, Japan

2.1 Introduction

The central idea of hetero-epitaxy of GaN is the multilayer buffer structure, in whicheach stacked layer has a different role, that is, alignment of crystal axis, dislocationfiltering, and balancing or compensation of a large thermal mismatch between the sub-strate and overlayers. The actual tactics depends on the substrate material. The hetero-epitaxial growth is described in the following chapters, vapor phase reaction and theoutcomes of suppressing parasitic reaction are highlighted in this chapter.

As an introduction, two major rector configurations are explained, vertical and hor-izontal flow reactor. Fig. 2.1 illustrates two types of vertical flow reactors: shower headreactor and a high rotational disc reactor. In the vertical flow reactor, the flow in thereactor is designed so as to establish a stagnation point flow above the wafers. If anideal stagnation point flow is established, uniform distribution of mass transportover the wafer holder is mathematically predicted. This can be understood if you ima-gine a flow in a semi-infinite space. An ideal stagnation point flow is realized in a flowthat impinges perpendicular to the plane in a semi-infinite space. At the symmetricalcenter of the flow, the gas flow is stagnated. In the very vicinity of the stagnation point,the mass transport as well as the heat transport is uniform. If the geometry is symmet-rical and semi-infinite half plane, the flow condition is essentially identical in the vi-cinity of the stagnation point. In a realistic reactor, there is a peripheral of the plane,then the actual flow differs somehow from the ideal stagnation point flow. A compre-hensive review of the stagnation point flow can be found in the report by Wahl.1

High-speed rotation disc

500 ~ 1500 rpm

Cooled shower head

Figure 2.1 Two types of vertical flow reactors: shower head reactor and a high rotational discreactor illustrated.

Nitride Semiconductor Light-Emitting Diodes (LEDs). https://doi.org/10.1016/B978-0-08-101942-9.00002-2Copyright © 2018 Elsevier Ltd. All rights reserved.

https://doi.org/10.1016/B978-0-08-101942-9.00002-2

Shower head gas distributor closely facing the hot wafer holder is often used for chem-ical vapor deposition to realize a stagnation point flow. However, in GaAs, excessamount of arsenic precursor AsH3 is supplied to establish a sufficient over pressureof arsenic. Because the shower head is radiatively heated from the hot wafer holder,the shower head must be cooled by water. As a result, thick arsenic byproducts aredeposited on the surface of the shower head. Therefore, other methods to realize a stag-nation point flow were developed for IIIeVs metal organic chemical vapor deposition(MOCVD). One was to use a tapered down flow, which eventually evolved into aconcentrated multinozzle configuration. Other was to employ a high speed rotationdisc, which was distantly located from the shower head.2 By matching the totalflow rate and the rotational speed, you can establish a stagnation point flow over thedisc. Example of the high speed rotation disc reactor is shown in Fig. 2.1. The typicalrotational speed of the disc is from 500 to 1500 rpm. Because nitrogen does notcondense on top of the cool shower head, the cooled shower head configuration inFig. 2.1 can be employed for nitride semiconductor MOCVD.3

Another popular reactor configuration is a horizontal flow reactor (Fig. 2.2).Because the precursor is consumed along the flow direction, it is essential to rotatethe substrate to spatially average the materials supply over the wafer.4 This type reactorwith a planetary rotation of wafers has a long history since Kern in 1968.5

If you look at the mechanism of obtaining a uniform film growth, a stagnation pointflow looks more sophisticated than a horizontal flow design. However, there are severalproblems in the vertical flow design as follows. First, in nitride semiconductors, thehetero-epitaxial growth on a foreign substrate is common, which results in a significantbowing of the wafer during growth. If you look at a multiwafer growth on a large disc,you cannot control the intrawafer temperature gradient by adjusting the heater powerinput. You can only control a circularly symmetric power input over the whole disc. Sec-ond, because the growth temperature of GaN and AlGaN is around 1000�C, most of themetalorganics are decomposed into reactive intermediate byproducts in the vicinity of thewafer surface. J. R. Creighton observed GaN nanoparticles piling up on top of the ther-mal boundary layer by laser scattering under a high supply rate condition and AlN nano-particles for an order of magnitude smaller supply rate.6 The difference between GaN andAlN comes from a critical size of nanoparticles for stable growth that is too small forAlN. On top of the surface of the thermal boundary layer, these reactive precursors sticktogether to form large particles. Large particles are pushed out toward the cooler regionalong the temperature gradient by thermophoretic force. During the thermophoretic mo-tion, particles effectively gather more precursors on top of their surface because the

Figure 2.2 A horizontal flow reactor.


movement of precursors and particles is in the counterdirection with each other. You canpartially mitigate the growth of particles by employing a higher carrier gas flow rate todilute the precursors as well as employing a higher rotational speed to increase masstransport efficiency to compensate the decrease of precursor’s concentration. Using avery short distance between the wafer and the shower head can also partially mitigatethe growth of particles.7 However, precursor supply nozzles would be heated up andchoked in a long growth time eventually. In both reactors, the lower pressure growthis useful to reduce particles. Although high rotational speed and low pressure are effec-tive to reduce particles, this may bring about high carbon incorporation. We will get backto this issue later.

In a horizontal flow reactor, the thermophoretic motion of clusters is perpendicularto the flow direction, which partially reduces the probability of clusters to gather moreprecursors on top of their surface. However, the particles need to be evacuated quicklybefore they grow large. Because it is easy to employ high flow speed as well as reason-able growth efficiency in a horizontal reactor, a horizontal flow configuration is usefulto reduce particle formation in vapor phase.8 However, we need to take care of thedownfall of the deposit from the top ceiling plate. In Fig. 2.3, an example of a horizon-tal flow production reactor (Taiyo Nippon Sanso UR25k) is shown. In this reactor,high growth rate GaN and AlN is realized by high-speed gas flow.9,10 By employinga triple gas injection nozzle to avoid a parasitic gas phase reaction as well as a highflow speed, particle formation in vapor phase is reduced in this reactor. As a result,the growth rate of GaN is independent of the growth pressure (Fig. 2.4).11 The GaNgrowth rate at an atmospheric pressure is the same as that at 40 kPa. Regarding the

Wafer plate

Quartzceiling plate

*6-zone heater controlled separately

Tri-layergasnozzle

Quartzceiling plate

Susceptor

Gas nozzle 6-zone heater

Waferplate

H6 H5 H4 H3 H2 H1

Figure 2.3 An example of a horizontal flow production reactor (Taiyo Nippon Sanso UR25k) isshown.

MOCVD growth of nitride semiconductors 27

growth on a bowing wafer, uniform growth of AlGaN was obtained by employingwafer rotation and heater power distribution. By adjusting the flow condition andthe heater power input distribution, we can control the surface temperature of the waferpocket for either concave or convex shape wafer.12 In order for the easy exchange ofthe deposited top ceiling plate, the top ceiling plate can be automatically exchanged bythe robot arm.

In the following sections, growth mechanisms, Mg doping, visible multi quantumwell (MQW) growth, and ultraviolet (UV) materials growth by using high flow speedreactor are described.

2.2 Growth mechanism

In this section, elementary growth chemistry is described by quantum chemical calcu-lation. Parasitic reaction for AlN is especially highlighted.

Fig. 2.5 illustrates the possible molecular configuration of trimethyl-gallium(TMG)-NH3 adduct transition state associated with two NH3.

13 Because metal-alkylhas a vacant orbital at the center metal, it attracts electron. NH3 has a lone orbitalwith an excess electron pair at nitrogen. By transferring the excess electron pair of ni-trogen to the vacant orbital of Ga of TMG, TMG and NH3 form an adduct molecule.Under an excess NH3 ambience usually encountered in GaN MOCVD, two NH3 mol-ecules participate to form adduct. In Fig. 2.5, one of the hydrogen of NH3 bridges be-tween nitrogen and carbon of methyl-ligand of TMG. This is a transition state toeliminate methane from TMG. The transition state energy is smaller for two NH3 asso-ciated state than one NH3 associated state as is shown in a reaction diagram inFig. 2.6.13 Then, high V/III ratio enhances the decomposition of TMG. There is a

8

7

6

5

4

3

2

1

0–100 –50 0 50 100

Pg = atmospheric pressure

Pg = 40 kPa

Downstream Upstream

Position in wafer (mm)

Gro

wth

rate

of G

aN (μ

m/h

)

Figure 2.4 GaN growth rate distribution along the flow direction over the stationary wafer isshown both at an atmospheric pressure and at 40 kPa.11 Both data are identical, whichdemonstrates that the vapor phase reaction is well controlled in this reactor.


discussion about what is the most probable path of decomposition of TMG throughadduct or a single molecule decomposition of TMG.6 It is known that in the case ofGaAs, stable isotope labeling study showed that an adduct mechanism betweenTMG and AsH3 is the dominant path of decomposition of TMG in GaAs growth.14

In the case of aluminum and indium, bimolecular adduct path to form oligomers isimportant. Fig. 2.7 illustrates the reaction diagram of bimolecular decomposition paththrough two adducts.15 In the case of TMG, transition state energy of two adducts ishigher than the individual molecules. Therefore, the adduct of TMG and NH3 is morelikely to crack rather than form a bimolecular transition state. However, in the case oftrimethyl-aluminum (TMA), the transition state energy is lower than an individualmolecule. Therefore, the bimolecular transition state path is preferentially going on.It is also noted that bimolecular reaction between adducts emits one methane; as aresult, the number of molecules before and after the reaction is the same. It means en-tropy with regard to the number of molecules is unchanged.

N

M

C

H

1.310

1.4152.526

2.267

1.997

Figure 2.5 The possible molecular configuration of trimethyl-gallium (TMG)-NH3 adducttransition state associated with two NH3.

13 Through this transition state, methane is eliminated.

TS1g TS4g14.83 13.04

–41.03

+NH3

+NH3

+NH3

–CH4 –CH42g

1g

–18.92

0.00

–22.02 –22.621g: Ga(CH3)32g: (CH3)3Ga•NH33g: Ga(CH3)2NH27g: H3N•Ga(CH3)3•NH38g: H3N•Ga(CH3)2•NH2

7g 3g

8g

Figure 2.6 Reaction diagram for TMG-NH3 adduct formation, transition state with one NH3

and two NH3 association and methane elimination is shown.13


Table 2.1 summarizes the transition energy of each combination of metals.15 Youcan see that adduct which contains aluminum or indium has a tendency to lower thetransition energy to form oligomers.

As a result of the low-energy barrier to form oligomers by TMA, dramaticreduction of the growth rate of AlN under atmospheric pressure is observed ataround 500�C, at which the onset of the oligomerization in vapor phase issuggested.15

Oligomers

(CH3)3M, NH3

(CH3)3M–NH3Cracking

Bonding

TS1Ga/Ga

A1/A1

CH4 release

Go over TS

Figure 2.7 The reaction diagram of bimolecular decomposition path through two adducts isillustrated.15

Table 2.1 The transition energy of each combination of metals for bi-molecular adducts reaction path is summarized.15 Potential energyof (CH3)3M:NH3 system for bimolecular reaction mechanism(kcal/mol)

(CH3)3M:NH3D(CH3)3M0:NH3 / (CH3)3M:NH2M0(CH3)2:NH3DCH4

M/M0 Adduct TS Product

Al/Al �46.34 �10.56 �68.82

Al/Ga �42.09 �2.90 �61.63

Ga/Al �42.09 �5.51 �63.80

Ga/Ga 37.83 �2.39 �56.74

Ga/In �37.14 �3.19 �56.70

In/Ga �37.14 �2.93 �55.67

In/In �36.45 �2.78 �56.15

The energy of the (CH3)3M and NH3 is set to zero.


2.3 Carbon incorporation and Mg doping of GaN

Carbon is a major impurity in MOCVD. In most of the optical device applications bynitride semiconductors, it is important to reduce carbon impurity in the film, becausecarbon forms a deep level and thereby compensates a donor and also an acceptor inGaN and AlGaN.16 In GaAs, carbon forms a shallow acceptor level. Therefore, carbonis sometimes used as an acceptor in a diode laser or an emitter impurity in an n-p-nhetero-bipolar-transistor. The advantage of carbon as an acceptor in GaAs is its smalldiffusion constant in a solid as well as its low resistivity for a highly doped layer. Incontrast, in nitride semiconductor electron devices, carbon is often intentionally dopedto form a high resistivity layer.9

Carbon incorporation in various growth conditions for GaN is shown in Fig. 2.8.17

High V/III ratio, higher growth pressure, or higher temperature reduces carbon incor-poration.9,17,18 At a constant V/III ratio, carbon incorporation is independent of thegrowth rate. It is also notable that carbon incorporation is determined by V/III ratioindependently of the reactor size. Carbon incorporation in GaN showed saturation ataround 50 kPa. Under 50 kPa, carbon incorporation was increased monotonically asthe growth pressure was decreased. By adjusting the growth condition, we can controlcarbon concentration in GaN from 3 � 1015 to 1 � 1020 cm�3.

A detailed mechanism of the carbon incorporation is still open to discussion. Ifmethane ligand of organometals is eliminated through an adduct reaction path asdescribed in the former section, reaction byproduct in vapor phase is methane.Methane is rarely incorporated in the film due to its chemical stability. In highergrowth pressure, vapor phase reaction of eliminating methane would proceed. Becausehigher growth temperature reduces carbon incorporation, it is suggested that surfacereaction is one of the major path. Carbon incorporation efficiency is smaller in GaNgrown on a high-quality bulk GaN substrate than that in GaN grown on sapphire,which may also support that surface phenomena has some role in carbon incorporationinto the film.16 Probably both vapor phase and surface kinetics are contributing to thecarbon incorporation.

1.0E+19

1.0E+18

1.0E+17

1.0E+16

1.0E+15

C c

once

ntra

tion

(cm

–3)

100 1000 10000 100000

6" × 7Gr = 0.2~1.3 μm/h

6" × 7Gr = 3.6 μm/h

4" × 11Gr = 10 μm/h

2" × 3Gr = 2.3 μm/h

Tg = 850ºC, Pg = 50 kPa

Tg = 960ºC, Pg = 100 kPa

Tg = 1100ºC, Pg = 100 kPa

Pg = 100 kPaTg = 850ºC,

V/III ratio

Figure 2.8 Carbon incorporation in various growth conditions for GaN is shown.17


Next, by knowing the carbon incorporation conditions, we would like to considerMg doping in GaN. We need to consider several aspects of Mg doping for visible op-tical device application, solid phase diffusion, sheet resistance, and thermal budget forthe underlying layer. Because active region of visible optical device is composed ofthermally unstable InGaN, we need to minimize the thermal budget associated withgrowing Mg-doped p-clad layer. Mg also easily diffuses into the underlying layer dur-ing the growth at the growth temperature of a good quality nondoped GaN (nominally1115�C). Fig. 2.9 shows secondary ion mass spectroscopy (SIMS) depth profile of Mgdoped GaN at a growth temperature of 960 and 850�C. These layers were grown atatmospheric pressure by using UR25k. Mg showed an out-diffusion into the underly-ing layer at 960�C, but no out-diffusion at 850�C. Therefore, in terms of the thermalbudget and a sharp interface, the growth temperature of 850�C is better than 960�C.However, as we have seen in Fig. 2.8, lower growth temperature increases carbonincorporation thus resulting in a lower Mg activation.

Fig. 2.10 shows sheet resistance of Mg doped GaN under various growth condi-tions. By optimizing the Mg flow rate as well as a growth rate, we can achieve nearly2 U cm at the growth temperature of 960�C and 2.5 U cm at 850�C. These layers weregrown at atmospheric pressure to reduce carbon incorporation.

2.4 Blue and green MQW

In this section, optical property of blue and green MQW is described as a function ofvarious growth conditions. InGaN growth is affected largely by thermodynamics. In-dium incorporation is increased as the growth pressure is increased. Therefore, a highergrowth pressure is advantageous for a longer wavelength MQW. Because InGaN has a

1E+21

1E+20

1E+19

1E+18

1E+17

1E+16

Con

cent

ratio

n (a

tom

s/cm

3 ) Segregation?

Diffusion

Based onaccuracy of SIMS

p-GaN / n-GaNinterfacesTg = 850ºC

Tg = 960ºC

0.6 0.8 1 1.2Relative depth (μm)

Figure 2.9 Secondary ion mass spectroscopy (SIMS) depth profile of Mg-doped GaN at agrowth temperature of 960 and 850�C are shown.


large miscibility gap in a visible wavelength region, it is reported that substantial in-dium metal is segregated on top of the InGaN surface during growth and sometimesforms indium metal droplet.19 Due to an excess overpressure of indium, surface in-dium may be more piled up in a higher growth pressure InGaN. At any rate, weneed to evacuate an excess indium on the surface of InGaN before we form a barrierlayer on top of it. Fig. 2.11 shows a typical growth sequence of MQW. We haveramped up the growth temperature after the growth of InGaN well layer to evaporatean excess indium on top of it and also to improve GaN barrier crystal quality. In orderto avoid growth interruption before and after the well layer growth, we have inserted athin GaN layer in both the interfaces. We guess that excess indium atoms partiallyalloyed and climbed up to the surface during the growth of GaN cap.

Fig. 2.12 shows a photoluminescence (PL) intensity of MQW of blue spectrum re-gion under various growth conditions. In Fig. 2.12, it is shown that most of the MQWwere grown at atmospheric pressure by using UR25k. PL intensity tends to decrease atgrowth pressure less than 50 kPa due to the low growth temperature to maintain thesame wavelength. PL intensity is higher at the higher growth temperature than atthe lower growth temperature. We need higher overpressure of indium for the highergrowth temperature.

785°C

735°C

∆ = 50°C

Time

I1 I2 I2I1B1 B2 B2B3 B3B1 WWUnderlayer

Gro

wth

tem

pera

ture

Figure 2.11 A typical growth sequence of multi quantum well growth is illustrated.

7000

2

4

6

8

10

12

800 900 1000 1100Growth temperature (°C)

Ele

ctric

resi

stiv

ity (Ω

cm

)(mg) = 3.7E+19

(Mg) = 5.3E+19

(Mg) = 6.6E+19

(Mg) = 2E+19(Mg) = 1E+19(cm−3)

Normal GaN growthtemp. ==>

Figure 2.10 Sheet resistance of Mg-doped GaN under various growth conditions are shown.


However, there is a trade-off between uniformity of wavelength and the PL inten-sity. Fig. 2.13 shows wavelength mapping of MQW grown at 735�C with trimethylindium (TMI)/(TMG þ TMI) ¼ 0.4 and at 740�C with TMI/(TMG þ TMI) ¼ 0.8.1s variation of the wavelength at 455 nm was 1.4 nm for the sample grown at735�C and 3.1 nm for that grown at 740�C.

0

500

1000

1500

2000

2500

700 710 720 730 740 750

Inte

grat

ed P

L in

tens

ity (a

.u.)

Growth temperature of InGaN (ºC)

TMI / MO = 0.8

TMI / MO = 0.4

80 nm/h

62 nm/h

37 nm/h188 nm/h

123 nm/h

62 nm/h

λPL = 460 ± 3 nm (He-Cd: 3.0 mW)

*TMI / MO = 0.8Pg = 27 KPa

Pg = 50 KPa

Figure 2.12 A photoluminescence intensity of multi quantum well of blue spectrum regionunder various growth conditions is shown.

SU325-3 (6-inch wafer) SU240-3 (6-inch wafer)

* 6-inch wafer

Ave.: 456.5 nmstd(1σ): 3.1 nm

Ave.: 455.4 nmstd(1σ): 1.4 nm

TMI/III = 0.4TMI/III = 0.8

Peak lambdaPeak lambda

* 1 Color = 1 nm

Nanometrics Nanometrics468.0

464.8

461.6

458.4

455.2

452.00.0 29.6 59.2 88.8 118.4 148.0

nm

Distance (mm)0.0 29.6 59.2 88.8 118.4 148.0

466.0

462.8

459.6

456.4

453.2

450.0

nm

Distance (mm)

nmnm466.0464.0462.0460.0458.0456.0454.0452.0450.0

466.0464.0462.0460.0458.0456.0454.0452.0

468.0

Figure 2.13 Wavelength mapping of multi quantum well grown at 735�Cwith trimethyl indium(TMI)/(TMG þ TMI) ¼ 0.4 and at 740�C with TMI/(TMG þ TMI) ¼ 0.8 is shown,respectively.


Fig. 2.14 summarizes the effect of the cap layer of green MQW on the PL spectrum.As a preliminary result, the combination of thin GaN and AlGaN cap resulted in a nar-rower PL spectrum as well as a higher PL intensity at high injection condition, thedetail of which are still under investigation. The effect of AlGaN barrier for amberor red MQW is discussed in Refs. 20,21.

2.5 UV materials growth

In the following section, we describe AlGaN and AlN growth in the UV region. Asdescribed in Section 2.2, TMA and NH3 adduct easily forms oligomers in vapor phasewith small excitation energy, which work as the nuclei of the growth of particles. Inorder to avoid this parasitic reaction, AlN and AlGaN are grown under low pressureat 10e70 kPa. The optimum growth pressure, the growth temperature, and V/III ratioare selected in terms of growth rate, carbon incorporation, dopant activation, andaluminum composition. Here, vapor phase reaction must be well controlled to suppressparasitic reaction. The layer was grown by using 200 � 3 high-temperature MOCVD(Taiyo Nippon Sanso SR4000HT).

Fig. 2.15 shows carbon concentration in AlN grown at 1220e1300�C with V/IIIratio of 10.22 By increasing the growth temperature, carbon concentration is decreasedas in the case of GaN. Carbon concentration in AlN is also decreased from 1 � 1017 to6 � 1016 cm�3 by increasing V/III ratio from 10 to 200. Because carbon acts as a colorcenter in AlN to absorb UV light shorter than 300 nm, carbon concentration must becontrolled to less than 1 � 1017 cm�3 to avoid absorption loss of UV light in light-emitting diode.23

It is known that crystal quality of AlN is critical to improve the internal quantumefficiency of AlGaN MQW on top of it. X ray diffraction full width at half maximum(XRD FWHM) of (002) and (102) of underlying AlN of AlGaN MQW must be lessthan 300 arcsec and 500 arcsec, respectively.24 Fig. 2.16 shows a typical X ray diffrac-tion measurement result of 3-mm-thick AlN, which was grown at 13 kPa with a growthrate of 3.6 mm/h and V/III ratio of 220. XRD FWHM of (002) and (102) of AlN wasuniformly distributed and less than 250 and 450 arcsec, respectively, over the whole 2-inch-diameter sapphire substrate.22 It is notable that relatively high V/III ratio of 220was realized at the growth rate of 3.6 mm/h. Generally speaking, this amount of V/IIIratio for AlN is difficult for the reactor in which vapor phase reaction is not wellcontrolled.

Fig. 2.17 shows a growth rate of Al0.6Ga0.4N as a function of the input metalorganicprecursors.22 Al0.6Ga0.4N growth rate was linearly increased up to 6 mm/h. Fig. 2.18shows the aluminum composition over the 2-inch-diameter wafer for samples grownwith the growth rate of 4, 5, and 6 mm/h. Aluminum composition variation was lessthan 2% independent of the growth rate.

Fig. 2.19 shows carbon concentration in AlxGa1 � xN with x ¼ 56% and 67% as afunction of the susceptor temperature from 1000 to 1160�C.22 Carbon concentrationmonotonically decreased as the growth temperature was increased or V/III ratio was


He-Cd excitation(low power)

YAG excitation(high power)

GaN-Cap: 1.3 nmGaN-Cap: 0.65 nm

& AlGaN-Cap: 1.3 nmGaN-Cap: 0.33 nm

& AlGaN-Cap: 1.3 nm

λ= 526.0 nmFWHM= 65.4 nm

λ= 520.0 nmFWHM= 47.1 nmλ= 497.7 nm

FWHM= 70.4 nm

400 450 550 600

Inte

nsity

(V)

Wavelength (nm)500

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7 YAG: 2.0 mWYAG: 1.0 mWYAG: 0.1 mWYAG: 0.05 mW

Inte

nsity

(V)

Wavelength (nm)

YAG: 2.0 mWYAG: 1.0 mWYAG: 0.1 mWYAG: 0.05 mW

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0400 450 500 550 600

Inte

nsity

(V)

Wavelength (nm)

YAG: 2.0 mWYAG: 1.0 mWYAG: 0.1 mWYAG: 0.05 mW

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0400 450 500 550 600

3.0

2.0

1.0

0.0450 500 550 600

Inte

nsity

(V)

He-Cd: 3.0 mWHe-Cd: 1.0 mWHe-Cd: 0.5 mW

Wavelength (nm)

3.0

2.0

1.0

0.0450 500 550 600

Inte

nsity

(V)

Wavelength (nm)


3

2

1

0450 500 550 600

Inte

nsity

(V)

Wavelength (nm)


Figure 2.14 The effect of the cap layer of green multi quantum well on the photoluminescence spectrum is summarized.

36Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

1.E+18

1.E+17

1.E+161200 1250 1300 1350

Temperature (°C)

Car

bon

conc

entra

tion

(ato

m /

cm3 )

Figure 2.15 Concentration in AlN grown at 1220e1300�C with V/III ratio of 10 is shown.22

500

450

400

350

300

250

200

150

100

50

0–25 –20 –15 –10 –5 0 5 10 15 20 25

XR

C F

WH

M (a

rcse

c)

Distance from the center of a wafer

(0002)

(10–12)

Figure 2.16 A typical X ray diffraction measurement result of a 3 mm thick AlN, which wasgrown at 13 kPa with a growth rate of 3.6 mm/h and V/III ratio of 220, is shown.22


00.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

5 10 15

Total MO/total carrier gas × 10–3 (%)

AlxGa1–xN:0.6 μm

AlN:0.85 μm

Sapphire

GR

of A

lGaN

(μm

/h)

P = 20 kPaT = 1000°CV/III

=800∼4500

Figure 2.17 A growth rate of Al0.6Ga0.4N as a function of the input metalorganic precursors isshown.22

0 20 40 6030

35

40

45

50

55

60

65

70

Al c

ompo

sitio

n (%

)

Distance from the center of a wafer (mm)

A sample

B sample

C sample

Figure 2.18 The aluminum composition over the 2-inch-diameter wafer for samples grown withthe growth rate of 4, 5 and 6 mm/h is shown.22

1.E+18

1.E+17

1.E+16

1.E+19

1000 1050 1100 1150 1200Temperature on susceptor surface (°C)

Car

bon

conc

entra

tion

(cm

–3)

Al composition 56%, GR = 3.6 μm/h, V/III = 1488



Figure 2.19 Carbon concentration in AlxGa1 � xN with x ¼ 56% and 67% as a function of thesusceptor temperature from 1000 to 1160�C is shown.22

increased from 744 to 1488. Carbon concentration in Al0.67Ga0.33N grown at 1160�Cwas less than 1 � 1017 cm�3. Donor concentration of these layers was characterized byCeVmeasurement (Fig. 2.20). Donor concentration was more than 5 � 1018 cm�3 forthe samples with carbon concentration of less than 2 � 1017 cm�3. Fig. 2.21 shows asheet resistance of Al0.67Ga0.33N over the three wafers grown at a time with donor con-centration 1.2 � 1019 cm�3. Sheet resistance of 50 U/cm2 was obtained.

1.E+181.E+16 1.E+17 1.E+18

1.E+19

1.E+19

Carbon concentration (cm–3)

Nd-

Na

conc

entra

tion

(cm

–3)


Al composition 67%, GR = 3.6 mm/h, V/III = 1488

Al composition 67%, GR = 7.2 μm/h. V/III = 744

ND

Figure 2.20 Donor concentration of the layers shown in Fig. 2.15 were characterized by C-Vmeasurement.22

Si doped Al0.67Ga0.33N on AIN / sapphire

1–1

2–1

3–13–23–33–43–5

2–22–32–42–5

1–21–31–41–5

48.5949.2

48.9150.4647.4648.5349.1948.7250.3947.4448.5748.9550.4449.147.43

n-AI0.67Ga0.33N

ud-AI0.67Ga0.33N 0.3μm

2μm

Transition layerAIN

Sapphire

n-AIGaN structure

1–2

1–3

1–4 1–1 1–5

2–42–1

2–3

2–22–53–2 3–

43–13–

5 3–3

10mm

OF OFOF

Position Sheet resistance(Ω cm–2)

Figure 2.21 A sheet resistance of Al0.67Ga0.33N over the three wafers grown at a time withdonor concentration 1.2 � 1019 cm�3 is shown.22


Next, we look into Mg-doped AlGaN. As in the case of Mg-doped GaN, highergrowth temperature results in an out-diffusion of Mg into the underlying layer but car-bon concentration is decreased. Fig. 2.22 shows SIMS depth profile of Mg inAl0.2Ga0.8N grown at 1150 and 1025�C. Fairly abrupt interface was obtained for thesample grown at 1025�C, but Mg out-diffusion was observed for the sample grownat 1150�C. Because low growth temperature results in higher carbon incorporationand high resistivity of the layer, we have examined the resistivity of Mg doped AlGaNas a function of Mg flow rate and aluminum composition. We have obtained electricresistivity of less than 20 U cm for Al0.1Ga0.9N grown at 1025�C at 70 kPa, but theresistivity of Al0.2Ga0.8N was nearly 100 U cm. For the aluminum composition oflarger than 10%, it is difficult to obtain a low sheet resistance p-type layer becausethe Mg acceptor level becomes deep.

References

1. Wahl G. Thin Solid Films 1977;40:13e26.2. Wang CA, Groves SH, Palmateer SC, Weyburne DW. J Cryst Growth 1986;77:136e43.3. Pawlowski RP, Theodoropoulos C, Salinger AG, Mountziaris TJ, Moffat HK, Shadid JN,

Thrush EJ. J Cryst Growth 2000;221:622e8.4. Frijlink PM. J Cryst Growth 1988;93:207e15.5. Kern W. RCA Rev 1968:525e32.6. Creighton JR, Brelland WG, Coltrin ME, Pawlowski RP. Appl Phys Lett 2002;81:2626e8.7. Stellmach J, Pristovsek M, Savas O, Schlegel J, Yakovlev EV, Kneissl M. J Cryst Growth

2011;315:229e32.8. Ubukata A, Yano Y, Yamaoka Y, Kitamura Y, Tabuchi T, Matsumoto K. Phys Status Solidi

C 2013;10(11):1353e6. https://doi.org/10.1002/pssc.201300255.

1E+16

1E+17

1E+18

1E+19

1E+20

1E+21

0 0.1 0.2 0.3 0.4 0.5Depth (μm)

C

Mg

Con

cent

ratio

n (a

tom

s/cm

3 )C 1150°C 20 kPa C 1150°C 70 kPaC 1025°C 70 kPaMg 1150°C 70 kPa

Mg 1150°C 20 kPaMg 1125°C 70 kPa

Figure 2.22 Secondary ion mass spectroscopy depth profile of Mg in Al0.2Ga0.8N grown at1150 and 1025�C is shown.


https://doi.org/10.1002/pssc.201300255

9. Ubukata A, Yano Y, Shimamura H, Yamaguchi A, Tabuchi T, Matsumoto K. J CrystGrowth 2013;370(1):269e72.

10. Matsumoto K, Tokunaga H, Ubukata A, Ikenaga K, Fukuda Y, Tabuchi T, Kitamura Y,Koseki S, Yamaguchi A, Uematsu K. J Cryst Growth 2008;310:3850e952.

11. Matsumoto K, Yamaoka Y, Ubukata A, Arimura T, Piao G, Yano Y, Tokunaga H,Tabuchi T. Jpn J Appl Phys 2016;55:05FK04.

12. Yano Y, Tokunaga H, Shimamura H, Yamaoka Y, Ubukata A, Tabuchi T, Matsumoto K.Jpn J Appl Phys 2013;53:08JB06.

13. Nakamura K, Makino O, Tachibana A, Matsumoto K. J Organomet Chem 2000;611:514e24.

14. Stringfellow GB. Prog Cryst Growth Charact 1989;19:115e23.15. Matsumoto K, Tachibana A. J Cryst Growth 2004;272:360e9.16. Armstrong A, Arehart AR, Moran B, DenBaars SP, Mishra UK, Speck JS, Ringel SA. Appl

Phys Lett 2004;84:374e6.17. Piao G, Ikenaga K, Yano Y, Tokunaga H, Mishima A, Ban Y, Tabuchi T, Matsumoto K.

J Cryst Growth 2016;456:137e9.18. Koleske DD, Wickenden AE, Henry RL, Twigg ME. J Cryst Growth 2002;242:55e69.19. Yamamoto T, Tamura A, Usami S, Mitsunari T, Nagamatsu K, Nitta S, Honda Y,

Amano H. Jpn J Appl Phys 2016;55:05FD03.20. Iida D, Lu S, Hirahara S, Niwa K, Kamiyama S, Ohkawa K. Jpn J Appl Phys 2016;55:

05FJ06.21. Hwang J-I, Hashimoto R, Saito S, Nunoue S. Apppl Phys Express 2014;7:071003.22. Ikenaga K, Mishima A, Yano Y, Tabuchi T, Matsumoto K. Jpn J Appl Phys 2016;55:

05FE04.23. Kumagai Y, Kubota Y, Nagashima T, Kinoshita T, Dalmau R, Schlesser R, Moody B, Xie J,

Murakami H, Koukitu A, Sitar Z. Appl Phys Express 2012;5:055504.24. Hirayama H. Proc SPIE 2010;7617:76171G. https://doi.org/10.1117/12.845512.


https://doi.org/10.1117/12.845512


GaN on sapphire substrates forvisible light-emitting diodes 3Jae-Hyun Ryou 1, Wonseok Lee 2

1University of Houston, Houston, TX, United States; 2LED Business Unit, LG Innotek,Paju-si, Korea

III-nitride (III-N)-based visible blue and green light-emitting diodes (LEDs) currentlyin production are predominantly manufactured on sapphire substrates with a galliumnitride (GaN) buffer layer. Therefore, the platform of GaN on sapphire is the mostimportant technology currently in use for the applications of the LEDs and LED-based solid-state lighting (SSL). In this chapter, the fundamental, technical, andeconomic aspects of the GaN on sapphire substrates with a focus on the applicationsof visible LEDs are described.

3.1 Importance and historical backgrounds of GaNepitaxial growth and sapphire substrates

Epitaxial growth of GaN materials has a longer history than it is generally perceiveddonly slightly behind that of gallium arsenide (GaAs) and indium phosphide (InP)(w10 years later than GaAs epitaxial growth),1 even though the successful develop-ment of epitaxial growth technology followed by device development was quitebehind (w30 years later). This section gives an overview on the development historyof synthesis and epitaxy technology of the GaN materials based on the literature.

Sapphire is the most important and currently dominant substrate for the epitaxialgrowth of III-N-based photonic and electronic devices.2e5 Using sapphire substratesfor the epitaxy of IIIeV semiconductors (including III-N semiconductors) also has along history. The first epitaxial single-crystal GaN films were grown on a sapphiresubstrate.6,7 Manasevit, who is acknowledged as an inventor of metalorganic chemicalvapor deposition (MOCVD) process,8,9 also used sapphire along with other substratessuch as spinel (MgAl2O4), beryllium oxide (BeO), and thorium oxide (ThO2) for theepitaxial growth of IIIeV and IIeVI materials.10,11

3.1.1 Development history of epitaxial GaN on sapphiresubstrates for device-quality materials

Development history of GaN material technology can be divided into several phasesby the maturity of the technology, including early development, powder GaN, thin-film GaN, and device-quality epitaxial GaN.

Nitride Semiconductor Light-Emitting Diodes (LEDs). http://dx.doi.org/10.1016/B978-0-08-101942-9.00003-4Copyright © 2018 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/B978-0-08-101942-9.00003-4

3.1.1.1 Early development

GaN materials in the early development stage were powder forms prepared mostly bychemists. The studies were focused on the synthesis processes of nitrides of variousmetallic compounds, also including boron nitride (BN) and aluminum nitride (AlN),and their fundamental properties. The synthesis of GaN materials is traced back tothe 1930s.

Johnson et al.12 reported the formation of GaN material using metallic gallium (Ga)and ammonia (NH3) gas, which are similar precursors for the process of gas-sourcemolecular beam epitaxy (MBE) of these days. The reaction occurred by passingNH3 gas on metallic Ga at 900e1000�C by the following suggested reaction:

2Gaþ 2NH3/2GaNþ 3H2 (3.i)

In the same report, they also described an unsuccessful attempt to react the metallicGa with nitrogen gas (N2). In addition, very stable thermal and chemical properties ofGaN in the solutions of HCl, HF, HNO3, and hot aqua regia were reported.

12 Hahn andJuza13 also prepared GaN by thermal decomposition of ammonium hexafluorogallate((NH4)3GaF6). Later, Renner

14 prepared GaN by a van Arkel-de Boer process (crystalbar process) by the following reaction using the precursors similar to hydride vaporphase epitaxy (VPE) of these days:

GaCl3 þ NH3/GaNþ 3HCl (3.ii)

3.1.1.2 Powder GaN

In the 1960s, the application of GaN as one of the wide-bandgap semiconductorsstarted to be sought, especially for phosphor applications mainly in cathode-ray tubes.Hence, the GaN materials were still powder forms, as opposed to thin-film epitaxialmaterials for active devices in today’s applications. Taking several examples,Addamiano15 used powders of gallium phosphide (GaP) and GaAs in a stream ofNH3 gas at 1000e1100�C for the following reaction:

GaPþ NH3/GaNþ 14P4 þ 3

2H2 (3.iii)

In the same study, AlN was also synthesized and these AlN and GaN powders weremixed in a heated and sealed quartz container in an attempt to prepare ternary AlGaNmaterials. Even though it was not successful, this was the first study for the synthesis ofternary III-N materials in the literature. Lorenz and Binkowski16 used the reductionand nitridation of gallium trioxide (Ga2O3) to prepare GaN powder using NH3 bythe following reaction:

Ga2O3 þ 2NH3/2GaNþ 3H2O (3.iv)


3.1.1.3 Thin-film GaN

Thin-film GaN materials were later prepared by gas discharge on a quartz plate byPasternak and Souckova17 but the deposited film appeared to be poly-crystalline.

2GaCl3 þ N2/2GaNþ 3Cl2 (3.v)

The first “single-crystalline” epitaxial GaN films were reported by Kosicki andKahng6 andMaruska and Tietjen,7 and both were grown on sapphire (0001) substrates.Kosicki and Kahng6 used remote gaseous discharge to dissociate molecular N2 intoatomic nitrogen (N) and elemental Ga, which was similar to the method used byPasternak and Souckova,17 to obtain single-crystal epitaxial GaN layer at temperaturesabove 550�C. This paper, for the first time, identified 30 degrees rotation of basalhexagon of GaN along the c-axis with respect to sapphire to minimize the mismatchbetween them:

ð0001ÞGaN��ð0001ÞAl2O3

and�1210

�GaN

��1100

�Al2O3

(3.vi)

Maruska and Tietjen7 deposited the film by hydride VPE using HCl and NH3 by thefollowing simplified reactions:

Gaþ HCl /GaClþ 12H2 and GaClþ NH3/GaNþ HClþ H2 (3.vii)

Structural and electronic properties including lattice constants of a ¼ 3.189 Å andc ¼ 5.185 Å and a bandgap energy of Eg ¼ 3.39 eV at room temperature, which arevery close to currently accepted values, were measured. High electron concentrationof n ¼ 1 � 5 � 1019 cm�3 with a mobility of m ¼ 125e150 cm2/V-s was obtained,possibly due to high densities of defects such as point defects (vacancies andimpurities), line defects (dislocations), and even planar defects (grain boundaries).However, there was no description on surface morphology in this report, which wassuspected to be rough, not specular.

Since the first demonstration of single-crystalline thin-film GaN layers, manystudies were followed to further investigate the quality of the films using varioussynthesis methods. Faulkner et al.18 prepared thin-film GaN layers using vapor depo-sition by reacting GaCl3 and NH3 on silicon carbide (SiC) substrates. Manasevitet al.19 reported MOCVD growth of GaN and AlN on sapphire and SiC substratesin his publications on the growth using “metalorganics” via following simplifiedchemical reaction:

GaðCH3Þ3 þ NH3/GaNþ 3CH4 (3.viii)

This paper reported on the surface morphology of a GaN layer on a sapphiresubstrate, showing a rough surface including whiskers, which are essentially same

GaN on sapphire substrates for visible light-emitting diodes 45

as nanorods used in nanotechnology of these days. Wickenden et al.20 also reported onthe epitaxial growth of GaN on both sapphire and SiC substrates using both hydrideVPE and trichloride VPE. This study especially compared the growth of GaN onsapphire substrates with various orientations including (0001),

�1010

�, and

�1120

�,

which is a similar approach with current nonpolar and semipolar growth of GaN.The surfaces were not smooth on the (0001) sapphire substrate in this report. In parallelwith the development efforts of the epitaxial growth for high-quality GaN thin-filmmaterials,21 the investigations on optical, chemical, and electrical properties of newlydeveloped GaN layers were carried out.22e24 Especially, Pankove et al.25 reported, forthe first time, electroluminescence (EL) in GaN. They observed blue emission fromspots near a negative electrode at w2.6 eV (l w 475 nm) for a Zn-doped GaN layeron a sapphire substrate.

Besides the development of thin-film GaN growth, Zetterstrom26 synthesizedsingle-crystal GaN “needles” with dimension up to 1 mm thick and 5 mm in lengthfrom a specially treated GaN powder in a stream of NH3.

3.1.1.4 Device-quality GaN epitaxial growth

Although there had been extensive studies on GaN materials and their properties forpotential applications of visible light-emitting devices, the progress toward the“device-quality” GaN epitaxial layer was delayed, until the development of abuffer-layer concept for heteroepitaxial growth. Yoshida et al.27 and Amano et al.28

reported, for the first time, on the growth of device-quality GaN epitaxial layers onsapphire substrates employing a “buffer” layer. This buffer layer for strained heteroe-pitaxial growth is probably the most important progress in the development ofIII-N-based materials and device technology, which is further described in Section 3.3.

3.2 Sapphire substrates

Synthetic sapphire, whose chemical formula is Al2O3, is distinguished from alumina(aluminum(III) oxide or Al2O3) in the fact that sapphire is a single-crystalline material,whereas alumina is a generally poly-crystalline material. As described, single-crystalline sapphire substrates are the most dominantly used platform for the growthof GaN buffer layers in the LED heteroepitaxial structures. Besides the applicationsfor substrates of III-N LEDs, the sapphire is a popular choice for various semicon-ductor, electronics, and optics applications, due to the combination of favorablechemical, electrical, mechanical, optical, and thermal properties including thefollowing detailed properties29:

Resistant to chemical attack by a wide range of chemicals.High electrical resistivity even at elevated temperatures, for example, typically >1011 U-cmatw300K. However, this property along with low thermal conductivity (thermal conductiv-ity coefficient of <30 W/(mK) near room temperature) is not necessarily beneficial in theapplication of LEDs as a substrate.


High dielectric strength with dielectric constant of 11.5 and 9.3 in the directions parallelwith (//c) and perpendicular to (tc) c-axis, respectively, at 298K in the frequency rangeof 103e109 Hz.High compression strength ofw2 GPa (w3 � 105 psi), not necessarily high tensile strength(275e400 MPa), and high bending strengths (w1.03 GPa and w758 MPa in //c and tc,respectively).High hardness: Knoop hardness are1900 and 2200 kg/mm2 in //c and tc, respectively.High degree of refractoriness.

Besides these properties, in this section, the properties of sapphire relevant as aforeign substrate for III-N strained heteroepitaxial growth are described.

3.2.1 Properties of sapphire for substrates of III-N materials

3.2.1.1 Structural properties

Unlike gemstone sapphires having characteristic colors by the trace of other elements(i.e., impurities), synthetic sapphire used in electronics applications is produced in thepurest form of single-crystalline Al2O3, with no porosity and grain boundaries. Thefinal form of crystalline Al2O3 is also called a-alumina or corundum. The a-alumina(a-Al2O3) is one of the phases along with other metastable or transitional phases (b, g,h, q, k, c, etc.) in aluminum oxides and is the most thermodynamically stable phase.The corundum is also a representative name of the crystalline structure of a-Al2O3. Inthe corundum structure, the O2� anions form layers of a hexagonal closest-packedplane in parallel with a relative rotation angle ofw64.3 degrees (not ideal 60 degrees)and the Al3þ cations are located in two-thirds of octahedral sites between the adjacentO2� layers. The coordination number of Al3þ is 6 neighboring with 3 O2� in upperlayer and 3 O2� in lower layer, hence, filling a hollow in an octahedron consistingof 6 O2� anions. The Al3þ ions and remaining octahedral hollows also compose ahexagonal closest-packed plane. The corundum has a rhombohedral lattice structure(a ¼ b ¼ c and a ¼ b ¼ g s 90 degrees), if considering a primitive unit cell (10atoms in the cell); however, an equivalent hexagonal lattice unit cell, which containsthree primitive cells (30 atoms), is more widely used, and it is more convenient todescribe the epitaxial growth of hexagonal GaN along c-axis [0001]. Such hexagonalgeometry of atomic configuration on the basal plane of sapphire is similar to that onc-plane of thermodynamically stable GaN single crystal, which has a wurtzitestructure.

3.2.1.2 Chemical and thermal properties

Hexagonal atomic arrangement makes sapphire as a naturally preferred choice offoreign substrates for GaN heteroepitaxy. However, the hexagonal geometry is notthe only reason for making sapphire such a popular choice for the epitaxy of III-Nmaterials and visible LEDs. In fact, there are many other single-crystalline materialswith hexagonal lattices that can also serve as a substrate of III-N epitaxy. Some ofthem offer smaller lattice mismatch with GaN, as further described in Section 3.2.2.


In addition to the benefit of the similar crystalline structure, sapphire is a chemicallyand thermally stable material. With a melting temperature of 2323K (2030�C) andboiling temperature of 3253K (2980�C), sapphire maintains its stability at very hightemperatures of GaN buffer epitaxial growth higher than 1000�C. In particular,sapphire is very resistant to chemical attack by a wide range of chemicals even atelevated temperatures. In a typical GaN epitaxy by MOCVD, hydrogen (H2) iscommonly introduced as a carrier gas and a byproduct of hydride cracking. H2 canalso be generated in some MBE techniques using NH3. H2 at elevated temperaturesenhances the “thermal” etching of materials, resulting in decomposition of substratesurface. Therefore, for the growth of GaN on foreign substrates, maintaining thermaland chemical stability of the substrates at high temperature in H2 ambience is the firstrequirement as a viable platform of the epitaxial growth. In fact, even for the sapphiresubstrate, minor degree of surface decomposition is suspected to occur, releasingoxygen out of the sapphire surface during the heat up. This released oxygen is reincor-porated during the initial stage of GaN epitaxy, until it is depleted from the growthchamber, forming a thin GaN layer autodoped with oxygen near the sapphiresubstrates. Impurity and dopant depth profile for GaN layers on sapphire substratemeasured by secondary-ion mass spectrometry showed an oxygen concentration [O]tailing with a peak [O] w 1 � 1018 cm�3 decreasing to a detection limit of [O] <1 � 1016 cm�3 over a thickness range of w0.5 mm.30

In summary, important materials properties of sapphire together with GaN semicon-ductor are listed in Table 3.1.

3.2.2 Comparison to other substrates

Bulk GaN substrates32,33 are not readily available at a reasonable cost; hence, thedevelopment and manufacturing of GaN-based materials and structures have been

Table 3.1 Materials properties of sapphire and GaN7,31

Sapphire (a-Al2O3) GaN (hexagonal)

Crystal structure Rhombohedral Wurtzite

Lattice parameter, a 4.758 Å 3.189 Å

Lattice parameter, c 12.99 Å 5.185 Å

Thermal expansion coefficient a// 8.1 � 10�6K�1a 5.59 � 10�6K�1b

Thermal expansion coefficient at 7.3 � 10�6K�1a 3.17 � 10�6K�1c

7.75 � 10�6K�1d

Thermal conductivity 0.3 W/cm K 2.3 W/cm K

aMean thermal expansion coefficients in the range of 200e800�C.bMean thermal expansion coefficients in the range of 27e627�C (300e900K).cMean thermal expansion coefficients in the range of 27e427�C (300e700K)).dMean thermal expansion coefficients in the range of 427e627�C (700e900K).


focused on the heteroepitaxy of GaN on different materials that can be prepared in theform of high-quality single-crystalline substrate with a large surface area. For theselection of foreign substrates to be employed in the heteroepitaxial growth of GaN,several properties and conditions are generally considered, such as similarity incrystalline structure, in-plane lattice mismatch on a growth surface, thermal expansioncoefficient mismatch, chemical and thermal stability at elevated temperatures of GaNepitaxial growth, maturity of the substrate technology, manufacturability and scalabil-ity of the substrates considering maximum wafer size, production volume, and price(and also intellectual property of developed technology).

In this section, several alternative substrates to sapphire for the epitaxy of GaN andLED heterostructures are described with a focus on aforementioned characteristics. Wewill limit the description to the substrates for wurtzite GaN and III-N materials (notcubic III-N materials), with potential applications to LEDs (not to transistors or otherapplications).

3.2.2.1 SiC

SiC substrates are currently the second popular substrate next to the sapphire in III-N-based heterostructures and devices, especially for visible blue, green, and whiteLEDs34 and heterostructure field-effect transistors (also known as high-electronmobility transistors). Among many polytypes of SiC materials, 4H- and 6H-SiCsubstrates, mostly on a Si-face surface, are used for the heteroepitaxy of wurtziteGaN. They have hexagonal crystal structures (the same space group as wurtzite)and their in-plane lattice constants are closer to GaN (a ¼ 3.073 Å and a ¼ 3.081 Åfor 4H- and 6H-SiC, respectively) than that of sapphire. These lattice constantsmake in-plane lattice mismatch between GaN and SiC significantly smaller(w�3.5%) than that between GaN and sapphire (w14%). This lattice mismatch iseven further mitigated using an AlN buffer layer between the GaN layer and theSiC substrate, which is typical for the growth of GaN heteroepitaxy on SiC,35,36 asthe lattice constant of AlN is between the values of GaN and SiC. For the GaN on6H-SiC with an AlN buffer layer, the lattice mismatch is split into �0.96%and �2.45% for AlN on 6H-SiC and GaN on AlN, respectively. For the case ofGaN directly grown on SiC substrates, achieving layer-by-layer two-dimensionalgrowth is difficult possibly due to wetting problem of GaN on SiC or the onset ofStranskieKrastanow growth mode.37,38

The GaN epitaxial film grown on hexagonal SiC using the AlN buffer layer main-tains simple crystallographic relations of parallel basal planes and parallel faces ofhexagons,39 unlike those of GaN and sapphire:

ð0001ÞGaN==ð0001ÞAlN==ð0001ÞSiC (3.ixa)

�1010

�GaN

��1010

�AlN

��1010

�SiC or

�1120

�GaN

��1120

�AlN

��1120

�SiC

(3.ixb)


For the high-quality GaN materials on SiC, the growth condition and quality of AlNbuffer layer40,41 and the surface treatment of SiC substrates42e44 were reported to beimportant and the dislocation density in the GaN film employing optimized growthconditions was estimated to be slightly lower (on the order of as low as 107 cm�2)than that on sapphire substrates.45 Thermal expansion coefficient of SiC is generallysmaller than those of AlN and GaN. This difference in thermal expansion coefficientbetween III-N and SiC makes the lattice mismatch at growth temperature larger inaddition to lattice mismatch at room temperature. A GaN or AlN film grown at highertemperature should be under higher biaxial compressive stress. An AlN layer grown onSiC at temperature of higher than 1400�C was pilled-off from the substrates mainly bythe mismatch of thermal expansion coefficients.46 The strain status of a GaN layer withvarious thicknesses on SiC substrates employing an AlN buffer layer is very compli-cated,47 possibly related to growth mode and strain relaxation mechanism associatedwith the buffer layer.38

SiC materials have significantly higher thermal conductivities than sapphire. Thisproperty offers an important benefit of better heat dissipation; hence, improvedthermal management especially for high-power devices. In addition, unlike sapphireas an insulator, the energy bandgaps of 4H- and 6H-SiC fall on wide bandgapsemiconductors, where effective doping is possible for both conductive and semi-insulating substrates. These conductive substrates allow the implementation of back-side contact on the substrate for vertical geometry devices in the case of diodes suchas LEDs and laser diodes (LDs). However, the AlN buffer layer commonly used inGaN heteroepitaxy on SiC is quite resistive, if not insulating, even with controlleddoping for n-type conductivity. To avoid this resistive layer in vertical devices, aconducting Si-doped Al0.1Ga0.9 N layer between n-SiC substrate and n-GaN layerhas been developed.48

As described, SiC substrates have many advantageous characteristics and propertiesover sapphire substrates; however, in terms of substrate economy, SiC substrates havesignificant drawbacks compared to sapphire substrates. The SiC substrates arecommercially available only from selected manufacturers at relatively high prices.The maximum size of available substrates, as of the time of this writing, is 600 in diam-eter, but mainstream substrates are still limited up to 400.

3.2.2.2 Silicon (Si)

When it comes to substrate economy and maturity of substrate technologydnot thematurity of epitaxial growth of GaN on silicon (Si)dSi is the best choice for thesubstrates of III-N heteroepitaxy. Si substrates with a very large size (larger than1200 (maximum of 1800)) are available at a very low cost. Their quality in terms ofcrystalline perfection and surface condition is better than any other substrate materialsused in IIIeV materials and structures due to their matured technology. Si has a dia-mond crystal structure that belongs to face-centered cubic lattice in Bravais lattice sys-tem. A (111) plane, which is a preferred plane for the epitaxy of wurtzite-structure


materials, provides a hexagonal closest-packed plane on the growing surface.However, the crystalline quality of GaN epitaxial layers and performance characteris-tics of LEDs on Si substrates are still inferior to those on sapphire substrates, whileextensive efforts are being made to grow high-quality GaN films and to fabricatehigh-brightness LEDs on Si (111) substrates.

3.2.2.3 Other foreign substrates based on oxides, sulfides, andmetals

Several single-crystalline oxide materials other than sapphire have been considered asa candidate for III-N heteroepitaxy. Some of them such as ZnO and LiGaO2 havesmaller lattice mismatch between them and GaN, which was the compelling motiva-tion for the investigation. First, ZnO is the prime candidate for the substrate49,50 aswell as a compliant buffer layer51,52 and visible LED materials themselves.53 WurtziteZnO has the same crystalline structure as GaN. With a lattice constant of a ¼ 3.250 Å,its lattice mismatch (w1.9%) is significantly smaller than that of substrates describedabove.54 Additional mismatch induced by the difference in thermal expansion coeffi-cient is smaller than that of SiC substrates55 and this thermal mismatch inducedcompressive strain, compensating tensile strain from lattice mismatch at room temper-ature. The bulk ZnO substrates with large area (>300 diameter) from mass productionare currently not readily available. The development of GaN and LEDs on ZnOsubstrates may become important, considering continued and fast-pacing developmentof bulk ZnO to cope with increasing demands from optoelectronics and electronicsindustry.56 Possibly the biggest challenge of ZnO substrate for LEDs is associatedwith chemical and thermal stabilities. Decomposition of ZnO in the temperature rangeof typical GaN epitaxy byMOCVD andMBE is very severe, which can form Zn and Oimpurities in the subsequent GaN and LED structures.49 Also, many ZnO bulksubstrates contain high concentration of impurities that have generally high diffusioncoefficients. Such impurities in the heterostructure of LEDs by reincorporation anddiffusion during the growth the LED structure are suspected to provide high concen-tration of nonradiative recombination centers that limits the internal quantumefficiency (IQE) of III-N visible LEDs on ZnO substrates.

Another oxide substrate such as LiGaO257,58 shares the similar advantages

(i.e., small lattice match with Da ¼ 1.9% and Db ¼ �0.19%)59 and problems (i.e.,substrate economy and chemical and thermal stability).60 Other alternative oxide,sulfide, and even metallic substrates, including LiTaO3,

61 LiAlO2,62 LaAlO3,

63

(La,Sr) (Al,Ta)O3,64 NdGaO3,

65 Ca8La2(PO4)6O2,66 (Mn, Zn)Fe2O4,

67 MgAl2O4,68

MoS2,69 Hf,70 Cu,71 Ag,72 etc., have been investigated in attempts to find an alterna-

tive substrate with their own justification including better lattice and thermal matching,more economical substrates, better thermal dissipation, higher reflectivity, as so on.However, most studies failed to make breakthroughs to show them as a technologicallyviable substrate. No studies have shown any efficient light emitters except one study ofLDs on a spinel (MgAl2O4) substrate.

73


3.2.2.4 Native substrates

For “traditional” IIIeV semiconductors such as GaAs- and InP-based materials, theepitaxial growth generally begins with a homoepitaxial buffer layer grown on anative (hence, lattice-matched) substrate. The substrate is prepared from a single-crystal boule, which is a bulk material grown from a melt via either vertical gradientfreeze technique 74 or liquid-encapsulated Czochralski technique.75 In contrast, GaN-based materials are generally not grown on native GaN substrates. The bulk growthof GaN crystals in the same way that is employed in other IIIeV semiconductors(i.e., from a liquid melt) is extremely difficult due to very high vapor pressure ofnitrogen at the melting temperature of GaN.76 Alternate methods of GaN bulk crystalgrowth have been investigated, such as the ammonothermal crystal growth techniqueand the Na flux method.32,77,78 An alternative method of preparing GaN substrateshas also been developed employing a thick GaN layer grown on a foreign substrateby hydride VPE79 followed by removal of the foreign substrate and waferingprocesses.33 The GaN substrates prepared in such a way are called “free-standing”substrates to distinguish them from GaN substrates prepared by more traditionalbulk crystal growth technologies. Free-standing GaN substrates have relatively largediameter (w300 diameter) and low threading dislocation density as low as on the orderof 105 cm�280 and have become a dominant technology platform for the developmentof visible LDs in blu-ray players. “True bulk”GaN substrates from bulk GaN crystalsgrown using an ammonothermal process have recently been commercially developedwith a wafer diameter up to 200. However, current price and available wafer size arenot favorable for the adoption to the platform of LED technology despite signifi-cantly reduced dislocation densities. In addition, although reduced dislocationdensity must be beneficial in performance characteristics of LEDs,81 the degree ofbeneficial effect on IQE and efficiency droop of visible LEDs is stillcontroversial,82e88 even without considering the issues related to substrate economy.

In summary, various foreign and native substrates for GaN semiconductor materialfor the applications to LEDs are compared in Table 3.2.

3.3 Strained heteroepitaxial growth on sapphiresubstrates

This section describes a growth technique that finally overcomes one of the serioustechnical challenges in the epitaxial growth of GaN-based materials associated withlarge lattice mismatch between the GaN epitaxial layer and the sapphire substrate.This strained heteroepitaxy by two-step growth employing a buffer layer is probablythe most important technological progress in the development of III-N-based materialsand device technology. The section begins with development history of strainedheteroepitaxial growth by tenacious Japanese researchers90 and then discusses a mech-anism of growth evolution.


Table 3.2 Comparison of various foreign and native substrates for wurtzite III-N semiconductors for theapplications to LEDs59,61,68,89

Crystal structure

In-plane latticemismatch withGaNa

Thermal expansioncoefficient, a//

b

Thermalchemicalstability Substrate economyc

Sapphire(0001)

Rhombohedral 8.1 � 10�6K�1 Excellent Currently 3e400 diameter; 600 or 800developement in progress

4H-SiC(0001)

Hexagonal �3.6% Excellent High price; up to 400 (or 600) diameter;potentially intellectual property issue

6H-SiC(0001)

Hexagonal �3.4% Excellent

Si(111)

Diamond Medium Low price; large substrate (>1200diameter)

w-ZnO(0001)

Wurtzite 1.9% 6.5 � 10�6K�1 Problematic N/A (currently in research stage)

LiGaO2

(001)Orthorhombic 1.9% (a-direction)

�0.19% (b-direction)

Problematic N/A (currently in research stage)

MgAl2O4

(111)Spinel �10.3% N/A (currently in research stage)

GaN(0001)

Wurtzite 0% Same as GaN layer Excellent Very high price; up to 300 diameter

aIn-plane lattice mismatch is calculated using basal-plane lattice parameter, a (having hexagonal closest-packed plane) by a formula, (aS � aL)/aL, where aS and aL are bulk lattice parameters ofthe substrate and layer (before epitaxial growth), respectively. This formula yields negative value for biaxial in-plane compressive strain (aL > aS) and positive value for biaxial in-plane tensilestrain (aS > aL) applied in the epitaxial layer.bMean thermal expansion coefficients in the range of 200e800�C.cSubstrate economy includes maximum wafer size, production volume, price, and intellectual properties.

GaN

onsapphire

substratesfor

visiblelight-em

ittingdiodes

53

3.3.1 Concept development and demonstration of strainedheteroepitaxial growth

As briefly described in Section 3.1.1, the first “device-quality” GaN materials onsapphire substrates were developed by Japanese research groups using both MBEand MOCVD growth techniques.27,28 For the growth by reactive MBE using Al andGa molecular beams and NH3, a single-crystalline AlN layer grown at w1100�Cwas employed between a GaN epitaxial layer and a sapphire (0001) substrate insteadof growing a GaN layer directly on the surface of sapphire.27 Although this AlN layercan function as a buffer layer, Yoshida et al.27 did not use the term “buffer”. Instead,they used a term “AlN-deposited” sapphire. In this study, they focused on the latticemismatch and thermal expansion coefficient differences among sapphire, AlN, andGaN materials. From their previous experiments on AlN growth on sapphire sub-strates,91,92 they were able to produce single-crystalline AlN film having a smoothsurface without the formation of cracks or hillocks, even with a rather large in-planelattice mismatch and thermal expansion coefficient differences. This AlN layer (thick-nessw300 nm), having smaller lattice mismatch with GaN than the sapphire with GaNwas used as a buffer layer between the sapphire substrate and the GaN epitaxial layer toreduce the strain applied to the GaN layer. As described in Section 3.2.1, in-planelattice mismatch on (0001) plane isw2.4% andw13.9% for GaN on AlN and on sap-phire, respectively. In the same study, they reported significantly improved Hallmobility and luminescence peak from band-to-band transitions (at lw360 nm) bycathodoluminescence for the GaN on the AlN/sapphire compared to the GaN onsapphire, which supports the improved crystalline quality of GaN by the AlN bufferlayer.

3.3.1.1 Two-step strained heteroepitaxy

For the growth by MOCVD, Amano et al.28 reported, for the first time, an “opticallyflat” and “crack-free” GaN epitaxial layer on a sapphire substrate employing an AlN“buffer” layer between the substrate and the GaN epitaxial layer to improve thecrystalline quality of the GaN layer. Amano and Akasaki used the term a “buffer” layerin III-N epitaxial growth for the first time and they are credited for the development ofthe high-quality GaN layer by strained heteroepitaxy on foreign substrates.90 Althoughthis AlN buffer layer appears to be similar to that reported by Yoshida et al.27 usingreactive MBE, this layer is different in several aspects besides the use of differentgrowth technology. This AlN buffer layer was grown at relative low temperature toform fine crystallite and amorphous material in contrast to the single-crystallinematerial of the AlN grown at high temperatures as in reactive MBE and this low-temperature (LT) buffer layer is a layer of critical importance resulting in a smooth,specular, and crack-free surface of the GaN layer.93,94 In addition, its thickness(20e50 nm) was significantly thinner than the AlN by Yoshida et al.27 The reportespecially showed direct evidences of crystalline quality improvement with narrowfull width at half maximums (FWHMs) of the peak measured from symmetric andasymmetric X-ray rocking curves and optically smooth surface measured by scanning


electron microscopy. The strained heteroepitaxy on sapphire substrates consists of twosteps including a thin LT buffer, followed by a rather thick (>1 mm) high-temperature(HT-) GaN layer typically grown at 1000e1060�C; hence, it is often called “two-stepstrained heteroepitaxy” (“strained heteroepitaxial growth” or “strained heteroepitaxy”in short).

Subsequently, Nakamura95 demonstrated, for the first time, LT-GaN buffer layer(instead of LT-AlN buffer layer) grown between 450 and 600�C with a thickness of10e120 nm can play a similar role as the LT-AlN buffer layer. The study alsoreported a high-quality GaN layer in terms of smooth surfaces, high Hall mobility,and improved crystalline quality. This development of two-step strained heteroepi-taxial growth of GaN “template”, consisting of an LT-buffer (GaN or AlN) and anHT-GaN, paved a highway for further development of photonic and electronicheterostructures and devices based on III-N semiconductors, as these high-qualitytemplates are a platform for following epitaxial layer structures for operationaldevices.

3.3.2 Growth mechanism of GaN strained heteroepitaxialgrowth

After the demonstration of the GaN layer on sapphire substrates using the LT-bufferlayer, the growth mechanism of two-step strained heteroepitaxial growth was studiedin comparison to that of GaN growth directly on sapphire substrates. In the case ofdirect GaN growth, many hexagonal columns of crystalline GaN with different sizesand height are formed and they grow three dimensionally, resulting in rough surfaceand many pits at their boundaries. In contrast, the AlN (or GaN) LT-buffer layerused in the strain heteroepitaxial growth consists of fine hexagonal crystallites,amorphous-like structures, or zinc-blende crystallite with<111> direction, dependingon the deposition condition of LT-buffer, as measured by reflection high-energyelectron diffraction and cross-sectional transmission electron microscopy.96e99 Thezinc-blende crystallites contain high density of stacking faults and can partially beconverted to hexagonal crystallites after heating for HT-GaN growth.100 At thesame time, this LT-buffer layer promotes uniform surface coverage for the subsequentfilm growth. In the initial stage of GaN growth on the AlN LT-buffer, many truncatedhexagonal pyramidal mesas of GaN are formed. These mesas successively growquasilaterally, eventually resulting in a GaN flat surface.96 The role of the AlN bufferlayer is to decrease an interfacial free energy between the substrate and the epitaxialGaN layer and to supply nucleation sites for GaN with the same crystal orientationas the substrate.101 Therefore, the LT-buffer layer is also referred to as a “nucleationlayer”. Fig. 3.1 schematically shows the growth mechanism of the GaN heteroepitaxialgrowth on a sapphire substrate using an LT-buffer layer. The growth proceeds withfollowing steps: (1) nucleation of high-density GaN; (2) geometric selection of thecrystallographic direction of GaN columnar crystals; and (3) highly lateral growthvelocity of the trapezoid islands with c-face on top. This nonuniform growth ofGaN on AlN plays an important role in the realization of uniform growth and in obtain-ing high-quality GaN layers with significantly reduced defects.102 This growth


mechanism was also experimentally confirmed by in situ measurement of the growingsurface to avoid possible effects on surface morphology during the cool down forex situ measurement.103

3.3.3 Wafer bowing during the growth of GaN on sapphire

Strained heteroepitaxy of GaN on foreign substrates inevitably causes macroscopicstress in the epitaxial structures resulting in wafer bowing, misfit dislocation forma-tion, and sometimes crack formation. The bowing of the wafers occurs from twoorigins: (1) lattice mismatch between layers and substrate and (2) difference in thermalexpansion coefficients between III-N layer materials and foreign substrates. Excessiveresidual bowing in the wafers after the completion of the epitaxial growth can raiseadditional technical issues during the device fabrication and packaging processes,such as misalignment problem during photolithography and more challengingwafer-to-wafer bonding for flip-chip LEDs by wafer-level chip-scale packaging.104

Residual stress in GaN can be measured by ex situ wafer curvature measurement,105

(1)

(2)

(3)

(4)

(5)

(6)

AlN buffer layer

Nucleation of GaN

Geometric selection

Island growth

Lateral growth

Uniform growth

AlN

GaN

- Al2O3α

Trapezoid crystal

Dislocation

Sound-zone

Semi-sound-zone(∼150 nm)

Faulted-zone(∼50 nm)

AlN (∼50 nm)

Figure 3.1 Schematic diagrams showinggrowth process of GaN strainedheteroepitaxial growth on a sapphiresubstrate using low-temperature bufferlayer.102


high-resolution X-ray diffraction,106 and Raman spectroscopy.107 However, suchex situ techniques cannot measure the changes in stress and curvature during theepitaxial growth. The bowing during the growth causes a variation in the temperatureof growing surface of the wafer, which is one of the most important parameters inepitaxial growth, by localized thermal contact between the substrate and wafer holder(also called susceptor or wafer carrier).108 Lateral temperature variation will result innonuniform composition, thickness, and crystalline quality in the epitaxial wafers.109

In particular, the lateral uniformity of composition for InGaN quantum wells (QWs) invisible LEDs is strongly dependent on the temperature uniformity of the growingsurface, as the incorporation of indium into solid phase in MOCVD is very sensitiveto the temperature.110 Therefore, in situ monitoring of wafer curvature during thestrained heteroepitaxy using deflectometry became popular together with other insitu measurements including emissivity-corrected pyrometry for the measurement oftemperature and reflectometry for the measurement of layer thickness.111 In thissection, we will focus on in situ deflectometry.

Deflectometry employs multibeam optical sensors as schematically shown inFig. 3.2.112 The curvature in wafer (k) is measured from divergence (dd) of an arrayof initially parallel beams using the following equation:

k ¼ 1R¼ dd

d

cos a2L

(3.1)

The sign of the curvature and curvature radius (R) are usually defined positive andnegative for the concave and convex conditions, respectively. Fig. 3.3 shows thechange in curvature of a wafer consisting of III-N layers on a sapphire substrate duringheat-up, epitaxial growth, and cool-down steps. Concave bowing is observed duringthe initial heat-up of sapphire to w1000�C prior to the growth. This bowing is notassociated with the lattice and thermal expansion mismatches, as no film is deposited.It is caused by vertical temperature gradient across the substrate due to low thermalconductivity of sapphire.114 When AlGaN and InGaN layers are grown, the wafer

R > 0R < 0

d

Lα

δd

tfts

Film

Substrate

Wafer carrier

Figure 3.2 Schematic illustration on the principle of wafer curvature measurement.


shows more concave and convex bowing, respectively, which is in line with in-planetensile and compressive stress/strain of the layers on GaN. The degree of bowingbecomes more significant with increasing mole fractions of AlN and InN for AlGaNand InGaN layers, respectively.113 When the wafer is cooled down after the growth,convex bowing is induced due to smaller thermal expansion coefficient of GaN(aa,sapphire ¼ 5.59 � 10�6 K�1) than that of sapphire (aa,sapphire ¼ 8.3 � 10�6 K�1).

During the growth of GaN buffer on sapphire, concave bowing is observed. In situstress of the film can be calculated using Stoney’s equation:

sf tf ¼ Mst2s6

k (3.2)

where sf is the stress of the film, tf and ts are thickness of the film and substrate,respectively, and Ms is the biaxial modulus of the substrate. Because the measuredcurvature is a product of average stress (sf) and thickness (tf) of the film, the stress isevaluated based on the thickness of the film from in situ reflectometry. The GaN layeron sapphire is under tensile stress in the range of 0.14e0.29 GPa, regardless of thetypes of low-temperature layer (LT-GaN or LT-AlN) used in the strained hetero-epitaxy.115 The value of tensile stress is varied depending on the island coalescenceconditions during the strained heteroepitaxy.113,116 When the GaN layer is dopedwith Si, which is a conventional process for visible LED epitaxy, the induced tensilestress is higher than that of unintentionally doped GaN. If the doping level is veryhigh (e.g., 2 � 1019 cm�3), the Si-doped GaN film even forms cracks when it isgrown thicker than 2 mm,117 which accompanies high density of dislocations and

1200

1000

800

600

400

200

150

100

50

0

–50

–100

10,000 15,000 20,000Run time (s)

Cur

vatu

re (1

/km

)

T (°

C)

Satellite temperature

Curvature as measured

Curvature corrected(pure film stress bow)

Substrate bow

GaN AlGaN/GaN InGaNGro

wth

Des

orpt

ion

Figure 3.3 Transients of temperature and curvature during growth of an optoelectronic devicestructure on sapphire substrate.113


modification of surface structure.118 The wafer bowing behavior of GaN grown on aSi substrate is different from that of GaN on sapphire.119

3.3.4 Buffer layer grown by physical vapor deposition

A nucleation layer in strained heteroepitaxy is usually a thin GaN or AlN layer(20e50 nm) that is grown at low temperatures (500e800�C) by MOCVD. In thissection, such LT-buffer is referred to as an in situ nucleation/buffer layer becausethe layer is grown in the same batch of the MOCVD run with other layers grown athigh temperatures.

Recently, a new nucleation layer deposited by physical vapor deposition (PVD) wasdeveloped. The first PVD-deposited nucleation layer was demonstrated by Lee et al.120

using CrN on a sapphire substrate by reactive radio frequency (RF) sputtering with Crtarget and N2. An AlN layer was developed as an ex situ nucleation layer on a sapphiresubstrate for strained heteroepitaxy.121 A 25-nm-thick AlN layer was deposited by RFsputtering using AlN target. The sputtered AlN on sapphire was transferred toMOCVD for the growth of HT-GaN. The crystalline quality of GaN on sapphire usingthe ex situ sputtered AlN nucleation layer is significantly better than that using in situLT nucleation layers. The linewidths of the peaks from the crystalline plane of GaN aresignificantly narrower, suggesting reduced density of threading dislocations. Forexample, the FWHM are 101 arcsec and 110 arcsec for (002) and (102) GaN peaks,respectively, for a GaN layer on a patterned sapphire substrate (PSS) using theex situ nucleation layer, which are narrower than 230 arcsec and 240 arcsec for aGaN layer on a PSS using the in situ nucleation layer.122 As a result of defect reductionby ex situ nucleation layer, the LEDs show improved light output power and higherquantum efficiencies.122 Also the LEDs with ex situ AlN layer show lower reverseleakage currents and better resistance to the damage by electrostatic discharge.121

3.4 Epitaxial overgrowth of GaN on sapphire substrates

As described in Sections 3.1 and 3.2, epitaxial growth of GaN is generally not carriedout on native GaN substrates, unlike other semiconductors, epitaxial growth of whichgenerally begins with a homoepitaxial buffer layer grown on a lattice-matched same-material substrate. Instead, GaN-based epitaxial structures have been grown a GaNbuffer layer that is grown on a foreign substrate (sapphire, SiC, and Si) by strainedheteroepitaxial growth,28,36,123 which brought transformational changes in GaN-based materials and devices. However, even with this buffer layer, the dislocationdensity in the epitaxial materials is on the order of mid-108‒low-109 cm�2 due toboth lattice constant and thermal expansion mismatches, which is significantly higherthan those of other semiconductors with values on the order magnitude in 105 cm�2

and lower. Such dislocation density suggests that typical LED devices should containmore than 90,000 and 1,000,000 dislocations per device with dimensions of300 � 300 mm2 and 1 � 1 mm2, respectively, with an assumption that the dislocation


density is higher than 1 � 108 cm�2. Although this high dislocation density of III-Nepitaxial structures does not prevent devices from working as a photon emitter, it limitstheir full potential, as these defects act as current leakage paths and carrier traps,limiting the performance characteristics of transistors124,125 and providing avenuesof nonradiative recombination for light emitters.88,126

To further improve the crystalline quality of GaN-based epitaxial structures, reduc-tion in dislocation density is required. Several growth techniques using mainlyepitaxial overgrowth schemes for the reduction of dislocation density are describedin this section, including epitaxial lateral overgrowth (ELOG), pendeo-epitaxy, andPSS.

3.4.1 Selective area growth and epitaxial lateral overgrowthof GaN

Selective area epitaxial growth technique was reported and actively used in manymaterial structures and device applications based on traditional (GaAs- and InP-based) IIIeV semiconductors.127,128 For instance, higher order quantum-confinedstructures (than one-dimensionally confined quantum-well structures), such as quan-tum wires, quantum dots, and nanorods, can be fabricated using epitaxial growth inselected defined area. Buried heterostructures employing regrowth of lateral claddinglayers on etched surface for mesa definition are commonly used in LDs, commonlyreferred to as buried heterostructure LDs, for enhanced confinement of currents,photons, and carriers.129

The terms used in this section need to be clarified. Selective-area growth (SAG),which is also called selective growth (SG), selective epitaxy (SE), selective-areaepitaxial growth (SAEG), and selective epitaxial growth (SEG), involves epitaxialgrowth of materials on nonmasked (window) regions but does not necessarilyinclude lateral growth over the masked regions. When the lateral overgrowth overthe mask region is involved, the process is defined as lateral epitaxial overgrowth(LEO), lateral epitaxial growth (LEG), lateral overgrowth (LOG), ELOG or ELO,or selective-area lateral epitaxial overgrowth (SALEO). We use the term ELOG inthis chapter.

For III-N materials, Kato et al.130 and Kitamura et al.131 were the first group in theliterature who demonstrated the SAEG of GaN and AlGaN on stripe-patterned anddot-patterned windows using a SiO2 mask. In these reports, they demonstrated excel-lent selectivity of GaN growth only on GaN surface in window region (not on SiO2

surface) and self-limited stable formation of�1011

�facets for selectively grown

GaN materials. The growth selectivity of GaN (unlike AlGaN) is due to a differencein sticking coefficient, s, of gallium (Ga) adatoms (or its species) on the GaN surface(sw1) and the SiO2 surface (sw0). This sticking coefficient difference is related tobonding energy differences in SiO (799.6 kJ/mol), GaO (353.6 kJ/mol), SiN(439 kJ/mol). Therefore, a nucleation of GaN on SiO2 surface, replacing Si with Gais not energetically favorable; thus, is not likely to occur. However, those reportsfocused on SAG and no lateral epitaxial overgrowth was carried out.


Zheleva et al.132 and Nam et al.133 are credited for the first demonstration ofdislocation reduction effect by ELOG in III-N materials by MOCVD. They firstdemonstrated that lateral epitaxial growth occurs over the SiO2 mask after initialvertical growth in the window region, as GaN growth proceeds. Although threadingdislocations in underlying GaN in window region keep propagating through verticalgrowth area, lateral overgrowth region contains very low density of dislocations.Dislocations underneath the SiO2 are blocked by the mask due to termination of theepitaxial growth. Especially, Nam et al.133 investigated the effect of orientations ofstripe-patterned mask between

1120

(a-direction) and

1100

(m-direction) on a

(0001) c-plane GaN surface and demonstrated that rectangular GaN stripe having a(0001) top facet and

�1120

�side facets are grown for the mask opening oriented along

a1100

direction. This rectangular GaN can be further developed to a coalesced GaN

layer by merging two lateral growth fronts (with�1120

�side facets) from two

adjacent rectangular GaN stripes, as schematically shown in Fig. 3.4(a). This coalescedlayers results in a flat surface of GaN with reduced dislocation density in certain(laterally grown) area, which can offer a high-quality template for the subsequentgrowth of heteroepitaxial structures for devices. In a separate research, Usui et al.134

also reported dislocation reduction in a thick GaN layer (>140 mm) overgrown usinga SiO2 mask by hydride VPE. The density of dislocations decreased as low as6 � 107 cm�2 obtained in this film. However, the report did not describe the mecha-nism of dislocation reduction effect by ELOG, and the reduction in dislocation densitymight stem from a thick layer grown by hydride VPE.

GaN layers overgrown by SAG and/or ELOG can have different structures depend-ing on growth parameters and pattern geometries. The orientation of stripe maskpattern results in stripes with different stable facets. Such different structures originatefrom different growth rates of several planes depending on the orientation ofmasks.135,136 This growth rate of each facet can also be changed by growth tempera-ture and pressure, as schematically shown in Fig. 3.5.137,138 For the stripe masksoriented along

1120

direction, triangular cross-sectional stripes of GaN with stable�

1011�facets are formed independent of growth pressure and temperature, which

suggest that the growth rates on�1011

�planes are almost constant, not the function

MaskGaN

Foreign substrate

MaskGaN

Foreign substrate

WindowWing

Void

(a)

(b)

Figure 3.4 Schematic diagrams of (a) lateral epitaxial overgrowth and (b) pendeo-epitaxy ofGaN.


of pressure and temperature. However, for the stripe masks oriented along1100

direction, the formation of several facets, (0001),�1120

�, and

�1122

�, and their

resulting structures are a strong function of growth pressure and temperature.Triangular cross-sectional stripes with

�1122

�facet became more stable with higher

pressure and lower temperature, whereas rectangular cross-sectional stripes with(0001) and

�1120

�facets become more stable with lower pressures and higher

temperatures. These different structures are the result of different growth rate of severalfacets.

3.4.2 Pendeo epitaxy of GaN

Although ELOG technique brought innovative changes in reduction of dislocationdensity for GaN templates grown on foreign substrates, the beneficial effect of dislo-cation reduction is not uniform, but localized in selected areas. This areal dependenceof dislocation density makes the placement of devices confined only on laterallyovergrown region. To overcome such limitation of ELOG, a new approach of selec-tive growth was proposed for more uniform and large-area dislocation reduction.Pendeo (from the Latin, meaning “hangs on” or “suspends from”) epitaxy, a termcoined by Zheleva et al.139 and Linthicum et al.140 to describe lateral growth withoutcontact with a mask or substrate is an extension of previous ELOG to employ thesubstrate itself as a “pseudo-mask”. This approach, however, differs from previousELOG in that growth does not initiate through the open windows but begins onside-walls etched in the GaN layer which acts as a seed layer of pendeo. The firstpendeo-epitaxy employed a maskless approach only relying on preferred growthof certain facets of GaN.139 A GaN layer grown by strained heteroepitaxy wasstripe-patterned along

1100

direction and etched through the GaN layer into the

foreign substrate to form a seed GaN layer with top surface of (0001) facet andside walls of

�1120

�facets. Pendeo-epitaxy was carried out with a growth condition

of high temperature (1080�C) and low pressure (45 Torr) in order to facilitate thegrowth of preferred facets in lateral and vertical growth. The pendeo-epitaxial

Lower Tg

Lower PgHigher Pg

Lower Tg

Higher Pg

Higher Tg

Lower Pg

Higher Tg{1101} {1122} {1120}(0001)

[0001] [0001]

<1100><1100><1120>

<1120>

Figure 3.5 Schematic diagrams of evolution of selective area and lateral epitaxial overgrowthdepending on the orientation of stripe masks. Tg and Pg are temperature and chamber pressurefor the growth, respectively.


materials grow laterally as well as vertically and coalesce. During the growth ofpendeo-epitaxy, nucleation, and subsequent growth of GaN on the surface of theforeign substrate does not occur, which confirms the foreign substrate can act as apseudomask. This lateral growth in combination with no growth on the foreignsubstrate results in a “free-standing” (as it is not bound to the substrate) and disloca-tion reduced (as dislocations do not propagate laterally) layer from pendeo-epitaxy.The dislocation density of the lateral growth region is lower by the order of104e105 cm�2 of magnitude. However, the dislocation density of vertical growthregion is the same as typical GaN layer on foreign substrates (w109 cm�2), whichindicates that areal nonuniformity of dislocation density still exists.

To overcome the problem with remaining high dislocation density region, for thesecond pendeo-epitaxy demonstration, a SixN1�x mask was deposited by plasma-enhanced chemical vapor deposition (PE-CVD) on (0001) face of the GaNseed.140,141 No growth occurs on the SixN1�x mask and on exposed foreign substrates.Fig. 3.4(b) schematically shows structural development of GaN during the pendeo-epitaxy. For the initial stage of the pendeo-epitaxy with high growth temperatureand low pressure for enhancing lateral growth in combination with the effects ofSixN1 � x mask and foreign substrate pseudo-mask, only lateral growth on

�1120

�

facets are forced resulting in GaN grown by peodeo-epitaxy suspended from theside-walls of the GaN seed layer. These suspected GaN materials have exposedGaN (0001) facet that is not covered by the mask and vertical growth of GaN occurstogether with lateral growth with certain relative growth rates controlled by the growthparameters during pendeo-epitaxy. When the height of the GaN exceeds the thicknessof the SixN1�x mask, ELOG over the mask and eventual coalescence of pendeo-epitaxial grown GaN over the mask occur to complete the epitaxial growth with aflat continuous surface with uniform reduced dislocation density. The dislocationsare buried under the mask.

3.4.3 Growth of GaN on patterned sapphire substrates

Among many overgrowth techniques described in Section 3.4, the growth of GaN onPSS is a far more popular approach in the applications of the visible LEDs. The PSS,which is an inherently mask-free process, is a process free of potential contaminationfrom the mask and without the interruption during epitaxial growth. It can also reducethe threading dislocation density in the LED epitaxial structures. Moreover, possiblymore importantly, the LEDs on PSS show enhanced light extraction efficiency(LEE) compared to the LEDs on conventional flat sapphire substrates (FSS), due tooptical scattering effect of the patterns of sapphire/GaN interface.142 Many differenttypes of PSS technology with patterns of stripes,143 holes,144 microlens,145 and pyra-midal shapes146 using dry etching147 and wet etching148 have been studied andsuccessfully employed in the fabrication of high-brightness and high-power LEDs.The shapes, periodicity, aspect ratio of patterns significantly affect the LEE in relationwith LED device geometry, which is further discussed in Chapter 6 in the context ofLED fabrication and characterization. In this section, we focus on the characteristics ofthe GaN epitaxial growth on PSS.


Sapphire substrate patterning for the growth of GaN buffer layer was first reportedusing stripe-patterned deep trenches on the substrates.149,150 Ashby et al.149 employeda narrow stripe patterns with etched trenches for their coined term, cantilever epitaxyas a dielectric-mask-free alternative of the ELOG and pendeo-epitaxy. For the canti-lever epitaxy, the depth of the trenches is deep enough and the growth temperatureafter the LT buffer growth is high enough to achieve the coalescence of lateral growthfronts of unetched region before vertical growth in the etched trench region interferewith the cantilever growth. The density of dislocation in the cantilever region wassignificantly reduced to <107 cm�2. In addition, the voids at the coalescence frontsover the dielectric mask, which commonly occurs in the ELOG, were not observed.The growth of the LEDs on PSS was first reported by Tadatomo et al.151 The c-planesapphire substrates were patterned with parallel groves along two different directionsof

1120

sapphire and

1100

sapphire. Similar to the crystallographic relationship

between the GaN layer and sapphire substrate, as described in Section 3.2, GaN stripesalong the

1100

GaN and

1120

GaN were formed along the PSS with

1120

sapphire

and1100

sapphire patterns, respectively. The overgrowth evolution of GaN stripes

along different directions is very similar to that of ELOG shown in Fig. 3.5(a) and(b). In addition to the overgrowth behavior, Tadatomo et al.151 also reported reduceddensity of dislocations (1.5 � 108 cm�2 reduced from 4 � 108 cm�2 for FSS) andincreased light output power of near-UV LEDs.

Yamada et al.142 reported two-dimensional hexagon array of patterns (as opposedto one-dimensional array of stripe patterns), which is similar to a honeycomb structure,for visible LEDs. The side of hexagon was arranged in parallel with the

1120

sapphire

a-axis of the sapphire for improved planarization of GaN overgrowth buffer layer.Moderate improvement of lower dislocation density (4.8 � 108 cm�2 reduced from1 � 109 cm�2 for FSS) and significant improvement in external quantum efficiency(EQE) by w35% were reported, when the PSS is combined with a meshp-electrode in the LED device structures.

Cone, hemisphere, or lens shapes are most common patterns in PSS for visibleLEDs. The growth and characterization behaviors have been extensively studied. Inthis case, the nucleation and growth of GaN occur on the patterns and trenched flatsurface (i.e., (0001) plane) of the PSS. During the deposition of LT buffer, GaNwas dominantly nucleated on the flat surface and the GaN is grown as a film.152,153

On the patterns, truncated pyramidal GaN with facets of (0001) plane (top facet)and

�1011

�planes (six-side facets) were nucleated and grown.154 As the growth pro-

ceeds, GaN grown from the flat trench surface started to grow laterally to cover thepyramidal GaN from the patterns, which is a different process from that of cantileverepitaxy. As a result of the ELOG over the pattern, the density of threading dislocationis reduced.153,155

Optical properties of the GaN and visible LEDs on PSS in relation to defect forma-tion were studied.156,157 Reduced density in dark spots of luminescence was observedfor the film on PSS. Most dark spots were located on flat trench regions and apexes ofpatterns, which are related to threading dislocations or coalescence points. Wafer


bowing during the GaN on PSS is different from that on flat sapphire substrates. Acompressive stress (0.21 GPa) was observed during the lateral coalescence of GaN,followed by a tensile stress after the coalescence.158 This stress evolution is differentfrom the GaN growth on conventional FSS where consistent tensile stress wasobserved. This affects the residual stress of wafer after the growth on PSS.159

3.5 GaN growth on nonpolar and semipolar directions

Epitaxial growth of III-N films on sapphire substrates has been dominantly devel-oped on (0001) c-plane sapphire substrates for the layers and structures grown in apolar direction of <0001>. More specifically, the films are grown in the ½0001�direction to have a (0001) cþ-plane (a Ga-face plane) on top surface, instead of ac�-plane (an N-face plane) in the

�0001

�direction. The GaN film with Ga face is

relatively easier to achieve improved surface morphology and reduced crystalline de-fects by the stained heteroepitaxy on foreign substrates. As a result, the GaN bufferlayers on foreign substrates that have offered universal platforms for the heterostruc-tures of III-N light emitters and transistors are mostly the Ga-face GaN grown in the½0001� direction.

These III-N heterostructures grown in the polar direction, however, induce polar-ization charges near the interfaces by spontaneous and piezoelectric polarizations.160

The charges modify the electronic band structure profile in III-N heterostructures byinducing internal fields, which results in the formation of two-dimensional electrongas in AlGaN/GaN heterostructures without introducing modulation doping (this issometimes called “self-doping” effect) 161 and quantum-confine Stark effect(QCSE)162 in the QWs.163 For III-N visible LEDs, the QCSE in InGaN/GaN multipleQW (MQW) active region results in narrowing of effective bandgap of the QWs,blue-shift of the peak in photoluminescence and EL with increasing injections,and spatial separation of electron and hole wave functions.164 In particular, thespatial separation of the wave functions, leads to a reduced overlap between thewave functions of the two charged carriers which, in turn, reduces the oscillatorstrength for the recombination of the carriers.165,166 To improve the transition prob-ability for radiative recombination and IQE, the thickness of III-N QW has to bethinner than that of nonpolar QWs as in GaAs- and InP-based materials. Evenwith thinner QWs, the wavefunction separation cannot be completely suppressedand these thinner QWs may bring about other problems such as reduced capturingrate of electrons in combination with polarization-induced electronic bandbending167 and higher carrier density that could enhance nonradiative Auger recom-bination rates in the QWs.168

To address the technical challenges associated with the QCSE in InGaN QWs,nonpolar and semipolar heterostructures have been proposed for the LEDs andLDs,169 which requires the development of nonpolar and semipolar GaN. This sectiondescribes the growth of nonpolar and semipolar GaN on sapphire substrate withsurfaces other than c-planes in relation to their implication on polarization field anddevice performance characteristics.


The�1010

�,�1120

�, and

�1012

�are nonpolar or semipolar planes and designated

as m-, a-, and r-planes, respectively, in wurtzite structure. The epitaxial growth alongthose directions became important especially for light emitter applications withcompelling motivation described earlier. Interestingly, the epitaxial growth innonpolar directions has been carried out almost about the same time as the develop-ment of strained heteroepitaxial growth in an attempt to improve crystalline qualityof GaN film on foreign substrates,170 not necessarily for the mitigation of electrostaticfields in the QWs. Crystal orientation effect on the piezoelectric field and electronicproperties of III-N materials and QWs171,172 suggests that polarization-induced inter-nal field should be absent or mitigated depending on the growth plane. Driven by aninterest in studying the importance of these expected beneficial effects, the growth ofGaN, InGaN/GaN MQW, and LED structures with nonpolar and semipolar orienta-tions of the wurtzite crystals has been explored by the growth of III-N films on variousforeign substrates.

The nonpolar�1120

�a-plane GaN was grown on

�1102

�r-plane sapphire sub-

strates173,174 and on a-plane sapphire175 and SiC176 substrates. The nonpolar�1010

�

m-plane GaN was grown on m-plane sapphire177 and SiC178 substrates. For a-planeGaN on sapphire substrates, a two-dimensional layer-by-layer growth was obtainedwith following crystallographic relations:

�1120

�GaN

��1102

�sapphire and ½0001�GaN

��1101

�sapphire (3.x)

However, the epitaxial film contained high density of crystalline defects includingnot only high threading dislocation density of>2 � 1010 cm�2 but also high density ofstacking faults of 2 � 1010 cm�1.179 Higher density of crystalline defects in the filmgrown on nonpolar planes than that on a polar plane is related to anisotropicin-plane strain in the films.180 To reduce the density of crystalline defects,ELOG,181 insertion of metal interlayer,182 sidewall epitaxial overgrowth,183 in situSiN nanomask,184 and two-stage epitaxial lateral overgrowth185 have been employed.

In general, however, the nonpolar growth of GaN and InGaN materials via strainedheteroepitaxy on sapphire substrates suffered from poor crystalline quality with layerscontaining high densities of threading dislocations and stacking faults and rough sur-face morphologies. In the case of ELOG materials, the reduction in the threading dislo-cation densities was reported; however, stacking faults could not be eliminated.186 As aresult, meaningful device (with reasonable quantum efficiency) data from LEDs grownon nonpolar sapphire substrates have not been reported. Recently, LED structures weregrown on nonpolar and semipolar free-standing native GaN substrates. Stable peakpositions (i.e., no blue-shift) of EL with increasing current were demonstrated fromthe nonpolar visible LEDs.187,188 This low defect density substrate made a break-through in achieving device quality epitaxial materials for nonpolar growth. Externalquantum efficiencies of nonpolar LEDs grown on m-plane native GaN substrates weremeasured. The peak EQE of the LED operating at lw407 nm was reported to bew40%.189 A real benefit of nonpolar or semipolar structures may be found in theLDs. The long wavelength continuous wave lasing actions at lw481 nm (blueegreen)


and lw520 nm (green) were achieved at room temperature on nonpolar and semipolarGaN substrates.190,191

The results of LEDs and LDs demonstrated that native nonpolar and semipolar GaNsubstrates are encouraging, but there are technical and economic issues to be resolvedin order for these devices to be successful in the market. Even though quantumefficiency has been improved significantly, the reported state-of-the-art values ofquantum efficiency for nonpolar and semipolar LEDs are still lower than those ofwell-developed polar LEDs. In addition, the biggest hurdle for the commercializationtoward useful SSL devices, especially for LEDs, on native nonpolar substrates is theextremely high cost and small size of the substrates.

3.6 Outlook of LEDs on sapphire substrates

As described so far, the GaN on sapphire platform has been dominant in the applica-tion of LED components and modules and will continue to serve thanks to theircurrently optimized characteristics between technological maturity (as compared toGaN on Si platform) and advantages in volume and cost (as compared to GaN on otheralternative substrates including native GaN substrates). The size of sapphire substrateshas been increased from 200 diameter to 1200 diameter, while price has been decreased.Currently (as of time of this writing), successful transition to 400 diameter has beenmade by many LED manufacturers and the trend of moving larger wafer size toward600 and 800 diameter will continue to maintain price competitiveness of LEDs by savingmanufacturing cost. At the same time, GaN on sapphire platform as dominant technol-ogy in LEDs may face serious challenges from other alternative substrates, especiallyfrom the GaN-on-Si platform.

References

1. Dupuis RD, Krames MR. History, development, and applications of high-brightnessvisible light-emitting diodes. J Lightwave Technol 2008;26:1154e71.

2. Davis RF, Einfeldt S, Preble EA, Roskowski AM, Reitmeier ZJ, Miraglia PQ. Galliumnitride and related materials: challenges in materials processing. Acta Mater 2003;51:5961e79.

3. DenBaars SP. Gallium-nitride-based materials for blue to ultraviolet optoelectronicsdevices. Proc IEEE 1997;85:1740e9.

4. Mishra UK, Parikh P, Wu Y-F. AlGaN/GaN HEMTs e an overview of device operationand applications. Proc IEEE 2002;90:1022e31.

5. Pearton SJ, Ren F. GaN electronics. Adv Mater 2000;12:1571e80.6. Kosicki BB, Kahng D. Preparation and structural properties of GaN thin film. J Vac Sci

Technol 1969;6:593e6.7. Maruska HP, Tietjen JJ. The preparation and properties of vapor-deposited single-

crystalline GaN. Appl Phys Lett 1969;15:327e9.8. Manasevit HM. Recollections and reflections of MO-CVD. J Cryst Growth 1981;55:1e9.


9. Manasevit HM. (Rockwell International Corporation). Epitaxial composite and method ofmaking. US patent 4368098C; January 11, 1983.

10. Manasevit HM. Single-crystal gallium arsenide on insulating substrates. Appl Phys Lett1968;12:156e9.

11. Manasevit HM. The use of metalorganics in the preparation of semiconductor materials:growth on insulating substrates. J Cryst Growth 1972;13/14:306e14.

12. Johnson WC, Parsons JB, Crew MC. Nitrogen compounds of gallium III. Gallic nitride.J Phys Chem 1932;36:2651e4.

13. Hahn H, Juza R. Untersuchungen €uber die nitride von cadmium, gallium, indium undgermanium metallamide und metalinitride. Z Anorg Allgem Chem 1940;244:111e24.

14. Renner T. Herstellung der nitride von bor, aluminium, gallium und indium nach demaufwachsverfahren. Z Anorg Allgem Chem 1959;298:22e33.

15. Addamiano A. On the preparation of the nitrides of aluminum and gallium. J ElectrochemSoc 1961;108:1072.

16. Lorenz MR, Binkowski BB. Preparation, stability, and luminescence of gallium nitride.J Electrochem Soc 1962;109:24e6.

17. Pasternak J, Souckova L. Herstellung d€unner schichten von aluminium-, gallium, sowieindiumnitrid unter einer gasentladung. Phys Stat Sol 1963;3:K71e4.

18. Faulkner KR, Wickenden DK, Isherwood BJ, Richards BP, Scobey IH. Gallium nitrideformed by vapour deposition and by conversion from gallium arsenide. J Mater Sci 1970;5:308e13.

19. Manasevit HM, Erdmann FM, Simpson WI. The use of metalorganics in the preparation ofsemiconductor materials: IV. the nitrides of aluminum and gallium. J Electrochem SocSolid State Sci 1971;118:1864e8.

20. Wickenden DK, Faulkner KR, Brander RW, Isherwood BJ. Growth of epitaxial layers ofgallium nitride on silicon carbide and corundum substrates. J Cryst Growth 1971;9:158e64.

21. Hashimoto M, Amano H, Sawaki N, Akasaki I. Effects of hydrogen in an ambient on thecrystal growth of GaN using Ga(CH3)3 and NH3. J Cryst Growth 1984;68:163e8.

22. Ilegems M, Dingle R, Logan RA. Luminescence of Zn- and Cd-doped GaN. J Appl Phys1972;43:3797e800.

23. Khan MA, Skogman RA, Schulze RG, Gershenzon M. Electrical properties and ionimplantation of epitaxial GaN grown by low-pressure metalorganic chemical vapordeposition. Appl Phys Lett 1983;48:35e7.

24. Monemar B, Lagerstedt O. Properties of VPE-grown GaN doped with Al and some irongroup metals. J Appl Phys 1979;50:6480e91.

25. Pankove JI, Miller EA, Richman D, Berkeyheiser JE. Electroluminescence in GaN.J Lumin 1971;4:63e6.

26. Zetterstrom RB. Synthesis and growth of single crystals of gallium nitride. J Mater Sci1970;5:1102e4.

27. Yoshida S, Misawa S, Gonda S. Improvement on the electrical and luminescent propertiesof reactive molecular beam epitaxially grown GaN films using AlN-coated sapphiresubstrates. Appl Phys Lett 1983;42:427e9.

28. Amano H, Sawaki N, Akasaki I, Toyoda Y. Metalorganic vapor phase epitaxial growth of ahigh quality GaN film using an AlN buffer layer. Appl Phys Lett 1986;48:353e5.

29. Dobrovinskaya ER, Lytvynov LA, Pishchik V. Sapphire: material, manufacturing,applications. New York: Springer; 2009.


30. Lee W, Ryou J-H, Yoo D, Limb J, Dupuis RD, Hanser D, Preble E, Williams NM, Evans K.Optimization of Fe doping at the regrowth interface of GaN for applications to III-nitride-based heterostructure field-effect transistors. Appl Phys Lett 2007;90:093509-1e3.

31. YimWM, Paff RJ. Thermal expansion of AlN, sapphire, and silicon. J Appl Phys 1974;45:1456e7.

32. Denis A, Goglio G, Demazeau G. Gallium nitride bulk crystal growth processes: a review.Mater Sci Eng Rep 2006;50:167e94.

33. Kelly MK, Vaudo RP, Phanse VM, Gorgens L, Ambacher O, Stutzmann M. Large free-standing GaN substrates by hydride vapor phase epitaxy and laser-induced liftoff. Jpn JAppl Phys 1999;38:L217e9.

34. Cree. LED components and modules. 2013. Available from: http://www.cree.com/led-components-and-modules.

35. Lin ME, Strite S, Agarwal A, Salvador A, Zhou GL, Teraguchi N, Rockett A, Morkoc H.GaN grown on hydrogen plasma cleaned 6H-SiC substrates. Appl Phys Lett 1993;62:702e4.

36. Weeks Jr TW, Bremser MD, Ailey KS, Carlson E, Perry WG, Davis RF. GaN thin filmsdeposited via organometallic vapor phase epitaxy on a(6H)-SiC (0001) using high-temperature monocrystalline AlN buffer layers. Appl Phys Lett 1995;67:401e3.

37. Jeganathan K, Shimizu M, Okumura H. High-density self-assembled GaN nanoislands onSiC (0001) by molecular-beam epitaxy. Appl Phys Lett 2005;86:073106-1e073106-3.

38. Waltereit P, Brandt O, Trampert A, Ramsteiner M, Reiche M, Qi M, Ploog KH. Influenceof AlN nucleation layers on growth mode and strain relief of GaN grown on 6H-SiC(0001). Appl Phys Lett 1999;74:3660e2.

39. Ponce FA, Krusor BS, Major Jr JS, Plano WE, Welch DF. Microstructure of GaN epitaxyon SiC using AlN buffer layers. Appl Phys Lett 1995;67:410e2.

40. Fan JY, Rong G, Newman N, Smith DJ. Defect annihilation in AlN thin films by ultrahightemperature processing. Appl Phys Lett 2000;76:1839e2841.

41. Koleske DD, Henry RL, Twigg ME, Culbertson JC, Binari SC, Wickenden AE, Fatemi M.Influence of AlN nucleation layer temperature on GaN electronic properties grown on SiC.Appl Phys Lett 2002;80:4372e4.

42. Kawasuso A, Kojima K, Yoshikawa M, Itoh H, Narumi K. Effect of hydrogen etching on6H SiC surface morphology studied by reflection high-energy positron diffraction andatomic force microscopy. Appl Phys Lett 2000;76:1119e21.

43. Losurdo M, Capezzuto P, Bruno G, Brown A, Kim T-H, Yi C, Zakharov DN, Liliental-Weber Z. Interfacial reactions during GaN and AiN epitaxy on 4H-and 6H-SiC (0001).Appl Phys Lett 2005;86:021920-1e021920-3.

44. Torres VM, Edwards JL, Wilkens BJ, Smith DJ, Doak RB, Tsong IST. Influence of 6H-SiC (0001) substrate surface morphology on the growth of AlN epitaxial layers. ApplPhys Lett 1999;74:985e7.

45. Bassim ND, Twigg ME, Eddy Jr CR, Culbertson JC, Mastro MA, Henry RL, Holm RT,Neudeck PG, Trunek AJ, Powell JA. Lowered dislocation densities in uniform GaN layersgrown on step-free (0001) 4H-SiC mesa surfaces. Appl Phys Lett 2005;86:021902-1e21902-3.

46. Balakrishnan K, Fujimoto N, Kitano T, Bandoh A, Imura M, Nakano K, Iwaya M,Kamiyama S, Amano H, Akasaki I, Takagi T, Noro T, Shimono K, Riemann T, Christen J.Critical aspects of high temperature MOCVD growth of AlN epilayers on 6H-SiC sub-strates. Phys Stat Sol C 2006;3:1392e5.


http://www.cree.com/led-components-and-modules

http://www.cree.com/led-components-and-modules

47. Edwards NV, Bremser MD, Davis RF, Batchelor AD, Yoo SD, Karan CF, Aspnes DE.Trends in residual stress for GaN/AlN/6H-SiC heterostructures. Appl Phys Lett 1998;73:2808e11.

48. Yoo D, Limb J, Ryou J-H, Lee W, Dupuis RD. GaN full-vertical p-i-n rectifiers employingAlGaN: Si conducting buffer layers on n-SiC substrates. Appl Phys Lett 2006;88:193503-1e13.

49. Hellman ES, Buchanan DNE, Wiesmann D, Brener I. Growth of Ga-face and N-face GaNfilms using ZnO substrates. MRS Internet J Nitride Semicond Res 1996;1:16.

50. Matsuoka T, Yoshimoto N, Sasaki T, Katsui A. Wide-gap semiconductor InGaN andInGaAlN grown by MOVPE. J Electron Mater 1992;21:157e63.

51. Detchprohm T, Amano H, Hiramatsu K, Akasaki I. The growth of thick GaN film onsapphire substrate by using ZnO buffer layer. J Cryst Growth 1993;128:384e90.

52. Johnson MAL, Fujita S, Rowland Jr WH, Hughes WC, Cook Jr JW, Schetzina JF. MBEgrowth and properties of ZnO on sapphire and SiC substrates. J Electron Mater 1996;25:855e62.

53. Ozgur U, Alivov YI, Liu C, Teke A, Reshchikov MA, Dogan S, Avrutin V, Cho S-J,Morkoc H. A comprehensive review of ZnO materials and devices. J Appl Phys 1995;98:041301-1e103.

54. Morkoc H, Ozgur U. Zinc oxide: fundamentals, materials and device technology.Weinheim, Germany: Wiley-VCH Verlag; 2009.

55. Hamdani F, Yeadon M, Smith DJ, Tang H, Kim W, Salvador W, Botchkarev AE,Gibson JM, Polyakov AY, Skowronski M, Morkoc H. Microstructure and opticalproperties of epitaxial GaN on ZnO (0001) grown by reactive molecular beam epitaxy.J Appl Phys 1998;83:983e90.

56. Avrutin V, Cantwell G, Zhang J, Song JJ, Silversmith DJ, Morkoc H. Bulk ZnO: currentstatus, challenges, and prospect. Proc IEEE 2010;98:1339e50.

57. Doolittle WA, Kropewnicki T, Carter-Coman C, Stock S, Koh P, Jokerst NM,Metzger RA, Kang S, Lee KK, May G, Brown AS. Growth of GaN on lithium gallatesubstrates for development of a GaN thin compliant substrate. J Vac Sic Technol B 1998;16:1300e4.

58. Ishii T, Tazoh Y, Miyazawa S. Single-crystal growth of LiGaO2 for a substrate of GaN thinfilms. J Cryst Growth 1998;186:409e19.

59. Doolittle WA, Kang S, Brown A. MBE growth of high quality GaN on LiGaO2 for highfrequency, high power electronic applications. Solid State Electron 2000;44:229e38.

60. Kung P, Saxler A, Zhang X, Walker D, Lavado R, Razeghi M. Metalorganic chemicalvapor deposition of monocrystalline GaN thin films on b-LiGaO2 substrates. Appl PhysLett 1996;69:2116e8.

61. Tsuchiya Y, Kobayashi A, Ohta J, Fujioka H, Oshima M. Heteroepitaxial growth of GaNon atomically flat LiTaO3 (0001) using low-temperature AIN buffer layers. J Cryst Growth2006;293:22e6.

62. Xu K, Xu J, Deng P, Qiu R, Fang F. MOCVD growth of GaN on LiAlO2 (100) substrates.Phys Stat Sol A 1999;176:589e93.

63. Lee JJ, Park YS, Yang CS, Kim HS, Kim KH, Kang KY, Kang TW, Park SH, Lee JY.MBE growth of wurtzite GaN on LaAlO3 (100) substrate. J Cryst Growth 2000;213:33e9.

64. Sumiya M, Chikyow T, Sasahara T, Yoshimura K, Ohta J, Fujioka H, Tagaya S, Ikeya H,Koinuma H, Fuke S. Epitaxial growth of GaN film on (La,Sr)(Al,Ta)O3 (111) substrate bymetalorganic chemical vapor deposition. Jpn J Appl Phys 2002;41:5038e41.


65. Mamutin VV, Sorokin SV, Jmerik VN, Shubina TV, Ratnikov VV, Ivanov SV, Kopev PS,Karlsteen M, SoKdervall U, Willander M. Plasma-assisted MBE growth of GaN andInGaN on different substrates. J Cryst Growth 1999;201/202:346e50.

66. Yoshikawa A, Kochurikhin VV, Futagawa N, Shimamura K, Fukuda T. Growth ofCa8La2(PO4)6O2 single crystals as substrates for GaN epitaxial growth. J Cryst Growth1999;204:302e6.

67. Ohta J, Fujioka H, Takahashi H, Oshima M. Crystal growth of GaN on (Mn,Zn)Fe2O4

substrates. Phys Stat Sol A 2001;188:497e500.68. Kuramata A, Horino K, Domen K, Shinohara K, Tanahashi T. High-quality GaN epitaxial

layer grown by metalorganic vapor phase epitaxy on (111) MgAl2O4 substrate. Appl PhysLett 1995;67:2521e3.

69. Yamada A, Ho KP, Akaogi T, Maruyama T, Akimoto K. Layered compound substrates forGaN growth. J Cryst Growth 1999;201/202:332e5.

70. Beresford R, Paine DC, Briant CL. Group IVB refractory metal crystals as lattice-matchedsubstrates for growth of III nitrides by plasma-source molecular beam epitaxy. J CrystGrowth 1997;178:189e200.

71. Inoue S, Okamoto K, Matsuki N, Kim T-W, Fujiokaa H. Epitaxial growth of GaN oncopper substrates. Appl Phys Lett 2006;88:261910-1e261910-3.

72. Inoue S, Okamoto K, Nakano T, Ohta J, Fujiokaa H. Growth of single crystalline GaN onsilver mirrors. Appl Phys Lett 2007;91:201920-1e201920-3.

73. Nakamura S, Senoh M, Nagahama S-I, Iwasa N, Yamada T, Matsushita T, Kiyoku H,Sugimoto Y. InGaN multi-quantum-well structure laser diodes grown on MgAl2O4 sub-strates. Appl Phys Lett 1996;68:2105e7.

74. Gault WA, Monberg EM, Clemans JE. A novel application of the vertical gradientfreeze methods to the growth of high quality III-V crystals. J Cryst Growth 1986;74:491e506.

75. Mullin JB, Straughan BW, Brickell WS. Liquid encapsulation techniques: the used of aninert liquid in suppressing dissociation during the melt-growth of InAs and GaAs crystals.J Phys Chem Sol 1965;26:782e4.

76. Grzegory I. High pressure growth of bulk GaN from solutions in gallium. J Phys CondensMatter 2001;13:6875e92.

77. Avrutin V, Silversmith DJ, Mori Y, Jawamura F, Jitaoka Y, Morkoc H. Growth of bulkGaN and AlN: progress and challenges. Proc IEEE 2010;98:1302e15.

78. Ehrentraut D, Fukuda T. The ammonothermal crystal growth of gallium nitride e atechnique on the up rise. Proc IEEE 2010;98:1316e23.

79. Xu X, Vaudoa RP, Loria C, Salanta A, Brandesa GR, Chaudhuri J. Growth and charac-terization of low defect GaN by hydride vapor phase epitaxy. J Cryst Growth 2002;246:223e9.

80. Paskova T, Hanser DA, Evans KR. GaN substrates for III-nitride devices. Proc IEEE 2010;98:1324e38.

81. Dai Q, Schubert MF, Kim MH, Kim JK, Schubert EF, Koleske DD, Crawford MH,Lee SR, Fischer AJ, Thaler G, Banas MA. Internal quantum efficiency and nonradiativerecombination coefficient of GaInN/GaN multiple quantum wells with different disloca-tion densities. Appl Phys Lett 2009;94:111109-1e3.

82. Akita K, Kyono T, Yoshizumi Y, Kitabayashi H, Katayama K. Improvements of externalquantum efficiency of InGaN-based blue light emitting diodes at high current density usingGaN substrates. J Appl Phys 2007;101:033104-1e5.


83. Cao XA, LeBoeuf SF, D’Evelyn MP, Arthur SD, Kretchmer J, Yan CH, Yang ZH. Blueand near-ultraviolet light-emitting diodes on free-standing GaN substrates. Appl Phys Lett2004;84:4313e5.

84. Kyono T, Hirayama H, Akita K, Nakamura T, Ishibashi K. Effects of in composition onultraviolet emission efficiency in quaternary InAlGaN light-emitting diodes on free-standing GaN substrates and sapphire substrates. J Appl Phys 2005;98:113514-1e8.

85. Liu JP, Limb J, Lochner Z, Yoo D, Ryou J-H, Dupuis RD. Green light-emitting diodeswith p-InGaN: Mg grown on c-plane sapphire and GaN substrates. Phys Stat Sol A 2009;206:750e3.

86. Liu Z, Wei T, Guo E, Yi X, Wang L, Wang J, Wang G, Shi Y, Ferguson I, Li J. Efficiencydroop in InGaN/GaN multiple-quantum-well blue light-emitting diodes grown on free-standing GaN substrate. Appl Phys Lett 2011;99:091104-1e3.

87. Rozhansky IV, Zakheim DA. Analysis of processes limiting quantum efficiency ofAlGaInN LEDs at high pumping. Phys Stat Sol a 2007;204:227e30.

88. Schubert MF, Chhajed S, Kim JK, Schubert EF, Koleske DD, Crawford MH, Lee SR,Fischer AJ, Thaler G, Banas MA. Effect of dislocation density on efficiency droop inInGaN/GaN light-emitting diodes. Appl Phys Lett 2007;91:231114-1e3.

89. Norton DP, Heo YW, Ivill MP, Ip K, Pearton SJ, Chisholm MF, Steiner T. ZnO: growth,doping and processing. Mater Today June 2004;7:34e40.

90. Akasaki I. Lifetime of research on nitrides e alone in the wilderness. In: Takahashi K,Yoshikawa A, Sandhu A, editors.Wide bandgap semiconductors, Fundamental propertiesand modern photonic and electronic devices, vol. IXeXVII. Berlin: Springer; 2007.

91. Yoshida S, Misawa S, Itoh A. Epitaxial growth of aluminum nitride films on sapphire byreactive evaporation. Appl Phys Lett 1975;26:461e2.

92. Yoshida S, Misawa S, Fujii Y, Takada S, Hayakawa H, Gonda S, Itoh A. Reactivemolecular beam epitaxy of aluminium nitride. J Vac Sci Technol 1979;16:990e3.

93. Amano H, Akasaki I. Novel aspects of the growth of nitrides by MOVPE. J Phys CondensMatter 2001;13:6935e44.

94. Amano H, Kamiyama S, Akasaki I. Impact of low-temperature buffer layers on nitride-based optoelectronics. Proc IEEE 2002;90:1015e21.

95. Nakamura S. GaN growth using GaN buffer layer. Jpn J Appl Phys 1991;30:L1705e7.96. Akasaki I, Amano H, Koide N, Hiramatsu K, Sawaki N. Effect of AlN buffer layer on

crystallographic structure and on electrical and optical properties of GaN and Ga1-xAlxN(0<x�0.4) films grown on sapphire substrate by MOVPE. J Cryst Growth 1989;98:209e19.

97. Kapolnek D, Wu XH, Heying B, Keller S, Keller BP, Mishra UK, DenBaars SP, Speck JS.Structural evolution in epitaxial metalorganic chemical vapor deposition grown GaN filmson sapphire. Appl Phys Lett 1995;67:1541e3.

98. Kuznia JN, Khan MA, Olson DT, Kaplan R, Freitas J. Influence of buffer layers on thedeposition of high quality single crystal GaN over sapphire substrates. J Appl Phys 1993;73:4700e2.

99. Wickenden AE, Wickenden DK, Kistenmacher TJ. The effect of thermal annealing onGaN nucleation layers deposited on (0001) sapphire by metalorganic chemical vapordeposition. J Appl Phys 1994;75:5367e71.

100. Wu XH, Kapolnek D, Tarsa EJ, Heying B, Keller S, Keller BP, Mishra UK, DenBaars SP,Speck JS. Nucleation layer evolution in metal-organic chemical vapor deposition grownGaN. Appl Phys Lett 1996;68:1371e3.


101. Amano H, Akasaki I, Hiramatsu K, Koide N, Sawaki N. Effect of buffer layer in metal-organic vapor phase epitaxy of GaN on sapphire substrate. Thin Solid Film 1988;163:415e20.

102. Hiramatsu K, Itoh S, Amano H, Akasaki I. Growth mechanism of GaN growth on sapphirewith AlN buffer layer by MOVPE. J Cryst Growth 1991;115:628e33.

103. Nakamura S. In-situ monitoring of GaN growth using interference effects. Jpn J Appl Phys1991;30:1620e7.

104. Lee K-H, Asadirad M, Shervin S, Oh SK, Oh JT, Song J-O, Moon Y-T, Ryou J-H.Thin-film-flip-chip LEDs grown on Si substrate using wafer-level chip-scale package.IEEE Photon Technol Lett 2016;28:1956e9.

105. Kozawa T, Kachi T, Kano H, Nagase H. Thermal stress in GaN epitaxial layers grown onsapphire substrates. J Appl Phys 1995;77:4389e92.

106. Li W, Ni W-X. Residual strain in GaN epilayers grown on sapphire and (6H)SiC sub-strates. Appl Phys Lett 1996;68:2705e7.

107. Ager III JW, Suski S, Ruvimov S, Krueger J, Conti G, Weber ER, Bremser MD, Davis R,Kuo CP. Intrinsic and thermal stress in gallium nitride epitaxial films. Proc Mat Res SocSymp 1997;449:775e80.

108. Choi S, Kim HJ, Ryou J-H, Dupuis RD. Actual temperatures of growing surfaces ofIII-nitride-based materials depending on substrates and forced convection conditions inmetal organic chemical vapor deposition. J Appl Phys 2009;106:073512-1e6.

109. Hoffmann V, Knauer A, Brunner C, Einfeldt S, Weyers M, Trankle G, Haberland K,Zettler J-T, Kneissl M. Uniformity of the wafer surface temperature during MOVPEgrowth of GaN-based laser diode structures on GaN and sapphire substrate. J CrystGrowth 2011;315:5e9.

110. Bedair SM, McIntosh FG, Roberts JC, Piner EL, Boutros KS, El-Masry NA. Growth andcharacterization of In-based nitride compounds. J Cryst Growth 1997;178:32e44.

111. Belousov M, Volf B, Ramer JC, Armour EA, Gurary A. In situ metrology advances inMOCVD growth of GaN-based materials. J Cryst Growth 2004;272:94e9.

112. Taylor C, Barlett D, Chason E, Floro JA. Multibeam optical sensor (MOS)e a laser basedthin film growth monitor. Ind Phys 1998;4:25e6.

113. Brunner F, Knauer A, Schenk T, Weyers M, Zettler J-T. Quantitative analysis of in situwafer bowing measurements for III-nitride growth on sapphire. J Cryst Growth 2008;310:2432e8.

114. Brunner F, Hoffmann V, Knauer A, Steimetz E, Schenk T, Zettler J-T, Weyers M. Growthoptimization during III-nitride multiwafer MOVPE using real-time curvature, reflectanceand true temperature measurement. J Cryst Growth 2007;298:202e6.

115. Hearne S, Chason E, Han J, Floro JA, Figiel J, Hunter J, Amano H, Tsong IST. Stressevolution during metalorganic chemical vapor deposition of GaN. Appl Phys Lett 1999;74:356e8.

116. B€ottcher T, Einfeldt S, Figge S, Chierchia R, Heinke H, Hommel D, Speck JS. The role ofhigh-temperature island coalescence in the development of stresses in GaN film. Appl PhysLett 2001;78:1976e8.

117. Romano LT, Van de Walle CG, Ager III JW, G€otz W, Kern S. Effect of Si doping onstrain, cracking, and microstructure in GaN thin films grown by metalorganic chemicalvapor deposition. Appl Phys Lett 2000;87:7745e75.

118. Brunner F, Mogilatenko A, Knauer A, Weyers M, Zettler J-T. Analysis of doping inducedwafer bowing during GaN: Si growth on sapphire. J Appl Phys 2012;112:033503-1e5.


119. Krost A, Schulze F, Dadgar A, Strassburger A, Haberland K, Zettler T. Simultaneousmeasurement of wafer curvature and true temperature during metalorganic growth ofgroup-III nitrides on silicon and sapphire. Phys Stat Sol B 2005;242:2570e4.

120. Lee SW, Ha J-S, Lee H-J, Lee H-J, Goto H, Hanada T, Goto T, Fijii K, Cho MW, Yao T.Lattice strain in bulk GaN epilayers grown on CrN/sapphire template. Appl Phys Lett2009;94:082105-1e3.

121. Yen C-H, Lai W-C, Yang Y-Y, Wang C-K, Ko T-K, Hon S-J, Chang S-J. GaN-basedlight-emitting diodes with sputtered AlN nucleation layer. IEEE Photon Technol Lett2012;24:294e6.

122. Chen S-W, Yang Y, Wen W-C, Li H, Lu T-C. Significant improvement of GaN crystalquality with ex-situ sputtered AlN nucleation layers. Proc SPIE 2016;9768:9768D-1e9.

123. Dadgar A, Bl€asing J, Diez A, Alam A, Heuken M, Korst A. Metalroganic chemical vaporepitaxy of crack-free GaN on Si (111) exceeding 1mm in thickness. Jpn J Appl Phys 2000;39:L1183e5.

124. Hinoki A, Kikawa J, Yamada T, Tsuchiya T, Kamiya S, Kurouchi M, Kosaka K, Araki T,Suzuki T, Nanishi Y. Effects of traps formed by threading dislocations on off-statebreakdown characteristics in GaN buffer layer in AlGaN/GaN heterostructure field-effect transistors. Appl Phys Express 2008;1:011030-1e011030-3.

125. Lochner Z, Kim HJ, Lee YC, Zhang Y, Choi S, Shen S-C, Yoder PD, Ryou J-H,Dupuis RD. NpN-GaN/InxGa1-xN/GaN heterojunction bipolar transistor on free-standingGaN substrate. Appl Phys Lett 2011;99:193501-1e193501-3.

126. Rozhansky IV, Zakheim DA. Analysis of the causes of the decrease in the electrolumi-nescence efficiency of AlGaInN light-emitting-diode heterostructures at high pumpingdensity. Semiconductors 2006;40:839e45.

127. Caenegem TV, Moerman I, Demeester P. Selective area growth on planar masked InPsubstrates by metal organic vapour phase epitaxy (MOVPE). Prog Cryst Growth Charact1997;35:263e88.

128. Jastrzebski L. SOI BY CVD: epitaxial lateral overgrowth (ELO) processe review. J CrystGrowth 1983;63:493e526.

129. Coldren LA, Corzine SW, Mashanovitch ML. Diode lasers and photonic integratedcircuits. Hoboken: John Wiley & Sons; 2012. p. 27e8.

130. Kato Y, Kitamura S, Hiramatsu K, Sawaki N. Selective growth of wurtzite GaN andAlGaN on GaN/sapphire substrates by metalorganic chemical vapor deposition. J CrystGrowth 1994;144:133e40.

131. Kitamura S, Hiramatsu K, Sawaki N. Fabrication of GaN hexagonal pyramids on dot-patterned GaN/sapphire substrates via selective metalorganic vapor phase epitaxy. Jpn JAppl Phys 1995;34:L1184e6.

132. Zheleva TS, Nam O-H, Bremser MD, Davis RF. Dislocation density reduction via lateralepitaxy in selectively grown GaN structures. Appl Phys Lett 1997;71:2472e4.

133. Nam O-H, Bremser MD, Zheleva TS, Davis RF. Lateral epitaxy of low density defect GaNlayers via organometallic vapor phase epitaxy. Appl Phys Lett 1997;71:2638e40.

134. Usui A, Sunakawa H, Sakai A, Yamaguchi AA. Thick GaN epitaxial growth withlow dislocation density by hydride vapor phase epitaxy. Jpn J Appl Phys 1997;36:L899e902.

135. Kapolnek D, Keller S, Vetury R, Underwood RD, Kozodoy P, DenBaars SP, Mishra UK.Anisotropic epitaxial lateral growth in GaN selective area epitaxy. Appl Phys Lett 1997;71:1204e6.


136. Park J, Grudowski PA, Eiting CJ, Dupuis RD. Selective-area and lateral epitaxial over-growth of III-N materials by metalorganic chemical vapor deposition. Appl Phys Lett1998;73:333e5.

137. Hiramatsu K, Nishiyama K, Motogaito A, Miyake H, Iyechika Y, Maeda T. Recentprogress in selective area growth and epitaxial lateral overgrowth of III-nitrides: effects ofreactor pressure in MOVPE growth. Phys Stat Sol A 1999;176:535e43.

138. Hiramatsu K. Epitaxial lateral overgrowth techniques used in group III nitride epitaxy.J Phys Condens Matter 2001;13:6961e75.

139. Zheleva TS, Smith SA, Thomson DN, Linthicum KJ, Rajagopal P, Davis RF. Pendeo-epitaxy: a new approach for lateral growth of gallium nitride films. J Electron Mater1999;28:L5e8.

140. Linthicum K, Gehrke T, Thomson D, Carlson E, Rajagopal P, Smith T, Batchelor D,Davis R. Pendeoepitaxy of gallium nitride thin films. Appl Phys Lett 1999;75:196e8.

141. Gehrke T. Review of structured thin films in wide bandgap semiconductors: pendeo-epitaxy of GaN and AlGaN. J Nanophotonic 2008;2:021990-1e021990-12.

142. Yamada M, Mitani T, Narukawa W, Shioji S, Niki I, Sonobe S, Deguchi K, Sano M,Mukai T. InGaN-based near-ultraviolet and blue-light-emitting diodes with high externalquantum efficiency using a patterned sapphire substrate and a mesh electrode. Jpn J ApplPhys 2002;41:L1431e3.

143. Feng ZH, Lau KM. Enhanced luminescence from GaN-based blue LEDs grown ongrooved sapphire substrates. IEEE Photon Technol Lett 2005;17:1812e4.

144. Wuu DS, Wang WK, Shih WC, Horng RH, Lee CE, Lin WY, Fang JS. Enhanced outputpower of near-ultraviolet InGaNeGaN LEDs grown on patterned sapphire substrates.IEEE Photon Technol Lett 2005;17:288e90.

145. Choi HW, Liu C, Gu E, McConnell G, Girkin JM, Watson IM, Dawson MD. GaN micro-light-emitting diode arrays with monolithically integrated sapphire microlenses. Appl PhysLett 2004;84:2253e5.

146. Lee T-X, Gao K-F, Chien W-T, Sun C-C. Light extraction analysis of GaN-based light-emitting diodes with surface texture and/or patterned substrate. Opt Express 2007;15:6670e6.

147. Park S-H, Jeon H, Sung Y-J, Yeom G-Y. Refractive sapphire microlenses fabricated bychlorine-based inductively coupled plasma etching. Appl Opt 2001;40:3698e702.

148. Cuong TV, Cheong HS, Kim HG, Kim HY, Hong C-H, Suh EK, Cho HK, Kong BH.Enhanced light output from aligned micropit InGaN-based light emitting diodes usingwet-etch sapphire patterning. Appl Phys Lett 2007;90:131107-1e131107-3.

149. Ashby CIH, Mitchell CC, Han J, Missert NA, Provencio PP, Follstaedt DM, Peake GM,Griego L. Low-dislocation-density GaN from a single growth on a textured substrate. ApplPhys Lett 2000;77:3233e5.

150. Detchprohm T, Yano M, Sano S, Nakamura R, Mochiduki S, Nakamura T, Amano H,Akasaki I. Heteroepitaxial lateral overgrowth of GaN on periodically grooved substrates: anew approach for growing low-dislocation-density GaN single crystals. Jpn J Appl Phys2001;40:L16e9.

151. Tadatomo K, Okagawa H, Ohuchi Y, Tsunekawa T, Imada Y, Kato M, Taguchi T. Highoutput power InGaN ultraviolet light-emitting diodes fabricated on patterned sapphiresubstrates using metalorganic vapor phase epitaxy. Jpn J Appl Phys 2001;40:L583e5.

152. Huang X-H, Liu J-P, Kong J-J, Yang H, Wang H-B. High-efficiency InGaN-based LEDsgrown on patterned sapphire substrate. Opt Express 2011;19:949e55.


153. Shin H-Y, Kwon SK, Chang YI, Cho MJ, Park KH. Reducing dislocation density in GaNfilms using a cone-shaped patterned sapphire substrate. J Cryst Growth 2009;311:4167e70.

154. Song J-C, Lee S-H, Lee I-H, Seol K-W, Kannappan S, Lee CR. Characteristics comparisonbetween GaN epilayers grown on patterned and unpatterned sapphire substrate (0001).J Cryst Growth 2007;308:321e4.

155. Wuu DS, Wang WK, Shih WH, Horng RH, Lee CE, Lin WY, Fang JS. Enhanced outputpower of near-ultraviolet InGaNeGaN LEDs grown on patterned sapphire substrates.IEEE Photon Technol Lett 2009;17:288e90.

156. Lee K-S, Kwack H-S, Hwang J-S, Roh T-M, Cho Y-H, Lee J-H, Kim Y-C, Kim CS.Spatial correlation between optical properties and defect formation in GaN thin filmslaterally overgrown on cone-shaped patterned sapphire substrate. J Appl Phys 2010;107:103560-1e103560-4.

157. Lee K-S, Isnaeni, Yoo Y-S, Lee J-H, Kim Y-C, Cho Y-H. Influence of defect reductionand strain relaxation on carrier dynamics in IngaN-based light-emitting diodes on cone-shaped patterned sapphire substrates. J Appl Phys 2013;113:173512-1e173512-5.

158. Wang M-T, Brunner F, Liao K-Y, Li Y-L, Tseng SH, Weyers M. Optimization of GaNwafer bow grown on cone-shaped patterned sapphire substrates. J Cryst Growth 2013;363:103e12.

159. Lin H-C, Liu H-H, Lee G-Y, Chyi J-I, Lu C-M, Chao C-W, Wang T-C, Chang C-J,Chi SWS. Effects of lens shape on GaN grown on microlens patterned sapphire substratesby metalorganic chemical vapor deposition. J Electrochem Soc 2010;157:H304e7.

160. Bernardini F, Fiorentini V, Vanderbilt D. Spontaneous polarization and piezoelectricconstants of III-V nitrides. Phys Rev B 1997;56:R10024e7.

161. Ambacher O, Smart J, Shealy JR, Weimann NG, Chu K, Murphy M, Schaff WJ,Eastman LF, Dimitrov R, Wittmer L, Stutzmann M, Rieger W, Hilsenbeck J. Two-dimensional electron gases induced by spontaneous and piezoelectric polarizationcharges in N- and Ga-face AlGaN/GaN heterostructures. J Appl Phys 1999;85:3222e33.

162. Miller DAB, Chemla DS, Damen TC, Gossard AC, Wiegmann W, Wood TH, Burrus CA.Band-edge electroabsorption in quantum well structures: the quantum-confined Stark ef-fect. Phys Rev Lett 1984;53:2173e6.

163. Takeuchi T, Sota S, Katsuragawa M, Komori M, Takeuchi H, Amano H, Akasaki I.Quantum-confined Stark effect due to piezoelectric fields in GaInN strained quantumwells. Jpn J Appl Phys 1997;36:L382e5.

164. Ryou JH, Yoder PD, Liu JP, Lochner Z, Kim HS, Choi S, Kim HJ, Dupuis RD. Control ofquantum-confined Stark effect in InGaN-based quantum wells. IEEE J Select Topic QuanElectron 2009;15:1080e91.

165. Bernardini F, Fiorentini V. Macroscopic polarization and band offsets at nitride hetero-junction. Phys Rev B 1998;57:R9427e30.

166. Im JS, Kollmer H, Off J, Sohmer A, Scholz F, Hangleiter A. Reduction of oscillatorstrength due to piezoelectric fields in GaN/AlxGa1-xN quantum wells. Phys Rev B 1998;57:R9435e8.

167. Schubert MF, Schubert EF. Effect of heterointerface polarization charges and well widthupon capture and dwell time for electrons and holes above GaInN/GaN quantum wells.Appl Phys Lett 2010;96:131102-1e3.

168. Shen YC, Mueller GO, Watanabe S, Gardner NF, Munkholm A, Krames MR. Augerrecombination in InGaN measured by photoluminescence. Appl Phys Lett 2007;91:141101-1e141101-3.


169. Waltereit P, Brandt O, Trampert A, Grahn HT, Menniger J, Ramsteiner M, Reiche M,Ploog KH. Nitride semiconductors free of electrostatic fields for efficient white light-emitting diodes. Nature 2000;406:865e8.

170. Sasaki T, Zembutsu S. Substrate orientation dependence of GaN single crystal films grownby metalorganic chemical vapor phase epitaxy. J Appl Phys 1987;61:2533e40.

171. Park S-H, Chuang SL. Crystal-orientation effects on the piezoelectric field and electronicproperties of strained wurtzite semiconductors. Phys Rev B 1999;59:4725e37.

172. Park S-H. Crystal-orientation effects on electronic properties of wurtzite InGaN/GaNquantum wells. J Appl Phys 2002;91:9904e8.

173. Craven MD, Lim SH, Wu F, Speck JS, DenBaars SP. Structural characterization ofnonpolar

�1120

�a-plane GaN thin films grown on

�1102

�r-plane sapphire. Appl Phys

Lett 2002;81:469e71.174. Ng HM. Molecular-beam epitaxy of GaN/AlxGa1-xN multiple quantum wells on R-plane�

1012�sapphire substrates. Appl Phys Lett 2002;80:4369e71.

175. Paskova T, Darakchieva V, Valcheva E, Paskov PP, Monemar B, Heuken M. Growth ofGaN on a-plane sapphire: in-plane epitaxial relationships and lattice parameters. Phys StatSol B 2003;240:318e21.

176. Craven MD, Wu F, Chakraborty A, Imer B, Mishra UK, DenBaars SP, Speck JS.Microstructural evolution of a-plane GaN grown on a-plane SiC by metalorganic chemicalvapor deposition. Appl Phys Lett 2004;84:1281e3.

177. Matsuoka T, Hagiwara E. GaN growth on novel lattice-matching substrate: tilted m-planesapphire. Phys Stat Sol A 2001;188:485e9.

178. Gardner NF, Kim JC, Wierer JJ, Shen YC, Krames MR. Polarization anisotropy in theelectroluminescence of m-plane InGaNeGaN multiple-quantum-well light-emittingdiodes. Appl Phys Lett 2005;86:111101-1e111101-3.

179. Haskell BA, Wu F, Matsuda S, Craven MD, Fini PT, DenBaars SP, Speck JS,Nakamura S. Structural and morphological characteristics of planar nonpolar

�1120

�

a-plane gallium nitride grown by hydride vapor phase epitaxy. Appl Phys Lett 2003;83:1554e6.

180. Darakchieva V, Paskov PP, Paskova T, Valcheva E, Monemar B, Heuken M. Latticeparameters of GaN layers grown on a-plane sapphire: effect of in-plane strain anisotropy.Appl Phys Lett 2003;82:703e5.

181. Haskell BA, Wu F, Craven MD, Matsuda S, Fini PT, Fiji T, Fujito K, DenBaars SP,Speck JS, Nakamura S. Defect reduction in

�1120

�a-plane gallium nitride via lateral

epitaxial overgrowth by hydride vapor phase epitaxy. Appl Phys Lett 2003;83:644e6.182. Tavernier PR, Imer B, DenBaars SP, Clarke DR. Growth of thick

�1120

�GaN using a

metal interlayer. Appl Phys Lett 2004;85:4630e2.183. Imer BM, Wu F, DenBaars SP, Speck JS. Improved quality

�1120

�a-plane GaN with

sidewall lateral epitaxial overgrowth. Appl Phys Lett 2006;88:061908-1e061908-3.184. Chakraborty A, Kim KC, Wu F, Speck JS, DenBaars SP, Mishra UK. Defect reduction in

nonpolar a-plane GaN films using in situ SiNx nanomask. Appl Phys Lett 2006;89:041903-1e041903-3.

185. Ni X, Ozgur U, Fu Y, Biyikli N, Xie J, Baski AA, Morkoc H. Defect reduction in by two-stage epitaxial lateral overgrowth. Appl Phys Lett 2006;89:262105-1e262105-3.

186. Haskell BA, Baker TJ, McLaurin MB, Wu F, Fini PT, DenBaars SP, Speck JS,Nakamura S. Defect reduction in

�1100

�m-plane gallium nitride via lateral epitaxial

overgrowth by hydride vapor phase epitaxy. Appl Phys Lett 2005;86:111917-1e111917-3.


187. Liu JP, Limb J, Ryou J-H, Yoo D, Horne CA, Dupuis RD, Wu ZH, Fischer A, Ponce FA,Hanser AD, Liu L, Preble EA, Evans KR. Blue light emitting diodes grown on freestanding

�1120

�a-plane GaN substrates. Appl Phys Lett 2008;92:011123-1e011123-3.

188. Schmidt MC, Kim K-C, Sato H, Fellows N, Masui H, Nakamura S, DenBaars SP,Speck JS. High power and high external efficiency m-plane InGaN light emitting diodes.Jpn J Appl Phys 2007;46:L126e8.

189. Kim K-C, Schmidt MC, Sato H, Wu F, Fellows N, Saito M, Fujito K, Speck JS,Nakamura S, DenBaars SP. Improved electroluminescence on nonpolar m-plane InGaN/GaN quantum wells LEDs. Phys Stat Sol RRL 2007;1:125e7.

190. Okamoto K, Tanaka T, Kubota M. High-efficiency continuous wave operation of blue-green laser diodes based on nonpolar m-plane gallium nitride. Appl Phys Express 2008;1:072201-1e072201-3.

191. Yoshizumi Y, Adachi M, Enya Y, Kyono T, Tokuyama S, Sumitomo T, Akita K,Ikegami T, Ueno M, Katayama K, Nakamura T. Continuous-wave operation of 520 nmgreen InGaN-based laser diodes on semi-polar {20-21} GaN substrates. Appl Phys Express2010;2:092101-1e092101-3.


Gallium nitride (GaN) on siliconsubstrates for LEDs 4Matthew H. Kane 1, Nazmul Arefin 2

1Texas A & M University at Galveston, Galveston, TX, United States; 2University ofOklahoma, Norman, OK, United States

4.1 Introduction

III-nitride materials possess a number of properties that are simply not accessible in anyother semiconductors, which will continue to make them an active area of scientificand technological development. The unique capabilities of the III-nitrides include ahigh dielectric breakdown voltage and a bandgap that spans from the infrared to thedeep ultraviolet. However, there are some unique processing challenges related to thecrystal structure and bonding. They have long been of interest for implementation inapplications ranging from high-power transistors to a solid-state replacement fortraditional lighting. Widespread market implementation of solid-state lighting sourceswould result in at least a twofold reduction in current energy consumption for lighting(roughly 10% of overall energy demand) worldwide,1 with a corresponding drop indemand for foreign petroleum and a fall in greenhouse gas emissions.

The consumer implementation of these sources has been hindered by cost barriersfrom the initial material growth steps, which ultimately are related to the lack of athermally conductive, lattice-matched substrate. The implementation of GaN-basedLEDs has been extraordinarily successful, though GaN devices differ from other III-Vsemiconductors in that there are no suitable lattice-matched substrates for their growth.The best substrates, silicon carbide and sapphire, were selected for their chemicalstability in the aggressive metal-organic chemical vapor deposition (MOCVD) growthenvironment.One continuing issue forwidespread solid-state lighting is the need to lowerthe cost per lumen of solid-state lighting devices. Silicon is an extremely attractive optionin this regard because of the ready availability of cheap large-area substrates, but it hassome unique challenges in the growth of GaN-based devices. These challenges haveonly recently been addressed successfully.

4.2 An overview of gallium nitride (GaN) on siliconsubstrates

A key problem has been the lack of a suitable lattice-matched substrate for theIII-nitrides, as homoepitaxial substrates are not readily available and are extremely


https://doi.org/10.1016/B978-0-08-101942-9.00004-6

expensive. The most commonly used substrates for GaN growth are sapphire andsilicon carbide, but this choice is driven by the need for chemical stability under theaggressive conditions for nitride growth rather than lattice mismatch concerns. Theuse of sapphire as a substrate presents a number of complications for the implementa-tion of devices. One problem is that sapphire is an electrical insulator e instead of be-ing able to put contacts on the backside surface of the substrate, the electrical contactsmust all be placed on the topside of the structure. In addition to complicating thedevice processing, this results in the devices being dependent on a significant amountof lateral transport, even in nominally vertical devices. Given the high dislocationdensities present in III-nitride devices (w109 cm�2), this lateral transport can resultin a considerable amount of resistive heating within the material. The geometricconsiderations associated with contact placement of the p-and n-contacts can also resultin severe problems with current crowding in the devices. This effect is compounded bythe thermally insulating nature of sapphire. Excessive heating within the materialreduces device lifetimes. Moreover, the inability to remove heat effectively from adevice limits the drive current and emission power from high-power visible andultraviolet light-emitting diodes (LEDs).

The development of novel substrates for the III-nitrides has been explored since theinception of these devices. Though initial reports focused on sapphire as the substrateof choice, efforts focusing on exploring substrates with better lattice matches wereconducted in parallel. In order to understand the possible routes towards improveddevice performance in heteroepitaxial systems, it is useful to examine what haspreviously been reported for other techniques. Table 4.1 lists the materials parametersof several heteroepitaxial substrates that have been explored for GaN growth. Ingeneral, the most suitable materials for heteroepitaxial growth are oxides with close-packed oxygen sublattices. Most of these crystals, in contrast to GaN, can be readilygrown by the Czochralski method. Of the Czochralski-grown crystals, silicon hasachieved the highest level of perfection and lowest cost due to the ubiquitousimplementation of silicon-based devices.

4.3 Silicon overview

4.3.1 Advantages of silicon

One key parameter needed to drive solid-state lighting further into the marketplaceis lowering the cost per lumen. The luminous efficacy of these devices has increasedto well over the performance found in fluorescent and incandescent light bulbs andthe color quality has improved to render these devices suitable for most applications.Thus, the chief remaining hurdle has been the cost of GaNebased LED devices.The high cost of GaN LEDs is driven somewhat by the initial costs of themetal-organic precursors and the process expense of heating them to the growthtemperatures. Another significant cost is the price of the substrates traditionally


Table 4.1 Possible substrates for the epitaxial growth of GaN

Substrate Crystal structure Orientation

Lattice parameter

Lattice mismatch Polar/nonpolara (Å) b (Å) c (Å)

Al2O3 Corundum (0001) 4.758 4.758 12.988 16% Polar

6HeSiC Hexagonal (0001) 3.081 3.081 15.17 �3.4% Polar

g-LiAlO2 Tetragonal (100) 5.169 5.169 6.268 �1.4% Both

LiGaO2 Orthorhombic (001) 5.403 6.372 5.007 1.9; 0.2% Both

ZnO Wurtzite (0001), (11.0) 3.249 3.249 5.207 1.8; 0.4% Both

LSAT Perovskite (111) 7.735 7.735 7.735 <1% Polar

MnAl2O4 Spinel (111) w8 w8 w8 10.4% Polar

LiNbO3 Perovskite (111) 5.147 5.147 5.147 7.1% Polar

Si Diamond (111) 5.431 5.431 5.431 �16.9% Polar

Ge Diamond (111) 5.646 5.646 5.646 e Polar

ScMgAlO4 Rhombohedral (111) 3.236 3.236 3.236 1.8% Polar

LiTaO3 Perovskite (111) 5.154 5.154 e 7.2% Polar

MnAl6O10 Perovskite (111) 7.984 7.984 7.984 �11.5% Polar

GaN Wurtzite All 3.189 3.189 5.186 0% Both

Gallium

nitride(G

aN)on

siliconsubstrates

forLEDs

81

used for the growth of GaN-based materials. Silicon carbide and sapphire are bothtremendously expensive, and though they share some crystallographic similaritieswith GaN, they are still not ideal for growth. Homoepitaxial growth of large-areaGaN is still several years away. Switching to silicon instead of one of these other sub-strates could offer significant cost benefits, provided equivalent crystalline qualitycan be maintained.

There are several advantages in using silicon as a substrate for the growth of GaNLEDs, the most important of which is the low cost and high area of the available siliconsubstrates. The growth of silicon substrates has matured extensively over the past50 years enabling modern silicon technology. Silicon substrates of greater than 1200are routinely available, whereas sapphire and silicon carbide substrates are typicallyused in 200 or 400 sizes. Moreover, the cost of silicon is an order of magnitude lowerthan sapphire, and two to three orders of magnitude lower than silicon carbide ofthe same size. Thus, as efforts continue to lower the costs of solid-state lightingsources, the reduction of substrate costs can have a significant impact on the overallcost of a device. Substrate costs can be of the order of 80%2 of the total processingcost. In addition, silicon is still the leading material for microelectronic applications;thus if the growth of gallium nitride on silicon can be achieved, it would be possibleto monolithically integrate gallium nitride lighting devices with silicon micro-electronics. Another advantage of silicon as a substrate is that it can be very easilywet-etched, unlike sapphire or silicon carbide. Thus it should be possible to havedifferent device configurations such as vertical light-emitting diodes, and to removethe substrate completely and place the device on a high thermal conductivity heatsink; the latter can increase the lifetimes of GaN LEDs when operated at the highpowers needed for general illumination applications. Patterning of silicon is also easierbecause wet etching processes can be used, unlike sapphire, which requires a dry etchprocess.

4.3.2 Crystallography

Most of the effort spent on developing the growth of GaN on silicon has been for c-axisgrowth on Si (111) substrates. Gallium nitride has a wurtzite structure with an a-axislattice parameter of 3.189 Å and a c-axis lattice parameter of 5.185 Å. Silicon, on theother hand, crystallizes to a cubic diamond structure with a lattice parameter of5.430 Å. Thus, in order to match the symmetry of the gallium nitride layer with thesilicon substrate, most efforts to grow gallium nitride on silicon have been performedon the silicon (111) face. This results in an effective lattice parameter of 3.84 Å, whichgives a lattice mismatch of 16.9%. Thus, the critical thickness of the gallium nitridelayer on silicon is of the order of a few nanometers. However, this is not all thatproblematic because a typical sapphire substrate has a lattice mismatch of 16%. More-over, gallium nitride LEDs are known to work with dislocation densities of the order of1010 per cubic centimeter.


Gallium nitride grows on the silicon (111) surface with the following crystallographicorientation: GaN(0001)//Si (111), GaN [11�20]//Si[1�10]. In contrast, for GaN growon sapphire there is a 30 degrees rotation of the epilayer, yielding a 16% latticemismatch on sapphire. Thus, the magnitudes of the lattice mismatch between siliconand sapphire are similar; however, the layers end up in tensile strain on silicon andcompressive strain on sapphire, which has a huge impact on the necessary interlayergrowth schemes. The same issues that are found for c-axis-grown gallium nitride onsapphire substrates will also be found in GaN on silicon substrates, such as thequantum-confined Stark effect. More recent attempts3 have focused on the nonpolargrowth of gallium nitride on the Si (111) face byMOCVD, which has a lattice mismatchof 0.8% and 19% in the in-plane directions for basal plane growth.4 Fig. 4.1 shows therelative orientation of heteroepitaxial aluminum nitride on underlying atoms in thesilicon layer.

4.4 Challenges for the growth of GaN on siliconsubstrates

In addition to the crystallographic concerns, several challenges exist for the growth ofgallium nitride on a silicon substrate. First of all, the aggressive growth conditions andparticularly the hydrogen atmosphere used for MOCVD growth can cause severeproblems at the growth interface because chemical reactions occur at the surface ofthe silicon substrate. Meltback etching of the substrate can also occur, which results ina significant increase in the surface roughness and difficulty in nucleating layers.5 Inaddition, silicon is a fast diffuser, which results in the formation of impurities in the sub-sequently grown gallium nitride layers; gas phase diffusion of silicon can also occur.

4.4.1 Thermal expansion mismatch between GaN and silicon

Another severe problem with the growth of GaN materials is the thermal expansionmismatch between GaN and silicon, which can be as high as 54%. The thermalexpansion coefficient of gallium nitride is 5.59 � 10�6/K for gallium nitride in thea-axis direction, but only 3.59 � 10�6/K for silicon. Thus, at normal growth tempera-tures the GaN and silicon layers may grow appropriately, but as the sample cools thegallium nitride layer contracts significantly more than the silicon layer. This placesthe gallium nitride layer in tensile strain, which can result in cracking if the tensile stressis high enough. Images of such cracking are shown in Fig. 4.2. In addition, this strain inthe growth process can cause bowing of the silicon substrates at the growth temperature;this has a significant impact on sample and device uniformity, particularly for large-areasubstrates.

Though a film can be grown at elevated temperatures with relatively low disloca-tion densities, when the film cools the difference between the thermal expansion

Gallium nitride (GaN) on silicon substrates for LEDs 83

Si: 5.43 A

AIN: 5.39 A

AIN(11–20)

º

Si:

5.43

Aº

º

AIN

: 5.3

9 Aº

Si(–100) Si(001)

Si(0–10)

AIN(1–100)AIN(1–100) Si(–1–12) Si(111)

Si(–110)Si atomAI or N atom

Figure 4.1 Atomic orientation of AlN on Si (111) and Si (100) substrates.3

84Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

coefficients of the film and the substrate causes the gallium nitride layer to be placed intensile strain. Because gallium nitride is a tough ceramic material, this can often resultin severe cracking in the gallium nitride layers, rendering them useless for devices. Thedifference in lattice parameters can also result in bowing of a wafer, which can havedeleterious effects on the uniformity of an as-grown crystal.6

Thus strategies for growing gallium nitride on silicon must not only lower thedislocation densities to allow the growth, they must also take into account the tensilestrain that tends to result from the thermal expansion difference. Therefore, the mostcommon strategies involve placing either an aluminum nitride structure or aluminumnitride superlattice within a buffer layer to compensate for this thermal strain uponcooling down. As the lattice constant of aluminum nitride is lower than that of galliumnitride, the will cause the gallium nitride layer to be in compressive strain; if thisresidual compressive strain can be maintained, then it is possible to grow crack-freelayers as discussed in the following pages.

4.4.2 Thermal management

Another problem with the use of traditional substrates is the low thermal conductivity ofsapphire. Thus, when light-emitting diodes are produced from layers grown on sapphire,resistive heating causes premature failure. As the need for solid-state illumination andhigher-bandgap materials for extreme ultraviolet applications increases, the problemwith low thermal conductivity is exacerbated. Several failure mechanisms have beenproposed for GaN-based devices when the heat generated in the device is not properlydissipated, ranging from local structural modifications in the material,7 to depassivationand reaction of the dopant states in the p-type material.8 In order to control the junctiontemperature of the active region, several strategies have been applied. One method is touse substrates with a higher thermal conductivity than sapphire, such as diamond9,10;

Figure 4.2 Optical micrograph of a GaN layer grown directly on a Si (111) substrate showingcracking.


this is cost-prohibitive for solid-state lighting applications where cost-competitivenesswith traditional lighting applications is essential. Silicon has more than three timesthe thermal conductivity of sapphire, which would be beneficial for high-brightnessdiodes.

4.5 Buffer-layer strategies

4.5.1 Zinc oxide (ZnO), aluminum arsenide (AlAs) and othermaterials

The search for suitable buffer layers has not been limited to just the more traditionalAlN-based buffers and superlattice structures. GaN layers on silicon have also beengrown with GaAs, AlAs, BP and ZnO buffer layers. The results of these studies andthe advantages of each of these buffer layers will be discussed below.

GaAs was used in early studies on the growth of GaN on silicon through a two-stepprocess of molecular beam epitaxy (MBE) growth and nitridation.11 This allowed for thegrowth of layers, but it is not economical and limits the growth at high temperatures. Inresponse, Strittmatter et al.12 attempted to grow AlAs layers using MOCVD in serieswith the growth process. These AlAs layers were thought to be more thermally stable,conductive and yet still epitaxial to the underlying Si (111) substrate. X-ray diffraction(XRD) scans of these layers showed linewidths of 290 arcsec for the q � zq scan.Secondary ion mass spectrometry (SIMS) revealed no significant diffusion of eitherarsenic or silicon into a layer, but there was a pronounced yellow band emission inthe photoluminescence (PL), as seen in Fig. 4.3.

Another different buffer layer grown on silicon is ZnO. Park et al.13 demonstratedthat enhanced photoluminescence emission could be observed in GaN films grown on

Yellow band

T = 8 Kλexc. = 325 nm

Pho

tolu

min

esce

nce

inte

nsity

(a.u

.)

1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3.50Energy (eV)

Excitonic emission

Figure 4.3 PL spectra of GaN on an AlAs buffer on silicon.12


ZnO buffers on Si (0001); however, the films were polycrystalline and not devicequality. Films were heavily textured in the (0001) direction.

Boron phosphide (BP) was used as buffer layer by Nishimura et al.14 This materialhas a good lattice match with GaN and had been successfully grown on silicon. More-over, it opens up the possibility for the growth of cubic GaN. They used MOCVD togrow 0.1-mm layers of BP on silicon, prior to the growth of GaN. Scanning electronmicroscopy (SEM) images of the GaN layer show a smooth interface. The XRDspectra of GaN layers grown on (100) surfaces have cubic diffraction peaks. However,significant amounts of impurities were found in the GaN layer, as evidenced byultraviolet illumination.

Silicon nitride has also been used as a buffer layer in the growth of GaN films. Liuet al.15 used in situ nitridation of the silicon surface. After cleaning, they thermallyetched the silicon passivation layer and used a 5 � 10�4 Torr pressure to nitridatethe surface, and followed that with an overgrowth of a GaN buffer layer. X-ray photo-electron spectroscopy (XPS) showed peaks characteristic of silicon and Si3N4. Moreimportantly, yellow luminescence was suppressed in the photoluminescence spectrum.Huang et al.16 also used a similar pre-nitridation process to grow GaN on silicon. Theyobserved a similar reduction in the yellow luminescence in the PL spectrum, andattributed this to a reduction in the substitutional oxygen and silicon impurities inthe substrate. Limiting the diffusion of gas phase silicon through passivation of thesilicon surface can also be an important method for limiting diffusion. This is anothernovel buffer-layer scheme.

4.5.2 Aluminum nitride (AlN) buffer layer

As other growth buffer layers are unsuitable for long-term integration into large-areaGaN LEDs, the focus for the buffer layers quickly shifted to AlN. Due to a significantamount of lattice mismatch (17%) between c-plane GaN and Si (111), a buffer layerhas to be applied to reduce the strain on the epitaxial GaN layer. Although SiC hasthe best match in terms of lattice constant, SiC is expensive; hence, a comparativelycheaper buffer layer has been sought for a long time. Watanabe et al.17 first proposedthat a thin AlN buffer layer could be grown over the silicon before GaN deposition,after Takeuchi et al.18 suggested using SiC as a buffer-layer material. They usedmetal-organic vapor phase epitaxy (MOVPE) as the growth process under atmosphericpressure.

As a buffer layer, AlN has several advantages: (1) it possesses a similar wurtzitestructure as the GaN crystal structure, (2) when the AlN buffer layer is grownon Si (111), the tensile strain between GaN and silicon is converted to compres-sive strain and (3) AlN prevents the direct contact of Ga atoms with the siliconatoms, as this causes an SiN layer to grow on top of the silicon substrate as aresult of nitridation. There have been reports of growing high-temperature AlN(HT-AlN) and low-temperature AlN (LT-AlN), both of which have their ownadvantages. For example, when AlN is grown on Si (111), the AlN layer suffers


from tensile strain from the silicon layer. This can be reduced through decreasingthe AlN growth temperature,19 which results in GaN growth on LT-AlN undercompressive interlayer-induced strain. On the other hand, better crystalline qualityof the grown GaN layers is observed when HT-AlN is used as the buffer layer20

(Fig. 4.4).Thus it appears from the research that LT-AlN helps to prevent cracking of the

epitaxial GaN layer, while an HT-AlN buffer layer results in better crystalline qualityof the GaN epilayer. Arslan et al.21 used the advantages of both these techniques, withLT-AlN sandwiched between twoHT-AlN layers, followed by a gradedAlxGa1�xN layerbefore the desired GaN epilayer was deposited. This graded AlxGa1�xN layer wasintroduced to further reduce cracking in the GaN layer, which occurred even thoughthe AlN stack was thicker than 400 nm. The AlxGa1�xN composition was reducedfrom x ¼ 0.64 to 0.22 (in seven steps) to match well with the GaN layer on top.

4.5.3 Superlattice structures

A superlattice (SL) is a structure where two different materials are grown to aspecific thickness in alternating layers. In addition to LT-AlN and HT-AlN bufferlayers, superlattice structures are also very popular among researchers for engineer-ing the strain in the GaN epilayer. In addition, a superlattice helps to reduce the

Si(111)

Si(111)

Si(111)

X-r

ay in

tens

ity (a

.u.)

X-r

ay in

tens

ity (a

.u.)27 28 29 30 31 32 33 34 35 36

2h (degree)

2h (degree)

GaN(0002)

GaN(0002)

GaN(0002)

GaN(0004)

GaN(10–11)GaN(10–11) GaN(10–13) GaN(0004)

AIN buffer layer thickness ~100A º



30 40 50 60 70 80

Figure 4.4 Q-to-2QXRD patterns for (top left) an HT-GaN film grown on an LT-AlN coated Si(111) and (right) HT-GaN films grown on HT-AlN coated Si (111) with AlN buffer layerthickness of 100, 200 and 1000 Å.20


dislocation glide significantly. A very common approach for growing a GaN epilayerthat is smoother and free of strain-induced defects is to grow a GaN/AlN superlatticeright on top of the AlN buffer layer and then grow the GaN epilayer on top ofthe superlattice.22 The researchers used 10 to 20 periods of SL before theMOCVD growth of GaN (300 nm) on AlN nucleated silicon, keeping the othergrowth conditions constant. From Raman spectral analysis of the GaN films grownon different numbers of SL pairs (Fig. 4.5, top), it was concluded that the GaNepilayer with a 15-pair GaN/AlN SL had significantly reduced in-plane stress in

567.32 (E2 high)

567.31 (E2 high)

566.08 (E2 high)

Inte

nsity

(a.u

.)In

tens

ity (a

.u.)

(c) 20 pair

(b) 15 pair

(a) 10 pair

A1(LO)

(002)

560 580 600 620 640 660 680 700 720 740Raman shift (cm–1)

(b)

(c)

(a)

–3000 –2000 –1000 1000 2000 30000Rocking angle (arcsec)

Figure 4.5 Top: Raman spectra for GaN epilayers grown on Si (111) substrates with (a) 10-pair,(b) 15-pair and (c) 20-pair GaN/AlN SLs. Bottom: HRXRD rocking curves of GaN epilayersgrown on Si (111) substrates with (a) 10-pair, (b) 15-pair and (c) 20-pair GaN/AlN SLs.22


comparison to the 10 and 20-pair SLs. In addition, a high-resolution X-ray diffraction(HRXRD) scan (Fig. 4.5, bottom) showed that crystal quality is better for the 15-pairSL structure as the full width at half maximum (FWHM) of the 15-pair SL structurewas the lowest among the tested samples. Optical microscope images did not showany cracks in the structures with 10 to 15-pair SLs, while cracks were clearly visiblein a 20-pair SL structure. Ma et al.23 fabricated an improved GaN-based LED on Si(111) using AlN/GaN SLs and compared its performance with LT-AlN band andMT-AlN interlayer based devices. In terms of dislocation density reduction, PLintensity, output power and crystalline quality of the active region, the LED devicesfabricated using AlN/GaN SLs showed much better performance than the other twotechnologies.

Dadgar et al.24 used a superlattice structure of AlN/AlGaN and achieved a 10 timeslower dislocation density than a ‘traditional’ GaN on silicon sample. They used a15-fold SL structure of 0.9 mm thickness. Lin et al.25 used a multilayer buffer ofAlN (HT-AlN/LT-AlN/HT-AlN) on Si (111), growing the GaN on top of it to reducecrack formation on the edges of a 150 mm silicon wafer. The LT-AlN layer (grown at800�C) helped to reduce the propagation of dislocations from the first HT-AlN layer(grown at 1050�C); it was also used to reduce the formation of SixNy. The thirdlayer (HT-AlN at 1050�C) helped in growing a more relaxed AlN layer on top. LaterSaengkaew et al.26 adopted a similar HT-AlN/LT-AlN method to grow AlxGa1�xNlayers for ultraviolet lighting applications. Another SL approach was followed byXi et al.,27 who grew AlGaN using an AlxGa1�xN/AlyGa1�yN SL structure andcompared it to AlGaN with an LT-AlN interlayer. Their atomic force microscopy(AFM) scans showed that samples grown with a AlxGa1�xN/AlyGa1�yN SL had alower root mean square (RMS) roughness (0.4 nm) than the samples grown withan LT-AlN interlayer (0.5 nm) (Fig. 4.6). The AFM scans also revealed the reduceddislocation density in the AlxGa1�xN/AlyGa1�yN SL structure samples; these samplesalso had better electron mobility.

4.5.4 Atomic layer deposition (ALD) of Al2O3

A novel approach for GaN growth on large-area silicon has been to grow a thin Al2O3barrier layer by MOCVD to help alleviate the strain mismatch and chemicaldiffusivity issue of the silicon. Atomic layer deposition gives precise control ofthe thickness of layers through self-limiting surface reactions. Although the growthrates are slow due to the atomically controlled layer-by-layer growth, it has been usedin industrial processes where precise atomic-level thickness control is needed. Aschematic of the ALD process is shown in Fig. 4.7. Self-limiting surface reactionsallow for the control of the aluminum oxide layers through sequential pulses oftrimethylaluminum and water, at a temperature below the pyrolytic temperature oftrimethylaluminum.

Fenwick et al.28 performed the initial studies of ALD-grown interlayers. Theysystematically investigated the growth and annealing process for the as-grown layers.


Following substrate cleaning, ALD layers from 5 to 100 nm were grown andsubsequently annealed at temperatures ranging from 1000 to 1300�C in either an airor nitrogen atmosphere. Amorphous alumina layers were observed on the surfaceupon growth, which then crystallized during annealing. The layers became mirror-like upon annealing, but the thicker layers had surface pits, as shown in Fig. 4.8.Following the optimization of the growth process, GaN layers were grown onthe thinner layers using either a high-temperature AlN layer (Fig. 4.8) or a low-temperature GaN buffer layer. Layers grown with the LT-GaN layer were found tohave relatively large linewidths (14,720 arcsec for the GaN (0002) reflection), whilethe linewidths of the HT-AL were around 4000 arcsec. Further optimization of theprocess resulted in improvements to the XRD (002) and (102) GaN reflections to380 arcsec and 740 arcsec, respectively.

(a)

(b)

n-Al0.3Ga0.7N

n-Al0.3Ga0.7N

grown on superlattice

grown on LT interlayer

0.5 μm

0.5 μm

RMS roughness: 0.40 nm

RMS roughness: 0.50 nm

Figure 4.6 AFM images of samples: (a) grown on an AlGaN(AlN)/AlGaN SL, showing lowerdislocation density; (b) grown on an LT AlN interlayer, showing the higher dislocation density.27


Mag = 2.00 K X10µm EHT = 10.00kV

WD = 8 mmSignal A = InLensPhoto no. = 5187

Date :12 Mar 2008Time :13:39:54

Signal A = InLensPhoto no. = 1616

Date :21 Jun 2008Time :11:58:21Mag = 50.00 K X

100nm EHT = 10.00 kVWD = 7 mm

Figure 4.8 (Top) Surface pits in GaN grown on 100 nm ALD Al2O3. (Bottom) SEM image ofHT-AlN interlayer on ALD Al2O3.

28

Purge

Purge

Source A Source B

Figure 4.7 Series of reactant and purge pulses used for single layer control in atomic layerdeposition.

Jamil et al.29 further improved the growth of these layers and implemented theminto LEDs. Using the same atomic layer deposition on silicon described above, theygrew free-standing GaN LEDs and removed the substrate. Fig. 4.9 shows a plot ofthe internal quantum efficiency (IQE) of gallium nitride LEDs on sapphire andALD substrates. Fig. 4.10 shows electroluminescence (EL) spectra of these twosubstrates; similarities can be observed and there are FabryePérot oscillations forthe ALD substrate. This work is promising for the future growth on large-areasubstrates.

4.6 Device technologies

4.6.1 Early device efforts

The previous sections discussed the advantages of silicon, the challenges involved inattaining the objectives and the buffer-layer approaches used to resolve the issues. Thissection will focus on the LED technologies that have influenced the research arenathe most in the last 15 years. Sapphire and SiC continue to be the most dominantsubstrate materials in LED chip fabrication. In 1998, Guha and Bojarczuk30,31 produceddouble heterostructure LEDs grown using MBE on Si (111) substrates with an AlNnucleation layer and AlGaN buffer layers. Their samples had high threading(w5 � 109 cm�2) and planar defects (towards the 0001 growth direction). They alsodemonstrated that p-type doping is possible for AlGaN/GaN quantum well (QW)devices on silicon.

100

90

80

70

60

50

40

30

0 10050 150 200 250 300

Inte

rnal

qua

ntum

effi

cien

cy (%

)

LED on GaN/sapphire

LED on ALD-Al2O3/Si

Temperature (K)

Figure 4.9 IQE data for LEDs grown on sapphire and ALD-coated silicon substrates.29


Later Zhang et al.32 and Dadgar et al.24 fabricated blue LED chips on Si (111)using MOCVD, but these samples made during the late 1990s produced poor bright-ness compared to those made on sapphire or SiC substrates. Zhang et al. used anAlGaN/AlN buffer layer with InGaN/GaN QWs as the active region. They were

(a)

(b)

LED on sapphire

LED on ALD/Si

1 mA10 mA20 mA30 mA40 mA50 mA60 mA70 mA80 mA90 mA100 mA

1 mA10 mA20 mA30 mA40 mA50 mA60 mA70 mA80 mA90 mA100 mA

Ele

ctro

lum

ines

cenc

e in

tens

ity (a

.u.)

Ele

ctro

lum

ines

cenc

e in

tens

ity (a

.u.)

450 500 550 600 650350 400

450 500 550 600 650350 400Wavelength (nm)

Wavelength (nm)

Figure 4.10 EL spectra for LEDs grown (a) on sapphire and (b) on Si/ALDAl2O3.29


influenced by Guha and Bojarczuk30,31 to use an AlGaN/AlN buffer layer to reducethe strain-induced cracks from the thermal mismatch. Dadgar et al.,33 on theother hand, used a patterned Si (111) substrate (with a SixNymask). Both these devicetechnologies yielded LEDs with microwatt output, which was very low comparedto the more mature sapphire and SiC substrate devices and was not suitable forgeneral illumination applications. The use of superlattice structures also becamepopular34e36 to reduce the strain-induced cracks in the epilayers grown on a barenucleation layer. The devices with a superlattice had improved cracking and henceluminescence similar to LEDs on sapphire, but the output efficiency was very lowas they required 150 A/cm2 input current for an output of 400 mW (Fig. 4.11).Although the efficiency was not very impressive, the devices maintained a stableperformance during aging tests.

4.6.2 Progress in large-area substrates

It is useful to grow LEDs on large-area silicon substrates, i.e., of size 150, 200 or300 mm, as these are the standard sizes employed in the current silicon deviceindustry, and using these sizes obviously reduces costs.

As demonstrated by CamGaN Ltd, a University of Cambridge spin-off, the cost ofhigh brightness GaN LEDs can be reduced by 80% if 150-mm silicon wafers are usedas substrates.2 The cost of a 600 Si (111) wafer is around USD 25, while a 600 sapphirewafer costs at least USD 650. In addition, a 600wafer has 40% more usable areacompared to a 200 wafer, and the wafer processing cost for a 600 wafer is very closeto that of a 200 wafer but it can be used to make 10 times as many LED chips.Li et al.37 showed that the growth of an AlN/Si template followed by a thin gradedAlGaN (w50 nm) layer ultimately facilitates the growth of an LED on a large-areasilicon wafer. The AlGaN layer helps to reduce cracking and bowing of the waferdue to the thermal mismatch. A 2-mm n-doped GaN layer was grown on top of anAlGaN layer, followed by eight periods of an InGaN/GaN multiple quantum well(MQW), with a magnesium-doped p-type GaN layer on top. Characterization ofthe LEDs37 showed a forward bias of 4.1 V with a reverse leakage currentof 20 mA (at �20 V). Fig. 4.12 shows the IeV and LeI characteristics of the devicesalong with the EL spectrum of the blue LEDs grown on a 600 silicon wafer. Thedevices37 had an output power in the milliwatt range with input current injectionin the milliampere region, which is a significant improvement compared to devicesproduced by Egawa et al.36

Tripathy et al.38 used nanoscale silicon-on-insulator (SOI) substrates for InGaN/GaN blue LEDs. The SOI technique is useful for growing LEDs that can be easilylifted off through sacrificial etching of the SiO2 and the silicon overlayer. Also,a very thin SOI helps to produce more blue-green light (which is absorbed in thesilicon substrate of a vertical LED). SOI (111) substrates were prepared using theSIMOX (separation by implantation of oxygen) process. The rest of the LEDstructure is the same as that produced by Li et al.37 SOI substrates prepared using


SIMOX appear to have a highly reflective mirror-like surface beneath the AlN bufferlayer, due to the high contrast of the refractive indices in the interface. This results inmultiple interference peaks, yielding an increase in the aggregated EL intensitycompared to LEDs grown on an Si (111) substrate (Plate 1, see color plate section).

(a)

(b)

500

400

300

200

100

00 50 100 150 200

LED on sapphire

Out

put p

ower

(μW

)

Injected current density (A/cm2)

1.1

1.0

0.9

0.8

0.7

0.6

0.50 200 400 600 800 1000 1200

ACC: 20 mAat 27º C

at 80º C

Aging time (h)

Rel

ativ

e in

tens

ityLED on Si

25 mm

p-electrode

Figure 4.11 (a) Comparison of LeI characteristics of blue LEDs on sapphire and siliconsubstrates; (b) variation of relative intensity with aging time.36


(a)

(b)

180

160

140

120

100

80

60

20

40

0

0 3 6 9V (V)

λ = 495 nmVF = 4.1 V

1.5

1.2

0.9

0.6

0.3

0.00 40 80 120 160

1.0

0.5

0.0400 500 600

λ (nm)

/ (m

A)

/ (mA)

/ (a.

u.)L

(mW

)

492 nm

Figure 4.12 (a) IeV and (b) LeI curves for an InGaN/GaN MQW blue LED. Inset: (a) Opticalmicroscope image of the LED. Inset: (b) EL spectrum showing the dominant emission atl ¼ 492 nm.37


This shows that the energetic maxima of the EL interference peaks of LEDs on a SOI(111) are significantly stronger than those on Si (111).

Zhu et al.39 inserted (without an SixNy layer) a 100-pair n-AlN/GaN (5/20 nm)strained layer superlattice (SLS) on top of the buffer layer, before the n-GaNepilayer. They increased the n-GaN epilayer thickness, while keeping the otherlayers the same; this boosts the external quantum efficiency (EQE) and light outputpower of the LEDs, as shown in Fig. 4.13. In addition, the increase in the n-GaN

(a)

(b)

Inte

nsity

(a.u

.)

(0002)

(10–12)

Sample ASample B

Sample A

Current (mA)

Sample B

–3000 –2000 –1000 1000 2000 30000

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0 0.00 10050 150 200 250 300 350

2.0

1.6

1.2

0.8

0.4

Ligh

t out

put p

ower

(mW

)

Ext

erna

l qua

ntum

effi

cien

cy (%

)

Figure 4.13 (a) HRXRD rocking curves for GaN (0002) and (10e12) u-scans of samples A(n-GaN thickness 2 mm) and B (n-GaN thickness 1 mm), (b) EQE and light output power of thesamples.39


epilayer thickness results in an improved crystalline quality of the GaN layer,as shown in the HRXRD scan of the two samples. The EQE of the thicker GaN layer(2 mm) LED appeared to be almost double that of the LED with the 1-mm GaNepilayer, and had a higher input current value e 325 mA, compared to 200 mA forthe thinner GaN layer.

A big improvement in the IQE of LEDs occurs with the insertion of an SixNy layeron top of the buffer layer. Initially, the growth of GaN on Si (111) suffered frombowing and hence crack-generation due to the large thermal expansion coefficientmismatch (46%) between GaN and Si (111). Phillips and Zhu40 and Zhu et al.41

reported that Si (111) itself has a slight convex bow, which switches to a concavebow after heating and in situ annealing (due to the positive temperature gradientbetween the top and bottom of the wafer).

Although deposition of an AlN nucleation layer increases concave bowing, there isconvex bowing after the growth of the buffer layers and the n-GaN epilayer, increasingthe compressive stress. After the growth of the MQWs, barrier layers and ultimatelythe p-GaN layer on top, there is still significant convex bowing. It was found thatwith temperature optimization in the growth chamber, bowing can be controlled toyield a flat wafer.

The LEDs grown had higher dislocation densities, but Zhu et al.41 deposited a SiNx

layer on of top of the AlGaN/AlN buffer layer before the GaN epilayers were grownonto it. This layer helped the GaN epilayer to bend over the generated defectsand annihilate them, and thus reduced the threading dislocation density. Fig. 4.14 is

(0002) g-5g

500 nm 500 nm

m

(1120) g-4g_

GaN

SiNx

Al0.60Ga0.40N

Al0.75Ga0.25N

Al0.30Ga0.70N

Al0.42Ga0.58N

AIN

Si(111)

Figure 4.14 Weak-beam dark-field TEM images of GaN/SiNx/AlGaN/AlN/Si structures.The position of each interface is indicated by a horizontal arrow. Pure screw/mixed typedislocations are visible in the TEM image to the left, with loops labeled with ‘l’, whilepure edge/mixed type dislocations and disorientated grains (‘m’) are visible in the TEM imageto the right.41


a transmission electron microscope (TEM) image of a GaN/SiNx/AlGaN/AlN/Sistructure. It was demonstrated that the bending of the dislocations at the interfaceswas due to the compressive stress caused by the larger in-plane a lattice parameterof GaN compared with AlN. It was also suggested that the AlGaN buffer layers filterthe crystal disorientation arising from the AlN nucleation layer, thus improvingthe crystal quality of the GaN layer grown on top. Zhu et al.2 produced GaN LEDson large-area Si (111) substrates. They reported 58% IQE (at the blue 460 nm wave-length) for the MQWGaN LEDs on a 600 Si (111) wafer (a similar design as,41) and thisis the highest reported IQE so far. In terms of dislocation densities, very high numbersof threading dislocations (TDs) have been reported (8e9 � 108/cm2); this is strongenough to limit the IQE, in comparison to that for GaN LEDs grown on sapphire(IQE ¼ 70%). Their observation of the TDs using AFM and the PL spectra are shownin Fig. 4.15.

(a) (b)

(c)

1 µm 1 µm

12

10

8

6

4

2

0

2.2 2.4 2.6 2.8 3.0 3.2

300 K

20 K

PL

inte

nsity

(a.u

.)

Energy (eV)

Figure 4.15 Large-area AFM images of GaN films on Si (111): (a) with an SixNy interlayer,(b) without an SixNy interlayer. (c) Temperature-dependent PL of a 460 nm MQW structure.2


Thus, temperature optimization during the MOCVD growth of epilayers and theinsertion of SixNy as a TD blocking layer facilitate the growth of GaN LEDs onlarge-area silicon wafers. Zhu et al.2,41 produced n-GaN layers of thickness1e1.5 mm. Zhu et al.39 showed that there should be improvements through increasingthe thickness of the layer. The opposite of what was reported by Zhu et al.39 applies, asthe use of SixNy is strongly capable of reducing the TDs in devices and hence helpingto increase the EQE to attain a value than that reported.

4.6.3 Layer transfer

Layer transfer, which is commonly achieved using the laser lift-off process, has beenused extensively recently in the industrial production of light-emitting diodes. Themethod is used in transferring a prefabricated device from the growth substrate ontoa new host substrate for final active device processing. This ultimately integrates thebetter quality material to create a monolithically optimized system. In the laserlift-off process, high energy (more than the bandgap of GaN) excimer laser (KrF) lightis applied to the back of the substrate wafer; the laser is focused on the seed layerbeneath the GaN epilayer. This evaporates nitrogen from the back of the epilayer; athin layer of liquid gallium remains on the epilayer, which is cleaned by a wet chemicaletching process. The sapphire substrate is removed following post-growth treatmentusing an excimer laser.

Once the growth layer has been removed, it can be rebonded to another materialwith a higher thermal conductivity, which acts as a heat sink and lowers the junc-tion temperature of the device under bias. This is a relatively specialized processand requires an additional treatment step with high-power lasers. Moreover, thisprocess results in significant local heating of the sample at the places where the laserablation removes the substrate; this damage can be detrimental to the performanceof certain devices. The lift-off process has also been used to improve the efficiencyof III-nitride-based laser diodes.42

Ideally, there should be a process that can remove a substrate without thepost-processing laser lift-off step. Wet chemistry is commonly used to form a releaselayer for III-V,43,44 II-VI45 and IV-VI46 electronic devices. For GaN on silicon, astrong etchant like HNO3 þ HF þ CH3COOH þ H2O is used for etching Si (111)in a chemical lift-off process. However, it is hard to control the etchant withoutaffecting the GaN epilayer.

Although silicon is considered to be a very economical substrate for verticallygrown GaN LEDs, the large lattice and thermal expansion coefficient mismatchesdo not support good GaN crystal growth on silicon. This can be resolved using anAlN buffer layer on Si (111) prior to GaN growth, as mentioned in Section 4.5.2.However, the large band offset between the AlN buffer layer and the Si (111) substrateobstructs the current-voltage characteristics of vertically conducting LEDs.30,47

In addition, Si (111) substrates absorb vertically emitted visible light. To eliminate


these two drawbacks, Zhang et al.48 devised an approach to transfer vertically grownInGaN multiple QW LEDs, using a selective chemical lift-off process, onto a coppercarrier. This selective lift-off (SLO) approach helps to reduce the probability of crackgeneration from strain relaxation, which originates with the complete etching of asilicon substrate using conventional wet etching. In SLO, the InGaN MQW LEDstructure grown on silicon is bonded onto a copper carrier via metal-to-metal bonding.Then, the SLO process is used to remove the silicon substrate. The complete processis shown in Fig. 4.16.

Prior to SLO, first the p-type ohmic contacts are formed on the epilayer surfaceusing a 10 nm thick nickel layer and annealed at 600�C for 3 min in ambient air.Next, a highly reflective Al/Au layer is deposited onto the annealed nickel surface.The InGaN LED epiwafer is then bonded onto a gold-coated copper carrier at200�C using indium as the adhesive. Then the silicon substrate is mechanicallypolished down to 60 mm. Next, the silicon substrate is selectively removed using an

(a)

(b) (c)

(d) (e)

Si

Si

Cu

Si

Cu

LED epilayersNiAl/Au reflector

Polyimide

In

Lift-off region

Si Si Si Si Si

Cu

500 µm

Figure 4.16 Selective lift-off process flow: (a) A p-type ohmic contact is formed and ahigh-reflectivity metal reflector (Al/Au) is deposited onto the epitaxial surface of the LEDepiwafer. (b) The metal/InGaN MQW LED/silicon is bonded onto a copper carrier. (c) Thesilicon substrate is thinned by mechanical polishing. (d) The silicon substrate is selectivelyremoved by wet-chemical etching. (e) An image of the device after the SLO process.48


HNA solution (HF:HNO3:CH3COOH ¼ 1:1:1) and polyimide as the mask. Finally,the exposed region of the buffer layers is subjected to reactive ion etching (RIE)and Ti/Au contacts are deposited to complete the fabrication of the LED. The reporteddevice performance is shown in Fig. 4.17.

A double flip process was demonstrated by Wong et al.49 and Lau et al.50: the firstflip was used to bond the LEDs on silicon to a temporary substrate, followed by asecond flip where the devices were removed from the silicon and transferred toa copper substrate by electroplating. This results in LEDs where the p-side layer ison top, maintaining the as-grown order of the epilayers. In this process, afteran LED is grown onto a silicon substrate, the entire front device surface is spin-coated with a polyimide layer and baked up to 180�C for 4 h. This protects theLEDs during wet etching of the silicon. Next, the device side of the wafer is bondedtemporarily to a sapphire wafer using wax. The bonded structure is then put into anHNA solution (HF:HNO3:CH3COOH ¼ 1:2:3) for 40 min, which completelyremoves the silicon substrate. Following etching, Ti (5 nm)/Al (100 nm)/Ti(10 nm)/Au (100 nm) are deposited onto the backside of the LEDs using e-beamevaporation. Of these metal layers, aluminum serves as a reflective mirror andgold acts as a seed layer for the subsequent copper electroplating. The second flipnow takes place where the LED is flipped from the temporary sapphire wafer toa copper substrate. An 80 mm thick copper layer is electrodeposited as the newsubstrate for the LEDs. Then the temporary sapphire layer is debonded from theLEDs on a hot plate, and the wax is removed using trichloroethylene (TCE) cleaning.The polyimide is removed using an organic resist stripper at 70�C. Fig. 4.18 showsthe improvement of devices transferred onto copper after being grown on silicon.49

It can be concluded that this layer transfer approach is very helpful for fabricatinghigh-power LEDs, for large-area LED arrays subject to repeated high thermaloperation cycles, for the hybrid integration of GaN devices on silicon and for flip-chip bonding of LEDs.

4.6.4 GaN LEDs on patterned silicon substrates

Patterned silicon substrates were first used by Kawaguchi et al.51 to grow a fewmicron-sized GaN dots. However, Yang et al.52 were the first to use this idea tofabricate LEDs on silicon, although their devices had cracks. Dadgar et al.33,53,54

and Strittmatter et al.55 produced GaN LEDs on silicon without cracks using siliconsubstrate patterning. SEM images of the sample are shown in Fig. 4.19.

In this research, an AlGaN/GaN buffer layer was used to reduce the tensile strain.However, all the LEDs mentioned here have poor brightness compared to conven-tional sapphire substrate devices. Zhang et al.56 produced patterned silicon substrateLEDs where they used an HT-AlN nucleation layer as a dislocation filter, especially


(a)

(b)

(c)

Yellow band

n-GaN

101

102

103

2.2 2.4 2.6 2.8 3.0 3.2 3.4Energy (eV)

PL

inte

nsity

(a.u

.)

200

150

100

50

00 50 100 150 200

Opt

ical

pow

er (μ

m)

Current (mA)50

40

30

20

10

00 1 2 3 4 5 6

Cur

rent

(mA

)

Voltage (V)

Figure 4.17 (a) n-GaN PL spectra at room temperature, (b) LeI characteristics and (c) IeVcharacteristics of the LED before (dashed lines) and after (solid lines) substrate removal. Theinset in (c) shows an emission image of the LED fabricated on the substrate removal region.The image was taken at 0.5 mA under room light and microscope light conditions. The emissionwavelength is about 518 nm.48


(a)

(b)

(c)

201816141210

864

20

Cur

rent

(mA

)

0 1 2 3 4 5Voltage (V)

On siliconOn copper

On siliconOn copper

On siliconOn copper

60

50

40

30

20

10

0

80

70

350 400 450 500 550Wavelength (nm)

Out

put p

ower

(mW

)

8

7

6

5

4

3

2

1

00 20 40 60 80 100 120

Current (mA)

EL

spec

trum

(μW

/nm

)

Figure 4.18 (a) IeV characteristics, (b) EL spectra at 20 mA and (c) LeI characteristics ofLEDs on silicon and a copper wafer.49


for edge and mixed dislocations (Fig. 4.20). The device output was significantly lower(0.7 mW at 20 mA) compared to that of sapphire substrate devices (2.2 mW at 20 mA)at that time. In later research, Lau et al.50 grew the same device on patterned silicon,transferred it onto copper and showed that the PL intensity and the output light powerwere significantly higher.

Chiu et al.57 reduced the pattern size on the silicon wafer and showed that thissignificantly improved the LED device performance. They prepared 340 � 340 mmsized islands on a silicon wafer to create a micro-patterned silicon (MPSi) substrateand 200 nm diameter islands to create a nano-patterned silicon (NPSi) wafer.Fig. 4.21 shows the LED’s schematics and Fig. 4.22 shows the variation in PLintensity and EQE with output light power of the MPLED (an LED grown onMPSi) and NPLED (an LED grown on NPSi) devices. They reported also that theNPSi devices showed a significant improvement in terms of TDs, and better surfacemorphology and light emission resulting from better carrier confinement and a higherradiative recombination rate. In addition, a higher injection current (100 mA) wasobserved for NPSi devices with 20% less droop in EQE.

4.6.5 Semipolar and nonpolar GaN LEDs on silicon

The quantum-confined Stark effect (QCSE) causes strong polarization in InGaNMQWs when the active GaN layers are grown in the c-axis direction and arethicker than 3 nm. This reduces the efficiency of the LEDs and was the motivationbehind the search for semipolar or nonpolar GaN LEDs. Since the devices are grownin the semipolar or nonpolar direction, the QCSE-induced polarization field actslaterally through the active region and does not create any lack of carrier confine-ment in the QW. The semipolar or nonpolar growth of LEDs on a silicon substrate

Stripes Si<110> Stripes Si<110>⊥||Figure 4.19 SEM cross sections of GaN grown on a structured Si (111) substrate with differentstripe orientations.54


(a) (b)

(c) (d)

PL

inte

nsity

(a.u

.)

15 000

12 500

10 000

7500

5000

2500

0

350 360 370 380

No patternPaterned

Out

put p

ower

(mW

)

12

10

8

6

4

2

0

0 50 100 150 200 250

LED on siliconLED on sapphire

Injected current (mA)Wavelength (nm)

Figure 4.20 (a) Cross section and (b) top view of GaN on patterned Si (111). (c) PL spectra of GaN samples with a pattern and without a pattern on thesilicon substrate. (d) LeI characteristics of InGaN/GaN LEDs on a silicon substrate and a sapphire substrate (before being packaged).56

Gallium

nitride(G

aN)on

siliconsubstrates

forLEDs

107

involves selective-area growth and patterning of silicon wafers. Fig. 4.23 showsthe different planes in a GaN wurtzite structure.58 Fig. 4.24 shows the resultsfrom a theoretical simulation, which demonstrates the benefit of using the nonpolardirections for growing GaN LEDs (because it reduces the QCSE in the verticaldevice direction).58,59 Hikosaka et al.60 fabricated an InGaN/GaN active-regionLED using (1) GaN (1e101) growth on Si (001) and (2) GaN (11e22) growthon Si (113).60,61 Previously, they had grown nonpolar GaN via selective-area growth(SAG).62 Now they used the SAG epitaxial lateral overgrowth (SAG-ELO)technique. In this process, for example, a (111) face is opened on a (001) siliconwafer using KOH anisotropic etching and the other opened faces ((001) and(�1�11)) are covered using an SiO2 film. This silicon substrate is placed in a growthchamber to grow GaN by SAG on the open face. c-plane GaN grows on the cleavedopen face (normal to <111>) but at an inclined angle to the original silicon waferplane (001). Thus, a different plane of GaN (1�101) is observed as normal to theoriginal silicon wafer. The process is shown in Fig. 4.25.60 They also demonstratedthat to get a better quality GaN film, a deep groove followed by a narrow facetopening is helpful.

Ni et al.63 used a similar SAG-ELO growth process to grow m-plane (1�100)InGaN/GaN QW LEDs on a Si (112) wafer. They also showed that the TDs, by nature,propagate along the c-direction, and thus have a lesser impact on the device epilayersgrown along the m-direction. A schematic of their device, along with XRD scans, isshown in Fig. 4.26. To compare the m-plane and c-plane LEDs, the PL intensitiesof the devices were examined: the m-plane LED did not show any blue-shift comparedto the c-plane device. Additionally, it had a better IQE than the c-plane device, asshown in Fig. 4.2763: the IQE was 65% at a steady-state carrier density of1.2 � 1018/cm3, which appears to be almost twice that of the c-plane LED. Chiuet al.64 grew a (1�101) semipolar GaN LED on Si (001) achieving a low TD densityalong the growth direction, as did Ni et al.63 They also reported that semipolar GaNLEDs show less efficiency droop compared to polar (c-plane) LEDs at a high injectioncurrent.64 However, they also reported that there was a small blue shift in the PLspectrum of a semipolar LED.

p-GaN

InGaN/GaNMQWs X 5

n-GaN

AIN/AlGaN

u-GaNAlGaN

AIN

MPSi NPSi

Figure 4.21 Left: MPLED; right: NPLED.57


(a) (b)

(c) (d)

MPLEDs NPLEDs

PL

inte

nsity

(a.u

.)

PL

inte

nsity

(a.u

.)

0 10 20 30 40 50 0 10 20 30 40 50

Eb = 59 meV Eb = 87 meV

1000/T (K–1) 1000/T (K–1)

4

2

00 20 40 60 80 100

0

20

40

60

80

100

MPLEDs

Nor

mal

ized

EQ

E (%

)

Ligh

t int

ensi

ty (a

.u.)

Ligh

t int

ensi

ty (a

.u.)

4

2

00 20 40 60 80 100

Current density (A/cm2)Current density (A/cm2)

NPLEDs

100

80

60

40

20

0

Nor

mal

ized

EQ

E (%

)

Figure 4.22 PL intensity of (a) MPLEDs, (b) NPLEDs. Light intensity and EQE of (c) MPLEDs and (d) NPLEDs.57

Gallium

nitride(G

aN)on

siliconsubstrates

forLEDs

109

z (c-axis)

x (a-axis)

y (m-axis)

a-plane

c-plane

m-plane

Figure 4.23 Hexagonal prism representing a GaN crystal unit cell with nonpolar (a and m) andpolar (c) planes.58

(b)(a) Polar (0001)In0.15 Ga0.85N/GaN

3 nm/15 nm

Nonpolar (1120)In0.15 Ga0.85N/GaN

3 nm/15 nm

<0001>

PSP PSP PSP

PS

P

PS

P

PS

PPPZ

PP

Z

GaN

bar

rier

InG

aN

GaN

bar

rier

GaN

bar

rier

InG

aN

GaN

bar

rier

Ene

rgy

(eV

)

Ene

rgy

(eV

)

0.5

0

–0.5

–1.0

–3.0

–3.5

–4.0

–4.5–10 –5 5 100

2.60 eV

Fpol

DEC:DEV = 5:1 DEC:DEV = 5:1

Z (nm) Z (nm)

1.5

1.0

0.5

0

–2.0

–2.5

–3.0

–3.5–10 –5 0 5 10

2.81 eV

<1120>–

<0001>

1.74 MV cm–1

–

Figure 4.24 QW structures on (a) polar and (b) nonpolar orientations and their banddiagrams. Polarization charges appear at the interfaces of the polar-oriented QW andinduce electric fields that spatially separate electrons and holes in the QW. In nonpolarorientations, polarization charges do not affect the band structure. Because of the internalelectric fields in a polar-oriented QW, the transition energy is lower than that of a nonpolar-oriented QW (due to the QCSE). When a QW is embedded in a common þc-oriented LEDstructure, the internal electric fields increase as the LED positive bias increases.58,59

4.6.6 GaN nanowires and nanorods on silicon

For growing better GaN-based LEDs, only the epitaxial growth of GaN films has sofar been discussed. It is also possible to grow 3D wire-like GaN entities on siliconsubstrates, which are called nanorods, nanowires or nanocolumns. These structureshave significant advantages over a traditional epilayer grown on silicon: (1) nanorodshave zero or very few dislocations, (2) the larger surface area helps to outcouple lightto give a brighter LED, (3) nanorods have very high quality crystallinity along with astrong ability to manage the strain induced during growth and (4) nanorods do notcause bowing of the wafer on large-area substrates.65 On the other hand, they arenot free of constraints: (1) the growth method is more complicated than conventionalepilayer growth and (2) the large surface area is subject to the detrimental effects ofsurface states and possible surface degradation. In general, successful GaN nanoroddevices on silicon were initially produced using radio-frequency (RF) plasma-assisted MBE on n-type Si (111)66,67 or bulk Si (111).68 However, because of thepossibility of forming SixNy at the GaN-Si interface, Chen et al.67 and Callejaet al.69 grew ultrathin Si3N4 buffer layers using plasma nitridation on clean Si (111)7 � 7 surfaces, and then grew AlN, to avoid the possible loss of epitaxial orientation.

(a) (b)

(c)

p-contact (Ni/Au)

GaN:Mg (170 nm)

InGaN/GaN SQW (4/8 nm)

GaN: Si base layer

n-Si substrate

Al0.1Ga0.9N:Mg (20 nm)

n-contact (AuSb)

<1–101>

<11–22>

<0001>

<0001>

(001)Si substrate

(113)Si substrate

1 µm

1 µm

Figure 4.25 (a) (1e101) LED device structure. (b), (c) Cross-sectional SEM images of a(1�101) LED and a (11�22) LED.60 (SQW: single quantum well).


Kikuchi et al.66 grew GaN-based nanorods using an RF-MBE process. Theyfabricated p-n junction nanocolumn LEDs with InGaN/GaN multiple quantum disk(MQD) active layers. They increased the growth diameter of the nanorod LEDs inthe upper p-type GaN region, and ultimately all the nanorods were connected togetherin that region. This coalescence on top of the nanorods helped to form a continuouslayer. This was an ELO process and the LEDs required only a simple contact formationon top. Fig. 4.28 shows an SEM cross section of the device66 and the schematic of thedevice. Chen et al.67 used a similar growth process to that of Kikuchi et al.,66 but theyused n-type Si (111) on which, prior to GaN nanorod growth, an ultrathin b-Si3N4

buffer layer was formed by plasma nitridation; this prevented the silicon surfacefrom reacting with the nitrogen atoms. They grew two different lengths of nanorods:0.4 and 1.0 mm. A field emission scanning electron microscope (FE-SEM) scan and

(0001)GaN

(1100)GaN

(1120)GaN

(1101)GaN

(111)Si

(112)Si

(1101) m-plane GaN

Dislocations

Si(111)

GaN(1100)

Si(111)

Si(112) substrate

71º

AIN

(110)Si

SiO2

Inte

nsity

(a.u

.)

Rocked toward a-axisFWHM: 9 arcmin

Rocked toward c-axisFWHM: 27 arcmin

15 16 17 18 19Omega (degrees)

(a)

(b)

Figure 4.26 (a) Growth of selective area m-plane GaN on patterned Si (112). (b) X-ray rockingcurves of a GaN film grown toward the m-plane showing broadening due to tilting during lateralgrowth in both þc and �c directions.63


room-temperature (RT) PL spectra of a 1.0 mm GaN nanorod sample67 are shown inFig. 4.29. From the RT-PL spectral analysis, Chen et al.67 reported that the devicehas strong near-band-edge ultraviolet photoluminescence at 3.40 eV with an FWHMof 50 meV. No defect-related emissions except for a very weak broadband yellow

380 400 420 440 460 480 500Wavelength (nm)

PL

inte

nsity

(a.u

.)P

L in

tens

ity (a

.u.)

5.1 kW/cm2

0.05 kW/cm2

390 420 450 480 510Wavelength (nm)

IQE

1.0

0.5

0.00.0 0.4 0.8 1.2

m-plane InGaN DHon Si (112)

c-plane ref

n (1018 cm–3)

(a)

(b)

Figure 4.27 Room-temperature PL spectra (measured with a HeCd laser at different excitationdensities) for 6 nm thick InGaN double heterostructure LED active layers on (a) c-GaN onsapphire and (b) m-GaN on Si (112). The excitation power densities in both plots were 0.05,0.15, 0.52, 1.0, 2.0, 2.5 and 5.1 kW/cm 2. The inset in (b) shows the IQEs of both samplesextracted from the excitation dependence of the PL intensity using a titanium-sapphire laser(370 nm).63 (DH: double heterostructure).


emission were observed at room temperature, as the GaN nanorods grown under theoptimized conditions efficiently suppress yellow luminescence; this is an importantstep toward obtaining pure green LEDs.

The MOCVD growth of GaN nanorods was demonstrated by F€undling et al.70

using a pre-patterned Si (111) substrate. They grew hexagonal silicon pillars vialithography (e-beam or photo). The LED structures were grown on 3D siliconmicro-structures (5 � 5 mm) and nano-structures (700 � 700 nm), which werepatterned with hexagonal columns. Fig. 4.30 shows a schematic of the LED on siliconpillars. F€undling et al.70 also reported that the funnel-type growth of p-GaN, as shown

Transparent p-electrode (Ni/Au)

p-GaN

InGaN/GaNMQD active

InGaN/GaNMQD active

InGa liquid metal

n-GaNnanocolumn

n-electrode (Ti)n-(111) Si substrate

n-(111) Si substrate

Cu heat sink

p-GaN

i-GaN

i-GaN

n-GaN

500 nm

(a)

(b)

Figure 4.28 (a) InGaN/GaN nanorod LEDs on Si (111); (b) SEM of a device cross section witha single nanorod.66


in Fig. 4.31, is highly influenced by the incorporation of magnesium ions during thegrowth of the p-GaN layer. This results in enhanced lateral growth of the ELO GaNlayer. They also found additional growth in the {1e101} planes, which ultimatelycreated a 3D pyramidal structure in the LED with a hexagonal top facet, as seen inthe cathodoluminescence (CL) images. Fig. 4.31 shows the CL intensity due to lumi-nescence from different facets, top, side and bottom. As the CL peaks are differentwhen different facets are illuminated, it was estimated that this was due to changesin indium content and/or growth rates in the {1�101} or (0001) planes. When CLmeasurements were taken at both room temperature and cryogenic temperature, ablue shift was observed. When the bottom layer was illuminated, it output a broadluminescence around 570 nm (the yellow band) with a low intensity, indicating there

Inte

nsity

(a.u

.)

298 K

350 400 450 500 550 600 650 700Wavelength (nm)

Inte

nsity

(a.u

.)

3.1 3.2 3.3 3.4 3.5 3.6

2 LO

1 LO

Energy (eV)

~50 meV

3.40 eV

1 μm

(a)

(b)

Figure 4.29 (a) FE-SEM (cross-section view) and (b) RT-PL spectra of 1.0 mm GaN nanorodsample.67


p-type GaN (200 nm)

5 x InGaN/GaN MQWs

GaN (1.6 µm)

AIN (100 nm)

Si pillar (8 µm)

5 µm 5 µm

10 µm

(a) (b)

(c) (d)

Figure 4.30 SEM images of Si (111) pillars after dry etching: (a) 5 � 5 mm (nominal diameter � distance) array, (b) 700 � 700 nm array, without aphotoresist mask on the top (0001) plane. (c) Cross section of the silicon/GaN 3D LED structure. (d) FE-SEM image of GaN structures grown on apatterned silicon substrate showing a 5 � 5 mm array.70

116Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

was a high defect density in the bottom layer. In another report, Hasegawa et al.71

observed a more vertical and uniform growth of GaN nanorods on Si (001) whenSiO2 was grown as an intermediate layer between a GaN nanorod and the Si (001) sur-face. This also assists in preventing the formation of Si3N4.

4.7 Conclusion

The development of solid-state lighting technologies for the immediate future willfocus on methods for reducing the cost of the solid-state lighting sources. One ofthe quickest ways to reduce the cost of these devices is to reduce the cost of the sub-strate by switching from sapphire to silicon. Provided that the same quality of light canbe achieved and the same output levels and efficiency are realized, this will present avery promising avenue for future research and industrial development. This chapter

350 400 450 500 550 600

CL at RT, 20 keVWhole pyramidTop facet Side facetBottom

Wavelength (nm)

366 nm 412 nm 500 nm

CL

inte

nsity

(PM

T co

unts

)

Figure 4.31 Room-temperature CL spectrum of a GaN LED structure made by exciting thewhole pyramid compared with those made by exciting its top facet, one of its side facets andthe deposition on the bottom substrate. The upper CL images were taken at the peak wave-lengths of 366, 412 and 500 nm.70


has briefly described some of the methods used to grow GaN on silicon substrates andthe challenges associated with this growth. The latter half of this chapter focused onsome of the recent developments for LED devices on silicon. Future efforts are sureto push the boundaries in this rapidly developing field.

References

1. Craford MG. LEDs for solid state lighting and other emerging applications: status, trends,and challenges. Proc SPIE 2005;5941:594101.

2. Zhu D, McAleese C, Haberlen M, Kappers MJ, Hylton N, et al. High-efficiency InGaN/GaNquantum well structures on large area silicon substrates. Phys Stat Sol A 2012;209(1):13e6.

3. Dadgar A, Schulze F, WieneckeM, Gadanecz A, Bl€asing J, et al. Epitaxy of GaN on siliconeimpact of symmetry and surface reconstruction. New J Phys 2007;9:389.

4. Reiher F, Dadgar A, Blasing J, Wieneke M, Muller M, et al. InGaN/GaN light-emittingdiodes on Si(1 1 0) substrates grown by metaleorganic vapour phase epitaxy. J Phys DAppl Phys 2009;42:055107.

5. Ishikawa H, Yamamoto K, Egawa T, Soga T, Jimbo T, et al. Thermal stability of GaN on (11 1) Si substrate. J Cryst Growth 1998;189/190:178e82.

6. Krost A, Schulze F, Dadgar A, Strassburger G, Haberland K, et al. Simultaneous mea-surement of wafer curvature and true temperature during metalorganic growth of group-IIInitrides on silicon and sapphire. Phys Stat Sol B 2005;242(13):2570e4.

7. Meneghesso G, Levada S, Zanoni E, Podda S, Mura G, et al. Failure modes and mechanismsof DC-aged GaN LEDs. Phys Stat Sol C 2002;194:389e92.

8. Meneghini M, Trevisanello L, Meneghesso G, Zanoni E, Rossi F, et al. High-temperaturefailure of GaN LEDs related with passivation. Superlattices Microstruct 2006;40:405e11.

9. Song JH, Huang JL, Lu HH, Sung JC. Investigation of wurtzite (B,Al)N films prepared onpolycrystalline diamond. Thin Solid Films 2007;516:223e7.

10. May PW, Tsai HY, Wang WN, Smith JA. Deposition of CVD diamond onto GaN.Diamond Relat Mat 2006;15:526e30.

11. Nakamura S, Senoh M, Nagahama S, Iwasa N, Yamada T, et al. InGaN/GaN/AlGaN-basedlaser diodes with modulation-doped strained-layer superlattices grown on an epitaxiallylaterally overgrown GaN substrate. Appl Phys Lett 1998;72:211.

12. Strittmatter A, Krost A, Strassburg M, T€urck V, Bimberg D, et al. Low-pressure metalorganic chemical vapor deposition of GaN on silicon(111) substrates using an AlAsnucleation layer. Appl Phys Lett 1999;74:1242.

13. Park DC, Fujita S, Fujita S. Growth of polycrystalline GaN on silicon (001) substrates byRF plasma chemical vapor deposition with ZnO buffer layer. Phys Stat Sol A 1999;176:579.

14. Nishimura S, Matsumoto S, Terashima K. GaN on Si for LED and LD applications. PhysStat Sol C 2004;1:238.

15. Liu H, Ye Z, Zhang H, Zhao B. Wurtzite GaN epitaxial growth on Si(111) using SiN as aninitial layer. Mat Res Bull 2000;35:1837.

16. Huang J, Ye Z, Wang L, Yuan J, Zhao B, et al. Comparison of GaN epitaxial films onsilicon nitride buffer and Si(111). Solid-State Electron 2002;46(8):1231e4.

17. Watanabe A, Takeuchi T, Hirosawa K, Amano H, Hiramatsu K, et al. The growth of singlecrystalline GaN on a Si substrate using AlN as an intermediate layer. J Cryst Growth 1993;128:391e6.


18. Takeuchi T, Amano H, Hiramatsu K, Sawaki N, Akasaki I. Growth of single crystallineGaN film on Si substrate using 3C-SiC as an intermediate layer. J Cryst Growth 1991;115:634.

19. Blasing J, Reiher A, Dadgar A, Diez A, Krost A. The origin of stress reduction by low-temperature AlN interlayers,. Appl Phys Lett 2002;81(15):2722e4.

20. Gong J, Yeh M, Wang C. Growth and characterization of GaN and AlN films on (111) and(001) Si substrates. J Cryst Growth 2003;247:261e8.

21. Arslan E, Ozturk M, Teke A, Ozcelik S, Ozbay E. Buffer optimization for crack-free GaNepitaxial layers grown on Si(111) substrate by MOCVD. J Phys D Appl Phys 2008;41:155317.

22. Kim T, Yang S, Son J, Hong Y, Yang G. Growth of a GaN epilayer on a Si (111) substrateby using and AlN/GaN superlattice and application to a GaN microcavity structure withdielectric-distributed Bragg reflector. J Kor Phys Soc 2007;50(3):801e5.

23. Ma J, Zhu X, Wong K, Zou X, Lau KM. Improved GaN-based LED grown on silicon (111)substrates using stress/dislocation-engineered interlayers. J Cryst Growth 2012;370:265e8.

24. Dadgar A, Christen J, Riemann T, Richter S, Blasing J, et al. Bright blue electrolumines-cence from an InGaN/GaN multiquantum-well diode on Si(111): impact of an AlGaN/GaNmultilayer. Appl Phys Lett 2001;78:2211.

25. Lin K, Chang E, Hsiao Y, Huang W, Li T, et al. Growth of GaN film on 150 mm Si (111)using multilayer AlN/AlGaN buffer by metal-organic vapor phase epitaxy method. ApplPhys Lett 2007;91:222111.

26. Saengkaew P, Dadgar A, Blasing J, Hempel T, Veit P, et al. Low-temperature/high-temperature AlN superlattice buffer layers for high-quality AlxGa1�xN on Si(111).J Cryst Growth 2009;311:3742e8.

27. Xi Y, Chen K, Mont F, Kim J, Schubert E, et al. Comparative study of n -type AlGaN grownon sapphire by using a superlattice layer and a low-temperature AlN interlayer. J CrystGrowth 2007;299:59e62.

28. Fenwick W, Melton A, Xu T, Li N, Summers C, et al. Metal organic chemical vapordeposition of crack-free GaN-based light emitting diodes on Si (111) using a thin Al2O3

interlayer. Appl Phys Lett 2009;94:222105.29. Jamil M, Xu T, Melton A, Jampana B, Zaidi T, Liu S, et al. , Free-standing GaN-based

LEDs with ALD-Al 2O3/Si substrate removed by wet etching. In: Proceedings of SPIE7784, Tenth International Conference on Solid State Lighting, 77840E (18 August 2010);2010.

30. Guha S, Bojarczuk NA. Ultraviolet and violet GaN light emitting diodes on silicon. ApplPhys Lett 1998;72:415e7.

31. Guha S, Bojarczuk NA. Multicolored light emitters on silicon substrates. Appl Phys Lett1998;73:1487e9.

32. Zhang BJ, Egawa T, Ishikawa H, Nishikawa N, Jimbo T, et al. InGaN multiple-quantum-well light emitting diodes on Si(111) substrates. Phys Stat Sol A 2001;188(1):151e4.

33. Dadgar A, Alam A, Riemann T, Blasing J, Diez A, et al. Crack-free InGaN/GaN lightemitters on Si(111). Phys Stat Sol A 2001;188(1):155e8.

34. Tran CA, Osinski A, Karlicek Jr RF, Berishev I. Growth of InGaN/GaN multiple-quantum-well blue light-emitting diodes on silicon by metalorganic vapor phase epitaxy. Appl PhysLett 1999;75:1494e6.

35. Dadgar A, Poschenrieder M, Blasing J, Fehse K, Diez A, et al. Thick, crack-free blue light-emitting diodes on Si(111) using low-temperature AlN interlayers and in situ SixNy

masking. Appl Phys Lett 1999;80:3670e2.


36. Egawa T, Zhang B, Ishikawa H. High performance of InGaN LEDs on (111) silicon sub-strates grown by MOCVD. IEEE Elec Dev Lett 2005;26(3):169e71.

37. Li J, Lin J, Jiang H. Growth of III-nitride photonic structures on large area silicon substrates.Appl Phys Lett 2006;88:171909.

38. Tripathy S, Lin V, Teo S, Dadgar A, Diez A, et al. InGaN/GaN light emitting diodes onnanoscale silicon on insulator. Appl Phys Lett 2007;91:231109.

39. Zhu Y, Watanabe A, Lu L, Chen Z, Egawa T. High performance of GaN-based lightemitting diodes grown on 4-in. Si(111) substrate. Jap J Appl Phys 2011;50:04DG08.

40. Phillips A, Zhu D. UK cracks GaN-on-silicon LEDs. Compd Semicond 2009;19.41. Zhu D, McAleese C, H€aberlen M, Salcianu C, Thrush T, et al. InGaN/GaN LEDs grown on

Si(111): dependence of device performance on threading dislocation density and emissionwavelength. Phys Stat Sol C 2010;7(7e8):2168e70.

42. Wong WS, Kneissl M, Mei M, Treat DW, Teepe M, et al. Integration of InxGa1�xN laserdiodes with dissimilar substrates by laser lift-off. MRS Proc 2001;639. 12e2.

43. van Niftrik ATJ, Schermer JJ, Bauhuis GJ, Mulder P, Larsen PK, et al. HF species anddissolved oxygen on the epitaxial lift-off process of GaAs using AlAsP release layers.J Electrochem Soc 2008;155:35e9.

44. Kyono CS, Ikossi-Anastasiou K, Rabinovich WS, Bowman SR, Katzer DS, et al. GaAs/AlGaAs multiquantum well resonant photorefractive devices fabricated using epitaxial lift-off. Appl Phys Lett 1994;64:2244e6.

45. Chou HC, Rohatgi A, Jokerst NM, Kamra S, Stock SR, et al. Approach toward high effi-ciency CdTe/CdS heterojunction solar cells. Mat Chem Phys 1996;43:178e82.

46. Li YF, Sow A, Yao C, McCann PJ. Transfer of IV-VI multiple quantum well structuresgrown by molecular beam epitaxy from Si substrates to copper. Thin Solid Films 2005;488:178e84.

47. Ishikawa H, Zhang B, Egawa T, Jimbo T. Valence-band discontinuity at the AlN/Siinterface. Jpn J Appl Phys 2003;42:6413e4.

48. Zhang B, Egawa T, Ishikawa H, Liu Y, Jimbo T. Thin-film InGaN multiple-quantum-welllight-emitting diodes transferred from Si (111) substrate onto copper carrier by selective lift-off. Appl Phys Lett 2005;86:071113.

49. Wong K, Zuo X, Chen P, Lau KM. ,Transfer of GaN-based light-emitting diodes fromsilicon growth substrate to copper. IEEE Elec Dev Lett 2010;31(2):132e4.

50. Lau KM, Wong K, Zuo X, Chen P. Performance improvement of GaN-based light-emittingdiodes grown on patterned Si substrate transferred to copper. Opt Exp 2011;19:A956e61.

51. Kawaguchi Y, Honda Y, Matsushima H, Yamaguchi M, Hiramatsu K, et al. Selective areagrowth of GaN on Si substrate using SiO2 mask by metalorganic vapor phase epitaxy. Jap JAppl Phys 1998;37:L966e9.

52. Yang J, Lunev A, Simin G, Chitnis A, Shatalov M, et al. Selective area deposited blueGaNeInGaN multiple-quantum well light emitting diodes over silicon substrates. ApplPhys Lett 2000;76:273e5.

53. Dadgar A, Poschenrieder M, Blasing J, Fehse K, Riemann T, et al. MRS Fall Meeting, vol.14.7; 2001. Boston, MA.

54. Dadgar A, Strittmatter A, Bl€asing J, Poschenrieder M, Contreras O, et al. Metalorganicchemical vapor phase epitaxy of gallium-nitride on silicon. Phys Stat Sol C 2003;0(6):1583e606.

55. Strittmatter A, Rodt S, Reissmann L, Bimberg D, Schr€oder H, et al. Maskless epitaxiallateral overgrowth of GaN layers on structured Si(111) substrates. Appl Phys Lett 2001;78:727e9.


56. Zhang B, Liang H, Wang Y, Feng Z, Ng KW, et al. High-performance III-nitride blue LEDsgrown and fabricated on patterned Si substrates. J Cryst Growth 2007;298:725e30.

57. Chiu C, Lin C, Deng D, Lin D, Li J, et al. Optical and electrical properties of GaNebasedlight emitting diodes grown on micro-and nano-scale patterned Si substrate. IEEE J QuanElec 2011;47(7):899e906.

58. Masui S, Nakamura S, DenBaars S, Mishra U. Nonpolar and semipolar III-nitride light-emitting diodes: achievements and challenges. IEEE Trans Elec Dev 2010;57(1):88e100.

59. Chichibu SF, Uedono A, Onuma T, Haskell BA, Chakraborty A, et al. Origin of defect-insensitive emission probability in In-containing (Al,In,Ga)N alloy semiconductors. NatMater 2006;5:810e6.

60. Hikosaka T, Tanikawa T, Honda Y, Yamaguchi M, Sawaki N. Fabrication and properties ofsemi-polar (1e101) and (11e22) InGaN/GaN light emitting diodes on patterned Si sub-strates. Phys Stat Sol C 2008;5(6):2234e7.

61. Tanikawa T, Hikosaka T, Honda Y, Yamaguchi M, Sawaki N. Growth of semi-polar(11e22)GaN on a (113)Si substrate by selective MOVPE. Phys Stat Sol C 2008;5(9):2966e8.

62. Hikosaka T, Narita T, Honda Y, Yamaguchi M, Sawaki N. Optical and electrical propertiesof (1e101)GaN grown on a 7 off-axis (001)Si substrate. Appl Phys Lett 2004;84:4717.

63. Ni X, Wu M, Lee J, Li X, Baski A, et al. Nonpolar m-plane GaN on patterned Si(112)substrates by metalorganic chemical vapor deposition. Appl Phys Lett 2009;95:111102.

64. Chiu C, Lin D, Lin C, Li Z, Chang W, et al. Reduction of efficiency droop in semipolar(1101) InGaN/GaN light emitting diodes grown on patterned silicon substrates. Appl PhysExp 2011;4:012105.

65. Waag A, Wang X, F€undling S, Ledig J, Erenburg M, et al. The nanorod approach: GaNnanoLEDs for solid state lighting. Phys Stat Sol C 2011;8(7e8):2296e301.

66. Kikuchi A, Kawai M, Tada M, Kishino K. InGaN/GaN multiple quantum disk nanocolumnlight-emitting diodes grown on (111) Si substrate. Jpn J Appl Phys 2004;43(12A):L1524e6.

67. Chen H, Lin H, Shen C, Gwo S. Structure and photoluminescence properties of epitaxiallyoriented GaN nanorods grown on Si(111) by plasma-assisted molecular-beam epitaxy. ApplPhys Lett 2006;89:243105.

68. Park Y, Park C, Fu D, Kang T, Oh J. Photoluminescence studies of GaN nanorods on Si(111) substrates grown by molecular-beam epitaxy. Appl Phys Lett 2004;85:5718e20.

69. Calleja E, Ristic J, Fernandez-Garrido S, Cerutti L, Sanchez-Garcia M, et al. Growth,morphology, and structural properties of group-III-nitride nanocolumns and nanodisks.Phys Stat Sol B 2007;244:2816e37.

70. F€undling S, Li S, S€okmen €U, Merzsch S, Hinze P, et al. Three-dimensionally structuredsilicon as a substrate for the MOVPE growth of GaN nanoLEDs. Phys Stat Sol A 2009;206(6):1194e8.

71. Hasegawa S, Seo J, Uchida K, Tambo H, Kameoka H, et al. Influence of native siliconoxides on the growth of GaN nanorods on Si (001). Phys Stat Sol C 2009;6(S2):S570e3.



Phosphors for white LEDs 5Zhanchao Wu1,2, Zhiguo Xia 2

1Qingdao University of Science and Technology, Qingdao, People’s Republic of China;2University of Science and Technology Beijing, Beijing, China

5.1 Introduction

White light-emitting diodes (wLEDs), the new-generation illumination light sourcereplacing the conventional incandescent and fluorescent lamp, have attracted moreand more attention from researchers, merchants and customers due to the followingadvantages, such as low energy consumption, high efficiency, long lifetime, environ-mental friendliness and so on.1e5 As it is reasonably estimated, white light solid-statelighting on the basis of LEDs-excited inorganic phosphors will bring a revolutionarychange to the lighting and display industries and enable light sources and systems topossess unmatched performance and/or functionality.6,7 There are commonly threekinds of processes to fabricate wLEDs: (1) single-phased yellow or mixed greenand red phosphors are excited by a blue LED chip to realizing white light; (2) thenear ultraviolet (n-UV) LED chips are used to excite the red, green, and blue phosphorsto producing white light; and (3) combination of red, green, and blue three individualmonochromatic LED chips forms white light.8,9 Because the third approach encoun-ters lots of troubles, such as complicated electrics, high cost and mismatched agingproperties (different thermal and driving behaviors), etc., then the former two fabrica-tion schemes making use of phosphors have become the main trend in the academicresearches and practical applications. wLEDs (wLEDs) have made an amazing andexciting progress in the past 20 years, since the first commercially available wLEDscame into being in Nichia Corporation via combining the blue InGaN LED chipand the yellow garnet Y3Al5O12:Ce

3þ (YAG:Ce) phosphor.10 At present, the luminousefficiency of commercially available phosphor-converted wLEDs devices is raised to200 lm/W.11

Phosphors, namely luminescence materials, consisting of a matrix (crystalline host)and an activator (luminescent center), are essential components of LEDs devices andplay a crucial role in determining the quality of wLEDs. Phosphors applied in LEDsare excited by the emission of LED chips and such a kind of luminescence from phos-phors is known as the photoluminescence (PL).12 As an indispensable component ofwLEDs devices, phosphors with different emission colors are being explored anddeveloped for use in lighting and backlit display sources. Also, this is considered tobe one of the most critical and urgent challenges in the lighting field. There are appar-ently many methods to design and discover the new LEDs phosphors, including (1) theselection of suitable activators (such as broadband emitting Eu2þ, Ce3þ, and Mn2þ

ions; line-emitting rare earth ions Ln3þ and Mn4þ; and so on), and (2) the exploration


https://doi.org/10.1016/B978-0-08-101942-9.00005-8

and evaluation of different host compounds (such as garnets, sulfides, (oxo)nitrides,silicates, aluminates, borate, phosphates, and so on).2,3

Herein, this chapter is structured as follows: first, we briefly present the require-ments for phosphors applied in wLEDs, including principle on the fabrication ofwLEDs and key parameters for phosphors. Second, we describe four currently com-mercial phosphors for wLEDs, which are YAG:Ce3þ phosphor and its modifications,(oxy)nitride phosphors, Silicates phosphors and Mn4þ-activated fluoride phosphors,respectively. Third, we propose the design strategy to discover new phosphors fromtwo different perspectives, i.e., structural design in different host systems and codopedactivators via energy transfer (ET). Finally, we discuss the topics of structureeproperty relationships which are important for future development.

5.2 Requirements for phosphors used in wLEDs

5.2.1 Principle on the fabrication of wLEDs

As mentioned earlier, there are commonly three kinds of processes to fabricatewLEDs. The single-phased yellow or mixed green and red phosphors excited by ablue LED chip, or the red, green, and blue phosphors pumped by the n-UV LED chips,have been extensively studied.2e5 The first wLEDs lamp was fabricated by the com-bination of a blue LED chip and YAG:Ce yellow-emitting phosphor, which is the mostcommon method in the application of lighting and display areas. However, YAG:Ceyellow phosphor suffers some weaknesses, such as a poor color rendering index(CRI) and low stability of color temperature. In addition, the lighting color of this de-vice changes with the drive voltage and the phosphor coating thickness, and therefore,it is difficult to fabricate stable wLEDs in industrial production. To optimize the colorrendering properties of white and multicolor emitting phosphors, the current focus oftheir fabrication shifts gradually from the YAG:Ce-based to the redegreeneblue(RGB) emitting color phosphors excited by n-UV LEDs. As human eyes are not sen-sitive to n-UV, the color obtained by n-UV LEDs only depends on the phosphors. Sophosphors play a decisive role in wLEDs.

Despite the differences in the two fabrication schemes, we also call both of themphosphor converted wLEDs (pc-wLEDs). The pc-wLEDs involve two luminescenceprocesses. One is the electroluminescence process of LED chip, which is anonthermal generation of light resulting from the application of a voltage to a sub-strate. In the electroluminescence process, excitation is accomplished by recombina-tion of charge carriers of contrary sign (electron and hole) injected into an inorganicor organic semiconductor in the presence of an external circuit.12 The other is the PLprocess of phosphors, which absorb all or part of the light emitted by the chip andemit light with longer wavelength. Ultimately, the white light is realized bycombining the emission from LED chips and phosphors or by the emission of justphosphors.

There are several basic concepts referred to characterize fundamental aspects of pc-wLEDs, including correlated color temperature (CCT), the Commission Internationale


de I’Eclairage (CIE) and CRI or Ra. Therefore, it is necessary to list these basic con-cepts as follows.2,3,13,14

1. CCT is the temperature of a black body whose chromaticity most nearly resembles that of alight. Low CCT implies warmer (more yellow-red) light, whereas high color temperature ap-pears to be a colder (more blue) light. It is important to install an electrical lighting systemthat emits warm or cold light as needed.

2. CIE is the most widely used method to describe the composition of any color in terms of threeprimaries (RGB). Artificial “colors”, denoted by X, Y, Z, also called tristimulus values, canbe added to produce real spectral colors. By a piece of mathematic legerdemain, it is neces-sary only to quote the quantity of two of the reference stimuli to define a color because thethree quantities (x, y, z) are made always to sum to 1. The x, y, z, i.e., the ratios of X, Y, Z ofthe light to the sum of the three tristimulus values, are the so-called chromaticity coordinates.(x, y) is usually used to represent the color.

3. The color rendering index (CRI or Ra) definition is based on comparing the color of testobjects when illuminated by the light source under test, to the colors of the objects illumi-nated by a reference source. The introduction of this parameter is based on the fact thatobjects may look quite different in color under lamps that look quite alike in successionbut are different in spectral distribution. When CRI is calculated, it can be rated on a scalefrom 0 to 100. A CRI of 100 would represent that all color samples illuminated by a lightsource in question would appear to have the same color as those same samples illuminatedby a reference source.

4. Luminous efficacy is a figure of merit for light source. The luminous efficacy of a light sourceis defined as the ratio of the total luminous flux (lumens) to the power (watts or equivalent).The luminous efficacy is always in contradiction with CRI, because a high CRI value requiresproper spectral dispersion over all the visible range, which would make the luminous efficacyfar below 683 lm per W (the theoretically attainable maximum value).

5. Quantum yield involved in most references and reports is referred to the absolute quantumyield, i.e., the ratio between the number of emitted photons and the number of absorbed pho-tons, which is an intrinsic property of the luminescence conversion process.

5.2.2 Key parameters for LEDs phosphors

Phosphors used in wLEDs play an important role in determining the quality of whitelight. In general, phosphors consist of a matrix (crystalline host) and an activator (lumi-nescent center). In order to obtain a high-quality solid-state light source, phosphorsused in LEDs should contain the following basic characteristics: (1) Excitation spec-trum, matching well with the pumping LED chips, shows a large absorption intensityof n-UV (360e420 nm) or blue light (420e480 nm). (2) Emission spectrum, combi-nation with the emission of LED chips and phosphors, produces a pure white emissionwith a specific CRI or Ra and corresponding color temperature (CCT). (3) Efficientluminescence with high quantum efficiency (QE). (4) Good physical, chemical andthermal stability under application conditions, e.g., high stability against moisture, ox-ygen, and heating. (5) Mild synthesis condition including nonhazardous cost, reason-able production and facile control over particle morphology.

Concerning these criterion, there are many methods to design and develop theLEDs phosphors, including (1) investigation and evaluation of different host

Phosphors for white LEDs 125

compounds (such as garnets, (oxo) nitride, aluminate, borate, silicate, sulfides, phos-phates, and so on), and (2) the selection of suitable activators (such as broadbandemitting Eu2þ, Ce3þ, and Mn2þ ions, line-type emitting activators, Ln3þ andMn4þ, and so on).

Furthermore, with respect to the PL control of new LED phosphors, the followingissues should be considered in order to discover and optimize suitable phosphors forLED applications, which include (1) the selection of a crystal structure, (2) the modi-fication of the chemical composition, (3) the coupling of activators to the host latticeand (4) the design of an ET process. First, a suitable crystal structure should beselected. A proper structural model can help to realize the expected excitation andemission spectra after doping of the activators. For example, Ce3þ shows long-wavelength emission in a few crystal structures, such as garnet-type compounds.15

Eu2þ shows narrow-band red emission in UCr4C4-type nitrides.16 Moreover, goodPL thermal stability can be realized in some ABPO4-type (A ¼ alkali metal; andB ¼ alkaline earth metal)17 or Ba9RE2Si6O24-type (RE ¼ Sc, Y and Lu) com-pounds.18 That is to say, obtaining an understanding of the crystal and local struc-tures, and the exploration of a suitable structural model will be the first step in thecontrol of PL; we can also refer to this as crystal-phase engineering, from the view-point of the investigation into solid-state materials.19 Second, modification of thechemical composition for a given phosphor system plays an important role in PLtuning and luminescence optimization.20,21 Especially, the fed transitions repre-sented by Ce3þ and Eu2þ can be modified by different crystal field strengths, sothat the emission can be tuned over a wide range in the visible light region. Herein,the modification of chemical compositions, for example, by cation/anion substitu-tion, can lead to red-shifting or blue-shifting of the emission peaks, which is ascribedto the crystal field effect mentioned earlier, as observed in the emission shifts in(Sr,Ba)2Si5N8:Eu

2þ or (Ca,Sr)AlSiN3:Eu2þ phosphors.22,23 Third, coupling of the

activators to the host lattice is another important factor in the modification or opti-mization of PL to match an LED application. In this regard, a strategy called as“crystal-site engineering” has been proposed and developed in several phosphor sys-tems recently.24e26 For example, Sato et al. discovered deep-red luminescence inCa2SiO4:Eu

2þ and also found that the red emission from Ca2�xEuxSiO4 was stronglyrelated to the peculiar coordination environments of Eu2þ in two types of Ca sites.Lin et al. have presented a facile method to break down the geometrical restrictionson the valence state of Eu in CaGdAlO4 through the replacement of Al3þeGd3þ bySi4þeCa2þ, which produced tunable emission colors over a wide range, includingwhite light, which indicated a potential application in wLEDs.27 Fourthly, the designof the ET process is also an effective way to tune the PL and aid the discovery of newLED phosphors. Based on previous reports on ion-ion interactions, different acti-vator couples, such as Ce3þeMn2þ, Ce3þeTb3þ, Eu2þeMn2þ, Ce3þeEu2þ,Ce2þeTb3þeMn2þ, and so on, have been used to tune the PL in many host systems,and an ET process with induced tunable emission can be realized.28e30 The tunableemission from such ioneion interactions also enriches the large family of LED phos-phors and fulfills different LED applications.31


5.3 The state-of-the-art phosphors for wLEDs

5.3.1 YAG:Ce phosphor and its modification

Today, most commercial pc-wLEDs devices are still based on YAG:Ce, as the yellow-emitting luminescence material, which match well with the high efficient blue LEDchips.32e35 Generally speaking, YAG:Ce and its composition modifications (Y,Gd)3(Al,Ga)5O12:Ce phosphors are still the optimal ones for the fabrication of thecommercial LEDs devices in order to realize the combination of high luminous effi-cacy and excellent comprehensive performance. Herein, the typical photolumines-cence excitation (PLE) and PL spectra of representative YAG:Ce phosphors areshown in Fig. 5.1. As shown in Fig. 5.1, the PLE spectrum is mainly composed oftwo broad bands that are centered at 346 and 440 nm, which are assigned to the4fe5d electronic transitions of Ce3þ and correspond to the transitions to the twolowest energy levels of the 5d orbital. The strong and broad absorption of blue lightby YAG:Ce suggests that it matches very well with the blue LED chip. The PL spec-trum shows a very broad yellow band with a peak at about 532 nm and a full width athalf maximum (FWHM) of w130 nm. In fact, the emission peak varies in a broadrange from 520 to 580 nm, depending on the variation of the minor compositions ofthe host lattice. Except for YAG:Ce phosphor, many other garnet-type LED phosphorshave been reported recently36e38 and their chemical compositions are derived fromcation/anion substitutions based on the aluminate/gallate/silicate garnet model. It isclear that Ce3þ-doped garnet-type LED phosphors (YAG:Ce and its modifications)will occupy the main stream at present and in the near future, due to their excellent

300 350 400 450 500 550 600 650 700Wavelength (nm)

PL

inte

nsity

(arb

. uni

t)

λem = 560 nmλex = 460 nm

2F5/2,7/2

2F5/2,7/2 2F5/2,7/2 2D5/2

2D3/2 2D3/2

Figure 5.1 Photoluminescence excitation and photoluminescence spectra of typical Ce3þ-doped Y3Al5O12 (YAG:Ce) phosphor.

8

Modified from Chen L, et al. Light converting inorganic phosphors for white light-emittingdiodes. Materials 2010;3(3):2172e95. https://doi.org/10.3390/ma3032172.


https://doi.org/10.3390/ma3032172

luminescence properties and tunable emission, enabling them to be adapted to differentapplications.39,40 However, the main drawbacks for the YAG based wLEDs areascribed to poor color rendering for the lack of red emission component and poor ther-mal quenching luminescence critical for high power application. So it is still a chal-lenge to develop new phosphor for wLEDs.

5.3.2 (Oxy)nitride phosphors

Presently, the most developed red-emitting phosphors suitable for commercial LEDsdevices are still Eu2þ-activated nitridosilicates, such as (Ba,Sr)2Si5N8:Eu

2þ and(Ca,Sr)AlSiN3:Eu

2þ.41e45 Both of the two types of commercially used red phosphorsare very important in the practical application and have many advantages, such asbroad excitation bands, tunable emission colors, good thermal behavior, and a high ef-ficacy. Herein, we consider (Ca,Sr)AlSiN3:Eu

2þ to demonstrate the state of the art ofthe nitride red phosphors. The typical PLE and PL spectra of CaAlSiN3:Eu

2þ phosphorcan be seen from Fig. 5.2(a). Due to the high covalency and strong crystal fields of thehost lattices, most Eu2þ-activated nitridosilicate phosphors can well match the blue-LED chips. Moreover, the thermal stability of the PL is generally good. However, itstill has two fundamental drawbacks for the red emitting nitride phosphors exceptfor the critical synthesis conditions (high synthesis temperature, high N2 pressureand air-sensitive starting chemical reagents).46 First one is the serious reabsorptionphenomenon owing to the spectral overlap of broadband absorption in the visible light

200 300 400 500 600 700 800

Wavelength (nm)

1.0

0.8

0.6

0.4

0.2

0.0

Inte

nsity

(arb

. uni

t)

Inte

nsity

(arb

. uni

t)

250 300 350 400 450 500 550 600 650 700 750 800

Wavelength (nm)

(a) (b)

Figure 5.2 Photoluminescence excitation and photoluminescence spectra typical nitridephosphors CaAlSiN3:Eu

2þ (a)41 and Sr[Mg3SiN4]:Eu2þ (b).42

Modified from (a) Xie RJ, Hirosaki N. Silicon-based oxynitride and nitride phosphors for whiteLEDs e a review. Sci Technol Adv Mater 2007;8(7e8):588e600. https://doi.org/10.1016/j.stam.2007.08.005. (b) Schmiechen S, et al. Toward new phosphors for application inillumination-grade white pc-LEDs: the nitridomagnesosilicates Ca[Mg3SiN4]:Ce

3þ, Sr[Mg3SiN4]:Eu

2þ, and Eu[Mg3SiN4]. Chem Mater 2014;26(8):2712e19. https://doi.org/10.1021/cm500610.


https://doi.org/10.1016/j.stam.2007.08.005

https://doi.org/10.1016/j.stam.2007.08.005

https://doi.org/10.1021/cm500610

https://doi.org/10.1021/cm500610

region of the nitride phosphor and the emission of the mixed green or yellow phos-phors. Thus it causes color offset and, also, decreases the luminous efficacy becausethe phosphor absorption partly covers the visible spectral range. Second, the enhance-ment of the CRI originating from such a kind of red phosphors also mean compro-mising on achievable luminous efficacy caused by the broad emission bands ofEu2þ at wavelengths longer than 700 nm, that is beyond the human eye.11

Just recently, Sr[LiAl3N4]:Eu2þ was invented as a kind of narrow-band red-

emitting phosphor suggesting an important progress compared to broadband red phos-phors represented by (Ba,Sr)2Si5N8:Eu

2þ or (Ca,Sr)AlSiN3:Eu2þ mentioned earlier

and this nitride phosphor showed a significant increase of luminous efficacy of wLEDsdevices.47,48 Accordingly, some similar phosphors with narrow-band emission havebeen reported, such as, Ca[LiAl3N4]:Eu

2þ,16 M[Mg3SiN4]:Eu2þ (M ¼ Ca and Sr)42

and M[Mg2Al2N4]:Eu2þ (M ¼ Ca, Sr, Ba).49 The PLE and PL spectra of representa-

tive Sr[Mg3SiN4]:Eu2þ are demonstrated in Fig. 5.2(b), and the character of the broad-

band absorption and narrow-band red emission can be clearly found, as compared tothe PL and PLE spectra of CaAlSiN3:Eu

2þ phosphor shown in Fig. 5.2(a). Thesenewly developed phosphors could not only act as the red component of the wLEDsfor illumination, but they also show great potential for display application owing tothe narrow emission character. They have become one of the hot issues in the recentdevelopment of the LED phosphors, especially the discovery of the new nitride phos-phors. However, there are also some challenges needed to be considered in the futureinvestigations. First, the harsh synthesis conditions for nitride phosphors, especiallyLi-containing compounds, still restrict their general develop and increase the cost priceof such systems. Second, the small band gap of narrow-band red phosphors oftenenables significant thermal quenching of the luminescence. Finally, the stabilitytoward humidity and ambient atmosphere should be evaluated and improved in thenear future.

5.3.3 Silicates phosphors

The family of silicate compounds plays an important role in the design of new LEDphosphors. Especially, silicates that crystallize in the orthorhombic structure, (A,B)2SiO4 (A, B ¼ Ca, Sr, and Ba) and their modifications, have received great interestwith respect to potential use.21,50,51 As basic structural units, silicates [SiO4]

4� canconstitute relatively complex crystal structures, which often contain a wide varietyof structurally complex anion groups, via different connection methods, formingislands, rings, chains, or layered structures. Among these, (A, B)2SiO4, A3SiO5, Li2A-SiO4, and Ca3Sc2Si3O12 phases are common and their luminescence properties and ap-plications as LED phosphors after doping have been extensively studied.2 Fig. 5.3gives the crystal structures of several typical silicate hosts and the coordination ofthe cations and [SiO4] tetrahedra. Among these, Sr2SiO4 has two polymorphs:b-Sr2SiO4 at low temperatures and a-Sr2SiO4 at high temperatures (>85�C); however,the addition of Ba or Eu in Sr2SiO4 stabilizes the a phase at room temperature.2 There-fore, Fig. 5.3(a) only shows the crystal structure of the commonly used a-Sr2SiO4.Becauseorthorhombic a-Sr2SiO4 has the same structure as Ba2SiO4, the


SrxBa2�xSiO4:Eu2þ solid solution phosphor maintains the same orthorhombic struc-

ture for all Sr concentrations. All samples can be efficiently excited with both blue(lmax ¼ 450 nm) and near-UV (lmax ¼ 405 nm) light and emit in the green to yellowregion, making these materials ideal for white lighting applications and also forcommercialization. A recent paper indicated that an intermediate composition with46% Sr, with a more rigid crystal structure compared to the end members, has the high-est resistance to thermal quenching of luminescence, remaining stable up to 413K andmaintaining an emission efficiency of 75%.51 Other important binary alkaline earth sil-icates are M3SiO5 (M ¼ Ca, Sr, and Ba), with different crystal structures, viz., mono-clinic (space group Cm) for Ca3SiO5, tetragonal (space group P4/ncc) for Sr3SiO5, andtetragonal (space group I4/mcm) for Ba3SiO5.

52,53 As an example, Fig. 5.3(b) showsthe crystal structure and coordination diagram for Sr3SiO5, and it has 9- and 10-coordinate Sr sites. (Ba, Sr)3SiO5:Eu

2þ phosphors have a broad excitation bandfrom 350 to 450 nm with an emission peak between 570 and 590 nm, depending onthe different Ba/Sr ratios. More recently, blue-light-excited yellow-emitting Sr3SiO5:Ce3þ, Liþ phosphors have also been reported and the emission from these phosphorsis broader than that of their Eu2þ-activated counterpart. As well as the two binaryalkaline earth silicate phosphors mentioned earlier, garnet-type green-emittingCa3Sc2(SiO4)3:Ce

3þ phosphors excited by blue LEDs were also developed for

(SiO4)

Sr2Sr1

b

c

a

Li

Sr

(SiO4)b

c

a

LiSrSiO

(SiO4)

ScCa

Sr1 Sr2SrSiO

SrSiO

(SiO4) a

bc

CaScSiO

a

cb

(a) (b)

(c)(d)

Figure 5.3 Crystal structures of several typical silicate hosts and the coordination of the cationsand [SiO4] tetrahedra: (a) a-Sr2SiO4, (b) Sr3SiO5, (c) Ca3Sc2(SiO4)3, and (d) Li2SrSiO4.

2

From Xia ZG, Xu ZH, Chen MY, Liu QL. Recent developments in the new inorganic solid-stateLED phosphors. Dalton Trans 2016;45:11214e32. https://doi.org/10.1039/C6DT01230B.


https://doi.org/10.1039/C6DT01230B

application in wLEDs.54 Fig. 5.3(c) shows the typical crystal structure of theCa3Sc2(SiO4)3 phase, and Ca, Sc, and Si are coordinated by 8, 6, and 4 oxygen atoms,respectively. Orange-yellow emitting Eu2þ-activated Li2SrSiO4 phosphors (with anemission peak at 562 nm) have been reported and can be effectively excited by theInGaN blue LED chip.55,56 Fig. 5.3(d) demonstrates the crystal structure of Li2SrSiO4,in which there is only one Sr site coordinated to eight oxygen atoms, Liþ is coordinatedto four oxygen atoms, and Si4þ is coordinated to four oxygen atoms.

As demonstrated earlier, alkaline earth silicates are determined as promisingmatrices in the use of phosphor-converted wLEDs, due to their excellent chemicalstability and abundant emission colors. Recently, there were several other impor-tant discoveries in the new silicate LED phosphors. For example, Ce3þ/Eu2þ-acti-vated Ba9RE2Si6O24 (RE ¼ Sc, Y, and Lu) exhibits tunable emissions rangingfrom n-UV to blueegreen. Ba9RE2Si6O24 crystallizes in a rhombohedral structurethat consists of corner-shared SiO4eREO6eSiO4 layers.18,56 The RE atoms pro-vide only one crystallographic site coordinated by six oxygen atoms, formingREO6 octahedra, which are arranged in a nearly hexagonal array, sharing cornerswith the SiO4 tetrahedra, and the REO6 octahedra are distorted. There are threeindependent barium sites in this structure, which are indicated by Ba(1), Ba(2),and Ba(3). The Ba(1) and Ba(2) ions are alternately arrayed in chains in the inter-layer gap and link the LueSieO layers that contain chains of Ba(3) ions. The three

250 300 350 400 450 500 550 600 650λ (nm)

77Kλex = 394 nm

298Kλex = 394 nmλem = 480 nm

λem = 480 nm

100 150 200 250 300 350 400 450 500T(K)

Intensity

0

5 x 104

1.00

0.75

0.50

0.25

0.00

650

575

500

425

Ba9Y2Si6O24:Ce3+

YAG:Ce3+

λ (n

m)

Rel

ativ

e P

LQY

(a) (b)

(c)

Figure 5.4 (a) Room temperature photoluminescence excitation and photoluminescent spectraof the Ba9Y1.94Ce0.06Si6O24 phosphor, (b) temperature dependence of the relative integratedphotoluminescent intensity for Ba9Y1.94Ce0.06Si6O24 (squares) and the commercial YAG:Ce(circles) as a comparison, (c) minimal red shift and a decrease in emission intensity is observedat temperatures below 295K, whereas a blue shift of the emission intensity is observed withincreasing temperature.56

Modified from Brgoch J, Borg CK, Denault KA, Mikhailovsky A, DenBaars SP, Seshadri R. Anefficient, thermally stable cerium-based silicate phosphor for solid state white lighting. InorgChem 2013;52(14):8010e16. https://doi.org/10.1021/ic400614r.


https://doi.org/10.1021/ic400614r

barium sites, Ba(1), Ba(2), and Ba(3), are coordinated with 12, 9, and 10 oxygenatoms, respectively, forming three different distorted polyhedra. Fig. 5.4(a) givesthe room temperature (298K) PLE and PL spectra of Ba9Y1.94Ce0.06Si6O24. Theexcitation spectrum, collected using lem ¼ 480 nm, extends from 250 to 420 nmand contains three peaks. The emission spectrum, collected using lex ¼ 394 nm,is broad, extending from 400 to 675 nm. Fig. 5.4(b) gives the comparative temper-ature dependent PL quantum yields (PLQY) of Ba9Y1.94Ce0.06Si6O24 and industrystandard YAG:Ce phosphors. It is found that the PLQY values of Ba9Y1.94Ce0.06-Si6O24 decrease by only 25% of the room temperature efficiency at 500K. Thissmall drop indicates that this phosphor host is more thermally robust than manyother silicates, such as M2SiO4:Eu

2þ (M ¼ Sr, Ba) (50% of room temperature ef-ficiency around 400K). It is believed that the connectivity in Ba9Y1.94Ce0.06Si6O24

is closer to that of YAG:Ce3þ, containing corner-shared tetrahedra-octahedra,which limits the vibrational degrees of freedom, improving the quenching charac-teristics. In addition to the decrease in the PLQY, there is a blue shift of approx-imately 15 nm with increasing temperature, illustrated in Fig. 5.4(c). This isascribed to the fact that higher temperatures cause an increase in the unit cell di-mensions and the corresponding CeeO bond distances, which decreases crystalfield splitting (CFS).

As a summary, silicate phosphors, represented by SrxBa2�xSiO4:Eu2þ, have always

shown great potential and have been used in practical LED devices. However, the poorthermal stability of the PL restricts its further application, especially in high-power orlong-lived devices. Some new silicate LED phosphors with improved luminescenceproperties are still needed, such as the emerging Ba9Y1.94Ce0.06Si6O24 discussedearlier.

5.3.4 Mn4þ-activated fluoride phosphors

Recently, many researches focused on a kind of important line-type-emitting red phos-phors, represented by Mn4þ-activated fluoride phosphor, such as K2SiF6:Mn4þ andBaSiF6:Mn4þ,58e64 which also act as the most promising candidates for improvingthe color rendering for wLED. Herein, the PLE and PL spectra of typical K2SiF6:Mn4þ phosphor are shown in Fig. 5.5 and this kind of phosphor can absorb theblue light and show line-type red emission originated from the ded transition ofMn4þ.58 Since 2008, Adachi’s group has reported a series of A2MF6:Mn4þ (A ¼ K,Na, Cs; M ¼ Si, Ge, Ti, Zr) phosphors, which were prepared by wet chemical etchingvia mixing the precursors and silicon in the hydrogen fluoride (HF) solution.61e68

However, it did not arouse enough attention for the application in LEDs at thattime. The main reason is that the previous method by Adachi contains the complexexperimental operation and expensive starting materials. In 2014, Zhu et al. success-fully synthesized K2TiF6:Mn4þ by coprecipitation method. The high luminous effi-cacy and excellent optical performance were found for practical wLEDs devicefabricated by commercial YAG:Ce and as-prepared line-emitting red phosphor


K2TiF6:Mn4þ.69 Owing to its high emission intensity and the characteristic line-typered emission of Mn4þ in these fluorides, it will not only enhance the CRI of the devicebut also demonstrate application in the back-lighting device. Presently, the Mn4þ

doped hexaflorometallates phosphors have been extensively studied by many groups,focusing on the new fluoride structural types, different synthesis methods, intrinsicluminescence mechanism and wLEDs device evaluation and application.70e79 Webelieve that this is a promising phosphor type for the wLEDs to be able to fulfill therequirement of the practical device package. However, the main problem is that thiskind of hexaflorometallates systems were generally prepared by the solution-basedmethod, which will consume plenty of water, HF, and oxidizer, that provides possiblepollution in the mass production. Moreover, Mn ion is highly sensitive to different re-action conditions and it can exist in multiple valance states including Mn2þ, Mn3þ,Mn4þ, Mn6þ, and Mn7þ, so that proper synthesis conditions should be controlled,otherwise it will affect the quality of the final phosphor products.

5.4 New advances of future phosphors for wLEDs

As summarized earlier, many promising phosphors for commercial use have beendiscovered. However, there are still a lot of drawbacks including low CRI, bad thermalstability, poor resistance to humidity and harsh preparation conditions and so on.Therefore, the methodology used in the discovery of new phosphors is still relevantand useful for the accelerated development of wLED techniques.2e5 From a strategic

300 400 500 600 700Wavelength (nm)

PLE

(arb

. uni

ts)

PLE PL

4.0 3.5 3.0 2.5 2.0

PL

(arb

.uni

ts)

4A2→4T1

4A2→4T2

Photon energy (eV)

Figure 5.5 Photoluminescent (lex ¼ 450 nm) and photoluminescence excitation (PLE)(lem ¼ 630 nm) spectra for K2SiF6:Mn4þ phosphor at 300K.58

Modified from Takahashi T, Adachi S. Mn4þ-activated red photoluminescence in K2SiF6phosphor. J Electrochem Soc 2008;155(12):E183e8. https://doi.org/10.1149/1.2993159.


https://doi.org/10.1149/1.2993159

perspective, the development of new fluorescent powder method can be divided intotwo categories. One is discovery of new phosphor via structural design in differenthost systems. The other is design of new phosphors via the codoped activators viaET. In this section, we focus on these two methods.

5.4.1 New LEDs phosphors with different host systems

The 5d levels of the activators (such as Eu2þ and Ce3þ) in the doped phosphors aregreatly affected by the surrounding environment and, therefore, they can be tunedthrough host structure interactions. The host structure may consist of several elementsfrom the alkali and alkaline earth metals, rare earth elements, and actinides composingoxides, aluminates, silicates, borates, phosphates, nitrides, fluorides, chlorides, andcombinations thereof. The varying degrees of covalency, connectivity, and local coor-dination environment of activator ion sites all affect the energetics of the incorporatedactivator ion and determine phosphor properties such as excitation and emission wave-lengths, luminescence efficiency and resistance to thermal quenching effects. As aresult, the host structure and coordination environment around the activator ion playa key role in determining the optical characteristics of phosphor materials. Here, weclassify the new phosphors by hosts, such as aluminate, borate, silicate, nitride andso on, and discuss the discovery of new phosphor via the modification of the chemicalcompositions’ and structural design.

5.4.1.1 Aluminates

The aluminate phosphors have drew much attention due to high quantum conversionefficiency, good color rendering and wide excitation range. The most famous phos-phate phosphors are Y3Al5O12:Ce

3þ and its ramification. And the first pc-wLEDwas fabricated by combining the yellow-emitting Y3Al5O12:Ce

3þ (YAG:Ce) phosphorwith a blue-emitting InGaN chip. These have been already simply introduced in thefront section. Therefore, we don’t repeat them again.

However, we will focus on the ways to modify YAG via structural design. Asshown in Fig. 5.6(a), YAG consists of three different cation-oxygen polyhedra:[YO8] dodecahedron, [AlO6] octahedron, and [AlO4] tetrahedron. Therefore the PLproperties of YAG:Ce can be adjusted by cation substitution. For example, (Y, Gd)AG:Ce phosphors with different Gd3þ content were prepared by vacuum solid-statereaction to modify the PL properties of YAG:Ce.80 Gd3þ ion has a stable electronicstructure, and when larger Gd3þ substitutes Y3þ, the host environment has little effecton the energy levels of Gd3þ ions and crystal field around Ce3þ ions becomes strong.As shown in Fig. 5.6(b), with the increase of Gd3þ content, the peak wavelength forYAG:Ce phosphors shifts from 543 nm (x ¼ 0.00) to 561 nm (x ¼ 0.20) and the colorchanges from yellow green to orange. However, the emission intensity greatly de-creases due to the change of structure and ET between luminescence ions. Moreover,the high lowest excitation energy of Gd3þ causes the direct ET from Ce3þ to Gd3þ tobe difficult. Therefore, the PL intensity also decreased by the lattice distortion. In


addition, YAG:Ce, Gd phosphors with high concentration of Gd3þ ions have a goodthermal stability and potential advantage for commercial application on wLED.

As a comparison, the emission band of Ce3þ shows blue shift when Y3þ wassubstituted by smaller Lu3þ. As an example, the PL emission spectra of[(Gd1�xLux)0.99]3Al5O12:0.03Ce phosphor are shown in Fig. 5.6(c).

81 Lu3þ doping af-fects the centroid position and CFS of Ce3þ 5d energy levels. More Lu3þ incorporation

x = 0.80x = 0.60x = 0.40x = 0.20x = 0.00

500 550 600 650 700Wavelength (nm)

Inte

nsity

(arb

. uni

t)

x = 0

x = 0.10

x = 0.20

500 550 600 650 700 750Wavelength (nm)

λex = 460 nm

PL

emis

sion

inte

nsity

( ar

b. u

nit)

O2–

Y3+

AI3+a

bc

700600500400300200

1000

PL

inte

nsity

(arb

. uni

t)

(Y0.99Ce0.01)AG

(Lu0.99Ce0.01)AG

500 550 600 650 700 750 800Wavelength (nm)

x = 0.1x = 0.2x = 0.3x = 0.4x = 0.5

(a)

(c)

(b)

(d)

Figure 5.6 (a) Crystal structure of Y3Al5O12 and the Y3þ and Al3þ coordination surroundingsin the lattice.34 (b) Photoluminescent (PL) emission spectra of [(Gd1�xLux)0.99]3Al5O12:0.03Ce phosphor (lex ¼ 457 nm).81 (c) PL spectra of (Y0.98�xGdx)3Al5O12:Ce0.06 phosphorsupon excitation at 460 nm.80 (d) Room temperature emission spectra (lex ¼ 470 nm) of(Y0.97Ce0.03)3Al5�xSixO12�xNx.

36

Modified from (a) Shang MM, Fan J, Lian HZ, Zhang Y, Geng DL, Lin J. A double substitutionof Mg2þeSi4þ/Ge4þ for Al(1)3þeAl(2)3þ in Ce3þ-doped garnet phosphor for white LEDs.Inorg Chem 2014;53:7748e55. https://doi.org/10.1021/ic501063j. (b) Li JK, Li JG, Liu S, Li X,Sun X, Sakka Y. The development of Ce3þ-activated (Gd, Lu)3Al5O12 garnet solid solutions asefficient yellow-emitting phosphors. Sci Technol Adv Mater 2013;14(5):054201. https://doi.org/10.1088/1468-6996/14/5/054201. (c) Shi HL, Zhu C, Huang JQ, Chen J, Chen DC, Wang WC,Wang FY, Cao YG, Yuan XY. Luminescence properties of YAG: Ce, Gd phosphors synthesizedunder vacuum condition and their white LED performances. Opt Mater Express 2014;4(4):649e55. https://doi.org/10.1364/OME.4.000649. (d) Setlur AA, Heward WJ, Hannah ME,Happek U. Incorporation of Si4þ-N3-into Ce3þ-doped garnets for warm white LED phosphors.Chem Mater 2008;20(19):6277e83. https://doi.org/10.1021/cm801732d.


https://doi.org/10.1021/ic501063j

https://doi.org/10.1088/1468-6996/14/5/054201

https://doi.org/10.1088/1468-6996/14/5/054201

https://doi.org/10.1364/OME.4.000649

https://doi.org/10.1021/cm801732d

produces a rigid host lattice, from which increased phonon energy and the lumines-cence could be improved.82

Anion substitution and the recent reported chemical unit cosubstitution strategy byXia et al.83 can also tune the PL of YAG:Ce phosphor. Setlur et al. studied the effect ofSi4þeN3� incorporation on RE3Al5O12:Ce

3þ (RE ¼ Lu3þ, Y3þ, or Tb3þ) garnetphosphors.36 Fig. 5.6(d) shows that the emission band of (Y0.97Ce0.03)3Al5�xSix-O12�xNx shifts toward red because the Ce3þeN3� bonds have higher covalency andpolarizability, compared to Ce3þeO2� bonds. And Si4þ replaces Al3þ on the tetrahe-dral sites in order to compensate for the charge. By N/O replacement, the lowest 5dexcited state of Ce3þ was at lower energy compared to typical garnets, results in astrong red component appears in the emission spectra under blue excitation, whichmakes the as-obtained phosphors possible for creating warm wLEDs. However,when O2� replaced by N3�, the low energy Ce3þ ions have stronger luminescentquenching compared with typical Ce3þ ions in garnets and this could affect both thelamp efficacy and color in practical applications.

In YAG, besides the cosubstitution of [Si4þeN3�] for [Al3þeO2�] couple, chem-ical unit cosubstitution also can take effect on the [Y3þeAl3þ] couple, [Al3þeAl3þ]couple and [Y3þeAl3þeAl3þ] group. For instance, CaY2Al4SiO12 via [Ca2þeSi4þ]cosubstitution, Y3Mg2AlSi2O12 via [Mg2þeSi4þ] cosubstitution and Ca3Sc2Si3O12

via [Ca2þeSc3þeSi4þ] cosubstitution were synthesized and studied. CaY2Al4SiO12:Ce3þ exhibits a broad band emission in the range 460e750 nm.84 The emissionband maximum is blue-shifted in comparison to Y3Al5O12:Ce

3þ and shifts from542 to 560 nm by increasing the Ce3þ concentration (from 0.1% to 3%). Due to theincorporation of Ca2þeSi4þ pairs into the host it might be that different Ce3þ centersare appeared in the CaY2Al4SiO12 by the inherent disorder in the host. Therefore, thereis ET process from the high-energy centers to low-energy centers. Such energy migra-tion becomes much more likely at higher Ce3þ concentrations, thus causing the red-shift of emission. For Y3Mg2AlSi2O12, the incorporation of Mg2þ and Si4þ ions inthe octahedral and tetrahedral sites, respectively, could lead to red-shifted emissionin comparison with conventional YAG:Ce garnet phosphors.85 However, strong con-centration and thermal quenching have been observed. It is also noticed that Ca3Sc2-Si3O12:Ce

3þ phosphor also has a garnet-type structure as Y3Al5O12, and it could beapplicable as a green-emitting phosphor for wLEDs.86 This phosphor absorbs bluelight around 450 nm and emits green luminescence, with a peak wavelength around505 nm. It is a promising candidate for application in wLEDs as quenching of thephosphor at 150�C was smaller than that of Y3Al5O12:Ce yellow phosphor.

Sr2.975�xBaxCe0.025AlO4F (SBAF:Ce3þ) were another new phosphors in the alumi-nate family, which were first reported by Seshadri’s group.87 The phosphor displayshighly efficient green emission under 400 nm excitation. Fig. 5.7(a) show the crystalstructure of SBAF: Ce, in which Sr (8-coordination) and Ba (10-coordination) sites areboth substituted by Ce ions. The room temperature PL spectra of SBAF:Ce3þ(x ¼ 0)with various Ce3þ concentration (y) at 400 nm excitation are shown in Fig. 5.7(b). Theemission spectrum is broad, and displays two emission bands, at 460 nm and at502 nm due to Ce3þ ions occupy both two sites (of Sr1 and Sr2). With increasingCe3þ concentration, the spectral shape changes, a decrease of the blue component


around 460 nm and an increase of green around 502 nm. And because Ce3þ ions weresmaller than Sr2þ ions, the crystal field around Ce3þ changed, that results in the emis-sion band shifts to longer wavelength. The near 100% QE and efficient broad greenemission of the new SBAF:Ce3þ phosphor suggests that it has great potential on fabri-cating high luminous efficacy wLEDs.

Mn4þ-doping aluminates were another promising systems in order to discoveryline-type red emitting phosphors, such as CaAl12O19:Mn4þ,88 Sr2MgAl22O36:Mn4þ,89 SrMgAl10O17:Mn4þ 90, and so on. BaMgAl10O17 is well-known aluminatephosphor host and also can be modified through the substitution of Al3þeAl3þ byMn4þeMg2þ to find new narrow-band phosphors. Wang et al. reported that BaM-gAl10�2xO17:xMn4þ, xMg2þ (x ¼ 0.005e0.050) phosphors are good narrow-band(FWHM w30 nm) red-emitting (peaking at 660 nm) phosphors and have high colorpurity and an excellent color stability against heat.91 In this host, the incorporationof Mg2þ reduces number of Mn4þeMn4þeO2� pairs which replace a couple of neigh-boring Al3þ sites. And the retarding of concentration quenching and the luminescence

400 500 600 700λ (nm)

1.0

0.5

0.0

Nor

. int

ensi

ty y

(a)

(b)

Figure 5.7 (a) Unit cell representation of the fully ordered crystal structure of Sr2BaAlO4F(SBAF). Light gray, dark gray, blue, orange, and green spheres represent Sr, Ba, Al, O, and Fatoms, respectively. (b) Emission spectra of Sr3�yCeyAlO4F under 400 nm excitation sourcewith varying Ce3þ concentration y.87

Modified from Im WB, Brinkley S, Hu J. Sr2.975�xBaxCe0.025AlO4F: a highly efficient green-emitting oxyfluoride phosphor for solid state white lighting. Chem Mater 2010;22:2842e9.https://doi.org/10.1021/cm100010z.


https://doi.org/10.1021/cm100010z

enhancement phenomenon can be attributed to the reduction in energy loss channelsbetween Mn4þ ions.

As mentioned earlier, the cation/anion substitution or chemical unit cosubstitutioncan effectively modify the PL properties of aluminate phosphors. Except for that, thecrystal-site engineering approaches can lead to the contraction or expansion of the ac-tivator’s sites, and the change of the occupied sites by increasing doping level of theactivators.27,92,93 Accordingly, the customization of the luminescence in the specifichosts could be achieved and then it will result in the discovery of the new phosphors.For example, based on the coordination environment modification, we can control thecontraction or expansion of the activator polyhedrons. Thus the Eu2þ and Eu3þ cansimultaneously exist in one host, as a result, the luminescence changes from blue-green emission (Eu2þ) to the red emission (Eu3þ) can be observed. Liu’s group re-ported that by appropriate dopant incorporation, the valence state of Eu can be tunedfrom Eu3þ to Eu2þ in phosphors due to the replacement enlarge the activator site.92 InCa12Al14O32F2:Eu

3þ, Si4þeO2� are incorporated to substituted Al3þeF� in order torelease the geometry restriction of the Eu site and then enables Eu3þ to be reduced intoEu2þ due to the contraction of the (Al, Si)O4 tetrahedra and relaxation of the Ca

2þ site.Fig. 5.8(a) shows the PL spectra of Ca12Al14�zSizO32þzF2�z:Eu (z ¼ 0e0.5) at roomtemperature under 254 nm excitation. With increasing z, besides the emission of Eu3þ,a surprising observation which contains the appearance of a broad band peak at440 nm, it can be attributed to 5d/4f emission of Eu2þ. Interestingly, when z valueincreases, the emission of Eu3þ within 570e700 nm disappears and broadband emis-sion at 440 nm emerges simultaneously, suggesting that Eu3þ is reduced to Eu2þ in thelattice. With reference to the tendency for refined bond length of AleO, CaeO, andCaeF, the expanded of Ca2þ site could be proved by replacement of Al3þeF� bySi4þeO2�. Eu3þ can therefore be transformed to Eu2þ in the CASOF lattice. Duringvalence change of Eu, the CIE coordinates obtained upon 254 nm excitation are regu-larly shifted from red to blue region, as shown in Fig. 5.8(b), and in the inset, theproposed crystal structure variations and photographs of each compositions irradiatedunder a 254 nm UV lamp are depicted. This phenomenon clearly demonstrates thechange from Eu3þ-activated phosphor to Eu2þ- or Eu3þ/Eu2þ-activated phosphors.

Recently, Lin’s group applied crystal-site engineering approach to modify theCaYAlO4 host by the substitution of Al3þeY3þ by Si4þeCa2þ, that would lead tothe shrinkage of AlO6 octahedrons accompanied by the expansion of CaO9 polyhedronand, thus, enable the partial of Eu3þ reduced to Eu2þ.93 Fig. 5.8(c) shows the localstructural coordination of mixed-valence state Eu ions in the CaYAlO4:0.01Eu

3þ

and Ca0.99þxY1�xAl1�xSixO4:Eu0.01 (x ¼ 0e0.30) lattices. The CaYAlO4 frameworkconsists of AlO6 octahedrons and (Ca/Y)O9 polyhedrons connected by sharing theO2� vertices. Eu3þ would randomly occupy the Ca/Y sites in the host lattice. Becauseof the AlO6 octahedron sloping and highly compressed Ca/Y polyhedrons, Eu3þ isdifficult to reduce to Eu2þ and this results in the red emission of Eu3þ. When Al3þeY3þ is substituted by Si4þeCa2þ, the attractive force of the Ca2þeO2� becomesweaker; meanwhile, the attractive force of the Si4þeO2� becomes stronger, ascompared to the substitution of Y3þeO2� and Al3þeO2� in CaYAlO4. The bond


length of CaeO becomes longer (L3 > L1) and the bond length of SieO becomesshorter (L4 < L2). Thus the Ca2þ sites could be expanded through such a substitutionof Al3þeY3þ by Si4þeCa2þ. Accordingly, we can observe the tunable PL from red toblue emission with increasing x as illustrated in the CIE diagram in Fig. 5.8(d). Theemission of Eu2þ and Eu3þ covers the whole visible region with comparable intensity,leading to a light-yellow emission, which is located near the white light zone, indi-cating that this phosphor can be used for wLEDs. This crystal-site engineering

300 400 500 600 700

z = 00.1

0.20.3

0.40.5

ReplacementSi4+

Ca2+

AI3+

Y3+

(a) (b)

(d)(c) λ (nm)

Figure 5.8 (a) Emission spectra (lex ¼ 254 nm), (b) the Commission Internationale deI’Eclairage (CIE) chromaticity coordinates and photographs with varying z value inCa11.9Al14�zSizO32þzF2�z:Eu0.1 (z ¼ 0e0.5) under a 254 nm ultraviolet lamp.92 (c) Localstructural coordination of Eu ions in CaYAlO4:0.01Eu

3þ and Ca0.99þxY1�xAl1�xSixO4:Eu0.01(x ¼ 0e0.30) series.93 (d) Photoluminescent spectra and photographs ofCa0.99þxY1�xAl1�xSixO4:Eu0.01 (x ¼ 0e0.30) samples with lex ¼ 300 nm.93

Modified from (a and b) Huang KW, ChenWT, Chu CI, Hu SF, Sheu HS, Cheng BM, Chen JM,Liu RS. Controlling the activator site to tune europium valence in oxyfluoride phosphors. ChemMater 2012;24(11):2220e7. https://doi.org/10.1021/cm3011327. (c and d) Zhang Y, Li X, Li K,Lian H, Shang M, Lin J. Crystal-site engineering control for the reduction of Eu3þ to Eu2þ inCaYAlO4: structure refinement and tunable emission properties. ACS Appl Mater Interfaces2015;7(4)2715e25. https://doi.org/10.1021/am508859c.


https://doi.org/10.1021/cm3011327

https://doi.org/10.1021/am508859c

approach is promising for obtaining novel phosphor materials because only a singleactivator Eu can generate the multiband emission by optical combination of differentvalences of Eu.

Except for the crystal-site engineering controlling for Eu, the redistribution of thesites occupied by Ce has been also realized in Sr6(Y1�xCex)2Al4O15 phosphors.94

The Sr6Y2Al4O15 crystal structure is formed by corner connected YO6 octahedraand AlO4 tetrahedra, and Sr2þ ions coordinated by eight or nine oxide anions. Inthe Sr6Y2Al4O15 structure, Ce3þ can be doped into three different sites, includingtwo Sr2þ sites correspond to the SrO8 and SrO9 polyhedrons and one Y3þ site ofYO6 octahedra. The emission color of Sr6(Y1�xCex)2Al4O15 phosphors, dependingon concentration of Ce3þ, is tunable from blue to orange-red that is achieved by con-trolling the Ce3þ doping level in the Sr6Y2Al4O15 lattice and the phosphor exhibits theorange-red emission with maximum at 600 nm under blue light excitation, withincreasing Ce3þ concentration. This obvious difference can be explained by thechange of the Ce3þ sites in the Sr6Y2Al4O15 lattice that is indicated by the existenceof three kinds of the excitation and emission centers according to the three differenttypes of the crystallographic environment of Ce3þ. Because the blue emission bandhas two centers at 396 and 450 nm, the orange-red emission band has one center at600 nm, and the crystal field strength of O2� around Ce3þ in the Sr2þ site is weakerthan that of Y3þ site. Therefore, Ce3þ ions are preferably occupied the Sr2þ site inthe host Sr6Y2Al4O15 lattice and this center provides the blue emission. When Ce3þ

concentration increase above the limit for replacement of the Sr2þ site, Ce3þ ionswill occupy Y3þ sites in this host and the obvious red emission originating fromCe3þ at Y3þ site can be observed.

5.4.1.2 Silicates

Rare-earth-activated silicates are widely used as wLEDs phosphors because they haveversatile chemical compositions and crystal structures, tunable luminescence properties,high PL QE, low cost, and so on.2,3,95 But, unsatisfactory thermal stability and the ex-pected spectral positions of the excitation/emission bands limit their applications. In or-der to overcome these disadvantages, many silicate phosphors have been developed. Asthe examples, the structure design and modification originated from the (A,B)2SiO4,Ca3Sc2Si3O12, A3B2C3O12, M5(Si3O9)2, and Ca3Si2O7 phosphors have been studied,as also discussed previously. In this section, we will discuss their structures, lumines-cence properties and applications as LED phosphors after doping.

Herein, Eu2þ or Ce3þ doped orthosilicates A2SiO4 (A ¼ Ca, Sr, Ba) phosphors canbe regarded as a representative model because of their versatile polymorphs and chem-ical compositions. As we know, crystal site engineering can be used to tune the lumi-nescence properties by changing the coordination environment for phosphorsemploying Ce3þ or Eu2þ ions characterized by d-f transitions. Among them, the nitri-dation of the orthosilicate phosphors has demonstrated great potential.50,96e104 Forexample, Sohn reported the Sr2SiO4�xN2x/3:Eu

2þ phosphors.96 Gu reported the Ndoped Sr2SiO4:Eu

2þ phosphors.97 Black found the new LaSrSiO3N and LaBaSiO3Nphases, and the doped Eu2þ showed orange-red emission in this system.104 Herein,


we will give a typical example on the chemical unit cosubstitution of [Lu3þeN3�] for[Sr2þeO2�] in Sr2SiO4:Eu

2þ, and the yellow emitting LuxSr2�xSiNxO4�x:Eu2þ phos-

phors with tunable PL have been reported.50 Fig. 5.9(a) clearly demonstrate the pro-posed chemical unit cosubstitution mechanism between Sr1.97Eu0.3SiO4 andSr1.965Eu0.3Lu0.005SiO3.995N0.005. Two ions Sr2þ and O2�, which together have zerosum of charge, are substituted by two ions Lu3þ and N3�, which also have zerosum of charge. This scheme can be seen as: Sr2þ þ O2� / Lu3þ þ N3�. FromFig. 5.9(a), we can know that the component replacement impacts the lattice environ-ment of the luminescence center, and further affects the emission spectrum of phos-phors.50 The as-measured and normalized PL spectra of LuxSr1.97�xSiNxO4�x:0.03Eu2þ (x ¼ 0, 0.0025, 0.00375, 0.005) phosphors under 365 nm excitation areshown in Fig. 5.9(b) and (c), respectively. The emission spectra consist of an asym-metric broad band centered at around 570 nm, and the emission peaks give a red-shift from 563 nm to 583 nm with increasing [Lu3þeN3�] content. Fig. 5.9(d)comparatively gives the typical images of LuxSr1.97�xSiNxO4�x:0.03Eu

2þ (x ¼ 0,0.0025, 0.00375 and 0.005) phosphors under a 365 nm UV lamp and natural light,respectively. With increasing [Lu3þeN3�] content for the cosubstitution of [Sr2þe

450 500 550 600 650 700 750Wavelength (nm)

x = 0x =0.0025x =0.00375x =0.005

x = 0x =0.0025x =0.00375x =0.005

563–583 nm

Nor

mal

ized

inte

nsity

(arb

. uni

t)

LuxSr1.97–xSiNxO4–x:0.03Eu2+

b

ca b

ca

Lu3+Sr2+ O2– N3–

Eu2+

Sr1.97Eu0.03SiO4

Sr2+

(SiO4)4–

Sr1.965Eu0.03Lu0.005SiO3.995N0.005LuxSr1.97–xSiNxO4–x:0.03Eu2+

Inte

nsity

(arb

. uni

t)

λex = 365 nm

λex = 365 nm

400 450 500 550 600 650 700 750 800Wavelength(nm)

Natural light

UV light

x = 0 x = 0.0025 x = 0.00375 x = 0.005

(a) (b)

(c)

(d)

Figure 5.9 (a) The proposed chemical unit cosubstitution strategy of [Lu3þeN3�] for[Sr2þeO2�] to highlight the possible atoms’ transfer between Sr1.97Eu0.3SiO4 andSr1.965Eu0.3Lu0.005SiO3.995N0.005. The as-measured (b) and normalized (c) photoluminescentspectra of LuxSr1.97�xSiNxO4�x:0.03Eu

2þ (x ¼ 0, 0.0025, 0.00375, and 0.005) phosphorsunder 365 nm excitation. (d) Comparative images of LuxSr1.97�xSiNxO4�x:0.03Eu

2þ phos-phors under natural light and a 365 nm ultraviolet lamp.50

Modified from Xia Z, Miao S, Molokeev MS, Chen M, Liu Q. Structure and luminescenceproperties of Eu2þ doped LuxSr2�xSiNxO4�x phosphors evolved from chemical unitcosubstitution. J Mater Chem C 2016;4:1336e44. https://doi.org/10.1039/c5tc04222d.


https://doi.org/10.1039/c5tc04222d

O2�], we can clearly find the emission color change from yellow green to yellow. Aswe know, the N3� has less electronegative and more polarizable than O2� and its intro-duction in Sr1.97Eu0.03SiO4 increases the covalent character of the bonds with the Sr

2þ/Eu2þ, and we can infer that some of the Eu2þ ions in the LuxSr1.97�xSiNxO4�x:0.03Eu2þ phosphor are coordinated with nitrogen, so that we can find the obviousred-shift behavior. The PL emission reveals a red-shift and the emission color evolu-tion from yellow green to deep yellow with increasing [Lu3þeN3�] substitution con-tent. The tunable luminescence properties demonstrate that LuxSr2�x�ySiNxO4�x:yEu2þ phosphors can be applied as potential yellow phosphors for wLEDs. It isbelieved that such a strategy can be successfully used to discover the new phosphormaterials and tune the luminescence properties.

In garnets with the general formula A3B2C3O12, atoms A, B and C occupy the po-sitions 24c, 16a, and 24d, respectively.4 As an example for Lu3�xYxMgAl3SiO12:Ce

3þ

phosphor, the C site was occupied by Al3þ/Si4þ ions, the B site was occupied by Al3þ/Mg2þ ions, and the A site was occupied by Lu3þ/Y3þ/Ce3þ with fixed occupanciesaccording to their nominal chemical formulas. Ji et al.38 reported this new garnet phos-phors, Lu3�xYxMgAl3SiO12:Ce

3þ(x ¼ 0e3), which were developed using the struc-ture design strategy combining the chemical unit substitution and the cationsubstitution, starting from Lu3Al5O12:Ce

3þ. First, by cosubstitution of the Mg2þeSi4þ

pair for the Al3þeAl3þ pair in Lu3Al5O12:Ce3þ, they designed the Lu3(MgAl)(Al2Si)

O12:Ce3þ phosphor. This process can also be regarded as chemical units of MgO6/

SiO4 replacing AlO6/AlO4 polyhedra. Then, the Lu atoms are gradually substitutedby bigger Y atoms, forming (Lu3�xYx)(MgAl)(Al2Si)O12:Ce

3þ phosphors(x ¼ 0e3). The PL properties, on aspects of excitation/emission spectra, QE, and ther-mal emission stability were characterized regarding their potential application in blueLED chip-based wLED lighting. Furthermore, the red-shift tuning of the emission wasdiscussed and correlated with the local coordinating environment evolution aroundCe3þ in this series. The prepared samples show yellow color emission under daylight,as shown in Fig. 5.10(a). As expected for a host with a general garnet structure, thelowest Ce3þ 4f-5d absorption transition is in the blue spectral region, leading fromgreen to yellow color emission, which is readily useable in the applied blue chip basedpc-wLEDs fabrication. The PL spectra of the Lu3�xYxMgAl3SiO12:Ce

3þ phosphorsunder lex ¼ 450 nm are shown in Fig. 5.10(b). All phosphors present a broad asym-metric emission band attributed to the 5d-4f transitions of Ce3þ. The maxima of theemission peak are 575, 588, 594, and 597 nm. At the same doping concentration ofCe3þ, an increasing Y3þ/Lu3þ ratio is a reason to induce the red-shift of the emissionpeak from yellow to orange. At the same time, the peak intensity and the integratedintensity of the emission bands gradually decrease. The full-width at half-maximum(FWHM) values of the emission bands are 137, 140, 144, and 147 nm, respectively.All samples show the general cubic garnet structure. Lu/Y/Ce, Mg/Al, and Al/Sioccupy the positions 24c, 16a, and 24d, forming the (Lu/Y/Ce)O8 square antiprism,(Mg/Al)O6 octahedron, and (Al/Si)O4 tetrahedron, respectively. With the biggerY3þ ions substituting Lu3þ, the unit cell enlarges and the (Lu/YeO) bond lengths in-crease but the growth rates is different. Additionally, the CeO8 polyhedron has jointedges with (Lu/Y)O8 polyhedra, and the enlargement of (Lu/Y)O8 with x leads to


shrinkage of CeO8, consequently, d(CeeO) decreases. This should be the reason for thestronger CFS around Ce3þ and the red-shift of the peaking emission wavelength.Moreover, the distortion of the (Lu/Y/Ce)O8 square antiprism becomes larger andthis increases the diversity of local coordination environments of Ce3þ, which explainswhy the FWHM values of the emission bands become bigger as shown in Fig. 5.10(c)and (d). The new phosphors are excited efficiently at 450 nm and emit yellow color

Inte

nsity

(arb

. uni

t)

X = 1X = 0

X = 3X = 2

CeO8

CeO8

(AI/Si)O4 (AI/Mg)O6

CeO8

(Lu/Y)O8

450 500 550 600 650 700 750 800Wavelength (nm)

x = 0x = 1x = 2x = 3

CeO8

LuO8x = 0 x = 1

x = 3 x = 2

YO8

(a) (b)

(c) (d)

Figure 5.10 (a) Phosphor images and (b) emission spectra of Lu3�xYxMgAl3SiO12:Ce3þ

(x ¼ 0e3) phosphors under lex ¼ 450 nm, (c) local coordination environment around CeO8 inthe crystal structure of Lu3�xYxMgAl3SiO12:Ce

3þ compounds, and (d) shrinkage mechanismof the CeO8 polyhedron with Y3þ substituting Lu3þ in Lu3�xYxMgAl3SiO12:Ce

3þ:(x ¼ 0e3).38

Modified from Ji H, Wang L, Molokeev MS, Hirosaki N, Huang Z, Xia ZG. New garnetstructure phosphors, Lu3�xYxMgAl3SiO12: Ce

3þ(x¼ 0-3), developed by solid solution design. JMater Chem C, 2016;4:2359e66. https://doi.org/10.1039/c6tc00089d.



with relatively high luminescence intensity and room temperature QE, making themvery promising candidates for application in practical wLEDs. The successful prepa-ration of the (Lu3�xYx)MgAl3SiO12:Ce

3þ garnet phosphors suggests that more phos-phors can be developed within the garnet structure, by using this structure designstrategy.

As well as the two binary alkaline earth silicate phosphors mentioned earlier, a newgreen emitting phosphor Ca3Sc2Si3O12:Ce

3þ excited by blue LEDs were also devel-oped for application in wLEDs, which had high emission intensity and high thermalstability superior to YAG:Ce3þ.86 In 2012, the crystal structure and optical propertiesof the solid solution Ca3�(xþ0.06)LuxCe0.06Sc2�yMgySi3O12 have been reported, as wellas an original interpretation for the Ce3þ emission shift and thermal quenching mech-anisms.105 However, the relationship between local environments and the PL proper-ties was not studied in detail.106e109 Pan et al.110 reported the solid solutionCa2.97�xYxCe0.03Sc2�xMgxSi3O12 (x ¼ 0e1) and systematically investigated theirstructural and optical properties. The cosubstitution of Y3þeMg2þ pairs into theCa3Sc2Si3O12 lattice for Ca2þeSc3þ sites can enable the phase formation ofCa3�xYxSc2�xMgxSi3O12. The effect of Y3þeMg2þ pair incorporation into Ca2þeSc3þ sites can be analyzed by the interatomic distances of d(Ca/Y)eO in the first sphereand d(Sc/Mg)-O in the second sphere obtained from Rietveld refinement analysis. As x inCa2.97�xYxSc2�xMgxSi3O12:0.03Ce

3þ increases, the distance of d(Ca/Y)eO becomesshorter, which can be assigned to the substitution of smaller Y3þ for Ca2þ. However,the d(Sc/Mg)eO decreases due to the smaller Mg2þ substituted into the Sc3þ site simul-taneously. The shorter bond length, d(Ca/Y)eO and d(Sc/Mg)eO prove the coexistentshrinkage effects of the first sphere and the second sphere, and that there must exista competition between the two shrinkage effects. The contractive of the first sphere re-sults in shorter Ce3þeO2� bonds and a stronger CFS effect. However, the shrinkage ofthe second sphere will hinder the contraction of the [Ca/YO8] dodecahedra, thus weak-ening the CFS effect and causing the red shift of the excitation and emission wave-length only in a short range. This may be a reason why the emission wavelengths ofthe Ca2.97�xYxSc2�xMgxSi3O12:0.03Ce

3þ series can only be tuned from 505 nm forx ¼ 0e546 nm for x ¼ 1 as observed. Besides, the Stokes shift is another key factoraccounting for the red shift of the Ce3þ emission, which can be obtained from the en-ergy difference between the lowest 4fe5d excitation peak and the highest 5de4f emis-sion peak. The calculated Stokes shift of Ca2.97�xYxSc2�xMgxSi3O12:0.03Ce

3þ seriesshows a rapid increasing trend with the increase of x values. Thus for theCa2.97�xYxSc2�xMgxSi3O12:0.03Ce

3þ series, the red-shifted emission could beascribed to both the stronger CFS and the larger Stokes shift.

Li’s group111 present new insight into a changing Eu2þ crystallographic site pref-erence in Eu-doped M5(Si3O9)2 (M ¼ Sr, Ba, Y, Mn) phosphors. In their work,Sr2.97�xBaxEu0.03Y2(Si3O9)2 (Ba series) solid solutions (0 � x � 1.59) were prepared.Sr3Y2(Si3O9)2:Eu

2þ consists of Sr/Y/Eu atom layers and Si3O9 ring layers. As the Baconcentration increase should imply an increasing cell volume. This situation is un-usual in this case and it may lead to an unexpected change in the coordination environ-ment of Eu2þ. For x ¼ 0 in Sr2.97�xBaxEu0.03Y2(Si3O9)2, the PLE consists of a broadband from 250 to 430 nm with the maximum at 365 nm. The PL spectrum covers a


broad range from 425 to 575 nm, centered at 474 nm. By substitution of Ba2þ for Sr2þ,the lattice expansion is restricted to specific cation sites, resulting in the abrupt blueshifted emission of Eu2þ ions. With the x increasing from 0 to 1.59, the large blue shiftemission from 468 to 438 nm occurred. This shift is attributed to random occupation oflarger Ba2þ over Sr2þ sites, in view of the size-difference between Ba2þ ions and Sr2þ

ions, the coordination environment of the cations would change with the Ba2þ doping,the lattice environment around Eu2þ became looser, and thus the average EueO bondlength increased. Generally, the crystal field strength is proportional to 1/d5. Therefore,Eu2þ ions in the looser sites with the longer bond length will possess a higher energyemission and it will generate a blue shift of the emission band.

The newly developed solid solution phosphors Ca2.985�yEu0.015MgySi2O7 may findpotential application in wLEDs. This is a typical example to develop new phosphorsby cation substitution. Singh et al.112 tuned the Eu2þ emission by crystal chemical iso-valent substitution of Mg2þ in Ca2þ site. The PL emission spectrum ofCa2.985�yEu0.015MgySi2O7 (y ¼ 0, y ¼ 0.25, y ¼ 0.50, y ¼ 0.75 and y ¼ 1) excitedby 400 nm near UV photons is shown in Fig. 5.11(a). As discussed earlier, whenCa2þ is substituted by a smaller Mg2þ ion, the crystal site of Eu2þ is expanded, andthe magnitude of the crystal field decreases. Thus the 5d band of Eu2þ is increasedand there is a continuous blue shift in the emission along with the increasing Mg2þ

content. In addition, as mentioned earlier that the Mg2þ size is smaller than that ofCa2þ ion. Hence, Mg2þ has the highest attractive force towards O2�, when Ca2þ issubstituted by the smaller Mg2þ ion, the bond length between Eu2þ and O2� becomeslonger and the magnitude of the crystal field strength decreases, thus, resulting in the

450 500 550 600 650 700 750Wavelength (nm)

y = 0y = 0.25y = 0.5y = 0.75y = 1

1.0

0.8

0.6

0.4

0.2

0.0

Nor

mal

ized

inte

nsity

λexc = 400 nm0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8x

y

54

32

1

(a) (b)

Figure 5.11 (a) The photoluminescence excitation spectrum of Ca2.985�yEu0.015MgySi2O7

(y ¼ 0e1, insteps of 0.25) under lex ¼ 400 nm. (b) The Commission Internationale deI’Eclairage (CIE) color coordinate diagram for Ca2.985�yEu0.015MgySi2O7 (y ¼ 0e1, in stepsof 0.25) phosphor and inset shows the digital images under 365 nm.112

Modified from Singh K, Vaidyanathan S. Eu2þ luminescence in Ca3Si2O7 and spectral wideningand tuning of Eu2þ emission color (orangish-red to green) by crystal chemical substitution. RSCAdv 2016;6:98652. https://doi.org/10.1039/c6ra24258hK.


https://doi.org/10.1039/c6ra24258hK

blue shift in the emission. Then, the yellowish-orange-red emission (y ¼ 0) tuned togreen color (y ¼ 1). Accordingly, due to the Mg2þ in the Ca2þ site of the Ca3Si2O7

host lattice, the corresponding CIE coordinates of Ca2.985�yEu0.015MgySi2O7

(y ¼ 0e1) change from (0.5647, 0.4202) to (0.3540, 0.5932), due to the variation ofthe Mg2þ composition in the Ca2þ site of the Ca3Si2O7 host lattice, shown inFig. 5.11(b). These results indicate that the Ca2.985�yEu0.015MgySi2O7 (y ¼ 0e1, in-steps of 0.25) phosphor may have broad application prospects for NUV based wLEDs.

5.4.1.3 Sulfides

Sulfides have attracted significant concerns due to their capability for the lighting anddisplay applications although sulfide phosphors have met predicament in terms ofrapid degradation on its luminescent intensity due to a relatively unstable thermalchemical property.2 In order to overcome the potential drawbacks, such as relativelystrong concentration quenching, limited stability with moisture and serious thermalquenching even near room temperature, several kinds of new sulfide phosphors withimproved luminescence properties have been designed for the potential applicationin wLEDs.

Wu et al. reported a spectral blue shift from 553 to 590 nm (yellowish orange light)to 440e470 nm (blue light) in Y1.98Ce0.02(Ca1�ySry)F4S2 (0 � y � 1) when graduallyreplacing Ca2þ with Sr2þ (Fig. 5.12(a)), which is related to the change of the crystalfield strength.113 Fig. 5.12(c) shows the exact 1 � 1 � 1 unit cell crystal structure ofthe YCFS lattice viewed from the [010] and the Y atomic sites. The above replacementleads to the expansion of the lattice volume because of the larger Sr2þ ions, thus chang-ing four CeeS and four CeeF bonds to 286.94 and 262.41 pm within the internalYF4S5 polyhedra. The bond lengths of apical CeeS, 284.63 and 284.71 pm are almostthe same in Y1.98Ce0.02SrF4S2 and Y1.98Ce0.02CaF4S2. In such cases, the Ce3þ ion ex-periences a weaker CFS, which is ascribed to the expansion of YF4S5 polyhedra in theY1.98Ce0.02SrF4S2 system; therefore, it is reasonable to reveal the blue-shifted excita-tion and emission spectra. A brief scheme for the luminescent mechanism (seeFig. 5.12(b)), electrons are excited from VB via Ce (4f) to CB via Ce (5d) inYCFS:Ce3þ and YSFS:Ce3þ in Fig. 5.16(b), and then through the nonradiative Stokeshift relaxation to the lower stage is shown. Finally, the electron goes back to the VB;such a process may result in luminescence or it may lost thermally.

It is generally required that the red-emitting phosphor should have good color puritywith the CIE coordinates close to (0.67, 0.32) for obtaining full-color emission. Chen’sgroup reported a novel red phosphor a-(Y,Gd)FS:Ce3þ.114 Fig. 5.13 presents the PLspectrum of a-(Y0.99�xCe0.01Gdx)FS (x ¼ 0e0.3) excited at 450 nm. It was foundthat the Ce3þ emission had been enhanced and shifted from 660 to 672 nm ina-(Y0.69Ce0.01Gd0.3)FS, when Y3þ ions were partly replaced by Gd3þ ions.Demourgues et al. reported that the unit cell parameters decreased as a function ofthe rare earth ionic size at the same time that the bond length of four equatorial Ln-S and four Ln-F became smaller and the one apical Ln-S length was almost thesame (Ln ¼ LaeGd).115 In a-YFS host lattice, the substitution of Gd3þ ion (1.10 Å,CN ¼ 9) for Y3þ (1.07 Å, CN ¼ 9) caused an anisotropic expansion of the atomistic


structure together with a local compression within the internal (Y, Ce, Gd)F4S5 poly-hedron. With an increasing ratio in Gd3þ substitution, the diffraction peaks shiftedslightly toward the lower angle in accordance with the size effect. The CIE chroma-ticity coordinates of a-(Y, Gd)FS showed that their (x, y) value change from (0.65,0.35) to (0.66, 0.32) depending on the doped Gd3þ content, proving the potential toserve as a good red phosphor.

The optical properties of Ca1�xSrxLaGa3S6O:0.05Eu2þ solid solution phosphor is

also reported.116 A tunable emission with a suitable peak position and color coordi-nates in the 560-540 nm range were realized by adjusting the Sr content(x) in theCa1�xSrxLaGa3S6O:0.05Eu

2þ phosphors. The changes in the lattice constants a andc and the cell volume V as a function of the compositional ratio x in

y = 1

y = 0.75

y = 0.5

y = 0.25

y = 0.1

y = 0

400 500 600 700 800 900Wavelength (nm)

Y SrF S :Ce Y CaF S :Ce

OrangeYellow

Ce 5dY 4d

Ce 4f

s 3p

553 575 590

Blueshif Redshif

PL

inte

nsity

( ar

b. u

nit)

2.84

2.81

2.55

2.55

a

c

(a)

(b)

(c)

Figure 5.12 (a) Excitation and emission spectra of Y1.98Ce0.02(Ca1�ySry)F4S2 (y ¼ 0, 0.1, 0.25,0.5, 0.75, and 1). The insets are the corresponding photos taken under 365 nm excitation. (b) Aplausible mechanism of electronic transition in Y2CaF4S2:Ce

3þ and Y2SrF4S2:Ce3þ system.

The arrows represent the electronic transitions from Ce (4f) to Ce (5d) and photoemission,respectively. (c) Schematic unit cell crystal structure of Y1.98Ce0.02CaF4S2 and coordinationenvironment around YF4S5 and CaF8. Pink (dark gray in print versions), blue (black in printversions), green (gray in print versions), and yellow (light gray in print versions) spherical ballsdescribe Y/Ce, Ca, F, and S atoms.113

Modified from Wu YC, Chen YC, Chen TM, Lee CS, Chen KJ, Kuo HC. Crystal structurecharacterization, optical and photoluminescent properties of tunable yellow to orange-emittingY2(Ca, Sr)F4S2:Ce

3þ phosphors for solid-state lighting. J Mater Chem 2012;22(16):8048e56.https://doi.org/10.1039/C2JM16882K.


https://doi.org/10.1039/C2JM16882K

Ca1�xSrxLaGa3S6O:0.05Eu2þ is revealed in Fig. 5.14(a). The value of a, c, and V grad-

ually increase as the Sr concentration x from 0 to 1 owing to the ionic radius of Ca2þ

(1.12 Å) is smaller than that of Sr2þ (1.26 Å). The lattice constants of CaLaGa3S6O:0.10Eu2þ (a ¼ 9.296 Å, c ¼ 6.040 Å, V ¼ 521.95 Å3) are slightly smaller than thoseof SrLaGa3S6O:0.10Eu

2þ (a ¼ 9.377 Å, c ¼ 6.092 Å, V ¼ 535.65 Å3). Comparedwith previous results (CaLaGa3S6O:a ¼ 9.271 Å, c ¼ 6.035 Å; SrLaGa3S6O:a ¼ 9.347 Å, c ¼ 6.089 Å), the lattice constants of the samples are reasonable.117

On the other hand, the unit cell volume V of the Eu2þ-doped oxysulfide phosphor isslightly larger than that of the CaLaGa3S6O or SrLaGa3S6O itself without exception.It may be owned to the fact that the divalent Eu ions are expected to occupy the Sr orCa sites in this phosphor because of the slightly larger ionic radius of Eu2þ (1.25 Å)compared with those of Sr2þ or Ca2þ. Thus the doping of Eu2þ leads to the enlarge-ment of the lattice volume. The lattice constant linearly increases with increasing xmatching well with the Vegard’s law. Fig. 5.14(b) reveals the excitation spectra ofthe CaLaGa3S6O:0.05Eu

2þ, Ca0.6Sr0.4LaGa3S6O:0.05Eu2þ, and SrLaGa3S6O:

0.05Eu2þ phosphors. The excitation spectra of the three phosphors reveal similarbroad band absorptions in the 200e500 nm range, from UV to visible light. The broadexcitation bands match well with the emissions of UV LEDs (350e410 nm) and blueLEDs (430e500 nm). Fig. 5.14(c) shows the normalized emission spectra ofCa1�xSrxLaGa3S6O:0.05Eu

2þ (x ¼ 0, 0.2, 0.4, 0.6, 0.8, and 1.0) under 450 nm excita-tion. The emission spectra of Ca1�xSrxLaGa3S6O:0.05Eu

2þ phosphor show the char-acteristic broadband Eu2þ emission with the emission peak shifting about 20 nm. The

600 700 800 900 1000 1100Wavelength (nm)

PL

inte

nsity

(arb

. uni

t)x = 0.2

0.10.050

0.3

λex = 450 nm

0.6

0.3

0.4

0.7CIE x

CIE

y

Planckian(0.65, 0.35)

(0.66, 0.32)

α–Y0.99–xCe0.01GdxFS

365 nm

Figure 5.13 The emission spectra of a-(Y0.99�xCe0.01Gdx)FS (x ¼ 0, 0.05, 0.1, 0.2, and 0.3)(lex ¼ 450 nm). The figure shows the variation of Commission Internationale de I’Eclairage(CIE) chromaticity coordinates as a function of Gd3þ dopant content and the photo of red (grayin print versions) a-YFS:Ce3þ taken under 365 nm excitation.114

Modified fromWu YC, Chen YC, Wang DY, Lee CS, Sun CC, Chen TM. a-(Y, Gd)FS: Ce3þ: anovel red-emitting fluorosulfide phosphor for solid-state lighting. J Mater Chem 2011;21(39):15163e6. https://doi.org/10.1039/C1JM12819A.


https://doi.org/10.1039/C1JM12819A

dependence of FWHM and the relative intensity on the Sr content (x) inCa1�xSrxLaGa3S6O:0.05Eu

2þ is revealed in Fig. 5.14(d). As shown in line (a), theFWHM values range from 55 to 38 nm. Compared with the common FWHM valueof Eu2þ ions in most phosphors (50e100 nm), the observed values are smaller,demonstrating a weak interaction of the Eu2þ ions with the host. Increasing the Sr con-tent in Ca1�xSrxLaGa3S6O:0.05Eu

2þ phosphors causes a decrease in the FWHM,which owning to the decrease in phonon energy of the host. The FWHM is character-istic of a phonon-broadened emission that can be depicted by a single configurationcoordinate model. The phonon energy is basically influenced by the sizes of the MII

cations, which decreases slightly with increasing size of the divalent cation in the orderCa, Eu, Sr, and Ba.118,119 Line (b) in Fig. 5.14 shows the change in the emission

0.0 0.2 0.4 0.6 0.8 1.0

Sr content (x)

1: X = 0.02: X = 0.23: X = 0.44: X = 0.65: X = 0.86: X = 1.0

475 500 525 550 575 600 625 650 675 700Wavelength (nm)

Em

issi

on p

eak

(nm

)

Nor

mal

ized

inte

nsity

(arb

. uni

t)

%Sr

654321

Experimental dataFitted line



535

530

525

520

x

9.40

9.35

9.30a (A

ngst

rom

)

6.10

6.08

6.06

6.04

c (A

ngst

om)

Nor

mal

ized

inte

nsity

(arb

. uni

t) 1.0

0.8

0.6

0.4

0.2

0.0250 300 350 400 450 500

Wavelength (nm)

(a)

(a)

(b)

(b)

(c)

(c)

λ = 450 nm

Sr LaGa S O:0.05Eu

V (A

ngst

rom

)

CaLaGa S O:0.05Eu

Ca Sr LaGa S O:0.05Eu

SrLaGa S O:0.05Eu

(a) (b)

(c) (d)In

tens

ity (a

rb. u

nit)

Sr content (x)

1.50

1.25

1.00

0.75

0.50

0.25

0.0 0.2 0.4 0.6 0.8 1.0

606550454035302520

FWH

M (n

m)

(a) FWHM

(a)

(b)%Sr

(b) λ = 450 nm

Figure 5.14 (a) Dependence of the lattice constants a and c and the cell volume V on thecompositional ratio (x) in the Ca1�xSrxLaGa3S6O:0.05Eu

2þ phosphors. (b) Excitation spectraof Ca1�xSrxLaGa3S6O:0.05Eu

2þ with different x (x ¼ 0, 0.4, and 1) (a: lem ¼ 560 nm, b:lem ¼ 555 nm, c: lem ¼ 540 nm). (c) Emission spectra of Ca1�xSrxLaGa3S6O:0.05Eu

2þ withdifferent x (lex ¼ 450 nm). (d) Dependence of full width half maximum and relative intensityon the Sr content (x) in Ca1�xSrxLaGa3S6O:0.05Eu

2þ.116

Modified from Yu R, An Y, Wang C, Wang H, Wu Y, Zhang J. Tunable yellowish-green togreen (Ca1�xSrx)LaGa3S6O: Eu

2þ phosphors for potential LED application. Electrochem SolidSt 2012;15(1):J1e5. https://doi.org/10.1149/2.017201esl.


https://doi.org/10.1149/2.017201esl

450 475 500 525 550 575 600 625 650 675 700Wavelength (nm)

(1)x = 0.01(2)x = 0.03(3)x = 0.06(4)x = 0.10(5)x = 0.15(6)x = 0.18(7)x = 0.20

λex = 405 nm

Inte

nsity

(arb

. uni

t)

Ca1–xLaGa3S7:xEu2+

5

6

432

17

0.00 0.05 0.10 0.15 0.20Eu2+ content (x)

Inte

nsity

(a.u

.)

Figure 5.15 Dependence of photoluminescent intensities of Ca1�xEuxLaGa3S7 atlex ¼ 405 nm at different Eu2þ concentrations. The upper inset shows the influence of theconcentration on the emission intensity of Ca1�xEuxLaGa3S7 phosphor (x ¼ 0.01, 0.03, 0.06,0.10, 0.15, 0.18, 0.20).120

Modified from Yu RJ, Li HJ, Ma HL, Wang CF, Wang H, Moon BK, Jeong JH.Photoluminescence properties of a new Eu2þ-activated CaLaGa3S7, yellowish-green phosphorfor white LED applications. J Lumin 2012;132(10):2783e7. https://doi.org/10.1016/j.jlumin.2012.05.004.

1.2

1.0

0.8

0.6

0.4

0.2

0.0250 350 450 550 650 750 850 400 500 600 700 800

Wavelength (nm) Wavelength (nm)

20 mA50 mA100 mA150 mA200 mA250 mA300 mA350 mA

(a) (b)

Inte

nsity

(arb

. uni

t)

Inta

nesi

ty (a

rb. u

nit)

Figure 5.16 (a) Photoluminescence excitation (lem ¼ 465 nm) and photoluminescent(lex ¼ 700 nm) spectra of Sr4(PO4)2O:Eu

2þ (red (gray in print versions) and Ca4(PO4)2O:Eu2þ (blue (dark gray in print versions)).132 (b) Electroluminescent spectra of white light-emitting diode using a 460 nm blue (dark gray in print versions) chip combined with Sr2SiO4:Eu2þ and Ca3.93(PO4)2O:0.07Eu

2þ under various applied currents. The inset shows an imageof the light-emitting diode lamp package.129

Modified from (a) Komuro N, Mikami M, Saines PJ, Akimoto K, Cheetham AK. Deep redemission in Eu2þ-activated Sr4(PO4)2O phosphors for blue-pumped white LEDs. J Mater ChemC 2015;3:7356e62. https://doi.org/10.1039/c5tc01151e. (b) Deng D, Yu H, Li Y, Hua Y, Jia G,Zhao S, Wang H, Huang L, Li Y, Li C, Xu S. Ca4(PO4)2O: Eu

2þ red-emitting phosphor forsolid-state lighting: structure, luminescent properties and white light emitting diode application.J Mater Chem C 2013;1:3194e9. https://doi.org/10.1039/c3tc30148f.


https://doi.org/10.1016/j.jlumin.2012.05.004

https://doi.org/10.1016/j.jlumin.2012.05.004

https://doi.org/10.1039/c5tc01151e

https://doi.org/10.1039/c3tc30148f

intensities as increasing Sr content under 450 nm excitation. The emission intensity de-creases with increasing Sr content until it reaches about 26% of the intensity of CaL-aGa3S6O:0.05Eu

2þ. These characteristics demonstrate that Ca1�xSrxLaGa3S6O:0.05Eu2þ can be good yellowish-green to green phosphor candidates for wLEDs.

A yellowish-green Eu2þ-activated CaLaGa3S7 chalcogenide phosphor was syn-thesized by a two-step solid-state reaction,120 which has the similar structure as CaL-aGa3S6O. Because the sulfur has the smaller electronegativity than that of oxygen,the stronger nephelauxetic effect of doped Eu2þ ions appeared and the absorptionmay extend to the visible region (400e500 nm) and its emission is shifted to longerwavelength region. The dependence of PL emission intensity of Ca1�xLaGa3S7:xEu2þ phosphors on Eu2þ concentrations (x ¼ 0.01, 0.03, 0.06, 0.10, 0.15, 0.18,0.20) is shown in Fig. 5.15. With UV excitation (lex ¼ 405 nm), all of the phosphorsexhibited yellowish-green emission. It can be noted from the figure that the inten-sities increase with the Eu2þ content. The maximum occurs when x ¼ 0.15 thenthey decrease due to concentration quenching caused by the resonant ET amongEu2þ ions.

5.4.1.4 Phosphate

Recently, many promising phosphate LED phosphors have also been reported.2,3

Among them, phosphate phosphors represented by the apatite-type Ca5(PO4)3Cl:Eu2þ have a long history of use in the lighting and display industry. However, the op-timum absorption of Eu-doped phosphate phosphors rarely matches the emission ofblue LEDs. That is to say, most of these could only be excited by n-UV LED chipsin the wavelength range 350e420 nm. For example, Eu2þ-doped orthophosphateABPO4-type (A ¼ alkali metal, such as Li, Na, and K; B ¼ alkaline earth metal,such as Ca, Sr, and Ba) phosphors usually emit blue luminescence with a broadband under ultraviolet excitation at 360 nm, but most of them have excellent thermallystable PL at high temperatures.121 Among these, LiSrPO4:Eu

2þ shows significant redshifts in its excitation and emission bands with an emission peak at 445 nm, which issignificantly different from those of KSrPO4:Eu

2þ and KBaPO4:Eu2þ. However, the

excitation and emission wavelengths of the series KSrPO4:Eu2þ and KBaPO4:Eu

2þ

are very similar, which should be ascribed their similar crystal structures.Another type of emerging phosphate phosphors possess the b-Ca3(PO4)2-type crys-

tal structures, and typical hosts include Sr1.75Ca1.25(PO4)2, Ca9Ln(PO4)7 (Ln ¼ Y, Gd,Lu, Sc), Sr8ZnSc(PO4)7, and Sr8MgLn(PO4)7 (Ln ¼ Y,La).122e125 In general, thesecompounds crystallize as a hexagonal structure with space group R3c, similar to theb-Ca3(PO4)2 phase. As examples, Eu2þ-activated Sr8MgY(PO4)7 (SMYP) andSr8MgLa(PO4)7 (SMLP) phosphors show broad band excitation and absorption inthe 250e450 nm near-ultraviolet region, which meets the application requirementsfor n-UV LED chips.125 Upon excitation at 400 nm, both the SMYP:Eu2þ andSMLP:Eu2þ phosphors exhibit strong yellow emissions centered at 518, 610, and611 nm with better thermal stability than (Ba, Sr)2SiO4 (570 nm) phosphors.

Eulytite-type M3Ln(PO4) phosphate compounds (where M ¼ Sr, Ba and Ln ¼ rareearth) also act as the host materials for LED phosphors. These eulytite-type phosphate


compounds can incorporate various foreign ions and provide abundant crystal field en-vironments, which make it possible to finely tune the physical/chemical properties todesign new luminescent materials.126e128 For example, yellow-emitting Sr3Ce(PO4)3:Eu2þ phosphors give an extreme broadband emission peak at around 535 nm underexcitation at long wavelengths (>370 nm).130 A wLED device with a color-rendering index of 86.5, a color temperature of 5996K, and chromaticity coordinatesof (0.32, 0.38) was obtained by combining a 405 nm near-UV LED chip, phosphorblends of yellow-emitting Sr3Ce(PO4)3:Eu

2þ and the commercial blue-emitting BaM-gAl10O17:Eu

2þ phosphor.Nevertheless, to develop new phosphate phosphors that exhibit a good excitation

and emission performance when combined with blue LEDs remains a challenge. In2013, Xu’s group reported a red-emitting phosphor, Eu2þ-activated Ca4(PO4)2O,which exhibits strong red emission centered at 665 nm under blue light excitation at460 nm.129 After that, Li and Liu’s group reported the Ca4(PO4)2O:Eu

2þ, Ce3þ phos-phor.130 Furthermore, Cheetham’s group performed a detailed investigation on theorigin of the broadband absorption and deep red emission of a Ca4(PO4)2O:Eu

2þ phos-phor.131 Powder neutron diffraction was used to precisely detect the oxygen positionsin the crystal structure, and quite distorted coordination polyhedra around Ca siteswere revealed. The relation between the crystal structure and emission spectra was dis-cussed and the importance of the anion status and the distortion of the coordinationpolyhedron were highlighted. Moreover, they also discovered another deep red phos-phor, Sr4(PO4)2O:Eu

2þ, which has an excitation peak around 450 nm for blue LEDapplications. The crystal structure of Sr4(PO4)2O:Eu

2þ is found to be monoclinicP21 and is isotypic with Ca4(PO4)2O:Eu

2þ, which also shows deep red emissionwith a peak position at 680 nm. Fig. 5.16(a) comparatively shows the PLE and PLspectra of Sr4(PO4)2O:Eu

2þ and Ca4(PO4)2O:Eu2þ phosphors, and it is clearly seen

that they have a similar spectral profile.132 In order to demonstrate the applicationof this kind of phosphor in wLEDs, Fig. 5.16(b) shows EL spectra of a white LED us-ing a 460 nm blue chip combined with Sr2SiO4:Eu

2þ and Ca3.93(PO4)2O:0.07Eu2þ

phosphors under various applied currents. The EL spectra clearly show blue bandsat around 460 nm, green emitting bands at 540 nm, originating from Sr2SiO4:Eu

2þ

phosphor, and red emitting bands at around 665 nm, corresponding toCa3.93(PO4)2O:0.07Eu

2þ. As an example, when the applied current is 20 mA, thewhite LED has CIE color coordinates of (0.3135, 0.3316) for white light(Tc ¼ 6446K), an excellent Ra of 90.5 and a luminous efficiency of 41 lm/W.131

Cheetham’s group reported another yellow phosphor excited by blue light and n-UV light, Ca6BaP4O17:Eu

2þ.133 The new phase, Ca6BaP4O17, was found in the CaOeBaOeP2O5 phase diagram and its structure was solved from high resolution, synchro-tron X-ray powder diffraction data. As also reported by our group, Fig. 5.17(a) showsthe normalized emission spectra of Ca5.94�yLi0.03BaP4O17:0.03Ce

3þ, Eu2þ withdifferent Eu2þ concentrations under 365 nm excitation, and the emission spectraexhibit red-shift behavior from 463 nm, corresponding to Ce3þ emission in the Ca6B-aP4O17 host, to 528 nm, originating from Eu2þ emission in Ca6BaP4O17 via theincreasing Eu2þ concentration.134 Fig. 5.17(b) clearly depicts the chromaticity coordi-nates of Ca5.94�yLi0.03BaP4O17:0.03Ce

3þ, yEu2þ phosphors. It can be seen that the


1.2

1.0

0.8

0.6

0.4

0.2

0.0350 400 450 500 550 600 650 700 750 800

Wavelength (nm)

Inte

nsity

arb

. uni

t

20 mA (0.380, 0.408) Ra = 93

50 mA (0.381, 0.409) Ra = 93

100 mA (0.378, 0.407) Ra = 93

150 mA (0.375, 0.406) Ra = 93

200 mA (0.374, 0.406) Ra = 93

250 mA (0.373, 0.405) Ra = 93

300 mA (0.371, 0.404) Ra = 93

350 mA (0.370, 0.403) Ra = 93

400 450 500 550 600 650 700Wavelength (nm)

Nor

mal

ized

inte

nsity

(arb

. uni

t) Ca5.94–yLi0.03Ba(PO4)4O: 0.03Ce3+,yEu2+

λex = 365 nm

y = 0y = 0.005y = 0.01y = 0.03y = 0.05

(a)

(c)

(b)

Figure 5.17 (a) Normalized emission spectra and (b) Commission Internationale de I’Eclairage(CIE) chromaticity coordinates and digital images of Ca5.94�yLi0.03BaP4O17:0.03Ce

3þ, yEu2þ

phosphors under 365 nm excitation.134 (c) Electroluminescent spectra of a white light-emittingdiode (LED) consisting of the Ca6BaP4O17:Ce

3þ, Si4þ, Ca6BaP4O17:Eu2þ, CaAlSiN3:Eu

2þ

phosphor and an near ultraviolet LED. The inset shows an image of the white LED in oper-ation, the color coordinates and Ra at each operation current.135

Modified from (a and b) Chen M, Xia Z, Liu Q. Improved optical photoluminescence by chargecompensation and luminescence tuning in Ca6Ba(PO4)4O: Ce

3þ, Eu2þ phosphors. Cryst EngComm 2015;17:8632e8. https://doi.org/10.1039/C5CE01766A. (c) Komuro N, Mikami M,Shimomura Y, Bithell EG, Cheetham AK. Synthesis, structure and optical properties of cerium-doped calcium barium phosphatee a novel blue-green phosphor for solid-state lighting. J MaterChem C 2015;3:204e10. https://doi.org/10.1039/c4tc01835d.


https://doi.org/10.1039/C5CE01766A


color tone of the phosphors can be modulated from blue to green and yellow by simplyadjusting the doping amount of the Eu2þ ion and fixing the Ce3þ content in the Ca6B-aP4O17 host. Ca6BaP4O17:Ce

3þ shows blue-green emission upon near-UV LEDs oper-ating at 400 nm, and Ca6BaP4O17:Eu

2þ is an exceptional yellow phosphor.135 A whiteLED was then fabricated by using the Ca6BaP4O17:Ce

3þ, Si4þ, Ca6BaP4O17:Eu2þ,

CaAlSiN3:Eu2þ phosphor and a near-UV LED, and the EL spectra are given in

Fig. 5.17(c). The inset shows a photograph of the white LED in operation, the colorcoordinates and the Ra at each operation current. It was found that its CCT was4448K and the CIE color coordinates were (0.370, 0.403). The luminous efficacyand external QE of the LED at a high forward current of 350 mA were 45 lm/Wand 20.8%, respectively, which indicate potential suitability for future applications.Generally speaking, the blue-light-excited phosphate phosphors should be givenmore attention with respect to practical LED application, because they are chemicallystable and some of them could show highly thermally stable PL.

5.4.1.5 Borates

Borates also play important roles in the family of luminescence materials, and haveattracted much attention, due to their stability, potential low-cost synthesis and envi-ronmentally friendly characters, and a number of borates compounds with differentstructures can be selected.2,3 For example, in the systems of M2OeNOeB2O3

(M ¼ Li, Na, K; N ¼ Ca, Sr, Ba), one can find many new functional materials. Allthe atoms of borates are prefer to be close-packed, and then they are easy to crystallizein the cubic crystal system. Moreover, [BO3]

3� anionic groups are perpendicular toeach other, distributed along the (100) directions, beyond the isostructural novel com-pounds. For example, Wu et al.136 synthesized LiSr4(BO3)3, NaSr4(BO3)3, NaSrBO3,Na3SrB5O10, NaSrB5O9, and NaBa4(BO3)3 successfully, and one can find thatLiSr4(BO3)3, NaSr4(BO3)3, and NaBa4(BO3)3 are isostructural and the final refinementpatters are given in Fig. 5.18. Ce3þ doped MSr4(BO3)3 (M ¼ Li, Na) phosphors showblue emission at room temperature. The PLE and PL spectra of LiSr4(BO3)3:Ce

3þ andNaSr4(BO3)3:Ce

3þ are presented in Fig. 5.18(b-i) and (b-ii), respectively.137 It can beseen that the excitation spectral consist of three broad bands with peaks at about 265,290 and 345 nm (342 nm for M ¼ Na), which is ascribed to the CFS of Ce3þ 5d orbits.Under the UV excitation, the emission spectra of the samples show a broad asymmetricblue emission band for LiSr4(BO3)3 and NaSr4(BO3)3, which are attributed to the 4f-5dtransition of Ce3þ. In addition, there is a minor red shift for the emission wavelength ofCe3þ in the compound NaSr4(BO3)3 in comparison with that of Ce3þ in LiSr4(BO3)3compound, which can be understood in terms of the crystal field theory with the sub-stitution of Liþ by Naþ.

Gaussian peak fitting was carried out in order to analyze the asymmetric emissionspectra. Fig. 5.18(c-i) and (c-ii) shows Gaussian curve fitting of the emission spectra,and those could be disassemble into four Gaussian peaks those are bands I, II, III, andIV in LiSr4(BO3)3 and NaSr4(BO3)3, respectively. The center Ce (1) with a weakercrystal field is proposed to occupy the site of Sr (1), whereas the other center Ce (2)with a stronger crystal field due to enter the sites of Sr (2). For the same cations, the


large anion coordination number will lead to the strong crystal field strength, thus therewill be occur that CFS generally increasing. Therefore, the emission of Ce(2) shouldhave a red-shift in comparison with that of Ce (1) and Ce (2) occupy the site of Sr (2),where eight O2� anion coordinated one Sr2þ.

Generally speaking, by one kind of activator or by different activators at differentcrystal sites in a single phase, white light will be formed.138,139 Thus a proper crystalfield beyond the substitution sites for those activators is necessary to generate whitelight. Zhang et al.139 synthesized Ce3þ, Tb3þ, Mn2þ singly doped and codopedNaCaBO3:phosphors. The orthorhombic structure of NaCaBO3 belongs to the spacegroup Pmmn as shown in Fig. 5.19(a). Both the Naþ and Ca2þ cations are surroundedin two kinds of different chemical environment. From Fig. 5.19(b), one can also see theemission spectrum with a peak centered at about 425 nm of NaCaBO3:Ce

3þ, besides, it

LiSr4(BO3)3

NaSr4(BO3)3

NaBa4(BO3)3In

tens

ity (a

rb. u

nit)

350 400 450 500 550Wavelength (nm)

200 250 300 350 400 450 500 550 600Wavelength (nm)

I

I

III

III

II

II

IV

IV

(i)

(ii)

(i)

(ii)

λem = 427 nm

λem = 420 nm λex = 345 nm

λex = 342 nm

Inte

nsity

(arb

. uni

t)

Inte

nsity

(arb

. uni

t)

10 20 30 40 50 60 70 80 90 100 110 1202θ (deg.)

(a) (b)

(c)

Inte

nsity

(arb

. uni

t)In

tens

ity(a

rb. u

nit)

0.00 0.04 0.08

0.00 0.04 0.08

Concent of Ce

Concent of Ce

Figure 5.18 (a) Rietveld refinement plots of the three borate compounds.136 (b) Thephotoluminescence excitation and photoluminescent (PL) spectra of LiSr4(BO3)3:Ce

3þ (i) andNaSr4(BO3)3:Ce

3þ (ii). (c) Emission spectra excited at 345 nm at room temperature with fourdeconvoluted Gaussian peak, LiSr4(BO3)3:Ce

3þ (i) and NaSr4(BO3)3:Ce3þ (ii), Inset: the

dependence of PL intensity on the concentrations of Ce3þ.137

Modified from (a) Wu L, Chen XL, Li H, He M, Xu YP, Li XZ. Structure determination andrelative properties of novel cubic borates MM’4(BO3)3(M ¼ Li, M’ ¼ Sr; M ¼ Na, M’ ¼ Sr,Ba). Inorg Chem 2005;44:6409e14. https://doi.org/10.1021/ic050299s. (b and c) Guo CF, DingX, Seo HJ, Ren ZY, Bai JT. Luminescent properties of UV excitable blue emitting phosphorsMSr4(BO3)3:Ce

3þ (M ¼ Li and Na). J Alloys Compd 2011;509(14):4871e4. https://doi.org/10.1016/j.jallcom.2011.01.194.


https://doi.org/10.1021/ic050299s

https://doi.org/10.1016/j.jallcom.2011.01.194

https://doi.org/10.1016/j.jallcom.2011.01.194

exhibits a broad nonsymmetrical band in the wavelength range of 355 e600 nm.Because there are two Ca2þ sites, Ce3þ in a single precise lattice site usually showsa typical double emission band, then it can be fitted the emission spectrum into fourGaussian peaking at 387, 418, 427, and 468 nm, which are marked as band A, B,C, and D, respectively. The energy gap between the bands A and B is 1950 cm�1,and the bands C and D is 2015 cm�1, which is consistent with the theoretical value

250 300 350 400 450 500 550 600Wavelength (nm)

(λem = 460 nm)PLE

PL(λex = 347 nm)PLE

(λem = 427 nm)

Experimental curve

Deconvolution curveFitting curve

0.00 0.01 0.02 0.03 0.04 0.05

Ce3+ content

Rel

ativ

e in

tens

ity (a

.u.)

Rel

ativ

e in

tens

ity (a

rb. u

nit)

CaNaNaOB

b

a

b

a

(a)

(b)

AB

CD

Figure 5.19 (a) The crystal structure of NaCaBO3 viewed along the c-axis direction.(b) photoluminescent (PL) and photoluminescence excitation (PLE) spectra of NaCaBO3:0.01Ce3þ sample. Inset shows the PL intensity of the NaCaBO3:Ce

3þ samples as a function ofthe Ce3þ content.139

Modified from Zhang XG, Gong ML. Single-phased white-light-emitting NaCaBO3: Ce3þ,

Tb3þ, Mn2þ phosphor for LED applications. Dalton Trans 2014;43:2465e72. https://doi.org/10.1039/c3dt52328d.


https://doi.org/10.1039/c3dt52328d


of w2000 cm�1. From the emission spectra, it can be also deduced that there are twokinds of Ce3þ luminescent centers in the host lattice.

5.4.1.6 Narrow-band red nitride phosphors

Recently, novel narrow-band red LEDs phosphors become a hot issue as alsomentioned earlier. As we know, nitrides phosphors activated by Eu2þ are most likelyto have a large CFS and a low centroid level of Eu2þ 5d states, due to the high cova-lency coordination of N3�. Thence, the excitation and emission deriving from the tran-sition between the 4f and 5d states in Eu2þ-activated nitride phosphors tend to exhibitlong wavelengths, which is suitable for wLEDs. Until recently, Wang et al. reportedthat the narrow-band nitride phosphors of Eu2þ-doped is characterized by a large split-ting (DEs > 0.1 eV) between the two highest 4f7 bands of Eu2þ.140 Aimed to define aquantitative descriptor (DEs), they provided a screening strategy for narrow-bandEu2þ-activated emission according to a comparison of the electronic structure of re-ported narrow band (see Fig. 5.20(a)). They pointed out that the Eu2þ narrow-bandemission is the result of Eu2þ 4f orbitals crystal-field splitting in a cuboid or highlysymmetrical EuN9 environment. Fig. 5.20(b) shows that a large energy splitting(DEs > 0.1 eV) between the two highest 4f bands lead to narrow-band emission ofEu2þ-doped phosphors in wLEDs.141 Several potential narrow-band red phosphorswere obtained in this family, just as shown in Fig. 5.20(a), with the exception ofnarrow-band green-emitting phosphors BaLi2Al2Si2N6:Eu

2þ and b-SiAlON:Eu2þ.142 Nevertheless, the discovery and study of new nitride phosphors forimproving the luminescence performance of wLEDs is still very significant.

Herein, based on the typical Sr[LiAl3N4]:Eu2þ phosphors reported by Schnick’s

group,47 many novel narrow-band green-emitting phosphors for wLEDs have beenproposed recently. The excitation and emission spectrum of phosphors Sr[LiAl3N4]:Eu2þ are depicted in Fig. 5.21(a). Sr[LiAl3N4]:Eu

2þ shows an extremely broad absorp-tion band with peak wavelength at w466 nm. The maximum of the emission is atëem ¼ 654 nm, with an FWHM of only 1180 cm�1 (w50 nm). The grey curve inFig. 5.21(a) exhibits the emission spectrum of a commercially available CaAlSiN3:Eu2þ phosphor which is presently applied in warm-white high power pc-LEDs,with a resemble peak wavelength of ëem ¼ 649 nm. Compared with the CaAlSiN3:Eu2þ phosphor, the remarkably reduced FWHM of Sr[LiAl3N4]:Eu

2þ (Fig. 5.21(a))concentrates the emitting light in the visible spectral range while further optimizingthe chromatic saturation of the red spectra.

Isovalent substitutions on single cations’ or anions’ sites can change the chemicalcompositions of the isostructural hosts, and further acts as main reason for altering thecoordination environments of the activators, which then tune the PL of Eu2þ or Ce3þ

emission. Accordingly, nitride phosphor material Ca[LiAl3N4]:Eu2þ can be designed

by replacing Sr2þ with Ca2þ. Ca[LiAl3N4]:Eu2þ is also an interesting new narrow-

band red nitride phosphor material with potential for application in high-power pc-LEDs.16 As is shown in Fig. 5.21(b), with excitation by blue InGaN-based LEDs,the compound shows an emission maximum at 668 nm with an FWHM of only1333 cm�1 (w60 nm). More recently, Schnick group discovered and reported novel


nitride phosphors Ca18.75Li10.5[Al39N55]:Eu2þ.143 This phosphors exhibit intense red

luminescence upon the excitation of UV to green light. On account of Ca18.75L-i10.5[Al39N55]:Eu

2þ is just a representative compound in the solid-solution seriesCa(20�x)Li(8þ2x)[Al39N55]:Eu

2þ (x ¼ 0e2), the luminescence properties is expectedto be tuned by modifying the atomic ratio Ca:Li. Hence, several different host latticecompositions (x ¼ 0, 1.25 and 2) were investigated. Excitation of the representativecompound (x ¼ 1.25) at 450 nm exhibit an emission band with a maximum at647 nm and a remarkably narrow FWHM of 1280 cm�1 (w54 nm), with an internalquantum efficiency (IQE) of 11% (see Fig. 5.21(c)). Peak emission and FWHM valuesare very similar with Sr[LiAl3N4]:Eu

2þ (see Fig. 5.21(a)). There are two excitation

Lowest 5d

Narrow-band

ΔEs

–0.05

–0.10

–0.15

–0.20

SrLiAI 3N

4-1

SrLiAI 3N

4-2

SrMg 3S

iN 4

CaLiAI 3N

4

-SiAIO

N

BaLi 2A

I 2Si 2N

6

β

0.00

Aver

age

Eu2+

4ƒ

leve

ls (e

V)

4ƒ bands

Narrow-bandemission

ΔES > 0.1 eV

(a) (b)

Figure 5.20 (a) Average Eu2þ 4f band levels for five narrow-band phosphors. (b) Relationshipbetween emission bandwidth and Eu2þ 4f band levels.140

Modified from Wang Z, Chu IH, Zhou F, Ong SP. Electronic structure descriptor for thediscovery of narrow-band red-emitting phosphors. Chem Mater 2016;28:4024e31. https://doi.org/10.1021/acs.chemmater.6b01496.


https://doi.org/10.1021/acs.chemmater.6b01496


spectra bands: one in the range from UV to blue, peaking at w380 nm, and anotherbroader band in the blue to green spectral region (see Fig. 5.21(c), blue line). The exci-tation spectrum maximum peaked at w525 nm.

Moreover, the novel M[Mg2Al2N4] phosphors were recently reported by Schnicket al., which can be viewed as the lineal successor of the M[LiAl3N4] based on the

400 450 500 550 600 650 700 750 800Wavelength (nm)

350 400 450 500 550 600 650 700 750 800Wavelength (nm)

1.0

0.8

0.6

0.4

0.2

0.0350 400 450 500 550 600 650 700 750

Wavelength (nm)

1.0

0.8

0.6

0.4

0.2

0.0

Photolum

inescence intensity

Inte

nsity

arb

. uni

t

Inte

nsity

(arb

. uni

t)

Exc

itatio

n

(a) (b)

(c)

Figure 5.21 (a) Photoluminescence excitation (PLE) (Sr[LiAl3N4]:Eu2þ, blue (black in print

versions); CaAlSiN3:Eu2þ, light gray) and photoluminescent (PL) spectra (lex ¼ 440 nm) of

Sr[LiAl3N4]:Eu2þ (pink (gray in print versions)) and CaAlSiN3:Eu

2þ (dark gray). The dottedcurve indicates the upper limit of sensitivity of the human eye.47 (b) PLE (blue)(lem ¼ 668 nm) and PL (red (dark gray in print versions)) (lex ¼ 470 nm) spectra of Ca[LiAl3N4]:Eu

2þ.16 (c) PLE (blue (darkest gray in print versions) (lem ¼ 668 nm) and PL (pink)spectra of Ca18.75Li10.5[Al39N55]:Eu

2þ.143

Modified from (a) Pust P, Weiler V, Hecht C, T€ucks A, Wochnik AS, Henb AK, Wiechert D,Scheu C, Schmidt PJ, Schnick W. Narrow-band red-emitting Sr[LiAl3N4]: Eu

2þ as a next-generation LED-phosphor material. Nat Mater 2014;13(9):891e6. https://doi.org/10.1038/NMAT4012. (b) Pust P, Wochnik AS, Baumann E, Schmidt PJ, Wiechert D, Scheu C, SchnickW. Ca[LiAl3N4]: Eu

2þ-a narrow-band red-emitting nitridolithoaluminate. Chem Mater 2014;26(11):3544e9. https://doi.org/10.1021/cm501162n. (c) Wagatha P, Pust P, Weiler V, WochnikAS, Schmidt PJ, Scheu C, Schnick W. Ca18.75Li10.5[Al39N55]: Eu

2þ-supertetrahedron phosphorfor solid-state lighting. Chem Mater 2016;28:1220e6. https://doi.org/10.1021/acs.chemmater.5b04929.


https://doi.org/10.1038/NMAT4012

https://doi.org/10.1038/NMAT4012

https://doi.org/10.1021/cm501162n



cosubstitution of 2 Mg2þ ¼ Al3þ þ Liþ.49 Besides, M[Mg3SiN4] compounds can bedesigned based on M[Mg2Al2N4] compounds with the further substitution ofMg2þ þ Si4þ ¼ 2Al3þ.42 Accordingly, based on M[LiAl3N4], controllable substitu-tions including the typical Ga/Mg, Al/Mg, Li/Al, or Mg/Si pair on the differentcoordinated sites, can be introduced to develop new phosphor hosts. Crystal struc-tures of Sr[LiAl3N4], Sr[Mg2Al2N4], and Sr[Mg3SiN4] compounds are depicted inFig. 5.22(a)e(c), respectively. Sr[LiAl3N4] belongs to the triclinic space groupwith a highly condensed, rigid framework containing the ordered edge- andcorner-sharing AlN4 and LiN4 tetrahedra with channels of Vierer rings (it meansfour polyhedral connected to each other forming a ring) along [011] is observed,as shown in Fig. 5.22(a). Such a rigid structure can induce weak electronephononinteraction for the Eu2þ activator, which can significantly reduce the probability ofunwanted nonradiative relaxation processes in the excited Eu2þ ions and can alsoreduce the bandwidth. Similarly, the isoelectronic Sr[Mg2Al2N4] and Sr[Mg3SiN4],all isotypic to Sr[LiAl3N4], show a close structural relation to compounds crystal-lizing in the UCr4C4 structural type, with only difference in the network cation po-sitions distribution, as given in Fig. 5.22(b) and (c). The highly condensed networkof vertex- and corner-sharing (Mg/Al/Si)N4 tetrahedra is also formed in Sr[Mg2Al2N4] and Sr[Mg3SiN4] by forming Vierer ring channels along [001] and[100], respectively. All of these materials exhibit outstanding narrow-band red emis-sion, and a common structural feature of these narrow-band red phosphors is a cube-like coordination of the activator (Eu2þ) in the host lattice.

The excitation and emission spectra of phosphors Sr[Mg3SiN4]:Eu2þ are demon-

strated in Fig. 5.22(d).42 The excitation spectrum of Sr[Mg3SiN4]:Eu2þ exhibits a

broad band with the maximum at 450 nm, which can be excited very well by UV toblue light of commercial LEDs chips. The emission spectrum of Sr[Mg3SiN4]:Eu

2þ

(lex ¼ 440 nm) shows a narrow band peaking at 615 nm with an FWHM of onlyw1170 cm�1 (w43 nm). As far as we know, this is the narrowest emission forEu2þ phosphors in the red spectral region reported until now. It is very close to thetarget wavelength ranges of 620e630 nm and the FWHM of w30 nm of the idealnarrow-band red phosphors.

5.4.1.7 Mn4þ doped red phosphors

The emission peaks of the well-known nitride red phosphors, such as CaAlSiN3:Eu2þ

or Sr2Si5N8:Eu2þ, are often located at longer than 650 nm, therefore, the serious reab-

sorption behavior makes them unsuitable for ameliorating the efficiency of a device.Hence, Mn4þ-doped red phosphors, especially AMF6:Mn4þ (A ¼ alkali metal ion;M ¼ Si, Ge, Ti, etc.) with narrow band red emissions at approximately 630 nm,have recently attracted much research attention.2,3,57e79,144e147 As an example,Liu’s group reported narrow-band red emitting fluoride phosphor KNaSiF6:Mn4þ

for warm wLEDs.144 Wang’s group reported the HF-free hydrothermal synthesis ofK2SiF6:Mn4þ phosphors.145 As a comparison, the crystal structures of Na2SiF6:Mn4þ, KNaSiF6:Mn4þ and K2SiF6:Mn4þ as well as the corresponding SiF6

2� octahe-dron are shown in Fig. 5.23(a)e(c).144


SrAINLi

a

bc

a

c

250

1.0

0.8

0.6

0.4

0.2

0.0

300 350 400 450 500 550 600 650 700 750 800Wavelength (nm)

Inte

nsity

(arb

. uni

t)

b

a

(a) (b)

(c)(d)

Figure 5.22 Crystal structure of (a) Sr[LiAl3N4],48 (b) Sr[Mg2Al2N4], (c) Sr[Mg3SiN4] compounds,49 and (d) the excitation and emission spectra of Sr

[Mg3SiN4]:Eu2þ phosphors.42

Modified from (a) Tolhurst TM, Boyko TD, Pust P, Johnson NW, Schnick W, Moewes A. Investigations of the electronic structure and bandgap of thenext-generation LED-phosphor Sr[LiAl3N4]: Eu

2þ- experiment and calculations. Adv Opt Mater 2015;3(4):546e50. https://doi.org/10.1002/adom.201400558. (b and c) Pust P, Hintze F, Hecht C, Weiler V, Locher A, Zitnanska D, Harm S, Wiechert D, Schmidt PJ, Schnick W. Group (III) nitrides M[Mg2Al2N4] (M ¼ Ca, Sr, Ba, Eu) and Ba[Mg2Ga2N4] e structural relation and non-typical luminescence properties of Eu2þ doped samples. ChemMater 2014;26(21):6113e9. https://doi.org/10.1021/cm502280p (d) Schmiechen S, Schneider H, Wagatha P, Hecht C, Schmidt PJ, Schnick W.Toward new phosphors for application in illumination-grade white pc-LEDs: the nitridomagnesosilicates Ca[Mg3SiN4]: Ce

3þ, Sr[Mg3SiN4]: Eu2þ, and

Eu[Mg3SiN4]. Chem Mater 2014;26(8):2712e9. https://doi.org/10.1021/cm500610v.

Phosphors

forwhite

LEDs

161

https://doi.org/10.1002/adom.201400558

https://doi.org/10.1002/adom.201400558

https://doi.org/10.1021/cm502280p

https://doi.org/10.1021/cm500610v

NaSi2Si1F1F2F3

K

NaSiF1F2F3F4a

c b

K

SiF

b

c

a

b

ac

(a)

(b)

(c)

Figure 5.23 Crystal structures and the SiF62� octahedral of (a) Na2SiF6:Mn4þ,64 (b) KNaSiF6:

Mn4þ,146 and (c) K2SiF6:Mn4þ.144

Modified from (a, b and c) Jin Y, FangMH, Grinberg M, Mahlik S, Lesniewski T, Brik MG, LuoGY, Lin JG, Liu RS. Narrow red emission band fluoride phosphor KNaSiF6: Mn4þ for warmwhite light-emitting diodes. ACS Appl Mater Interfaces 2016;8:11194e203. https://doi.org/10.1021/acsami.6b01905.


https://doi.org/10.1021/acsami.6b01905


PL and PLE spectra of several typical Mn4þ doped phosphors are shown inFig. 5.24(a). All of the excitation spectrum contains two broad bands at w355 and460 nm (matching with the LED chips), consisted with the spin-allow transitions4A2g /

4T1g and 4A2g /4T2g of Mn4þ, respectively. The phosphor emits red-

emitting attributed to the d-d spin-forbidden transition 2Eg /4A2g with several nar-

row bands at w610e650 nm under 460 nm excitation. In order to understand theluminescence mechanism, we take phosphor KNaSiF6:Mn4þ as an example to analyzethe emission spectrum, as shown in Fig. 5.24(b) and (c).146,147 We can see that thestrongest emission is found at 629 nm. On the basis of group theory, there are six basicinternal vibronic modes, include í1, í2, í3, í4, í5, and í6, of the octahedral group with Oh

symmetry.69 Three narrow bands (longer than 620 nm) belong to Stokes í6, í4, and í3peaks, and two other narrow bands (shorter than 620 nm) are anti-Stokes í6 and í4peaks. The peak at about 620 nm is the zero-phonon line (ZPL) (Fig. 5.24(c)),which is the electronic dipole forbidden in the octahedral MnF6

2�. The MnF62�

octahedron is distorted because of :F1MnF2 ¼ :F1SiF2 ¼ 174.11 degrees and:F3MnF4 ¼ :F3SiF4 ¼ 179.29 degrees (Fig. 5.24(b)). As a comparison, as is shownin Fig. 5.24(c), only one type of MnF6

2� octahedron without distortion(:F1MnF1 ¼ :F1SiF1 ¼ 180 degrees) is found in K2SiF6. Thus no ZPL or veryweak is found in the PL spectrum of K2SiF6:Mn4þ phosphor in Fig. 5.24(a). By contrast,

350 400 450 500 550 600 650Wavelength (nm)

K2SiF6: Mn4+

54 nm

KNaSiF6: Mn4+

69 nm

Na2SiF6: Mn4+

75 nm

Inte

nsity

(arb

. uni

t)

λ = 460 nm

350 400 450 500 550 600 650 700Wavelength (nm)

580 590 600 610 620 630 640 650 660 670 680Wavelength (nm)

Inte

nsity

(arb

. uni

t)In

tens

ity (a

rb. u

nit)

KNaSiF : MnE → AA → T

A → T

λ =460 nm

λ = 629 nm

KNaSiF : Mn v

v

v

v v

ZPL

(a) (b)

(c)

Figure 5.24 (a) Photoluminescence (PL) and photoluminescence excitation (PLE spectra ofNa2SiF6:Mn4þ, KNaSiF6:Mn4þ and K2SiF6:Mn4þ.58,146,147 (b) PL and PLE spectra ofKNaSiF6:Mn4þ; (c) PL spectra of KNaSiF6:Mn4þ phosphor.144

Modified from (aec) Jin Y, Fang MH, Grinberg M, Mahlik S, Lesniewski T, Brik MG, Luo GY,Lin JG, Liu RS. Narrow red emission band fluoride phosphor KNaSiF6: Mn4þ for warm whitelight-emitting diodes. ACS Appl Mater Interfaces. 2016;8:11194e203. https://doi.org/10.1021/acsami.6b01905.




we can find two types of MnF62� octahedrons in Na2SiF6. Just as shown in Fig. 5.24(b), one

octahedron is without distortion (:F1MnF1¼ :F1SiF1 ¼ 180 degrees), but the other isdistorted (:F2MnF3 ¼ :F2SiF3 ¼ 171.48 degrees). Thus a strong ZPL is observedin KNaSiF6:Mn4þ phosphor. The position of the ZPL lies on the perturbation of thecation ion. The FWHM of the PLE provides more evidence about this distortion. Be-sides, the PLE spectrum is broadened. As the angles of K2SiF6, KNaSiF6, and Na2SiF6are 180, 174.11, and 171.48 degrees, respectively (just considering only larger distor-tions), the FWHM of the PLE should follow the order: K2SiF6 < KNaSiF6 < Na2SiF6.These characteristics are consistent with the following experimental results: 54, 69,and 75 nm for K2SiF6, KNaSiF6, and Na2SiF6, respectively (Fig. 5.24(a)).

When Si was replaced by Ge, Liu’s group et al. discovered the red narrow-bandphosphor K2GeF6:Mn4þ,70 and the photoluminescent evolution induced by structuraltransformation through thermal treating was discovered. The structural evolution sche-matic is shown in Fig. 5.25(a) during heat treatment of K2GeF6. The results showed nostructural transform occurs with pure P-3m1 (phase 1) when the temperature below200�C. With the temperature increasing to 250�C, P63mc (phase 2) generates, whereasP-3m1 is still the main phase at this time. Pure P63mc phase can be obtained at 400�C.Sequentially, when temperature is higher than 425�C, cubic Fm3m phase (phase 3) isobtained. Although the three symmetries have analogously coordination environmentsof MnF6

2� octahedron, the crystal structure different from each other significantly.Mn4þ ions show Oh symmetry in Fm3m K2GeF6 host, whereas the site symmetriesof Mn4þ ions in P-3m1 and P63mc hosts reduce to D3d and C6v, respectively. Samewith phosphor KNaSiF6 analyzed earlier, Oh have six fundamental internal vibronicmodes: í1(A1g), í2(Eg), í3(T1u), í4(T1u), í5(T2g), and í6(T2u). In D3d symmetry, the triplydegenerate modes í3, í4, í5, and í6 split into a doubly degenerate and a nondegeneratemode due to octahedral distortion.148,149 The antisymmetric vibronic modes coupled tothe ZPLs of 2E/ 4A2 results in the splitting phenomenon for emission lines, and re-sults in the broader emission spectra of K2GeF6:Mn4þ compared with sharp emissionlines of K2SiF6:Mn4þ. Luminescence spectra of the wLEDs (fabricated with blueInGaN chips, Y3Al5O12:Ce

3þ yellow phosphor, and K2GeF6:Mn4þ red phosphors)have been shown in Fig. 5.25(b). The CCTs of 3974 and 3363K are calculated whichconsists of red K2GeF6:Mn4þ phosphors with P-3m1 and P63mc symmetries, respec-tively. CRI values of 86 and 89 are acquired under a drive current of 15 mA for the twophases with different symmetries. These results show that the two K2GeF6:Mn4þ redphosphors exhibit great potential for commercial applications.

Except for the Mn4þ doped fluoride phosphors, some Mn4þ doped aluminate phos-phors for wLEDs have been reported. Recently, Wang et al. reported a Mn4þ-dopedbarium magnesium aluminate, BaMgAl10�2xO17:xMn4þ,xMg2þ (BMA,x ¼ 0.005e0.050).91 In consideration of the aluminate hosts like CaAl12O19,

88

Sr4Al14O25,150 and CaMg2Al16O27,

151 featuring [AlO6] octahedrons for Mn4þ substi-tution, in consideration of the Mn4þ moderate nephelauxetic effect and a peak emis-sion wavelength close to 650 nm as a consequence. Also, by introducingMn4þeMg2þ to instead Al3þeAl3þ in the lattice leads to the reduction of adverse non-radiative ET between Mn4þ ions due to the formation of Mn4þeMg2þ pairs replacethe Mn4þeMn4þeO2� clusters.152,153 Just as expected, under the excitation of blue


light, the synthesized BMA:xMn4þ,xMg2þ phosphors exhibit bright red luminescencepeaking at about 660 nm with a high color purity approaching 100% and a narrowFWHM of w30 nm.

Fig. 5.26(a) shows the normalized PLE and PL spectra of the BMA:0.02Mn4þ,0.02Mg2þ sample, which exhibits narrow-band red emission.91

Fig. 5.26(b) shows the calculated chromaticity value (x, y) of BMA:0.02Mn4þ,0.02Mg2þ phosphor in CIE 1931 chromaticity space is (0.723, 0.277),which demonstrated the high color purity. The PLE spectrum exhibits several excita-tion bands in the range between 250 and 550 nm, which can be decomposed into fourGaussian peaks, with the maxima at 465, 351, 335, and 300 nm due to the transitions

(a)

(b)

a

b c

Ge

K

Fm2m

P63mc

P63mc

P3m1

P3m1

400 500 600 700Wavelength (nm)

Inte

nsity

(arb

. uni

t)400°C

500°C

ba

c

F

Figure 5.25 (a) Schematic of structural evolution during heat treatment of K2GeF6. (b)Luminescence spectra of the white light emitting diode using K2GeF6:Mn4þ red phosphors.70

Modified from (a and b) Wei LL, Lin CC, Wang YY, Fang MH, Jiao H, Liu RS. Photoluminescent evolution induced by structural transformation through thermal treating in the rednarrow-band phosphor K2GeF6: Mn4þ. ACS Appl Mater Interfaces 2015;7:(20):10656e9.https://doi.org/10.1021/acsami.5b02212.



200 250 300 350 400 450 500 550 600 650 700 750 800Wavelength (nm)

1.0

0.8

0.6

0.4

0.2

0.0

λ em

= 6

60 n

m

λ ex

= 46

8 nm

4 A2g

→2 T

2g

4A2g→4T1g

2 Eg→

4 A2g

Nor

mal

ized

inte

nsity

(arb

. uni

t)

Mn-O CTB BMA:Mn,Mg660 nm

FWHM = 30nm

Small overlap

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.00.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

y

x

500

490

480

470460 380

520

540

560

580

600

620

700

(0.723, 0.277)

15002000

2500300040006000

10000

∞

Tc(°K)

YAG:Ce

BMA:0.02Mn4+,0.02Mg2+

(a)

(b)

4 A2g

→4 T

2g

Figure 5.26 (a) Photoluminescence excitation (ëem ¼ 660 nm) and photoluminescent(ëex ¼ 468 nm) spectra of the BMA:0.02Mn4þ,0.02 Mg2þ sample measured at roomtemperature, where the dotted lines represent data fit based on a Gaussian function. Emissionspectrum (yellow (light gray in print versions) solid line) of commercial YAG:Ce3þ is alsoprovided for comparison. (b) Chromaticity coordinate of the BMA:0.02Mn4þ,0.02 Mg2þ inCommission Internationale de I’Eclairage (CIE) diagram.91

Modified from (a and b) Wang B, Lin H, Huang F, Xu J, Chen H, Lin ZB, Wang YS. Non-rare-earth BaMgAl10�2xO17: xMn4þ, xMg2þ: a narrow-band red phosphor for use as a high-powerwarm w-LED. Chem Mater 2016;28:3515e24. https://doi.org/10.1021/acs.chemmater.6b01303.




from the Mn4þ ground state of 4A2g to the excited ones of4T2g,

2T2g, and4T1g and the

MneO charge transfer band (CTB), respectively. We noticed that the spectral overlapbetween the excitation spectrum of BMA:Mn4þ, Mg2þ and the emission spectra ofcommercial YAG:Ce3þ is rather small; thus, reduced the possible of photon reabsorp-tion that usually occurs between the nitride red phosphor and the YAG:Ce3þ yellowphosphor. Under the excitation of 468 nm, there is one dominant band with a peakat 660 nm (15,151 cm�1) ascribed to the 2Eg /

4A2g transition of Mn4þ in the[MnO6]

8� octahedral environment. Remarkably, the FWHM value of the red emissionband is as narrow as w30 nm. Such narrow-band emitting achieves the high colorpurity (98.3%) of BMA:Mn4þ, Mg2þ and contributes to improve the LE of the radia-tion because of the emission is concentrated mainly in the visible spectral region towhich the human eye is sensitive (<700 nm).154

5.4.1.8 Tungstates/molybdates based red phosphors

As we know, Eu2þ- and Ce3þ-doped phosphors become the main stream in the field ofwLEDs phosphors. Mn4þ doped narrow band red phosphors also draw much attentionas discussed previously. Eu3þ doped red phosphors are another series of importantphosphor systems for wLEDs, especially the tungstates/molybdates based red phos-phors. As an example, Vaidyanathan et al. reported the Eu3þ-substituted tungstatesor molybdates, and their solid solutions La1.95Eu0.05W2�xMoxO9 (x ¼ 0e2).155 ThePL excitation spectra of La1.95Eu0.05W2�xMoxO9 (x ¼ 0e2) is shown inFig. 5.27(a). The spectrum contains a broad absorption band ranging from w250 to400 nm and this band is attributed to oxygen to tungsten/molybdenum CT transition.That is to say, the observed CT band is ascribed to electronic transition from valenceband to conduction band. Hexavalent molybdenum and tungsten are interchangeableattribute to the almost exactly same ionic radii (Mo6þ ¼ 0.41, 0.59 Å andW6þ ¼ 0.42, 0.6 Å for coordination number 4 and 6, respectively). With theincreasing of the concentration of molybdenum in the host lattice, the CT band isobserved shift from 360 to 430 nm. This can be associated with the effective cationelectronegativity and the covalency of the d orbitals. In general, they observeddecrease in band gap when replacing 5d ion (W6þ) with less electropositive andmore covalent 4d ion (Mo6þ). They also point out that if the high electropositiveand less covalent W6þ is substituted by Mo6þ (La1.95Eu0.05W2�xMoxO9), the absorp-tion edge will shift towards n-UV region. The CT band of Eu3þeO2� is not clearlyfound in the excitation spectra. This could be attributed to possible overlap of theCT band with that of tungstate or molybdate group. Furthermore, the characteristicEu3þ excitation lines (5L6-

7F0,7F0-

5D2, and7F1-

5D1) are also observed in the spectra.But the relative intensity of the CT band and the excitation lines 7F0-

5L6 and7F1-

5D1

decreased, whereas the intensity of 7F0-5D2 line slightly increased. This may be attrib-

uted to the different crystal structure of the host lattice. The phosphorLa1.95Eu0.05W2�xMoxO9 (x ¼ 0) crystallizes in triclinic structure, nevertheless, it crys-tallizes in cubic structure when the substitution value x ¼ 0.3. Therefore, the observedresults show clearly that the slightly distortion in the crystal structure immediatelyimpact the luminescent properties in these systems. The PL emission spectrum of


La1.95Eu0.05W2�xMoxO9 (x ¼ 0e2) is shown in Fig. 5.27(b)e(d). The spectra containsharp lines at around 615 and 592 nm which are attributed to electric dipole (ED) andmagnetic dipole (MD) transitions of Eu3þ ion, respectively. Red emission attributed toEu3þ is observed under CT excitation, which clearly indicates the ET from tungstate ormolybdate group to Eu3þ levels. Nevertheless, the relative Eu3þ emission intensity isless with CT band excitation when compared with that due to Eu3þ excitation. Putanother way, presence of absorption band ascribe to tungstate or molybdate groupin the excitation spectra of Eu3þ, when monitored for Eu3þ emission (615 nm), clearlymanifests that the energy absorbed by tungstate or molybdate group is transferredto Eu3þ levels nonradiatively.156 This indicates that the energy from host lattice toEu3þ is not competent. Generally, the Eu3þ emission lines are highly sensitive tothe correlative crystal chemical environment. ED transition is dominant whendoped-Eu3þ occupies noncentrosymmetric site in the matrixes. The high relative inten-sity of ED transition can be observed. It clearly indicates that the Eu3þ ion occupiesnoncentrosymmetric site here. All compositions show red emission under CT band(Fig. 5.27(b)), 394 nm (Fig. 5.27(c)) and 465 nm (Fig. 5.27(d)) excitation but the rela-tive emission intensity with compositions is different.

200 250 300 350 400 450 500 550Wavelength (nm)

8x104 8x103

7x103

6x103

5x103

4x103

3x103

2x103

1x103

7x104

6x104

5x104

4x104

3x104

2x104

1x104

3.5x104 4.0x104

3.5x104

3.0x104

2.5x104

2.0x104

1.5x104

1.0x104

5.0x103

0.0

3.0x104

2.5x104

0

Inte

nsity

(arb

. uni

t)

Inte

nsity

(arb

. uni

t)

Inte

nsity

(arb

. uni

t)

Inte

nsity

(arb

. uni

t)

x = 0x = 0.3x = 0.6x = 0.9x = 1.2x = 1.5x = 1.8x = 2

x = 0x = 0.3x = 0.6x = 0.9x = 1.2x = 1.5x = 1.8x = 2

C - T Band

λem = 615 nm

λexc = 394 nm λexc = 465 nm

λexc = 350 nm

2.0x104

1.5x104

1.0x104

5.0x104

0.0500 550 600 650 700 500 550 600 650 700

Wavelength (nm) Wavelength (nm)

x = 0x = 0.3x = 0.6x = 0.9x = 1.2x = 1.5x = 1.8x = 2

x = 0x = 0.3x = 0.6x = 0.9x = 1.2x = 1.5x = 1.8x = 2

5 D0 -

7 F 2

5 D0 -

7 F 1

5 D0 -

7 F 3

5 D0 -

7 F 15 D

0 - 7 F 2

5 D0 -

7 F 3

7 F 1 - 5 D

1

7 F 0 - 5 D

2

5 D0 -

7 F 2

5 D0 -

7 F 1

5 D0 -

7 F 3

7 F 0 - 5 L 6

0

500 550 600 650 700Wavelength (nm)

(a) (b)

(d)(c)

Figure 5.27 The excitation spectra of La1.95Eu0.05W2�xMoxO9(x ¼ 0e2) (a), and the emissionspectra of La1.95Eu0.05W2�xMoxO9(x ¼ 0e2) under different excitation wavelengths,350 nm (b), 394 nm (c) and 465 nm (d).155

Modified from (aed) Vaidyanathan S, Jeon DY. A novel narrow band red-emitting phosphor forwhite light emitting diodes. Int J Appl Ceram Technol 2009;6:453e8. https://doi.org/10.1111/j.1744-7402.2009.02371.x.


https://doi.org/10.1111/j.1744-7402.2009.02371.x

https://doi.org/10.1111/j.1744-7402.2009.02371.x

5.4.2 New LEDs phosphors by codoped activators and theirenergy transfer

In Section 5.4.1, we focused on the types of the different hosts for the application inwLEDs, and the tunable PL can be realized by the modifications of the chemical com-positions and the crystal structures. As we know, activators also play important roles inthe PL tuning and the discovery of new LEDs phosphors with tunable emission for theapplication of wLEDs. Herein, we will first introduce some typical activators for thedifferent emission in the application of LEDs phosphors, as demonstrated in Table 5.1,and we have summarized some typical activators, the typical phosphors and theiremission colors. However, the ET is an efficient way to adjust the emission colorsof the codoped activators and to develop the new LEDs phosphors. In this section,we mainly discuss some examples of the typical LEDs phosphors by codoped activa-tors and their ET.

As is known to all, Ce3þ and Eu2þ can emit broad bands with different colors invarious hosts based on their 5de4f allowed transitions, because it contains the 5delectrons unshielded from the crystal field when they are located in the excited state.Therefore, the tunable emission color can be realized in Ce3þ or Eu2þ singly dopedphosphors by ion replacement methods.3,9 However, Ce3þ or Eu2þ can also serve asgood sensitizers to transfer their energy to activators such as Tb3þ, Eu3þ, Sm3þ,Mn2þ and Dy3þ to generate abundant and tunable emission colors by adjusting dopedion concentrations.10 For example, Fig. 5.28(a) demonstrated the CIE chromaticity co-ordinate diagram for Sr3Gd2(Si3O9)2:Ce

3þ, Tb3þ and Sr3Gd2(Si3O9)2:Ce3þ, Mn2þ

phosphors, respectively, and the PL tuning can be clearly found.

5.4.2.1 ET models with Ce3þ as sensitizers

Ce3þ often generates n-UV to green emission under UV/n-UV excitation in manyhosts owing to its 4fe5d spin-allowed transition.9,10 Because its emission spectra over-laps with the excitation spectra of Mn2þ, Tb3þ, Dy3þ or Eu2þ, it is possible to producethe ET from Ce3þ to these ions when they are co-doped into proper hosts.130,181e214

Therefore, tunable emission color can be reached in these systems.

Ce3þeMn2þ system 5.1

Mn2þ as luminescent center can emit different colors, including the green (tetrahe-dron field) and red (octahedron field) emission, depending on the different crystalfields around Mn2þ. Meanwhile, its excitation bands often locate in blue and greenareas, which may overlap with the blue and green emission spectra of Ce3þ. So theET from Ce3þ to Mn2þ may take place and generate tunable colors under appropriateexcitation. Zhou et al.181 prepared a blue to yellow/orange color emitting NaAlSiO4:Ce3þ, Mn2þ phosphor based on two main emission bands centered at 430 and590 nm and ET properties from the Ce3þ to Mn2þ ions. The 430 nm emission bandfrom Ce3þ can be decomposed into four bands peaking at 404, 438, 470 and519 nm, because Ce3þ can occupy two kinds of Naþ sites and its lowest 5d excitedstate can transfer to the 2F5/2 and

2F7/2 spin orbit 4f ground states. Fig. 5.28(b) presents


Table 5.1 Some typical singly doped activators and the designedphosphors for the white light-emitting diodes applications

Doped ionscategory

Main energylevelstransition Representative samples

Emissioncolors References

Eu2þ 4f65d1e4f7 Ca3Mg3(PO4)4:Eu2þ Blue 157

Ca3(PO4)2:Eu2þ Cyan 158

CaZrSi2O7:Eu2þ Green 159

Ca3Si2O7:Eu2þ Yellow 160

a-SrNCN:Eu2þ Orange 161

Ce3þ 5d1e4f1 Sr5(PO4)2(SiO4):Ce3þ Blue 162

Ca2GdZr2(AlO4)3:Ce3þ Green 163

(La,Gd)Sr2(Al,B)O5:Ce3þ Yellow 164

Y3Al5�xSixO12�xNx:Ce3þ Orange 165

Bi3þ Ca12Al14O32Cl2:xBi3þ Cyan 166

Ca3Al2O6:Bi3þ Green 167

LuVO4:Bi3þ Yellow 168

ScVO4:Bi3þ Orange 169

Mn2þ 4T1e6A1 NaZnPO4:Mn2þ Green 170

CdSiO3:Mn2þ Yellowishorange

171

Na2CaMg(PO4)2:Mn2þ Red 172

Tb3þ 5D4e7FJ (J ¼ 6,

5, 4, 3)CaMoO4:Tb

3þ Green 173

Dy3þ 4F9/2e6H15/2 Sr3RE2(BO3)4:Dy

3þ

(RE ¼ Y, La, Gd)Cyan 174

4F9/2e6H13/2 (Sr0.85Mg0.14)3(P1

�xSixO4)2:Dy3þ

White 175

Y2(MoO4)3:Dy3þ Yellow 176

Sm3þ 4G5/2e6H5/2,7/

2,9/2

SrMoO4:Sm3þ Red 177

Eu3þ 5D0e7FJ (J ¼ 0,

1, 2, 3, 4)Ba2CaWO6:Eu

3þ Orangered

178

RbZnPO4:Eu3þ Red 179


0.00.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

500

490

480

470460

y

520

540

560

580

600620

700

380

x

e2

g2g1

Ce3+:5dMn2+:4T1E

nergy

Energy

e1

Ce3+:2F7/2, 2F5/2Mn2+:6A1

30,000

25,000

20,000

20,000

10,000

5000

0Ce3+

2F7/2

4T2

4T24T1

4A1, 4E

2F5/2

350

nm

470

nm51

9 nm

438

nm40

4 nm

5d

Dipole-quadrupoleinteraction

590

nm

Mn2+

6A1

Ene

rgy

(cm

-1)

Blue light

ΔE2ΔE4{

UV light

M

CB

P

A

1

6

4

3

5

D

ΔE3

ΔE1

2

0.01Ce3+

0.01Ce3+,0.02Mn2+

0.01Ce3+,0.04Mn2+

0.01Ce3+,0.06Mn2+

0.01Ce3+,0.08Mn2+

0.01Ce3+,0.1Mn2+

0.01Ce3+,0.12Mn2+

0.01Ce3+,0.14Mn2+

0.01Ce3+,0.16Mn2+

0.08Mn2+

2000

30002500

15006000

10000

Tc(°K)

Tb3+

Mn3+

(a) (b)

(d)

(c)

Figure 5.28 (a) Commission Internationale de I’Eclairage (CIE) chromaticity coordinatediagram for Sr3Gd2(Si3O9)2:0.26Ce

3þ,yTb3þ and Sr3Gd2(Si3O9)2:Ce3þ,zMn2þ phosphors. (b)

Illustration of the energy transfer model for Ce3þe Mn2þ in the NaAlSiO4 host. (c)Commission International CIE chromaticity coordinates of the samplesSr7La3[(PO4)2.5(SiO4)3(BO4)0.5](BO2):0.01Ce

3þ, Sr7La3[(PO4)2.5(SiO4)3(BO4)0.5](BO2):0.08Mn2þ and Sr7La3[(PO4)2.5(SiO4)3(BO4)0.5](BO2):0.01Ce

3þ,xMn2þ. (d) Configurationalcoordinate diagram of the ground states of Ce3þ and Mn2þ and the excited states of Ce3þ andMn2þ.180e182

Modified from (a) Zhu Y, Liang Y, Zhang M, Tong M, Li G, Wang S. Structure, luminescenceproperties and energy transfer behavior of color-adjustable Sr3Gd2(Si3O9)2:Ce

3þ, Tb3þ/Mn2þ

phosphors. RSC Adv 2015;5:98350e60. https://doi.org/10.1039/c5ra20756h. (b) Zhou J, WangT, Yu X, Zhou D, Qiu J. The synthesis and photoluminescence of a single-phased white-emittingNaAlSiO4:Ce

3þ,Mn2þ, phosphor for WLEDs. Mater Res Bull 2016;73:1e5. https://doi.org/10.1016/j.materresbull.2015.08.006. (c and d) Ci Z, Sun Q, Sun M, Jiang X, Qin S, Wang Y.Structure, photoluminescence and thermal properties of Ce3þ, Mn2þ co-doped phosphosilicateSr7La3[(PO4)2.5(SiO4)3(BO4)0.5](BO2) emission-tunable phosphor. J Mater Chem C 2014;2:5850e6. https://doi.org/10.1039/c4tc00217b.


https://doi.org/10.1039/c5ra20756h

https://doi.org/10.1016/j.materresbull.2015.08.006


https://doi.org/10.1039/c4tc00217b

the ET model for Ce3þeMn2þ in a NaAlSiO4 host as the reference for the Ce3þeMn2þ system. Ce3þ absorbed UV light from the ground state to the 5d excited state,and then efficiently transferred the energy to the 4T2 level of Mn2þ. The excited freeelectron of Mn2þ then relaxed to the excited state of 4T1 (4G) through the 4E (4D),4T2 (4D), (4E, 4A1) (4G) and

4T2 (4G) intermediate energy levels by the process ofnonradiation. Then the characteristic optical transition 4T1e

6A1 of Mn2þ can be real-ized, which exhibited the characteristic emission of Mn2þ. Ci et al.182 obtained Ce3þ,Mn2þ codoped Sr7La3[(PO4)2.5(SiO4)3(BO4)0.5](BO2) phosphors with apatite crystalstructure (P63/m), and its emission color can be tuned from blue to pink withincreasing Mn2þ concentration (Fig. 5.28(c)). Moreover, they investigated the PLthermal quenching properties of Ce3þ and Mn2þ singly doped and Ce3þ, Mn2þ

codoped samples. It was found that a decline in emission intensity in singly Ce3þ

and Mn2þ doped Sr7La3[(PO4)2.5(SiO4)3(BO4)0.5](BO2) was lower than that of Ce3þ

and Mn2þ codoped samples. Novel Na4Ca4Si6O18:Ce3þ, Mn2þ183 and NaCa2Lu-

Si2O7F2:Ce3þ, Mn2þ Phosphors184 were also observed to emit blue to pink color under

UV excitation.As shown in Fig. 5.29(a), Ca4Si2O7F2:0.04Ce

3þ,yMn2þ samples was found torealize the color control from cyan to yellow under 365 nm excitation as y valuechanged.185 Jiao et al.186 and Li et al.187 prepared two novel Ce3þ, Liþ, Mn2þ codopedCa2SrAl2O6 and CaSr2Al2O6 phosphors with identical crystal structure and similar PLproperties were observed. The emission color was tuned from cyan to light orange pinkincluding white with increasing Mn2þ concentration (Fig. 5.29(b) and (c)). Moreover,Ce3þ and Mn2þ codoped Mg1.5Lu1.5Al3.5Si1.5O12 with garnet crystal structure canemit green to yellow color under 430 nm excitation by varying concentrations ofCe3þ and Mn2þ (Fig. 5.29(d)).188

Except for this kind of PL tuning, an attractive tunable full-color-emitting Ca3Sc2-Si3O12:Ce

3þ, Mn2þ phosphor was reported by Liu et al.189 They found that Mn2þ willoccupy not only the Ca2þ but also Sc3þ sites because there was only one site for bothCa2þ and Sc3þ, respectively, whereas two obvious emission peaks occurred in Mn2þ

singly doped Ca3Sc2Si3O12. Further, the emission intensities of the two Mn2þ ionswere enhanced and varied depending on Mn2þ concentration, which originates fromthe different ET efficiencies from Ce3þ to them in Ca3Sc2Si3O12:Ce

3þ, Mn2þ, aswell as compensation for the negative charge of Mn2þ substitution for Sc3þ in theform of Ce3þ substitution for Ca2þ in the presence of Ce3þ in Ca3Sc2eSi3O12:Ce3þ, Mn2þ. Then, they added rare earth elements Y, La, Gd, and Lu into the Ca3Sc2-Si3O12:Ce

3þ, Mn2þ phosphor190 and investigated the effects of them on the PL prop-erties in detail. It was found that the smaller the ionic radius, the larger the relativeintensity of Mn2þ. The reason is that the smaller the ionic radius is, the easier it isfor the Ln3þ at the Ca2þ site to act as a charge compensator, resulting in increasingMn2þ substitution at the Sc3þ site. It is interesting that the excitation band of theas-prepared samples ranged from 200 to 500 nm centered at 450 nm, which matchedwith commercial blue InGaN chips.

Ce3þeTb3þ system 5.2


0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

1 2 3 4 5 6

y

CSF:0.04Ce3+,yMn2+

1.y = 02.y = 0.043.y = 0.084.y = 0.125.y = 0.166.y = 0.20

1 23456

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8X

Y

1. n = 0.00

2. n = 0.01

3. n = 0.02

4. n = 0.05

5. n = 0.09

6. n = 0.13

1. y = 02. y = 0.013. y = 0.024. y = 0.035. y = 0.046. y = 0.057. y = 0.07

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

X

C.I.E 1931ChromaticityDiagram

a b c d ef g

(a) (b)

(c) (d)x

Y

Y

Figure 5.29 Commission Internationale de I’Eclairage (CIE) chromaticity diagram ofCa4Si2O7F2:0.04Ce

3þ,yMn2þ phosphors under 365 nm excitation (a), Ca2SrAl2O6:0.01Ce3þ,0.01Liþ,nMn2þ with the excitation wavelength of 355 nm (b), CaSr2Al2O6:0.03Ce3þ, 0.03Liþ,yMn2þ under 358 nm excitation (c) and Mg1.5Lu1.5Al3.5Si1.5O12:0.05Ce3þ,xMn2þ excited at 430 nm (d) The inset in (b) shows the corresponding photographsof the samples under a 365 nm UV lamp.185e188

Modified from (a) Lv W, Luo Y, Hao Z, Zhang X, Wang X, Zhang J. A new dual-emissionphosphor Ca4Si2O7F2:Ce

3þ,Mn2þ with energy transfer for near-UV LEDs. Mater Lett 2012;77:45e7. https://doi.org/10.1016/j.matlet.2012.02.095. (b) Jiao M, Jia Y, L€u W, Lv W, Zhao Q,Shao B, You H. A single-phase white-emitting Ca2SrAl2O6:Ce

3þ,Liþ,Mn2þ phosphor withenergy transfer for UV-excited WLEDs. Dalton Trans 2014;43:3202e9. https://doi.org/10.1039/c3dt52832d. (c) Li Y, Shi Y, Zhu G, Wu Q, Li H, Wang X, Wang Q, Wang Y. A single-component white-emitting CaSr2Al2O6:Ce

3þ,Liþ,Mn2þ phosphor via energy transfer. InorgChem 2014;53:7668e75. https://doi.org/10.1021/ic500963q. (d) Jia Y, Huang Y, Guo N, QiaoH, Zheng Y, Lv W, Zhao Q, You H. Mg1.5Lu1.5Al3.5Si1.5O12: Ce

3þ,Mn2þ: a novel garnetphosphor with adjustable emission color for blue light-emitting diodes. RSC Adv 2012;2:2678e81. https://doi.org/10.1039/c2ra00894g.

https://doi.org/10.1016/j.matlet.2012.02.095



https://doi.org/10.1021/ic500963q

https://doi.org/10.1039/c2ra00894g

Generally, Tb3þ acts as an efficient green-emitting activator based on its character-istic 5D4e

7FJ (J ¼ 6, 5, 4, 3) transition. Because the transition belongs to the 4fe4fspin-forbidden transition, the absorption spectrum and emission band are ratherweak and their widths are narrow. Therefore, it is desirable to enhance the emissionintensity via the ET effect. Ce3þ is an effective sensitizer for Tb3þ when Ce3þ emis-sion locates at the ultraviolet to cyan area to overlap Tb3þ excitation. Because theemission of Tb3þ locates at the green region, the emission color of Ce3þ, Tb3þ

codoped samples could be tuned from n-UV/blue/cyan to green with different dopingconcentrations based on the ET from Ce3þ to Tb3þ ions under UV/n-UV excitation.

In this series of phosphors, including the YBa3B9O18:Ce3þ, Tb3þ,191 KCaY(PO4)2:

Ce3þ, Tb3þ,192 Na2Gd2B2O7:Ce3þ, Tb3þ193, and NaCaPO4:Ce

3þ, Tb3þ,194 PL tuningfrom the ultraviolet to green color can be observed with increasing Tb3þ concentra-tions with fixed Ce3þ concentration. Taking Na2Gd2B2O7:Ce

3þ, Tb3þ as an example,we can see the variations of the PL spectra (lex ¼ 358 nm) of Na2Gd2B2O7:0.05Ce3þ,nTb3þ (n ¼ 0, 0.06, 0.12, 0.18, 0.25) with different Tb3þ concentrationsand their corresponding intensities in Fig. 5.30(a). The CIE diagram of these samplesexcited at 358 nm and digital PL photos under a 365 nm UV lamp excitation inFig. 5.30(b) illustrate the color variation with Tb3þ concentration change.193

Apatite structure phosphors have been demonstrated to be good hosts for rare earthion doping,195e200 such as La6(Sr/Ba)4(SiO4)6F2:Ce

3þ, Tb3þ, Ca4Y6(SiO4)6O:Ce3þ,

Tb3þ, MgGd4Si3O13:Ce3þ, Tb3þ, Sr2La8(SiO4)6O2:Ce

3þ, Tb3þ, and so on, whichwere reported to show good ET properties from Ce3þ to Tb3þ ions generating tunableemission color from blue to green under UV excitation. Our group201 also reported the

400 450 500 550 600 650 700

Wavelength (nm)

Inte

nsity

(arb

. uni

t)

λex = 358nm(1) n = 0.00(2) n = 0.06(3) n = 0.12(4) n = 0.18(5) n = 0.25

0.00 0.05 0.10 0.15 0.20 0.25

Contents of Tb3+ (n)

Ce3+

Tb3+

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8X

Y

Integrated intensity (a.u.)

1. n = 02. n = 0.063. n = 0.124. n = 0.185. n = 0.25

543

2

1

(a) (b)

1

2

345

5

4

3

2

1

Figure 5.30 (a) Photoluminescent spectra (lex ¼ 358 nm) of Na2Gd2B2O7:0.05Ce3þ,nTb3þ

(n ¼ 0, 0.06, 0.12, 0.18, 0.25) with corresponding Commission Internationale de I’Eclairage(CIE) diagram (b). Inset in (a) is the variation of integrated emission intensity of Ce3þ andTb3þ with different concentration of Tb3þ.193

Modified from (a and b) Guo C, Jing H, Li T. Green-emitting phosphor Na2Gd2B2O7:Ce3þ,Tb3þ

for near-UV LEDs. RSC Adv 2012;2:2119e22. https://doi.org/10.1039/c2ra00808d.


https://doi.org/10.1039/c2ra00808d

Ce3þ singly doped Y2Si2O7N2 phosphor, and its emission peak with a broad asym-metric band shifted from 438 to 500 nm under 365 nm excitation with increasingCe3þ doped concentration from 0.005 to 0.20 (Fig. 5.31(a)). The phenomenon wasexplained by the ET between the optically active Ce3þ ions in four Y3þ crystallo-graphic sites, as shown in Fig. 5.31(b), leading to concentration quenching. In addi-tion, the red shift of the Ce3þ emission correlated very well with the latticeexpansion. Therefore, tunable emission from cyan to green can be observed withincreasing Tb3þ concentration y in Y4Si2O7N2:0.005Ce

3þ, yTb3þ based on ET fromCe3þ to Tb3þ, as depicted in Fig. 5.31(c). The PL intensities decreased to 72.4% ofthe initial intensity for the Y4Si2O7N2:Ce

3þ,Tb3þ phosphor under 365 nm excitation,

400 450 500 550 600 650

Inte

grat

ed in

tens

ity

Inte

nsity

(arb

. uni

t) λex = 365 nm

25°C50°C100°C150°C200°C250°C

Wavelength (nm)

400 450 500 550 600 650Wavelength (nm)

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

(0.20,0.20)

(0.30,0.51)

x = 0.005x = 0.01x = 0.04x = 0.07x = 0.10x = 0.20

Inte

nsity

(arb

. uni

t)

a

b c

Y2Y3

Y4Y1

Y/CeSiNO/NO

(a)

(c)

(b)

(d)

λex = 365 nm

Figure 5.31 (a) Photoluminescent (PL; lex ¼ 365 nm) spectra of Y4Si2O7N2:xCe3þ with

varying Ce3þ concentrations. (b) The crystal structure description of Y4Si2O7N2 showing theSi(O,N)4 polyhedron and the coordination environment of the Y atoms with four differentcrystallographic sites. (c) A representation of the Commission Internationale de I’Eclairage(CIE) chromaticity coordinates for the Y4Si2O7N2:Ce

3þ,Tb3þ phosphors. (d) The temperature-dependent PL spectra of the Y4Si2O7N2:Ce

3þ,Tb3þ phosphor.201

Modified from (aed) Xia Z, WuW. Preparation and luminescence properties of Ce3þ and Ce3þ/Tb3þ-activated Y4Si2O7N2 phosphors. Dalton Trans 2013;42:12989e97. https://doi.org/10.1039/c3dt51470f.


https://doi.org/10.1039/c3dt51470f

https://doi.org/10.1039/c3dt51470f

corresponding to a temperature of 150�C (Fig. 5.31(d)), which indicated its good ther-

mal stability.

Ce3þeDy3þ system 5.3

ET originated from Ce3þ to Dy3þ are less than those of Ce3þeMn2þ or Ce3þeTb3þ systems. Li et al.202 prepared Ce3þ, Dy3þ codoped borate Sr3Y2(BO3)4 phos-phors. The emission spectrum of Ce3þ had a wide band centered at 420 nm under340 nm excitation, as shown in Fig. 5.32(a1), the excitation spectrum of Dy3þ moni-tored at 576 nm consisted of many excitation lines between 250 and 500 nm, exhibitedin Fig. 5.32(a2). The spectral overlap of Ce3þ emission and Dy3þ excitation impliedpossible ET from Ce3þ to Dy3þ ions. The similar excitation spectra monitored at420 nm of Ce3þ and 576 nm of Dy3þ in Fig. 5.32(a3) was used to prove the ETfrom Ce3þ to Dy3þ ions. The ET modes from Ce3þ to Dy3þ in single-phase white-emitting Ca20Al26Mg3Si3O68:Ce

3þ, Dy3þ in Fig. 5.32(b) was used to understand theET process from Ce3þ to Dy3þ ions.203 First, the electrons in the Ce3þ ground stateabsorbed the excited energy and jumped to the 5d excited state, some of them pro-duced the 5de4f transition, some of them in the lowest excited state transferred theirenergy to Dy3þ 4I13/2 þ 4F7/2 energy levels, resulting in Dy3þ 4F9/2-

6H15/2 and 4F9/

2-6H13/2 after the nonradiate transition from 4I13/2 þ 4F7/2 to

4F9/2 energy level. OtherCe3þeDy3þ ET systems such as Zn2P2O7:Ce

3þ, Dy3þ,204 KNaCa(PO4)2:Ce3þ,

200

100

08

4

120

60

0

0300 400 500 600 700

Wavelength (nm)

λex = 340nm

λex = 350nm

λex = 340nmλem = 420 nm

λem = 576 nm

λem = 420 nm

λem = 576 nm

Inte

nsity

(arb

. uni

t)

5d

ET

2F7/22F5/2

Ce3+

4K15/2

6P7/24I11/24G11/24I15/24F9/2

6F1/2

6H11/26H13/26H15/2

4I9/2 + 4G9/2

4I13/2 + 4F7/2

6F15/2 + 6H9/2

577

nm

475

nm

Dy3+

(1)

(2)

(3)

(a) (b)

Figure 5.32 (a) Photoluminescent (PL) and photoluminescence excitation spectra of (1)Sr3Y2(BO3)4:0.02Ce

3þ, (2) Sr3Y2(BO3)4:0.10Dy3þ and (3) Sr3Y2(BO3)4:0.02Ce

3þ,0.10Dy3þ

samples; insets are luminescent photos of the corresponding samples under 365 nm xenonlamp excitation.202 (b) Illustration of the energy transfer (ET) modes of Ce3þeDy3þ inCa20Al26Mg3Si3O68:0.08 Ce3þ, zDy3þ.203

Modified from (a) Li K, Chen D, Zhang R, Yu Y, Xu J, Wang Y. Enhanced luminescence inCe3þ/Dy3þ:Sr3Y2(BO3)4 phosphors via energy transfer. Mater Res Bull 2013;48:1957e60.https://doi.org/10.1016/j.materresbull.2013.01.042. (b) Yuan B, Song Y, Sheng Y, Zheng K,Zhou X, Ma P, Xu X, Zou H. Tunable color and energy transfer in single-phase white-emittingCa20Al26Mg3Si3O68:Ce

3þ,Dy3þ phosphors for UV white light-emitting diodes. J Solid StateChem 2015;232:169e77. https://doi.org/10.1016/j.jssc.2015.09.015.



https://doi.org/10.1016/j.jssc.2015.09.015

Dy3þ,205 GdOBr:Ce3þ, Dy3þ206 and Ca6La2Na2(PO4)6F2:Ce3þ, Dy3þ207 can also be

referenced with ET properties from Ce3þ to Dy3þ ions.

Ce3þeEu2þ system 5.4

Generally, Eu2þ acts as an activator in Ce3þ, Eu2þ codoped systems, whereas Ce3þ

acts as a sensitizer. Both of them are sensitive to the nearby crystal field. Therefore, theemission color can be realized from ultraviolet to blue and green, blue to green, yellow,and orange based on the ET from Ce3þ to Eu2þ ions in the subsequently discussedsystems.130,208e214 For example, Zhou et al.208 prepared Ce3þ, Eu2þ codoped chloridephosphate Sr5(PO4)3Cl with apatite structure and investigated the PL and cathodolu-minescence (CL) properties of the as-prepared samples. They found the emission spec-trum of Ce3þ was located in the ultraviolet area, which overlapped with the excitationspectrum of Eu2þ in this host. Therefore, Ce3þ may absorb the energy and transfer partof it to Eu2þ to enhance the Eu2þ emission. The emission intensity and fluorescentdecay lifetimes of Ce3þ monotonously decreased with increasing Eu2þ concentrationin Sr5(PO4)3Cl:Ce

3þ, Eu2þ phosphors, which verified the ET from Ce3þ to Eu2þ ions.When the emission band of Ce3þ located at the blue area, tunable emission color

from blue to green occurred in Ca7Mg(SiO4)4:Ce3þ, Eu2þ,210 Sr2Al2SiO7:

Ce3þ,Eu2þ,211 CaSrAl2SiO7:Ce3þ, Eu2þ,212 and Ba4Si6O16:Ce

3þ, Eu2þ 213 phosphorsbased on the ET from Ce3þ to Eu2þ. Fig. 5.33 shows the PL and corresponding CIEchromatic coordinates diagram of Sr2Al2SiO7:Ce

3þ, Eu2þ. The Ce3þ emission bandaround 415 nm decreased and the Eu2þ emission band around 510 nm increased

400 450 500 550 600 650Wavelength (nm)

(a) x = 0.02, y = 0

(b) x = 0.02, y = 0.001

(c) x = 0.02, y = 0.005

(d) x = 0.02, y = 0.01

(e) x = 0.02, y = 0.02(f) x = 0.02, y = 0.03

(g) x = 0.02, y = 0.04

(h) x = 0, y = 0.03

Inte

nsity

(arb

. uni

t)

a fg

ed

hb

c

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

654

3

2

1

(b)(a)

Figure 5.33 (a) Photoluminescent emission spectra and corresponding CommissionInternationale de I’Eclairage (CIE) chromaticity diagram (b) for Sr2-m-nAl2SiO7:mCe3þ,nEu2þ phosphors excited at 345 nm. [(1e4 and 6) m ¼ 0.02, n ¼ 0, 0.001, 0.005, 0.02 and0.03, (5) m ¼ 0, n ¼ 0.03.]211

Modified from Li G, Li M, Li L, Yu H, Zou H, Zou L, Gan S, Xu X. Luminescent properties ofSr2Al2SiO7:Ce

3þ, Eu2þ phosphors for near UV-excited white light-emitting diodes. Mater Lett2011;65:3418e20. https://doi.org/10.1016/j.matlet.2011.07.050.



with increasing Eu2þ concentration (Fig. 5.33(a)), which resulted in a shift in the CIEchromatic coordinates from blue to green (Fig. 5.33(b)).

Besides green, yellow and orange colors can also be obtained in some Ce3þeEu2þ

co-doped phosphors. Sr3Al2O5Cl2:Ce3þ, Eu2þ was investigated by Song et al.214

Fig. 5.34(a) shows the excitation and emission spectra of the Sr3Al2O5Cl2:

200 300 400 500 600 700Wavelength(nm)

λem = 444 nm

λem = 609nm

λex = 330nm

n = 0n = 0.0025n = 0.005n = 0.0075n = 0.01

Inte

nsity

(arb

. uni

t)

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8X

y

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.8

12

34 5 6 7

1

4

7

1

23 4

5678

450 500 550 600 650 700 750Wavelength (nm)

Ca4(PO4)2O:0.02Ce3+,0.12Eu2+

CIE: (0.376,0.346)

CCT: 3848 K

Inte

nsity

(a.u

.)

(a)

(c)

(d)

(b)

Figure 5.34 (a) Excitation and emission spectra of Sr3Al2O5Cl2:0.01Ce3þ,nEu2þ phosphors.

The inset is the excitation spectra monitored at 609 nm. (b) Commission Internationale deI’Eclairage (CIE) chromaticity diagram for Sr3Al2O5Cl2:m Ce3þ,nEu2þ excited at 330 nm. (c)CIE chromaticity diagram for Ca4(PO4)2O:0.02 Ce3þ,yEu2þ phosphors and inset shows thephosphor images excited at 380 nm. (d) Electroluminescence spectrum of Ca4(PO4)2O:0.02Ce3þ,0.012Eu2þ phosphor-based wLED under a current of 700 mA. The inset shows aphotograph of the LED package.130,214

Modified from (a and b) Song Y, Jia G, Yang M, Huang Y, You H, Zhang H. Sr3 Al2O5Cl2:Ce3þ,Eu2þ: a potential tunable yellow-to-white-emitting phosphor for ultraviolet light emittingdiodes. Appl Phys Lett 2009;94:091902e091902e3. https://doi.org/https://doi.org/10.1063/1.3094753. (c and d) Jia Y, Pang R, Li H, Sun W, Fu J, Jiang L, Zhang S, Su Q, Li C, Liu R-S.Single-phased white-light-emitting Ca4(PO4)2O:Ce

3þ,Eu2þ phosphors based on energy transfer.Dalton Trans 2015;44:11399e407. https://doi.org/10.1039/c5dt01018g.


https://doi.org/https://doi.org/10.1063/1.3094753

https://doi.org/https://doi.org/10.1063/1.3094753

https://doi.org/10.1039/c5dt01018g

0.01Ce3þ, nEu2þ phosphors. Similar excitation spectra monitored at 444 nm (Ce3þ)and 609 nm (Eu2þ) along with a decline in intensity of the Ce3þ emission band andincrease in Eu2þ emission intensity with increasing Eu2þ content illustrated the ETfrom Ce3þ to Eu2þ, resulting in the tunable emission color from blue to yellowincluding white corresponding to the CIE chromatic coordinate diagram inFig. 5.34(b). Moreover, the blueeorange phosphor Ca4(PO4)2O:Ce

3þ, Eu2þ canbe seen in Fig. 5.34(c). Ce3þ emitted a broad band centered at 460 nm under380 nm excitation and Eu2þ emitted a broad band around 650 nm under 460 nm exci-tation. Therefore, tunable emission from blue to orange via adjusting the doped ionconcentrations corresponded to the CIE chromatic coordinate diagram inFig. 5.34(c). The wLED lamp package emitted intense warm white light with aCCT of 4124K, a color coordinate of (0.359, 0.310), and Ra of 84 (Fig. 5.42(d)),which illustrated its potential as a single-composition white-emitting phosphor forwLED applications.130

Ce3þeTb3þeMn2þ system 5.5

Because Ce3þeTb3þ systems often produce tunable emission color from blue orcyan to green under UV/n-UV excitation, the supplement of a red component in themis necessary if white light is to be obtained. It is well known that Mn2þ can emit or-ange/red color when it is located in the octahedron field. Thus codoping Mn2þ inCe3þ, Tb3þ codoped systems is possible to obtain white light emission. For example,the PL properties of Ce3þ, Tb3þ, Mn2þ codoped oxyapatite Ca4Y6(SiO4)6O

215 andSr3.5Y6.5O2(PO4)1.5(SiO4)4.5

216 have been reported. Taking Ca4Y6(SiO4)6O:Ce3þ,

Tb3þ, Mn2þ as the example, we can see the emission spectra consisted of Ce3þ,Tb3þ and Mn2þ bands centered at the blue, green and red regions under 284 and358 nm excitation (Fig. 5.35(a) and (b)), therefore, it can generate white emissionwith appropriate doping contents of Ce3þ, Tb3þ and Mn2þ. In this condition, theET process from Ce3þ to both Tb3þ and Mn2þ ions is displayed with the correspond-ing energy level diagram in Fig. 5.35(c). Electrons in the Ce3þ 4f ground stateabsorbed excited energy to jump to the 5d excited state, then they nonradiativelyrelaxed to the lowest 5d excited state, some of them produced the 5de4f transition,and some transferred their energy to the Tb3þ 5D3 and Mn2þ 3d energy levels, result-ing in the 5D3e

7FJ (J ¼ 6, 5, 4, 3, 2) and 5D4e7FJ (J ¼ 6, 5, 4, 3) transitions of Tb3þ

and 4T1e6A1 transition of Mn2þ. The CL emission color in Fig. 5.35(d) appears to be

blue to green and red in the Ca4Y6(SiO4)6O:Ce3þ, Tb3þ and Ca4Y6(SiO4)6O:Ce

3þ,Mn2þ phosphors, respectively. And white light was produced after codoping Ce3þ,Tb3þ, Mn2þ into Ca4Y6(SiO4)6O, the best CIE chromatic coordinate reached(0.328, 0.331) for the Ca4Y6(SiO4)6O:10 mol%Ce3þ, 3 mol%Mn2þ, 2 mol%Tb3þ

sample. Moreover, Lin’s group also synthesized the Ca9Bi(PO4)7:Ce3þ, Tb3þ,

Mn2þ217 and Ca5(PO4)2SiO4:Ce3þ/Tb3þ/Mn2þ218 phosphors. Results showed that

white light can be generated under UV excitation. Good thermal quenching propertiesof these two samples indicated them to be good candidate phosphors for LEDs. Othersystems, such as NaCaBO3:Ce

3þ, Tb3þ, Mn2þ,139 Ca3Al2O6:Ce3þ, Tb3þ, Mn2þ,219

Ca9MgNa(PO4)7:Ce3þ/Tb3þ/Mn2þ220 and BaMg2(PO4)2:Ce

3þ, Tb3þ, Mn2þ221


combined the Ce3þ blue emission, Tb3þ green emission and Mn2þ red emission toacquire white light based on the ET from Ce3þ to Tb3þ and Mn2þ in a single host.

Ce3þeTb3þeEu3þ system 5.6

In some systems, Eu3þmay not be reduced to Eu2þ in a reductive atmosphere in thehost. However, the ET from Ce3þ to Eu3þ is not considered to effectively proceed

2500

2000

1500

1000

500

350 400 450 500 550 600 650 7000

Wavelength (nm)

green

green

red

red

λex = 284 nm

y z421114

3332

λex = 358 nm

y z421114

3332

blue

blue

PL

inte

nsity

(arb

. uni

t)

350 400 450 500 550 600 650 700Wavelength (nm)

1200

1000

800

600

400

200

PL

inte

nsity

(arb

. uni

t)

Energy transfer

UV

or e

letro

n be

amE

xcita

tion

sour

ce

23456

Tb3+

Tb3+

Ce3+

Tb3+

Mn2+

Mn2+

6A1

4A1

4T14E

4T1

5D3

5D4

4T2

4T,

4T2

7F2F7/2

2F5/2

3d

White light

350 400 450 500 550 600 650 700Wavelength (nm)

CL

inte

nsity

(a.u

)

Tb3+

Tb3+

Mn2+

Mn2+

Ce3+

Ce3+ Ce

Ce Mn

Ca Y (SiO ) O: Ce , Mn , Tb

5d

(a) (c)

(b) (d)

J

Figure 5.35 Photoluminescent spectra of Ca4Y6(SiO4)6O:10 mol%Ce3þ,y mol%Mn2þ,zmol%Tb3þ samples: (y ¼ 2, 3, 4; z ¼ 1, 2, 4) under different wavelengths of ultravioletexcitation: (a) 284 nm and (b) 358 nm. (c) Illustration of the energy transfer modes of Ce3þ -Tb3þ/Mn2þ and luminescent photographs of Ca4Y6(SiO4)6O:10 mol%Ce3þ,y mol%Mn2þ,zmol%Tb3þ samples: y ¼ 0, 1, 3, 5, 8, z ¼ 0, y ¼ 0, z ¼ 2, 8, y ¼ 3, z ¼ 1, 2, y ¼ 4, 5, z ¼ 2.The cathodoluminescence (CL) emission spectrum belongs to the representative sample ofCa4Y6(SiO4)6O:10 mol%Ce3þ,3 mol%Mn2þ,2 mol%Tb3þ.215

Modified from Li G, Zhang Y, Geng D, Shang M, Peng C, Cheng Z, Lin J. Single-compositiontrichromatic white-emitting Ca4Y6(SiO4)6O:Ce

3þ/Mn2þ/Tb3þ phosphor: luminescence andenergy transfer. ACS Appl Mater Interfaces 2012;4:296e305. https://doi.org/10.1021/am201335d.


https://doi.org/10.1021/am201335d

https://doi.org/10.1021/am201335d

because of the metal-metal charge transfer physical process of Ce3þ þ Eu3þeCe4þ þ Eu2þ. However, Tb3þ can transfer its energy to Eu3þ in many hosts, thereforeit can act as a bridge to form a terbium chain between Ce3þ and Eu3þ ions resulting ineffective ET from Ce3þ to Eu3þ ions, which is expressed as a Ce3þeTb3þ. . .Tb3þeEu3þ process. The emission color can be tuned from blue to yellow or orange based onthe combination of Ce3þ blue emission, Tb3þ green emission and Eu3þ red emission.Wen et al.222 successfully designed a terbium chain in the form of Ce3þe(Tb3þ)neEu3þ in a Na2Y2B2O7 host using the ET processes from Ce3þ to Tb3þ and Tb3þ toEu3þ ions. The ET properties of Ce3þ to Tb3þ and Tb3þ to Eu3þ were demonstratedby the decline of decay lifetimes of Ce3þ and Tb3þ together with the variations ofemission spectra in Ce3þ, Tb3þ and Tb3þ, Eu3þ codoped Na2Y2B2O7, respectively.The ET details are depicted in Fig. 5.36(a). First, Ce3þ ions can be effectively excitedby n-UV light and jump from the ground state (2F5/2) to the excited states (5d energylevels). Then, Ce3þ ions decay to the lowest vibrational level of the excited state andgive out the excess energy to their surroundings, subsequently returning to the state of2F7/2 or

2F5/2 simultaneously by a radiative process or ET to the 5D3 level of Tb3þ ions.

The probability for ET of Ce3þeTb3þ increased with increasing concentration of Tb3þ

attributed to more neighboring Tb3þ ions around the Ce3þ ions. Cross-relaxation5D3 þ 7F6 ¼ 5D4 þ 7F0 resulted in the characteristic emission of the 5D4e

7FJ(J ¼ 6, 5, 4, 3) transition. As the content of Tb3þ increased, the distances of Tb3þeTb3þ and Tb3þeEu3þ shortened, leading to the quenching of Tb3þ emission andthe enhancement of ET from Tb3þ to Eu3þ, which may be ascribed to the followingmechanism:223,224

Tb3þ(5D4) þ Eu3þ(7F0) þ DEph / Tb3þ(7F4) þ Eu3þ(5D0)5.7

Tb3þ(5D3) þ Eu3þ(7F0) / Tb3þ(7F3) þ Eu3þ(5D2) þ DEph 5.8

where DEph is the phonon energy and relatively low (�500 cm�1). Ultimately, theenergy level of 5D2 relaxes to

5D0 and emission of Eu3þ ions occurs because of thecharacteristic transition of 5D0 -

7FJ (J ¼ 1, 2, 3, 4) and the possibility of metalemetalcharge transfer effect disappears. After optimization, the emission colors of the as-prepared samples could be tuned from blue to orange by only adjusting the Tb3þ

concentration in Na2Y2B2O7: 1%Ce3þ, zTb3þ, 1%Eu3þ (AeL), as shown inFig. 5.36(b). Using the ET properties from Ce3þ to Tb3þ and then to Eu3þ, manyresearchers have obtained abundant emission colors including white light under UV/n-UV excitation in many recent systems such as BaY2Si3O10:Ce

3þ, Tb3þ, Eu3þ,225 (Y/Gd)2SiO5:Ce

3þ, Tb3þ, Eu3þ,226,227 GdBO3:Ce3þ, Tb3þ, Eu3þ.228

Ce3þeTb3þeSm3þ system 5.7

Because Sm3þ can emit characteristic red emission based on 4G5/2-6H7/2 transitions,

it is possible to obtain white light emission by codoping Sm3þ in Ce3þ, Tb3þ codoped


systems. The Ce3þeTb3þeSm3þ system has been successfully designed to obtaintunable emission color based on the ET from Ce3þ to Tb3þ and then to Sm3þ.Fig. 5.36(c) shows the PL emission spectra of as-prepared La5Si2BO13:0.01Ce3þ,yTb3þ, 0.01Eu3þ/Sm3þ (y ¼ 0.10, 0.30, 0.50) phosphors under 300 nm

Energy transfer

MMCT

J = 3J = 2J = 1J = 0

J = 4J = 3

J = 2

J = 0

J = 1

D

F

F

F

F

F

F

D

D

Exc

itatio

n5d

Ce Tb Eu

LSBO: 0.01CeLSBO: 0.01Ce ,0.02TbLSBO: 0.01Ce ,0.05TbLSBO: 0.01Ce ,0.08TbLSBO: 0.01Ce ,0.12TbLSBO: 0.01Ce ,0.15TbLSBO: 0.01Ce ,0.20TbLSBO: 0.01Ce ,0.30Tb ,0.01EuLSBO: 0.01Ce ,0.50Tb ,0.01EuLSBO: 0.01Ce ,0.30Tb ,0.01SmLSBO: 0.01Ce ,0.50Tb ,0.01Sm

LSBO: 0.01CeLSBO: 0.01Ce ,0.02Tb

LSBO: 0.01Ce ,0.05TbLSBO: 0.01Ce ,0.08Tb

LSBO: 0.01Ce ,0.12TbLSBO: 0.01Ce ,0.15Tb

LSBO: 0.01Ce ,0.20TbLSBO: 0.01Ce ,0.30Tb ,0.01Eu

LSBO: 0.01Ce ,0.50Tb ,0.01EuLSBO: 0.01Ce ,0.30Tb ,0.01Sm

LSBO: 0.01Ce ,0.50Tb ,0.01Sm 350 400 450 500 550 600 650Wavelength (nm)

Energy transfer30

25

20

15

10

5

0

438n

m

410n

m

491n

m54

5nm

586n

m62

2nm

563n

m60

0nm

646n

m4f

D

FF

DDD G

GG

6HFFFFF

6H6H6H6H

I

Tb3+Ce3+ Sm3+

Wav

enum

ber(

× 10

3 cm

-1)

5d

Na Y B :1%Ce(0.160, 0.081)

A Tb% = 1.5%(0.232, 0.316)

E Tb% = 10%(0.344, 0.501)

I Tb% = 60%(0.635, 0.361)

J Tb% = 75%(0.637, 0.360)

K Tb% = 90%(0.637, 0.360)

L Tb% = 98%(0.637, 0.360)

F Tb% = 15%(0.403, 0.487)

G Tb% = 30%(0.566, 0.399)

H Tb% = 45%(0.627, 0.364)

B Tb% = 2.5%(0.249, 0.368)

C Tb% = 5%(0.249, 0.461)

D Tb% = 7.5%(0.317, 0, 489)

Inte

nsity

(a.u

.)(a) (b)

(c) (d)

O

Figure 5.36 (a) Energy level model for the energy processes of Ce3þeTb3þeEu3þ in aNa2Y2B2O7 host. (b) Pictures of the samples of Na2Y2B2O7:1%Ce3þ and Na2Y2B2O7:1%Ce3þ,zTb3þ,1%Eu3þ (AeL) under a 365 nm ultraviolet (UV) box with the correspondingchromaticity coordinates.222 (c) PL spectra of La5Si2BO13:Ce

3þ,Tb3þ,Eu3þ/Sm3þ samplesunder 300 nm UV excitation.229 (d) Schematic energy-level diagram explaining the lumi-nescence process and the possible energy-transfer pathways for the CaYAl3O7:0.05Ce

3þ,yTb3þ, 0.04Sm3þ.230

Modified from (a and b) Wen D, Shi J. A novel narrow-line red emitting Na2Y2B2O7:Ce3þ,Tb3þ,Eu3þ phosphor with high efficiency activated by terbium chain for near-UV whiteLEDs. Dalton Trans 2013;42:16621e9. https://doi.org/10.1039/c3dt52214h. (c) Zhang X,Zhang J, Gong M. Luminescence and energy transfer of La5Si2BO13: A (A¼ Ce3þ/Tb3þ/Eu3þ/Sm3þ) phosphors under UV excitation. Mater Lett 2015;143:71e4. https://doi.org/10.1016/j.matlet.2014.12.056. (d) Yu H, Yu X, Xu X, Zhou D, Qiu J. Realization of enhanced sensitizationeffect in CaYAl3O7:Ce

3þ, Sm3þ phosphors via Tb3þ ions. ECS J Solid State Sci Technol 2014;3(12):R245e50. https://doi.org/10.1149/2.022406jss.


https://doi.org/10.1039/c3dt52214h



https://doi.org/10.1149/2.022406jss

excitation,229 which were used to verify the bridge of Tb3þ to sensitize Eu3þ/Sm3þ,herein the Eu3þ/Sm3þ content was fixed as 0.01 to alleviate the metal-metal chargetransfer effect. It was found that with a low Tb3þ concentration it is hard to form aterbium bridge to transfer Ce3þ energy to Eu3þ/Sm3þ because the metalemetal chargetransfer process is not suppressed. However, a strong narrow band of Eu3þ/Sm3þ redemission is exhibited when y ¼ 0.30 and 0.50, which illustrated that a terbium bridgewas successfully formed and efficient Ce3þeTb3þeEu3þ/Sm3þ ET was realized. Thered emission intensity increased with increasing Tb3þ content to 0.50 because theempirical saturation distance (Rc) is 6e7 Å for the Tb3þ bridge when the Tb3þ concen-tration is 0.50. Therefore, tunable color from blue to red was realized in this systemunder UV excitation. Yu et al.230 exploited this effect to add Tb3þ into CaYAl3O7:Ce3þ, Sm3þ to enhance the sensitization effect, they proposed a schematic energy-level diagram to explain the luminescence process and the possible ET pathways inCaYAl3O7:Ce

3þ, Tb3þ, Sm3þ, as shown in Fig. 5.36(d). The Ce3þ to Tb3þ ET processwas shown in the Ce3þeTb3þeEu3þ system earlier. In the Tb3þeSm3þ scheme, elec-trons in the excited state 5D4 level of Tb

3þ can transfer their energy to the 4G9/2 level ofSm3þ

first. Then, the Sm3þ ions at the 4G9/2 level relax nonradiatively to4G7/2 and

4G5/

2 and subsequently decay radiatively to the ground states 4H5/2,4H7/2 and

4H9/2 corre-sponding to the emission peaks at 563, 600 and 646 nm, respectively. Therefore, theemission color was tuned from blue to white based on the ET process ofCe3þeTb3þeSm3þ.

5.4.2.2 ET models using Eu2þ as sensitizers

Similar to Ce3þ, Eu2þ is another excellent activator ion which can emit broad bandswith different colors in various hosts based on its 4f65d1e4f7 allowed transitions.9,10

Many phosphors with different emission color varied from blue to red have been stud-ied by singly-doping Eu2þ, as also summarized in Table 5.1. However, the emissioncolor is generally unchangeable or barely changeable in certain systems. In order totune the emission color and obtain abundant colors, ET is a frequently used strategyduring phosphor preparation.

5.4.2.2.1 Eu2þeMn2þ systemThe Eu2þeMn2þ system is one of the most abundant systems for emission coloradjustment in wLEDs phosphors.158,231e251 Liu et al.231 synthesized a novel blue togreen phosphor g-AlON:Eu2þ,Mn2þ. They found the Eu2þ singly doped g-AlONcan emit a wide band centered at 470 nm corresponding to a blue color with an exci-tation band from 200 to 450 nm, and Mn2þ singly-doped g-AlON can produce a broadband peaking at 517 nm under 450 nm excitation. Both of them were considered tooccupy Al3þ sites, however, the mismatch of ionic radii of Al3þ (0.39 Å, 4CN) andEu2þ (1.17 Å, 4CN) limited the solubility of Eu2þ (Fig. 5.37(a)). Spectral overlapbetween Eu2þ emission and Mn2þ excitation was presented in the paper, whichgave a possibility of ET from Eu2þ to Mn2þ ions in Eu2þ, Mn2þ co-doped samples.Moreover, the similar excitation spectra of Eu2þ and Mn2þ emissions in Eu2þ,Mn2þ co-doped samples were used to demonstrate it. Therefore, the emission intensity


of Mn2þ can be enhanced many times compared to the Mn2þ singly-doped situation(Fig. 5.37(b)). Good thermal stability was deduced because the emission intensity ofg-AlON:Eu2þ, Mn2þ measured at 150�C remained at 80% of that at room temperature(Fig. 5.37(c)), which was 8% larger than that of g-AlON:Mn2þ. This may originate

(a) (b)

(c)

25000

20000

15000

10000

5000

0200 300 400 500 600

Wavelength (nm)

PL

inte

nsity

(arb

. uni

t)

1.0

0.8

0.6

0.4

0 50 100 150 200 250

Temperature (°C)

(Mn2+,5mol%)

(Mn2+,5mol%, Eu2+ = 2 mol%)γ-AION:Mn2+,Eu2+

ϒ-AION:Mn2+

ϒ-AION:Mn2+,Eu2+

γ-AION:Mn2+Rel

ativ

e P

L in

tens

ity

O N

AI Mg Mn EuAI

ab

c

470 nm 517 nm

Mn2+

Eu2+

Energy transfer

Figure 5.37 (a) Site occupation of Eu2þ and Mn2þ in crystal structure of ~a-AlON and (b)comparison of photoluminescence (PL) spectra of g-AlON:Eu2þ (blue (gray in printversions)), ~a-AlON:Mn2þ (red (dark gray in print versions)), and ~a-AlON:Mn2þ,Eu2þ (green(light gray in print versions)). (c) Temperature-dependent luminescence of ~a-AlON:Eu2þ,Mn2þ (Eu2þ ¼ 2 mol%, Mn2þ ¼ 5 mol%) and ~a-AlON:Mn2þ(Mn2þ ¼ 5 mol%). Exci-tation wavelength is 405 nm.231

Modified from Liu L, Wang L, Zhang C, Cho Y, Dierre B, Hirosaki N, Sekiguchi T, Xie R-J.Strong energy-transfer-induced enhancement of luminescence efficiency of Eu2þ- and Mn2þ-codoped gamma-AlON for near-UV-LED-pumped solid state lighting. Inorg Chem 2015;54:5556e65. https://doi.org/10.1021/acs.inorgchem.5b00683.


https://doi.org/10.1021/acs.inorgchem.5b00683

from a serious lattice distortion when Eu2þ and Mn2þ are co-doped into g-AlON withappreciable difference in ionic radii.

As to the Ca9Mg(PO4)6F2:Eu2þ, Mn2þ system,232 a new solid solution of

Ca9Mg(PO4)6F2 was obtained via a substitution of Mg for Ca with a similar crystalstructure to Ca5(PO4)3F. As is well known to us, Ca5(PO4)3F has the apatite hexag-onal structure belonging to the space group P63/m, in which there are two kinds ofcrystallographic lattices consisting of ninefold coordinated 4f sites with C3 pointsymmetry [denoted as Ca2þ(1)] and sevenfold coordinated 6h sites with Cs pointsymmetry [denoted as Ca2þ(2)] (Fig. 5.38(a)). Therefore, Eu2þ wants to occupyboth sites, resulting in an asymmetrically broad band which can be decomposedinto two Gaussian bands (Fig. 5.38(b)), and this is demonstrated by the differentfluorescent decay lifetimes for the two deconvoluted bands (Fig. 5.38(c)). Tunableemission color from blue to yellow including white along with CIE chromaticcoordinates was observed under 365 nm UV lamp excitation in Eu2þ, Mn2þ codopedCa9Mg(PO4)6F2 by adjusting the ratio of Eu2þ and Mn2þ concentrations(Fig. 5.38(d)). This phenomenon is based on the ET from the Eu2þ to Mn2þ ions,which was also demonstrated by the similar excitation spectra of Eu2þ and Mn2þ

and decline of fluorescent decay lifetimes of Eu2þ (Fig. 5.38(e)) in Ca9Mg(PO4)6F2:Eu2þ, Mn2þ. When four Ca atoms were substituted by two Y and two Na atoms tobalance the charge in Ca10(PO4)6F2, a similar phenomenon was also observed in theapatite crystal structured phosphor Ca6Y2Na2(PO4)6F2:Eu

2þ, Mn2þ.233 The substitu-tion of all Ca atoms by Sr atoms in this kind of crystal structure can also produce atunable emission color from blue to yellow in Sr5(PO4)3F:Eu

2þ, Mn2þ.234 Anothergood blueewhiteeyellow system Ca2YF4(PO4):Eu

2þ, Mn2þ was also investigatedby Geng et al.235

Huang et al.236 successfully combined as-prepared Ca4Si2O7F2:Eu2þ, Mn2þ phos-

phors with a 400 nm n-UV chip to fabricate a wLED, producing a good CIE chro-matic coordinate of (0.347, 0.338) and CCT of 4880K compared to a commercialwLED with a CIE chromatic coordinate of (0.302, 0.315) and CCT of 7272K.Commonly, as-prepared blue-yellow phosphors often lack a red component for whitelight generation, causing a cool white light, therefore, blue-orange and blue-redphosphors were investigated based on ET from Eu2þ to Mn2þ ions to obtain warmwhite light, such as Ca2PO4Cl:Eu

2þ, Mn2þ,237 Na(Sr,Ba)PO4:Eu2þ, Mn2þ,238

Sr1.7Mg0.3SiO4:Eu2þ,Mn2þ,239 Ca9Lu(PO4)7:Eu

2þ, Mn2þ,240 Ca6�x�yMgx�z(PO4)4:yEu2þ, zMn2þ,241 Ba2MgP4O13:Eu

2þ, Mn2þ.242 The good thermal stability ofCa2PO4Cl:Eu

2þ, Mn2þ is demonstrated, as shown in Fig. 5.39(a). Further, the CIEcolor coordinates of (0.3102, 0.3096), a CCT of 4296K, and CRI of 86(Fig. 5.39(b)) when Ca2PO4Cl:Eu

2þ, Mn2þ was fabricated with a 400 nm n-UVchip are also shown. As a cyan to pink phosphor, the tunable emission color ofKCaY(PO4)2:Eu

2þ, Mn2þ 243 can be generated based on the combination of emissionbands, centered at 480 nm (4f65d1e4f7 transition of Eu2þ) and 652 nm(4T1(4G)-

6A1(6S) transition of Mn2þ), as shown in Fig. 5.39(c) and (d). When theemission bands of Eu2þ in some hosts locate in the green area, it can be possibleto transfer energy to Mn2þ because the excitation bands of Mn2þ locate in greenregion.


Our group analyzed the Eu2þ occupation site and it indicated the preference of Eu2þ

to occupy Naþ sites in a NaScSi2O6 host.244 An intense green emission can beobserved with an emission peak at 533 nm upon 365 nm excitation. In Eu2þ,Mn2þ

codoped NaScSi2O6 phosphors, bright tunable color from green to yellow(Fig. 5.40(a)) with Eu2þ emission band centered at 533 nm and Mn2þ emission

Ca1/MgCa2/MgFOPa

b

PLE

PL

λem = 405nm, τ = 726.46 nsy = 0, τ = 495.93 nsy = 0.02, τ = 462.40 nsy = 0.06, τ = 438.42 nsy = 0.10, τ = 387.73 nsy = 0.18, τ = 304.28 nsy = 0.26, τ = 274.36 ns

λem = 520nm, τ = 393.31 nsλem= 454 nm

λex= 300 nm

Inte

nsity

(a.u

.)In

tens

ity (a

.u.)

Inte

nsity

(nm

)

Wavelength (nm)200 300 400 600500 700

1

0.1

0.01

0 1 2 3 4Decay times (μs) Decay times (μs)

1

0.1

0.01

1E–30 1 2 3 4

CMPF:0.18Eu2+,yMn2+

CMPF:0.18Eu2+,yMn2+H

G

F

E

D

C

B

A

0.0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

(a)

(b)

(c)

(d)

(e)

Figure 5.38 (a) Crystal structure of Ca9Mg(PO4)6F2 host. (b) Photoluminescence emission andexcitation spectra of Ca9Mg(PO4)6F2 (CMPF):0.18Eu2þ sample with corresponding decaycurves (monitored at 405 and 520 nm and excited at 300 nm), inset is the photograph of sampleexcited under a 365 nm ultraviolet (UV) lamp. (c) Commission Internationale de I’Eclairage(CIE) chromaticity coordination of CMPF:0.18Eu2þ,yMn2þ samples (AeH corresponds toy ¼ 0, 0.02, 0.06, 0.10, 0.18, 0.26, 0.34, 0.38, respectively) with corresponding photographs ofsamples excited under a 365 nm UV lamp on the right of the picture. (d) Decay curves andlifetimes of Eu2þ in representative samples CMPF:0.18Eu2þ,yMn2þ (monitored at 454 nmexcited at 300 nm).232

Modified from Li K, Geng D, Shang M, Zhang Y, Lian H, Lin J. Color-tunable luminescenceand energy transfer properties of Ca9Mg(PO4)6F2:Eu

2þ, Mn2þ phosphors for UV-LEDs. J PhysChem C 2014;118:11026e34. https://doi.org/10.1021/jp501949m.


https://doi.org/10.1021/jp501949m

band centered at 654 nm can be observed under 365 nm UV lamp excitation becauseof the ET from Eu2þ to Mn2þ ions through adjusting the concentration ratio of Eu2þ

and Mn2þ ions. In addition, we combined the as-prepared phosphor NaScSi2O6:0.05Eu2þ, 0.10Mn2þ and commercial blue BaMgAl10O17:Eu

2þ phosphors with a370 nm n-UV chip to produce a warm wLED with a CIE color coordinate of(0.358, 0.378), CCT of 4666K and good CRI of 92.2 (Fig. 5.40(b)). Another important

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

12

3 4 56

7

400 500 600 700Wavelength/(nm)

400 500 600 700Wavelength/(nm)

Rel

ativ

e in

tens

ity (a

rb. u

nit)

Inte

nsity

(arb

. uni

t)

Rel

ativ

e in

tens

ity (a

rb. u

nit) Ca2PO4CI:Eu2+,Mn2+

Ca2PO4CI:0.07Eu2+,0.2Mn2+

CIE (0.3102, 0.3096)CCT = 4296 K

20°C50°C100°C150°C200°C250°C

1.0

0.9

0.8

0.70 100 200N

or.in

ten.

/(a) YAG:Ce3+

CPL:Eu2+,

Eu2+

Mn2+

Mn2+

Temperature/(°C)

400

nm c

hip

350 400 450 500 550 600 650 700 750 800Wavelength (nm)

(a)

(c) (d)

(b)

Figure 5.39 (a) Temperature-dependent emission spectra of Ca2PO4Cl:0.07Eu2þ,0.2Mn2þ

(lex ¼ 370 nm). Inset: Normalized intensity of Ca2PO4Cl:0.07Eu2þ,0.2Mn2þ and YAG:Ce as

a function of temperature.237 (b) PL spectra of a series of KCaY(PO4)2:1%Eu2þ,x%Mn2þ

phosphors with different Mn2þ concentrations (x ¼ 0, 1, 2, 4, 5, 7, and 10 mol%) excited at365 nm.243 (c) Emission spectrum of a phosphor converted light emitting diode (wLED)(pc-LED) lamp fabricated with a 400 nm LED chip and warm, white-emitting phosphorCa2PO4Cl:0.07Eu

2þ,0.2Mn2þ.237 (d) CIE coordinates of KCaY(PO4)2:1%Eu2þ,x%Mn2þ

phosphors (x ¼ 0, 1, 2, 4, 5, 7, and 10). Insets show the phosphor images with different Mn2þ

doping concentrations excited at 365 nm in the ultraviolet box.243

Modified from (a and c) Li P, Wang Z, Yang Z, Guo Q. A novel, warm, white light-emittingphosphor Ca2PO4Cl: Eu

2þ, Mn2þ for white LEDs. J Mater Chem C 2014;2:7823e9. https://doi.org/10.1039/c4tc01055h. (b and d) Liu W-R, Huang C-H, Yeh C-W, Tsai J-C, Chiu Y-C, YehY-T, Liu R-S. A study on the luminescence and energy transfer of single-phase and color-tunable KCaY(PO4)2:Eu

2þ,Mn2þ phosphor for application in white-light LEDs. Inorg Chem2012;51:9636e41. https://doi.org/10.1021/ic3007102.


https://doi.org/10.1039/c4tc01055h

https://doi.org/10.1039/c4tc01055h

https://doi.org/10.1021/ic3007102

400 500 550 600 650 700 750450Wavelength (nm)

Rel

ativ

e in

tens

ity (a

rb. u

nit)

(c)

(d)

(b)(a)

λex = 355 nmn = 0n = 0.005n = 0.01n = 0.02n = 0.03n = 0.04n = 0.05n = 0.06n = 0.08


370 nmn-UV chip

NaScSi2O6:

BaMgAI10O17:Eu2+

EU2+ Mn2+,

Inte

nsity

(a.u

.)

n = 0

n = 0.005

n = 0.01

n = 0.02

n = 0.03

n = 0.04

n = 0.06

n = 0.08

1.

2.

3.

4.

5.

6.

7.8.

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

X

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8x

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Y

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

y

SLuP:0.005Eu2+,nMn2+

Color-tunable

1 8

5

4

3

2

1

1 2 34 5

Figure 5.40 (a) Commission Internationale de I’Eclairage (CIE) chromaticity diagram and aseries of digital photographs of the selected NaScSi2O6:0.05Eu

2þ,xMn2þ phosphors under365 nm ultraviolet lamp excitation. (b) Electroluminescence (EL) spectrum of the white light-emitting diode (wLED) lamp based on the NaScSi2O6:0.05Eu

2þ,0.10Mn2þ and commercialblue BaMgAl10O17:Eu

2þ phosphors and driven by a current of 25 mA. The inset photographsare the wLED lamp package.244 (c) PL spectra of Sr3Lu(PO4)3:0.005Eu

2þ,nMn2þ phosphorson Mn2þ doping content (n) and (d) corresponding CIE chromaticity diagram for (point 1 to 8)excited at 355 nm.246

Modified from (a and b) Xia Z, Zhang, Molokeev MS, Atuchin VV. Structural and luminescenceproperties of yellow-emitting NaScSi2O6:Eu

2þ phosphors: Eu2þ site preference analysis andgeneration of red emission by codoping Mn2þ for white-light-emitting diode applications. JPhys Chem C 2013;117:20847e54. https://doi.org/10.1021/jp4062225. (c and d) Guo, Zheng Y,Jia Y, Qiao H, You H. Warm-white-emitting from Eu2þ/Mn2þ-codoped Sr3Lu(PO4)3 phosphorwith tunable color tone and correlated color temperature. J Phys Chem C 2012;116:1329e34.https://doi.org/10.1021/jp209891b.


https://doi.org/10.1021/jp4062225

https://doi.org/10.1021/jp209891b

examples are Eu2þ/Mn2þ codoped Sr3Y/Lu/Sc/Gd(PO4)3 systems.245e248 Forexample, as for the Sr3Lu(PO4)3:Eu

2þ, Mn2þ, the CIE chromatic coordinate variedfrom green to yellow including a white area (Fig. 5.40(c)) because the emission bandsconsisted of some appropriate blue, green and red components upon 355 nm excitation(Fig. 5.40(d)).

Huang et al. and Guo et al.249e251 investigated the PL properties and ET behaviorsof the (Ca/Sr/Mg)9(Y/La/Gd/Lu)(PO4)7:Eu

2þ, Mn2þ phosphors, and they found thatblue and red shifts occurred with increasing Mg2þ and Sr2þ concentrations, respec-tively, which is ascribed to the crystal field variation around Eu2þ when Ca2þ wassubstituted by smaller Mg2þ and larger Sr2þ in Ca9Y(PO4)7:Eu

2þ. Moreover, whenthe Mn2þ ions were introduced into Ca0.5Sr0.5)9Y(PO4)7:Eu

2þ, it showed tunable colorfrom green to red by adjusting the Eu2þ and Mn2þ concentration ratio. They also com-bined as-prepared (Ca0.5Mg0.5)9Y(PO4)7:0.007Eu

2þ and (Ca0.5Sr0.5)9Y(PO4)7:0.007Eu2þ, 0.02Mn2þ with a 380 chip to generate a wLED with a CCT of 6303K,a CRI of 87.4, and the CIE color coordinates x ¼ 0.314 and y ¼ 0.348. The othertwo systems Ca9La/Gd(PO4)7:Eu

2þ,Mn2þ with similar properties were also consid-ered to be good candidate phosphors for n-UV pumped wLEDs. Differently,Ca9Lu(PO4)7:Eu

2þ,Mn2þ presented tunable cyan-pink including white emission colorunder 355 nm excitation because of the slightly different emission bands of Eu2þ andMn2þ compared to the similar samples given earlier.

5.4.2.2.2 Eu2þeTb3þ systemGenerally, Tb3þ acts as an efficient green-emitting activator based on its characteristic5D4e

7FJ (J ¼ 6, 5, 4, 3) transition. Because the transition belongs to the 4fe4fspin-forbidden transition, the absorption spectrum and emission band are ratherweak and their widths are narrow. Therefore, it is desirable to enhance the emissionintensity via the ET effect. The sensitizer Eu2þ is considered to be a good candidateto transfer its energy to Tb3þ.30,252e254 In Eu2þ, Tb3þ codoped phosphors, the emis-sion color can be tuned from blue or cyan to green. Recently, Li et al. found that atunable blueegreen color can be produced via codoping Eu2þ and Tb3þ into thefamiliar b-Ca3(PO4)2 compound with high quantum yields under UV excitation.253

Ba3LaNa(PO4)3F:Eu2þ, Tb3þ with apatite structure produced excellent ET properties

from Eu2þ to Tb3þ ions, resulting in clear color variation from blue to green under365 UV lamp excitation (Fig. 5.41(a)) by adjusting the doped ion concentration ratio.30

The ET mechanism from Eu2þ to Tb3þ ions was demonstrated to be an EDequadrupole interaction using the InokutieHirayama (IeH) model (Fig. 5.41(b)).Zhang et al.254 developed a novel Sr3Y2(Si3O9)2:Eu

2þ, Tb3þ phosphor with corre-sponding structure reported. Tunable color from cyan to green can be observedfrom the emission spectra in Fig. 5.41(c) with increasing Tb3þ concentration inSr3Y2(Si3O9)2:0.01Eu

2þ, yTb3þ phosphors. A detailed schematic diagram of ET(Fig. 5.41(d)) from Eu2þ to Tb3þ ion was used to understand the ET process. It showedsome excited electrons in the Eu2þ 4f65d1 energy level transfer their energy to theTb3þ 5D4 energy level, resulting in a Tb3þ 5D4e

7FJ (J ¼ 6, 5, 4, 3) transition.


5.4.2.2.3 Eu2þeTb3þeMn2þ systemSimilar to the Ce3þeTb3þeMn2þ systems, the Eu2þeTb3þeMn2þ systems can alsobe candidates to obtain white emission because Mn2þ can supply the red componentin Eu2þ,Tb3þ systems. Therefore, many researchers have devoted their interest tothem. Lv et al.255 designed and prepared BaMg2Al6Si9O30:Eu

2þ, Tb3þ, Mn2þ phos-phors, they found two emission peaks at 376 and 450 nm occurred in Eu2þ singlydoped BaMg2Al6Si9O30 under 330 nm excitation since Eu2þ occupied two kindsof Ba2þ sites. Mn2þ can emit a band at around 610 nm under 407 nm excitation

400 450 500 550 600 650 700 750Wavelength (nm)

SYSO:0.01Eu2+,yTb3+

y = 0y = 0.01y = 0.05y = 0.13y = 0.14y = 0.17y = 0.21

Inte

nsity

(arb

. uni

t)

0

5000

10,000

15,000

20,000

25,000

30,000

35,000

40,000

0

5000

10,000

15,000

20,000

25,000

30,000

35,000

40,000

Ene

rgy

(cm

–1)

Ene

rgy

(cm

–1)

Eu2+ Tb3+

5D4

5D3

4f7(8S7/2)

4f65d

250∼

420

nm

474

nm

378

nm48

7 nm

541

nm58

4 nm

622

nm

ET

n = 0.15

n = 0.06

Slope = 0.360

n = 0.30 Slope = 0.373

Slope = 0.381

Log (t)

Log{

In( D

0(t)/

I D(t)

)}

–1.50–4–3–2–1

–4–3–2–1

0–4–3–2–1

0

0

–1.25 –1.00 –0.75 –0.50 –0.25 0.00 0.25

1. n = 0.002. n = 0.123. n = 0.204. n = 0.305. n = 0.605

43

2

1

0.0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

(d)(c)

(b)(a)

Figure 5.41 (a) Commission Internationale de I’Eclairage (CIE) chromaticity diagram forBa3LaNa(PO4)3F:0.01Eu

2þ,nTb3þ phosphors, together with their corresponding photographsunder a 365 nm ultraviolet lamp.30 (b) The PL spectra of Sr3Y2(Si3O9)2:0.01Eu

2þ,yTb3þ

phosphors (y ¼ 0, 0.01, 0.05, 0.13, 0.14, 0.17 and 0.21).254 (c) Experimental data plots of log{ln[ID0(t)/ID(t)]} versus log(t) of Eu2þ in Ba3LaNa(PO4)3F:0.01Eu

2þ,nTb3þ (n ¼ 0.06, 0.15,0.30) samples. The red (gray in print versions) lines indicate the fitting behaviors.30 (d) Theschematic diagram of energy transfer (ET) in Sr3Y2(Si3O9)2:Eu

2þ,Tb3þ.254

Modified from (a and c) Jiao M, Guo N, Lu W, Jia Y, Lv W, Zhao Q, Shao B, You H. Tunableblue-green-emitting Ba3LaNa(PO4)3F:Eu

2þ, Tb3þ phosphor with energy transfer for near-UVwhite LEDs. Inorg Chem 2013;52:10340e6. https://doi.org/10.1021/ic401033u. (b and d)Zhang M, Liang Y, Xu S, Zhu Y, Wu X, Liu S. Investigation of luminescence properties and theenergy transfer mechanism of tunable emitting Sr3Y2(Si3O9)2:Eu

2þ,Tb3þ phosphors.CrystEngComm 2016;18:68e76. https://doi.org/10.1039/c5ce01814e.


https://doi.org/10.1021/ic401033u

https://doi.org/10.1039/c5ce01814e

because of its occupancy of the Mg2þ site with sixfold coordinated oxygen. There-fore, tunable color from blue to green and red including white were generatedwhen Tb3þ and Mn2þ were codoped into it and the effective ET from Eu2þ toTb3þ and Mn2þ ions appears (Fig. 5.42(a)). In Eu2þ, Mn2þ codoped KCaGd(PO4)2

0.20

0.16

0.12

0.08

0.04

0.00

0.00 0.02 0.04 0.06 0.08 0.10

Tb content (y)(c) (d)

(a) (b)

Inte

nsity

(arb

. uni

t)

400 450 500 550 600 650 700

Wavelength (nm)

Tb3+

Eu2+Mn2+

Mn2+Eu2+Tb3+

1% Eu 10% Mn 1% Tb1% Eu 10% Mn 5% Tb1% Eu 10% Mn 10% Tb1% Eu 10% Mn 20% Tb

UV lightexcitation

Em

ission

Em

ission

Em

ission

Step 1

Step 3

Step 4

Step 2

Sensitize

ET

ET

Cascade model (P = pn)

Branch model

0.0

5.0 x 103

1.0 x 104

1.5 x 104

2.0 x 104

2.5 x 104

3.0 x 104

3.5 x 104

4.0 x 104

Ene

rgy

(cm

–1)

Mn

cont

ent (

z)

23456

412

nm54

3 nm

292

nm

416

nm

406

nm

646

nm

7FJ

5D4

5D3

6A1

4T2(4G)4T1(4G)

4T1(4P)

4T2(4D)4A1,

4E(4G)

4E(4D)

4f7

4f65d

ETET

n

Figure 5.42 (a) Photographs of the emission BaMg2Al6Si9O30:0.04Eu2þ,yTb3þ,zMn2þ

phosphors with different percent dopant contents (y and z) under excitation at 365 nm.255 (b)Schematic level diagram for energy transfer (ET) process from Eu2þ to Tb3þ and Mn2þ in theCa9Sc(PO4)7 host.

257 (c) Photoluminescent spectra of KCaGd(PO4)2:1%Eu2þ, 10%Mn2þ andy%Tb3þ phosphors (y ¼ 1, 5, 10, 20) excited at 365 nm.256 (d) ET models of the terbium chain(cascade model) and the terbium bridge (branch model).258

Modified from (a) LvW, Hao Z, Zhang X, Luo Y, Wang X, Zhang J. Tunable full-color emittingBaMg2Al6Si9O30:Eu

2þ, Tb3þ, Mn2þ phosphors based on energy transfer. Inorg Chem 2011;50:7846e51. https://doi.org/10.1021/ic201033e. (b) Jiang L, Pang R, Li D, Sun W, Jia Y, Li H, FuJ, Li C, Zhang S. Tri-chromatic white-light emission from a single-phase Ca9Sc(PO4)7:Eu

2þ,Tb3þ, Mn2þ phosphor for LED applications. Dalton Trans 2015;44:17241e50. https://doi.org/10.1039/c5dt02061a. (c) Liu WR, Huang CH, Yeh CW, Chiu YC, Yeh YT, Liu RS. Single-phased white-light-emitting KCaGd(PO4)2:Eu

2þ, Tb3þ, Mn2þ phosphors for LED applications.RSC Adv 2013;3:9023e8. https://doi.org/10.1039/c3ra40471d. (d) Xu S, Li P, Wang Z, Li T,Bai Q, Sun J, Yang Z. Luminescence and energy transfer of Eu2þ/Tb3þ/Eu3þ in LiBaBO3

phosphors with tunable-color emission. J Mater Chem C 2015;3:9112e21. https://doi.org/10.1039/c5tc01577d.


https://doi.org/10.1021/ic201033e

https://doi.org/10.1039/c5dt02061a

https://doi.org/10.1039/c5dt02061a

https://doi.org/10.1039/c3ra40471d



phosphors, the emission bands of Eu2þ and Mn2þ centered at 463 and 650 nm under365 nm excitation. Liu et al.256 added Tb3þ into Eu2þ, Mn2þ codoped KCaGd(PO4)2phosphors to compensate the green component, the green emission intensityincreased with increasing Tb3þ concentration (Fig. 5.42(b)), which made theemission color of KCaGd(PO4)2 Eu

2þ, Mn2þ shift from light red to the white regionunder 365 nm excitation. In the schematic level diagram for the ET processin Eu2þ,Tb3þ,Mn2þ codoped Ca9Sc(PO4)7 system,257 as a reference forEu2þ,Tb3þ,Mn2þ systems (Fig. 5.42(c)), we can see Eu2þ emitted a band around416 nm under 292 nm excitation, part of the energy was transferred from the Eu2þ

excited level to the 5D3 level of Tb3þ and 4A1 level of Mn2þ to enhance their corre-

sponding emission intensities.

5.4.2.2.4 Eu2þeTb3þeEu3þ systemIt is interesting that some Eu3þ was not reduced in the reductive calcined atmospherein Eu3þ, Tb3þ codoped LiBaBO3.

258 This phenomenon is not frequent. Xu et al.analyzed the ET properties from Eu2þ to Tb3þ and then to Eu3þ via the variationsof excitation spectra and fluorescent decay times in detail. The ET process of Eu2þe(Tb3þ)neEu3þ was relatively considerable, so they brought forward the branch modelto explain the process of ET for the terbium bridge, shown in Fig. 5.42(d). First, whenthe Eu2þ ions were excited by UV light, they gave out a blue emission and sensitizedTb3þ ions in the ground state. Then, the excited Tb3þ ions gave out a green emissionand transferred energy to Eu3þ, at the same time, the excited Tb3þ ion released energyin the way of cross-relaxation with another Tb3þ ion in the ground state, and the Tb3þ

which absorbed the energy from the excited Tb3þ also interacted with Eu3þ throughthe process of ET. They utilized this property to obtain color-tunable blueegreen tored emission under UV excitation.

5.4.2.2.5 Eu2þeTb3þeSm3þ systemAlthough the Ce3þeTb3þeSm3þ system has been successfully designed to obtaintunable emission color based on the ET from Ce3þ to Tb3þ and then to Sm3þ, thesystem of Eu2þeTb3þeSm3þ is rarely utilized to obtain abundant emission colors.Up to now, we’ve only found one report about Eu2þeTb3þeSm3þ system. Jiaet al.259 demonstrated Tb3þ can act as an ET bridge to connect Eu2þeSm3þ lumines-cent centers in Sr3Y(PO4)3, realizing the red emission of Sm3þ under near-UVexcitation.

5.5 Future development of wLEDs phosphors

wLED solid-state lighting technology has received extensive attention in these years,and phosphor materials act as one kind of key components in fabrication of wLEDsdevices. Accordingly, the related optical parameters, such as luminous efficiency,CRI and CCT are strongly dependent on the physical properties of the selected


phosphors. In this chapter, we summarized five fundamental requirements for phos-phors used in wLEDs and four commercial phosphors for wLEDs. Subsequently,we focused on the summary of design and discovery of novel phosphor host materialsvia the structurally design principles and ET viewpoint, which is generally efficient inthe optimization of the commercially used LED phosphors. For the design and discov-ery of novel phosphor via the structurally design principles, we described the latestresearch progress of the several most common phosphors systems for wLEDs,including aluminate, silicate, borate, sulfide, phosphate and the Mn4þ or Eu3þ dopedphosphors. In terms of ET to develop new phosphors, using Eu2þ or Ce3þ as the sensi-tizer, we described 12 kinds of ET systems, such as Ce3þeMn2þ, Ce3þeTb3þ, Ce3þeTb3þeMn2þ, Eu2þeMn2þ, Eu2þeTb3þ, Eu3þeTb3þeMn2þ and so on.

Looking forward, there are still many research areas that need further work,including but not limited to: (1) the modification of the present LED phosphors; (2)the discovery of new red LED phosphors; (3) methodology investigations in thediscovery of new LED phosphors; (4) The investigations of advanced inorganicsolid-state compounds and the optimization of their physical, e.g., band structure,and chemical properties. Anyway, a theoretical and experimental understanding ofthe intrinsic properties of the solid-state compounds is of great interest with respectto the design of the next generation of solid-state lighting LED phosphors.

Acknowledgments

This work was financially supported by the National Natural Science Foundation of China(51722202, 91622125, 51472132, 51572023 and 51272242), Qingdao Project of Science andTechnology (13-1-4-114-jch), and Natural Science Foundations of Beijing (2172036).

References

1. Schubert EF, Kim JK. Solid-state light sources getting smart. Science 2005;308:1274e8.2. Xia ZG, Xu ZH, Chen MY, Liu QL. Recent developments in the new inorganic solid state

LED phosphors. Dalton Trans 2016;45:11214e32.3. Xia ZG, Liu QL. Progress in discovery and structural design of color conversion phosphors

for LEDs. Prog Mater Sci 2016;84:59e117.4. Xia ZG, Meijerink A. Ce3þ-doped garnet phosphors: composition modification, lumi-

nescence properties and applications. Chem Soc Rev 2017;46:275e99.5. Ye S, Xiao F, Pan YX, Ma YY, Zhang QY. Phosphors in phosphor-converted white light-

emitting diodes: recent advances in materials, techniques and properties. Mater Sci Eng RRep 2010;71:1e34.

6. Nakamura S, Mukai T, Senoh M. Candela-class high-brightness InGaN/AlGaN double-heterostructure blue-light-emitting diodes. Appl Phys Lett 1994;64(13):1687e9.

7. Phillips JM, Coltrin ME, Crawford MH, Fischer AJ, Krames MR, Mueller-Mach R,Mueller GO, Ohno Y, Rohwer LES, Simmons JA, Tsao JY. Research challenges to ultra-efficient inorganic solid-state lighting. Laser Photon Rev 2007;1:307e33.


8. Chen L, Lin CC, Yeh CW, Liu RS. Light converting inorganic phosphors for white light-emitting diodes. Materials 2010;3(3):2172e95.

9. Li GG, Tian Y, Zhao Y, Lin J. Recent progress in luminescence tuning of Ce3þ and Eu2þ-activated phosphors for pc-WLEDs. Chem Soc Rev 2015;44(23):8688e713.

10. ShangMM, Li CX, Lin J. How to produce white light in a single-phase host. Chem Soc Rev2014;43:1372e86.

11. Meyer J, Tappe F. Photoluminescent materials for solid-state lighting: state of the art andfuture challenges. Adv Opt Mater 2015;3(4):424e30.

12. Setlur AA, Lyons RJ, Murphy JE, Kumar NP, Kishore MS. Blue light-emitting diodephosphors based upon oxide, oxyhalide, and halide hosts. ECS J Solid State Sci Technol2013;2:R3059e70.

13. Smet PF, Parmentier AB, Poelman D. Selecting conversion phosphors for white light-emitting diodes. J Electrochem Soc 2011;158(6):R37e54.

14. George NC, Denault KA, Seshadri R. Phosphors for solid-state white lighting. Annu RevMater Res 2013;43:481e501.

15. George NC, Pell AJ, Dantelle GR, Page K, Llobet A, Balasubramanian M, Pintacudaet G,Chmelka BF, Seshadri R. Local environments of dilute activator ions in the solid-statelighting phosphor Y3-xCexAl5O12. Chem Mater 2013;25(20):3979e95.

16. Pust P, Wochnik AS, Baumann E, Schmidt PJ, Wiechert D, Scheu C, Schnick W. Ca[LiAl3N4]: Eu

2þda narrow-band red-emitting nitridolithoaluminate. Chem Mater 2014;26(11):3544e9.

17. Zhang S, Nakai Y, Tsuboi T, Huang Y, Seo HJ. The thermal stabilities of luminescenceand microstructures of Eu2þ-Doped KBaPO4 and NaSrPO4 with b-K2SO4 type structure,.Inorg Chem 2011;50(7):2897e904.

18. Liu Y, Zhang J, Zhang C, Xu J, Liu G, Jiang J, Jiang H. Ba9Lu2Si6O24:Ce3þ: an efficient

green phosphor with high thermal and radiation stability for solid-state lighting. Adv OptMater 2015;3(8):1096e101.

19. Mao Z, Lu Z, Chen J, Fahlman BD, Wang D. Tunable luminescent Eu2þ-dopeddicalcium silicate polymorphs regulated by crystal engineering. J Mater Chem C 2015;3:9454e60.

20. Miao S, Xia Z, Molokeev MS, Chen M, Zhang J, Liu Q. Effect of Al/Si substitution on thestructure and luminescence properties of CaSrSiO4: Ce

3þ phosphors: analysis based on thepolyhedra distortion. J Mater Chem C 2015;3:4616e22.

21. Chen M, Xia Z, Molokeev MS, Liu Q. Structural phase transformation and luminescentproperties of Ca2-xSrxSiO4: Ce3þ orthosilicate phosphors. Inorg Chem 2015;54(23):11369e76.

22. Xie RJ. Optical properties of (oxy) nitride materials: a review. J Am Ceram Soc 2013;96(3):665e87.

23. Wang SS, Chen WT, Li Y, Wang J, Sheu HS, Liu RS. Neighboring-cation substitutiontuning of photoluminescence by remote-controlled activator in phosphor lattice. J AmChem Soc 2013;135(34):12504e7.

24. Ju LC, Cai C, Zhu QQ. Color tunable Sr2SiO4: Eu2þ phosphors through the modification of

crystal structure. J Mater Sci: Mater Electron 2013;24(11):4516e21.25. Zhang X, Wang J, Huang L, Pan F, Chen Y, Lei B, Peng M, Wu M. Tunable luminescent

properties and concentration-dependent, site-preferable distribution of Eu2þ ions in silicateglass for white LEDs applications. ACS Appl Mater Inter 2015;7(18):10044e54.

26. Sato Y, Kato H, Kobayashi M, Masaki T, Yoon DH, Kakihana M. Tailoring of deep-redluminescence in Ca2SiO4: Eu

2þ. Angew Chem Int Ed 2014;53(30):7756e9.


27. Zhang Y, Liu X, Li X, Li K, Lian H, Shang M, Lin J. Overcoming crystallographicallyimposed geometrical restrictions on the valence state of Eu in CaGdAlO4: realization ofwhite light emission from singly-doped Eu. Dalton Trans 2015;44:7743e7.

28. Xia Z, Liu RS. Tunable blue-green color emission and energy transfer of Ca2Al3O6F:Ce3þ, Tb3þ phosphors for near-UV white LEDs. J Phys Chem C 2012;116(29):15604e9.

29. Wu W, Xia Z. Synthesis and color-tunable luminescence properties of Eu2þ and Mn2þ-activated Ca3Mg3(PO4)4 phosphor for solid state lighting. RSC Adv 2013;3:6051e7.

30. Jiao M, Guo N, L€u W, Jia Y, Lv W, Zhao Q, Shao B, You H. Tunable blue-green-emittingBa3LaNa(PO4)3F: Eu

2þ, Tb3þ phosphor with energy transfer for near-UV white LEDs.Inorg Chem 2013;52(18):10340e6.

31. Lin CC, Liu RS. Advances in phosphors for light-emitting diodes. J Phys Chem Lett 2011;2:1268e77.

32. Blasse G, Bril A. A new phosphor for flying-spot cathode-ray tubes for color television eyellow-emitting Y3Al5O12-Ce

3þ. Appl Phys Lett 1967;11:53e5.33. Hoppe HA. Recent developments in the field of inorganic phosphors. Angew ChemInt Ed

2009;48:3572e82.34. Shang MM, Fan J, Lian HZ, Zhang Y, Geng DL, Lin J. A double substitution of

Mg2þeSi4þ/Ge4þ for Al(1)3þeAl(2)

3þ in Ce3þ-Doped garnet phosphor for white LEDs. InorgChem 2014;53:7748e55.

35. Bachmann V, Ronda C, Meijerink A. Temperature quenching of yellow Ce3þ lumines-cence in YAG: Ce. Chem Mater 2009;21:2077e84.

36. Setlur AA, Heward WJ, Hannah ME, Happek U. Incorporation of Si4þ-N3-into Ce3þ-Doped garnets for warm white LED phosphors. Chem Mater 2008;20(19):6277e83.

37. Zhong J, Zhuang W, Xing X, Liu R, Li Y, Liu Y, Hu Y. Synthesis, crystal structures, andphotoluminescence properties of Ce3þ-Doped Ca2LaZr2Ga3O12: new garnet green-emitting phosphors for white LEDs. J Phys Chem C 2015;119:5562e9.

38. Ji H, Wang L, Molokeev MS, Hirosaki N, Huang Z, Xia Z, Otmar M, Liu L, Xie R. Newgarnet structure phosphors, Lu3�xYx MgAl3SiO12: Ce

3þ (x¼ 0-3), developed by solidsolution design. J Mater Chem C 2016;4:2359e66.

39. Nikl M, Yoshikawa A, Vedda A, Fukuda T. Development of novel scintillator crystals.J Cryst Growth 2006;292:416e21.

40. Ueda J, Aishima K, Nishiura S, Tanabe S. Afterglow luminescence in Ce3þ-dopedY3Sc2Ga3O12 ceramics. Appl Phys Express 2011;4:042602.

41. Xie RJ, Hirosaki N. Silicon-based oxynitride and nitride phosphors for white LEDs e areview. Sci Technol Adv Mater 2007;8(7e8):588e600.

42. Schmiechen S, Schneider H, Wagatha P, Hecht C, Schmidt PJ, Schnick W. Toward newphosphors for application in illumination-grade white pc-LEDs: the nitridomagnesosili-cates Ca[Mg3SiN4]: Ce3þ, Sr[Mg3SiN4]: Eu

2þ, and Eu[Mg3SiN4]. Chem Mater 2014;26(8):2712e9.

43. Xie RJ, Hirosaki N, Li Y, Takeda T. Rare-earth activated nitride phosphors: synthesis,luminescence and applications. Materials 2010;3(6):3777e93.

44. Li YQ, van Steen JEJ, van Krevel JWH, Botty G, Delsing ACA, Disalvo FJ, de With D,Hintzen HT. Luminescence properties of red-emitting M2Si5N8: Eu

2þ (M¼Ca, Sr, Ba)LED conversion phosphors. J Alloys Compd 2006;417(1e2):273e9.

45. Uheda K, Hirosaki N, Yamamoto Y, Naito A, Nakajima T, Yamamoto H. Luminescenceproperties of a red phosphor, CaAlSiN3: Eu

2þ, for white light-emitting diodes. Electro-chem Solid State Lett 2006;9(4):H22e5.


46. Lin CC, Meijerink A, Liu RS. Critical red components for next-generation white LEDs.J Phys Chem Lett 2016;7:495e503.

47. Pust P, Weiler V, Hecht C, T€ucks A, Wochnik AS, Henb AK, Wiechert D, Scheu C,Schmidt PJ, SchnickW. Narrow-band red-emitting Sr[LiAl3N4]: Eu

2þ as a next-generationLED-phosphor material. Nat Mater 2014;13(9):891e6.

48. Tolhurst TM, Boyko TD, Pust P, Johnson NW, Schnick W, Moewes A. Investigations ofthe electronic structure and bandgap of the next-generation LED-phosphor Sr[LiAl3N4]:Eu2þ- experiment and calculations. Adv Opt Mater 2015;3(4):546e50.

49. Pust P, Hintze F, Hecht C, Weiler V, Locher A, Zitnanska D, Harm S, Wiechert D,Schmidt PJ, Schnick W. Group (III) nitrides M[Mg2Al2N4] (M ¼ Ca, Sr, Ba, Eu) and Ba[Mg2Ga2N4) - structural relation and non-typical luminescence properties of Eu2þ dopedsamples. Chem Mater 2014;26(21):6113e9.

50. Xia Z, Miao S, Molokeev MS, Chen M, Liu Q. Structure and luminescence properties ofEu2þ doped LuxSr2�xSiNxO4�x phosphors evolved from chemical unit cosubstitution.J Mater Chem C 2016;4:1336e44.

51. Denault KA, Brgoch J, Gaultois MW, Mikhailovsky A, Petry R, Winkler H, DenBaars SP,Seshadri R. Consequences of optimal bond valence on structural rigidity and improvedluminescence properties in SrxBa2�xSiO4: Eu

2þ orthosilicate phosphors. Chem Mater2014;26(7):2275e82.

52. Jang HS, Jeon DY. Yellow-emitting Sr3SiO5: Ce3þ, Liþ phosphor for white-light-emitting

diodes and yellow-light-emitting diodes. Appl Phys Lett 2007;90:041906.53. Kang EH, Choi SW, Su EC, Jisung J, Kwon S, Hong SH. Photoluminescence character-

istics of Sr3SiO5:Eu2þ yellow phosphors synthesized by solid-state method and Pechini

process. J Electrochem Soc 2011;158(11):J330e3.54. Shimomura Y, Honma T, Shigeiwa M, Akai T, Okamoto K. Photoluminescence and

crystal structure of green-emitting Ca3Sc2Si3O12 : Ce3þ phosphor for white light emitting

diodes. J Electrochem Soc 2007;154(1):J35e8.55. Saradhi MP, Varadaraju U. Photoluminescence studies on Eu2þ-Activated Li2SrSiO4 a

potential orange-yellow phosphor for solid-state lighting. Chem Mater 2006;18(22):5267e72.

56. Brgoch J, Borg CK, Denault KA, Mikhailovsky A, DenBaars SP, Seshadri R. An efficient,thermally stable cerium-based silicate phosphor for solid state white lighting. Inorg Chem2013;52(14):8010e6.

57. Arai T, Adachi S. Mn-activated Na2SiF6 red and yellowish-green phosphors: a compar-ative study. J Appl Phys 2011;110(6):063514.

58. Takahashi T, Adachi S. Mn4þ-activated red photoluminescence in K2SiF6 phosphor.J Electrochem Soc 2008;155(12):E183e8.

59. Kasa R, Adachi S. Mn-activated K2ZrF6 and Na2ZrF6 phosphors: sharp red and oscillatoryblue-green emissions. J Appl Phys 2012;112(1):013506.

60. Adachi S, Takahashi T. Direct synthesis of K2SiF6: Mn4þ red phosphor from crushedquartz schist by wet chemical etching. Electrochem Solid State Lett 2009;12(2):J20e3.

61. Adachi S, Takahashi T. Direct synthesis and properties of K2SiF6: Mn4þ phosphor by wetchemical etching of Si wafer. J Appl Phys 2008;104(2):023512.

62. Takahashi T, Adachi S. Synthesis of K2SiF6:Mn4þ red phosphor from silica glasses by wetchemical etching in HF/KMnO4 Solution. Electrochem Solid State Lett 2009;12(8):J69e71.

63. Adachi S, Takahashi T. Photoluminescent properties of K2GeF6: Mn4þ red phosphorsynthesized from aqueous HF/KMnO4 solution. J Appl Phys 2009;106(1):013516.


64. Xu YK, Adachi S. Properties of Na2SiF6: Mn4þ and Na2GeF6: Mn4þ red phosphorssynthesized by wet chemical etching. J Appl Phys 2009;105(1):013525e013525-6.

65. Adachi S, Adachi S, Abe H, Kasa R, Araial T. Synthesis and properties of hetero-dialkaline hexafluorosilicate red phosphor KNaSiF6: Mn4þ. J Electrochem Soc 2012;159(2):J34e7.

66. Hoshino R, Adachi S. Light-induced degradation in red-emitting ZnSiF6$6H2O: Mn4þ

hydrate phosphor. ECS J Solid State Sci Technol 2014;3(8):R144e9.67. Arai Y, Adachi S. Photoluminescent properties of K2SnF6$H2O: Mn4þ hydrate phosphor.

J Electrochem Soc 2011;158(3):J81e5.68. Arai Y, Takahashi T, Adachi S. Photoluminescent properties of K2SnF6$H2O: Mn4þ red

phosphor. Opt Mater 2010;32(9):1095e101.69. Zhu H, Lin CC, Luo W, Shu S, Liu Z, Liu Y, Kong J, Ma E, Cao Y, Liu RS, Chen X.

Highly efficient non-rare-earth red emitting phosphor for warm white light-emittingdiodes. Nat Commun 2014;5:4312.

70. Wei LL, Lin CC, Wang YY, Fang MH, Jiao H, Liu RS. Photo luminescent evolutioninduced by structural transformation through thermal treating in the red narrow-bandphosphor K2GeF6: Mn4þ. ACS Appl Mater Inter 2015;7(20):10656e9.

71. Sekiguchi D, Adachi S. Synthesis and photoluminescence spectroscopy of BaGeF6: Mn4þ

red phosphor. Opt Mater 2015;42:417e22.72. Zhou Q, Zhou Y, Liu Y, Luo L, Wang Z, Peng J, Yan J, Wu M. A new red phosphor

BaGeF6: Mn4þ: hydrothermal synthesis, photo-luminescence properties, and its applica-tion in warm white LED devices,. J Mater Chem C 2015;3(13):3055e9.

73. Lv L, Jiang X, Huang S, Chen X, Pan Y. The formation mechanism, improved photo-luminescence and LED applications of red phosphor K2SiF6: Mn4þ. J Mater Chem C2014;2(20):3879e84.

74. Lazarowska A, Mahlik S, Grinberg M, Lin CC, Liu RS. Pressure effect on the zero-phononline emission of Mn4þ in K2SiF6. J Chem Phys 2015;143(13).

75. Lee MJ, Song YH, Song YL, Han GS, Jung HS, Yoon DH. Enhanced luminous efficiencyof deep red emitting K2SiF6: Mn4þ phosphor dependent on KF ratio for warm-white LED.Mater Lett 2015;141:27e30.

76. Li XQ, Su XM, Liu P, Liu J, Yao ZL, Chen JJ, Yao H, Yu XB, Zhan M. Shape-controlledsynthesis of phosphor K2SiF6: Mn4þ nanorods and their luminescence properties. Crys-tEngComm 2015;17(4):930e6.

77. Ji HO, Kang H, Yun JE, Park HK, Do YR. Synthesis of narrow-band red-emitting K2SiF6:Mn4þ phosphors for a deep red monochromatic LED and ultrahigh color quality warmwhite LEDs. J MaterChem C 2015;3(3):607e15.

78. Lv L, Chen Z, Liu G, Huang S, Pan Y. Optimized photoluminescence of red phosphorK2TiF6: Mn4þ synthesized at room temperature and its formation mechanism. J MaterChem C 2015;3(9):1935e41.

79. Fang MH, Nguyen HD, Lin CC, Liu RS. Preparation of a novel red Rb2SiF6: Mn4þ

phosphor with high thermal stability through a simple one-step approach. J Mater Chem C2015;3(28):7277e80.

80. Shi HL, Zhu C, Huang JQ, Chen J, Chen DC, Wang WC, Wang FY, Cao YG, Yuan XY.Luminescence properties of YAG: Ce, Gd phosphors synthesized under vacuum conditionand their white LED performances. Opt Mater Express 2014;4(4):649e55.

81. Li JK, Li JG, Liu S, Li X, Sun X, Sakka Y. The development of Ce3þ-activated (Gd,Lu)3Al5O12 garnet solid solutions as efficient yellow-emitting phosphors. Sci Technol AdvMater 2013;14(5):054201.


82. Blasse G, Grabmaier BC. Radiative return to the ground state: Emission. LuminescentMater 1994:33e70.

83. Xia Z, Chonggeng Ma, Maxim S, Molokeev, Quanlin Liu, Karl Rickert, Kenneth R,Poeppelmeier. Chemical unit cosubstitution and tuning of photoluminescence in theCa2(Al1�xMgx)(Al1�xSi1þx)O7:Eu

2þ phosphor. J Am Chem Soc 2015;137:12494e7.84. Katelnikovas A, Sakirzanovas S, Dutczak D, Plewa J, Enseling D, Winklerd H, Kareiva A,

J€ustel T. Synthesis and optical properties of yellow emitting garnet phosphors for pc-LEDs. J Lumin 2013;136:17e25.

85. Katelnikovas A, Bettentrup H, Uhlich D, Sakirzanovas S, J€ustel T, Kareiva A. Synthesisand optical properties of Ce3þ-doped Y3Mg2AlSi2O12 phosphors. J Lumin 2009;129(11):1356e61.

86. Shimomura Y, Honma T, Shigeiwa M, Shigeiwa T, Okamoto K, Kijima N. Photo-luminescence and crystal structure of green-emitting Ca3Sc2Si3O12: Ce

3þ phosphor forwhite light emitting diodes. J Electrochem Soc 2007;154(1):J35e8.

87. Im WB, Brinkley S, Hu J. Sr2.975-xBaxCe0.025AlO4F: a highly efficient green-emittingoxyfluoride phosphor for solid state white lighting. Chem Mater 2010;22:2842e9.

88. Murata T, Tanoue T, Iwasaki M, Morinaga K, Hase T. Fluorescence properties of Mn4þ, inCaAl12O19, compounds as red-emitting phosphor for white LED. J Lumin 2005;114(3e4):207e12.

89. Cao R, Peng M, Song E, Qiu J. High efficiency Mn4þ doped Sr2MgAl22O36 red emittingphosphor for white LED. ECS J Solid State Sci Technol 2012;1:R123e6.

90. Meng L, Liang L, Wen Y. Deep red phosphors SrMgAl10O17: Mn4þ, M (M ¼ Liþ, Naþ,Kþ, Cl� ) for warm white light emitting diodes. J Mater Sci: Mater Electron 2014;25:2676e81.

91. Wang B, Lin H, Huang F, Xu J, Chen H, Lin ZB, Wang YS. Non-rare-earth BaMgAl10-2xO17: xMn4þ, xMg2þ: a narrow-band red phosphor for high-power warm w-LED. ChemMater 2016;28:3515e24.

92. Huang KW, Chen WT, Chu CI, Hu SF, Sheu HS, Cheng BM, Chen JM, Liu RS. Con-trolling the activator site to tune europium valence in oxyfluoride phosphors. Chem Mater2012;24(11):2220e7.

93. Zhang Y, Li X, Li K, Lian H, Shang M, Lin J. Crystal-site engineering control for thereduction of Eu3þ to Eu2þ in CaYAlO4: structure refinement and tunable emissionproperties. ACS Appl Mater Inter 2015;7(4):2715e25.

94. Kawano Y, Sun WK, Ishigaki T, Uematsu K, Toda K, Takaba H, Sato M. Site engineeringconcept of Ce3þ-activated novel orange-red emission oxide phosphors. Opt Mater Exp2014;4(9):1770e4.

95. Xia Z, Zhang Y, Molokeev MS, Atuchin VV, Luo Y. Linear structural evolution inducedtunable photoluminescence in clinopyroxene solid-solution phosphors. Sci Rep 2013;3(6161):995e8.

96. Sohn KS, Kwak JH, Yu SJ, Yan H, Reece MJ. Luminescence of Sr2SiO4�xN2x/3: Eu2þ

phosphors prepared by spark plasma sintering. J Electrochem Soc 2008;155(2):58e61.97. Gu Y, Zhang Q, Li Y, Wang H. Simultaneous tuning for excitation and emission of N

doped Sr2SiO4: Eu for white light LEDs. J Alloys Compd 2011;509(6):109e12.98. Lee SJ, Hong SH, Kim YJ. Synthesis and luminescent properties of (Sr, M)2Si(O1-xNx)4:

Eu2þ (M: Mg2þ, Ca2þ, Ba2þ. J Electrochem Soc 2012;159(5):J163e7.99. Zhao ZY, Shi Y, Wang C, Liu B. Red-emitting oxonitridosilicate phosphors Sr2SiNzO4-

1.5z: Eu2þ for white light-emitting diodes: structure and luminescence properties. J Mater

Chem C 2013;1:1407e12.


100. Park J, Lee SJ, Kim YJ. Evolution of luminescence of Sr2-y-zCazSi(O1-xNx)4: yEu2þ with

N3-, Eu2þ, and Ca2þ substitutions. Cryst Growth Des 2013;13(13):5204e10.101. Ju LC, Xu X, Hao LY, Lin Y, Lee MH. Modification of the coordination environment of

Eu2þ in Sr2SiO4: Eu2þ phosphors to achieve full color emission. J Mater Chem C 2015;

3(7):1567e75.102. Li X, Hua Y, Ma H, Deng D, Jia G. Luminescence properties and crystal structure of a’-

Sr2Si3x/4O2Nx: Eu2þ phosphors with different concentrations of N3� ions. RSC Adv 2015;

5(77):62659e69.103. Tian W, Song K, Zhang F, Zheng P, Deng J. Optical spectrum adjustment of yellow-green

Sr1.99SiO4�3x/2Nx: 0.01Eu2þ phosphor powders for near ultraviolet-visible light applica-

tion. J Alloys Compd 2015;638(25):249e53.104. Black AP, Denault KA, Or�o-Solé J, Go~ni AR, Fuertes A. Red luminescence and ferro-

magnetism in europium oxynitridosilicates with a b-K2SO4 structure. Chem Commun2015;51(11):2166e9.

105. Liu Y, Zhang X, Hao Z, Luo Y, Wang X, Zhang J. Crystal structure and luminescenceproperties of Lu3þ and Mg2þ incorporated silicate garnet [Ca3-(xþ0.06)LuxCe0.06](Sc2-yMgy)Si3O12. J Lumin 2012;132:1257e60.

106. Prasanth NK, Kishore MS, Chandran RG, Setlur A. Controlling Ce3þ emission color insilicate garnet phosphor. ECS Trans 2009;16:57e64.

107. Shimomura Y, Kurushima T, Shigeiwa M, Kijima N. Red shift of green photo-luminescence of Ca3Sc2Si3O12: Ce3þ phosphor by charge compensatory additives.J Electrochem Soc 2008;155:J45e9.

108. Kishore MS, Kumar NP, Chandran RG, Setlur AA. Solid solution formation and Ce(3þ)

luminescence in silicate garnets. Electrochem Solid State Lett 2010;13:J77e80.109. Zhang X, Liu Y, Hao Z, Wang X, Zhang J. Yellow-emitting (Ca2Lu1�xCex)(ScMg)

Si3O12 phosphor and its application for white LEDs. Mater Res Bull 2012;47:1149e52.

110. Pan F, Zhou M, Zhang J, Zhang X, WAng J, Huang L, WuM. Double substitution inducedtunable luminescent properties of Ca3�xYxSc2�xMgxSi3O12: Ce

3þ phosphors for whiteLEDs. J Mater Chem C 2016;4:5671e8.

111. Wei Y, Lin CC, Quan Z, Molokeev MS, Atuchin VV, Chan TS, Li G. Structural evolutioninduced preferential occupancy of designated cation sites by Eu2þ in M5(Si3O9)2 (M¼ Sr,Ba, Y, Mn) phosphors. RSC Adv 2016;6:57261e5.

112. Singh K, Vaidyanathan S. Eu2þ luminescence in Ca3Si2O7 and spectral widening andtuning of Eu2þ emission color (orangish-red to green) by crystal chemical substitution.RSC Adv 2016;6:98652.

113. Wu YC, Chen YC, Chen TM, Lee CS, Chen KJ, Kuo HC. Crystal structure character-ization, optical and photoluminescent properties of tunable yellow to orange-emittingY2(Ca, Sr)F4S2: Ce

3þ phosphors for solid-state lighting. J Mater Chem 2012;22(16):8048e56.

114. Wu YC, Chen YC, Wang DY, Lee CS, Sun CC, Chen TM. a-(Y, Gd)FS: Ce3þ: a novelred-emitting fluorosulfide phosphor for solid-state lighting. J Mater Chem 2011;21(39):15163e6.

115. Demourgues A, Tressaud A, Laronze H, Macaudiere P. Rare earth fluorosulfides LnSF andLn2AF4S2 as new colour pigments. J Alloys Compd 2001;323e324(01):223e30.

116. Yu R, An Y, Wang C, Wang H, Wu Y, Zhang J. Tunable yellowish-green to green (Ca1-xSrx)LaGa3S6O: Eu

2þ phosphors for potential LED application. Electrochem Solid StateLett 2012;15(1):J1e5.


117. Teske C, Teske Chr L. €Uber oxidsulfide mit Åkermanitstruktur CaLaGa3S6O, SrLa-Ga3S6O, La2ZnGa2S6O and Sr2ZnGe2S6O, Z. Anorg Allg Chem 1985;531:52e60.

118. Chartier C, Jabbarov R, Jouanne M, Morhange JF, Benalloul P, Barthou C, Gambarov E.Raman investigation of orthorhombic MII Ga2(S, Se)4 compounds. J Phys Condens Mat2002;14(14):13693e703.

119. Do YR, Ko KY, Na SH, Huh YD. Luminescence properties of potential Sr1�xCaxGa2S4:Eu green and greenish-yellow-emitting phosphors for white LED. J Electrochem Soc2006;153(7):H142e6.

120. Yu RJ, Li HJ, Ma HL, Wang CF, Wang H, Moon BK, Jeong JH. Photoluminescenceproperties of a new Eu2þ-activated CaLaGa3S7, yellowish-green phosphor for white LEDapplications. J Lumin 2012;132(10):2783e7.

121. Lin CC, Xiao ZR, Guo GY, Chan TS, Liu RS. Versatile phosphate phosphors ABPO4 inwhite light-emitting diodes: collocated characteristic analysis and theoretical calculations.J Am Chem Soc 2010;132:3020e8.

122. Ji H, Huang Z, Xia Z, Molokeev MS, Atuchin VV, Fang M, Huang S. New yellow-emitting whitlockite-type structure Sr1.75Ca1.25(PO4)2: Eu2þ phosphor for near-UVpumped white light-emitting devices. Inorg Chem 2014;53:5129e35.

123. Huang CH, Liu WR, Chen TM. Single-phased white-light phosphors Ca9Gd(PO4)7: Eu2þ,

Mn2þ under near-ultraviolet excitation. J Phys Chem C 2010;114:18698e701.124. Huang CH, Chiu YC, Yeh YT, Chan TS, Chen TM. Eu2þ-activated Sr8ZnSc(PO4)7: a

novel near-ultraviolet converting yellow-emitting phosphor for white light-emittingdiodes. ACS Appl Mater Inter 2012;4:6661e8.

125. Huang CH, Chen TM. Novel yellow-emitting Sr8MgLn(PO4)7: Eu2þ (Ln ¼ Y, La)phosphors for applications in white LEDs with excellent color rendering index. InorgChem 2011;50:5725e30.

126. Guo N, Huang Y, Jia Y, Lv W, Zhao Q, L€u W, Xia Z, You H. A novel orange-yellow-emitting Ba3Lu(PO4)3: Eu

2þ, Mn2þ phosphor with energy transfer for UV-excited whiteLEDs,. Dalton Trans 2013;42:941e7.

127. Wang Z, Xia Z, Molokeev MS, Atuchin VV, Liu Q. Blue-shift of Eu2þ emission in (Ba,Sr)3Lu(PO4)3: Eu2þ eulytite solid-solution phosphors resulting from release ofneighbouring-cation-induced stress. Dalton Trans 2014;43:16800e4.

128. Dai P, Lee SP, Chan TS, Huang CH, Chiang YW, Chen TM. Sr3Ce(PO4)3: Eu2þ: a

broadband yellow-emitting phosphor for near ultraviolet-pumped white light-emittingdevices. J Mater Chem C 2016;4:1170e7.

129. Deng D, Yu H, Li Y, Hua Y, Jia G, Zhao S, Wang H, Huang L, Li Y, Li C, Xu S.Ca4(PO4)2O: Eu

2þ red-emitting phosphor for solid-state lighting: structure, lumines-cent properties and white light emitting diode application. J Mater Chem C 2013;1:3194e9.

130. Jia Y, Pang R, Li H, Sun W, Fu J, Jiang L, Zhang S, Su Q, Li C, Liu RS. Single-phasedwhite-light-emitting Ca₄(PO₄)₂O: Ce3þ,Eu2þ phosphors based on energy transfer. DaltonTrans 2015;44:11399e407.

131. Komuro N, Mikami M, Saines PJ, Cheetham AK. Structureeproperty correlations in Eu-doped tetra calcium phosphate phosphor: a key to solid-state lighting application,. J Lumin2015;162:25e30.

132. Komuro N, Mikami M, Saines PJ, Akimoto K, Cheetham AK. Deep red emission in Eu2þ-activated Sr4(PO4)2O phosphors for blue-pumped white LEDs. J Mater Chem C 2015;3:7356e62.


133. Komuro N, Mikami M, Shimomura Y, Bithell EG, Cheetham AK. Synthesis, structure andoptical properties of europium doped calcium barium phosphate e a novel phosphor forsolid-state lighting. J Mater Chem C 2014;2:6084e9.

134. Chen M, Xia Z, Liu Q. Improved optical photoluminescence by charge compensation andluminescence tuning in Ca6Ba(PO4)4O: Ce

3þ, Eu2þ phosphors. CrystEngComm 2015;17:8632e8.

135. Komuro N, Mikami M, Shimomura Y, Bithell EG, Cheetham AK. Synthesis, structure andoptical properties of cerium-doped calcium barium phosphate e a novel blue-greenphosphor for solid-state lighting. J Mater Chem C 2015;3:204e10.

136. Wu L, Chen XL, Li H, He M, Xu YP, Li XZ. Structure determination and relativeproperties of novel cubic borates MM’4(BO3)3(M ¼ Li, M’ ¼ Sr; M ¼ Na, M’ ¼ Sr, Ba).Inorg Chem 2005;44:6409e14.

137. Guo CF, Ding X, Seo HJ, Ren ZY, Bai JT. Luminescent properties of UV excitable blueemitting phosphors MSr4(BO3)3:Ce

3þ (M ¼ Li and Na). J Alloys Compd 2011;509(14):4871e4.

138. Guo CF, Luan L, Xu Y, Gao F, Liang LF. White lightegeneration phosphor Ba2Ca(BO3)2:Ce3þ , Mn2þ for light-emitting diodes. J Electrochem Soc 2008;155:J310e4.

139. Zhang XG, Gong ML. Single-phased white-light-emitting NaCaBO3: Ce3þ, Tb3þ, Mn2þ

phosphor for LED applications. Dalton Trans 2014;43:2465e72.140. Wang Z, Chu IH, Zhou F, Ong SP. Electronic structure descriptor for the discovery of

narrow-band red-emitting phosphors. Chem Mater 2016;28:4024e31.141. Oh JH, Eo YJ, Yoon HC, Huh YD, Do YR. Evaluation of new color metrics: guidelines

for developing narrow-band red phosphors for WLEDs. J Mater Chem C 2016;4:8326e48.

142. Hirosaki N, Xie RJ, Kimoto K, Sekiguchi T, Yamamoto Y, Suehiro T, Mitomo M.Characterization and properties of green-emitting a

ˇ

-SiAlON: Eu2þ powder phosphors forwhite light-emitting diodes. Appl Phys Lett 2005;86:211905.

143. Wagatha P, Pust P, Weiler V, Wochnik AS, Schmidt PJ, Scheu C, Schnick W.Ca18.75Li10.5[Al39N55]: Eu2þ-supertetrahedron phosphor for solid-state lighting. ChemMater 2016;28:1220e6.

144. Jin Y, FangMH, Grinberg M,Mahlik S, Lesniewski T, Brik MG, Luo GY, Lin JG, Liu RS.Narrow red emission band fluoride phosphor KNaSiF6: Mn4þ for warm white light-emitting diodes. ACS Appl Mater Inter 2016;8:11194e203.

145. Huang L, Zhu YW, Zhang XJ, Zou R, Pan FJ, Wang J, Wu MM. HF-free hydrothermalroute for synthesis of highly efficient narrow-band red emitting phosphor K2Si1�xF6:xMn4þ for warm white light-emitting diodes. Chem Mater 2016;28:1495e502.

146. Arai T, Arai Y, Takahashi T, Adachi S. A yellow phosphor K2SiF6 activated by Mn2þ

ions. J Appl Phys 2010;108:063506-1e063506e063507.147. Nguyen HD, Lin CC, Liu RS. Waterproof alkyl phosphate coated fluoride phosphors for

optoelectronic materials. Angew Chem Int Ed 2015;54:10862e6.148. Xu YK, Adachi S. Properties of Mn4þ-activated hexafluorotitanate phosphors.

J Electrochem Soc 2011;158:J58e65.149. Nguyen HD, Lin CC, Fang MH, Liu RS. Synthesis of Na2SiF6: Mn4þ red phosphors for

white LED applications by Co-Precipitation. J Mater Chem C 2014;2:10268.150. Peng MY, Yin XW, Tanner PA, Brik MG, Li PF. Site occupancy preference, enhancement

mechanism, and thermal resistance of Mn4þ red luminescence in Sr4Al14O25: Mn4þ forwarm WLEDs. Chem Mater 2015;27:2938e45.


151. Wang B, Lin H, Xu J, Chen H, Wang YS. CaMg2Al16O27: Mn4þ-Based red phosphor: apotential color converter for high-powered warm W-LED. ACS Appl Mater Inter 2014;6:22905e13.

152. Brik MG, Pan YX, Liu GK. Spectroscopic and crystal field analysis of absorption andphotoluminescence properties of red phosphor CaAl12O19: Mn4þ modified by MgO.J Alloys Compd 2011;509:1452e6.

153. Pan YX, Liu GK. Enhancement of phosphor efficiency via composition modification. OptLett 2008;33:1816e8.

154. Huang XY. Solid-state lighting red phosphor converts white LEDs. Nat Photon 2014;8:748e9.

155. Vaidyanathan S, Jeon DY. A novel narrow band red-emitting phosphor for white lightemitting diodes. Int J Appl Ceram Technol 2009;6:453e8.

156. Sivakumar V, Varadaraju UV. Intense red-emitting phosphors for white light emittingdiodes. J Electrochem Soc 2005;152:H168e71.

157. Ju G, Hu Y, Chen L, Wang X, Mu Z. Blue persistent luminescence in Eu2þ dopedCa3Mg3(PO4)4. Opt Mater 2014;36:1183e8.

158. Li K, Chen D, Xu J, Zhang R, Yu Y, Wang Y. Phase transition and multicolor lumi-nescence of Eu2þ/Mn2þ-activated Ca3(PO4)2 phosphors.MaterRes Bull 2014;49:677e81.

159. RaoBandi V, Grandhe BK, Jang K, Kim SS, Shin DS, Lee YI, Lim JM, Song T. Lumi-nescent properties of a new green emitting Eu2þ doped CaZrSi2O7 phosphor for WLEDapplications. J Lumin 2011;131:2414e8.

160. Zhang X, Lu Z, Meng F, Lu F, Hu L, Xu X, Tang C. A yellow-emitting Ca3Si2O7: Eu2þ

phosphor for white LEDs. Mater Lett 2012;66:16e8.161. Krings M, Montana G, Dronskowski R, Wickleder C. �a-SrNCN:Eu2þ e a novel efficient

orange-emitting phosphor. Chem Mater 2011;23:1694e5.162. Tao Z, Huang Y, Seo HJ. Blue luminescence and structural properties of Ce(3þ)-activated

phosphosilicate apatite Sr5(PO4)2(SiO4). Dalton Trans 2013;42:2121e9.163. Gong X, Huang J, Chen Y, Lin Y, Luo Z, Huang Y. Novel garnet-structure Ca2GdZ-

r2(AlO4)3:Ce3þ phosphor and its structural tuning of optical properties. Inorg Chem 2014;

53:6607.164. Kim JH, Jung KY. Luminescence characteristics and optimization of (La, Gd)Sr2(Al,B)

O5:Ce phosphor for white light emitting diodes. J Lumin 2012;132:1376e81.165. Wang X, Zhou G, Zhang H, Li H, Zhang Z, Sun Z. Luminescent properties of yellowish

orange Y3Al5�xSixO12�xNx:Ce phosphors and their applications in warm white light-emitting diodes. J Alloys Compd 2012;519:149e55.

166. Yang Z, Zhao Z, Wen Y, Wang Y. Structure and luminescence properties of Bi3þ activatedCa12Al14O32Cl2 phosphors. J Alloys Compd 2013;559:142e5.

167. Ju H, Deng W, Wang B, Liu J, Tao X, Xu S. The structure and luminescence properties ofgreen Ca3Al2O6:Bi

3þ phosphors. J Alloys Compd 2012;516:153e6.168. Kang F, Peng M, Zhang Q, Qiu J. Abnormal anti-quenching and controllable multi-

transitions of Bi3þ luminescence by temperature in a yellow-emitting LuVO4:Bi3þ

phosphor for UV-converted. ChemEur J 2014;20:11522e30.169. Kang F, Zhang Y,Wondraczek L, Zhu J, Yang X, PengM. Processing-dependence and the

nature of the blue-shift of Bi3þ-related photoemission in ScVO4 at elevated temperatures.J Mater Chem C 2014;2:9850e7.

170. Haranath D, Mishra S, Yadav S, Sharma RK, Kandpal LM, Vijayan N, Dalai MK,Sehgal G, Shanker V. Rare-earth free yellow-green emitting NaZnPO4:Mn phosphor forlighting applications. Appl Phys Lett 2014;101:221905.


171. Lei B, Liu Y, Ye Z, Shi C. Luminescence properties of CdSiO3:Mn2þ phosphor. J Lumin2004;109:215e9.

172. Lv J, Du F, Zhu R, Huang Y, Seo HJ. Phase formations and tunable red luminescence ofNa2CaMg1�xMnx(PO4)2 (x ¼ 0.05e1.0). J Mater Chem 2011;21:16398e405.

173. Zhou X, Yang X, Xiao T, Zhou K, Chen T, Yan H, Wang Z. Luminescence properties andenergy transfer of host sensitized CaMoO4:Tb

3þ green phosphors. J Rare Earths 2013;31:655e9.

174. Rui Z, Xiang W. Preparation and luminescent characteristics of Sr3RE2(BO3)4:Dy3þ

(RE¼Y, La, Gd) phosphors for white LED. J Alloys Compd 2011;509:1197e200.175. Zhang Z, Mao Z, Song S, Zhang J, Liu L, Zhang W, Wang D. Preparation and investi-

gation of (Sr0.85Mg0.14)3(P1�xSixO4)2: Dy3þ single-phase full-color phosphor. Mater Lett

2013;90:1e3.176. Tian Y, Chen B, Tian B, Hu R, Sun J, Cheng L, Zhong H, Li X, Zhang J, Zhen Y, Yu T,

Huang L, Meng Q. Concentration-dependent luminescence and energy transfer of flower-like Y2(MoO4)3:Dy

3þ phosphor. J Alloys Compd 2011;509:6096e101.177. Lin X, Qiao X, Fan X. Synthesis and luminescence properties of a novel red SrMoO4:

Sm3þ, Rþ phosphor. Solid State Sci 2011;13:579e83.178. Sun X, Hao Z, Li C, He X, Qi H, Yu L, Luo Y, Zhang J, Gao J, Zhong R. Enhanced

orange-red emission by using Mo codoped in Ba2CaWO6:Eu3þ, Liþ phosphor under near

UV excitation. J Lumin 2013;134:191e4.179. Xin S, Wang Y, Zhu G, Ding X, Geng W, Wang Q. Structure- and temperature-sensitive

photoluminescence in a novel phosphate red phosphor RbZnPO4:Eu3þ. Dalton Trans

2015;44:16099e106.180. Zhu Y, Liang Y, Zhang M, Tong M, Li G, Wang S. Structure, luminescence properties and

energy transfer behavior of color-adjustable Sr3Gd2(Si3O9)2:Ce3þ, Tb3þ/Mn2þ phosphors.

RSC Adv 2015;5:98350e60.181. Zhou J, Wang T, Yu X, Zhou D, Qiu J. The synthesis and photoluminescence of a single-

phased white-emitting NaAlSiO4:Ce3þ,Mn2þ, phosphor for WLEDs. Mater Res Bull

2016;73:1e5.182. Ci Z, Sun Q, Sun M, Jiang X, Qin S, Wang Y. Structure, photoluminescence and thermal

properties of Ce3þ, Mn2þ co-doped phosphosilicate Sr7La3[(PO4)2.5(SiO4)3(BO4)0.5](BO2)emission-tunable phosphor. J Mater Chem C 2014;2:5850e6.

183. Wei D, Sun Y, Jiang L, Hua S, Lia D. Tunable luminescence and energy transfer propertiesof a novel Na4Ca4Si6O18:Ce

3þ,Mn2þ phosphor. New J Chem 2015;39:4753e8.184. Lv W, L€u W, Guo N, Jia Y, Zhao Q, Jiao M, Shao B, You H. Crystal structure and

luminescent properties of a novel high efficiency blue-orange emitting NaCa2LuSi2O7F2:Ce3þ, Mn2þ phosphor for ultraviolet light-emitting diodes. Dalton Trans 2013;42:13071e7.

185. Lv W, Luo Y, Hao Z, Zhang X, Wang X, Zhang J. A new dual-emission phosphorCa4Si2O7F2:Ce

3þ,Mn2þ with energy transfer for near-UV LEDs. Mater Lett 2012;77:45e7.

186. Jiao M, Jia Y, L€u W, Lv W, Zhao Q, Shao B, You H. A single-phase white-emittingCa2SrAl2O6:Ce

3þ,Liþ,Mn2þ phosphor with energy transfer for UV-excited WLEDs.Dalton Trans 2014;43:3202e9.

187. Li Y, Shi Y, Zhu G, Wu Q, Li H, Wang X, Wang Q, Wang Y. A single-component white-emitting CaSr2Al2O6:Ce

3þ,Liþ,Mn2þ phosphor via energy transfer. Inorg Chem 2014;53:7668e75.


188. Jia Y, Huang Y, Guo N, Qiao H, Zheng Y, Lv W, Zhao Q, You H. Mg1.5Lu1.5Al3.5-Si1.5O12: Ce

3þ,Mn2þ: a novel garnet phosphor with adjustable emission color for bluelight-emitting diodes. RSC Adv 2012;2:2678e81.

189. Liu Y, Zhang X, Hao Z, Luo Y,Wang X, Zhang J. Generating yellow and red emissions byco-doping Mn2þ to substitute for Ca2þ and Sc3þ sites in Ca3Sc2Si3O12:Ce

3þ greenemitting phosphor for white LED applications. J Mater Chem 2011;21:16379e84.

190. Liu Y, Zhang X, Hao Z, Luo Y, Wang X, Zhang J. Tunable full-color-emitting Ca3Sc2-Si3O12:Ce

3þ,Mn2þ phosphor via charge compensation and energy transfer. Chem Com-mun 2011;47:10677e9.

191. Han B, Zhang J, Li P, Li J, Bian Y, Shi H. Luminescence and energy transfer of emissiontunable phosphors YBa3B9O18:Ce

3þ,Tb3þ. Optik 2015;126:1851e4.192. Jiang T, Yu X, Xu X, Yu H, Zhou D, Qiu J. Tunable emitting-color and energy transfer of

KCaY(PO4)2:Ce3þ,Tb3þ phosphors for UV-excited white light-emitting-diodes. Mater

Res Bull 2014;51:80e4.193. Guo C, Jing H, Li T. Green-emitting phosphor Na2Gd2B2O7:Ce

3þ,Tb3þ for near-UVLEDs. RSC Adv 2012;2:2119e22.

194. Guo N, Song Y, You H, Jia G, Yang M, Liu K, Zheng Y, Huang Y, Zhang H. Opticalproperties and energy transfer of NaCaPO4:Ce

3þ,Tb3þ phosphors for potential applicationin light-emitting diodes. Eur J Inorg Chem 2010;29:4636e42.

195. Guo Q, Liao L, Molokeev MS, Mei L, Liu H. Color tunable emission and energy transferof Ce3þ and Tb3þ co-doped novel La6Sr4(SiO4)6F2 phosphors with apatite structure.Mater Res Bull 2015;72:245e51.

196. Guo Q, Liao L, Xia Z. Luminescence properties and energy transfer in La6Ba4(SiO4)6F2:Ce3þ,Tb3þ phosphors. J Lumin 2014;145:65e70.

197. Wen Y, Wang Y, Zhang F, Liu B. Near-ultraviolet excitable Ca4Y6(SiO4)6O: Ce3þ,Tb3þ

white phosphors for light-emitting diodes. Mater Chem Phys 2011;129:1171e5.198. Zhou X, Zhang Z, Wang Y. Ce3þ and Tb3þ singly-and co-doped MgGd4Si3O13 for ul-

traviolet light emitting diodes and field emission displays. J Mater Chem C 2015;3:3676e83.

199. Guo Q, Liao L, Mei L, Liu H, Hai Y. Color-tunable photoluminescence phosphors of Ce3þ

and Tb3þ co-doped Sr2La8(SiO4)6O2 for UV w-LEDs. J Solid State Chem 2015;225:149e54.

200. Li K, Shang M, Geng D, Lian H, Zhang Y, Fan J, Lin J. Synthesis, luminescence, andenergy-transfer properties of a

ˇ

-Na2Ca4(PO4)2(SiO4):A(A¼Eu2þ, Dy3þ, Ce3þ/Tb3þ)phosphors. Inorg Chem 2014;53:6743e51.

201. Xia Z, Wu W. Preparation and luminescence properties of Ce3þ and Ce3þ/Tb3þ-activatedY4Si2O7N2 phosphors. Dalton Trans 2013;42:12989e97.

202. Li K, Chen D, Zhang R, Yu Y, Xu J, Wang Y. Enhanced luminescence in Ce3þ/Dy3þ:Sr3Y2(BO3)4 phosphors via energy transfer. Mater Res Bull 2013;48:1957e60.

203. Yuan B, Song Y, Sheng Y, Zheng K, Zhou X, Ma P, Xu X, Zou H. Tunable color andenergy transfer in single-phase white-emitting Ca20Al26Mg3Si3O68:Ce

3þ,Dy3þ phosphorsfor UV white light-emitting diodes. J Solid State Chem 2015;232:169e77.

204. Xu M, Wang L, Jia D, Liu L. Enhancing the luminescent properites of Zn2P2O7: Ce3þ,

Dy3þ via efficient energy transfer. Mater Res Bull 2015;70:691e6.205. Yang Z, Liu P, Lv L, Zhao Y, Yu Q, Liang X. Fluorescence properties and energy transfer

of KNaCa2�x�y(PO4)2:xCe3þ,yDy3þ phosphors under ultraviolet excitation. J Alloys

Compd 2013;562:176e81.206. Li M, Xia Z, Wang Z. Color-tunable emission and energy transfer studies in GdOBr:Ce3þ

,Dy3þ phosphor. Opt Mater 2014;37:446e50.


207. Xie M, Ou S, Liang H, Hou D, Huang Y, Gao Z, Tao Y. White-emitting phosphorsCa6La2Na2(PO4)6F2:Dy

3þ and luminescence enhancement through Ce3þ/Dy3þ energytransfer. Mater Chem Phys 2013;142:339e44.

208. Zhou L, Liang H, Tanner PA, Zhang S, Hou D, Liu C, Tao Y, Huang Y, Li L.Luminescence, cathode luminescence and Ce3þ / Eu2þ energy transfer andemission enhancement in the Sr5(PO4)3Cl:Ce

3þ, Eu2þ phosphor. J Mater Chem C 2013;1:7155e65.

209. Chen H, Li C, Hua Y, Yu L, Jiang Q, Deng D, Zhao S, Ma H, Xu S. Influence of energytransfer from Ce3þ to Eu2þ on luminescence properties of Ba3Si6O9N4:Ce

3þ, Eu2þ

phosphors. Ceram Int 2014;40:1979e83.210. Jia Y, Qiao H, Zheng Y, Guo N, You H. Synthesis and photoluminescence properties of

Ce3þ and Eu2þ-activated Ca7Mg(SiO4)4 phosphors for solid state lighting. Phys ChemChem Phys 2012;14:3537e42.

211. Li G, Li M, Li L, Yu H, Zou H, Zou L, Gan S, Xu X. Luminescent properties ofSr2Al2SiO7:Ce

3þ, Eu2þ phosphors for near UV-excited white light-emitting diodes.MaterLett 2011;65:3418e20.

212. Miao S, Xia Z, Molokeev MS, Zhang J, Liu Q. Crystal structure refinement andluminescence properties of blue-green-emitting CaSrAl2SiO7:Ce

3þ, Liþ, Eu2þ phosphors.J Mater Chem C 2015;3:8322e8.

213. Chen M, Xia Z, Molokeev MS, Liu Q. Insights into Ba4Si6O16 structure and photo-luminescence tuning of Ba4Si6O16:Ce

3þ, Eu2þ phosphors. J Mater Chem C 2015;3:12477e83.

214. Song Y, Jia G, Yang M, Huang Y, You H, Zhang H. Sr3 Al2O5Cl2: Ce3þ,Eu2þ: a potential

tunable yellow-to-white-emitting phosphor for ultraviolet light emitting diodes. Appl PhysLett 2009;94. 091902-1e091902-3.

215. Li G, Zhang Y, Geng D, Shang M, Peng C, Cheng Z, Lin J. Single-compositiontrichromatic white-emitting Ca4Y6(SiO4)6O: Ce3þ/Mn2þ/Tb3þphosphor: luminescenceand energy transfer. ACS Appl Mater Inter 2012;4:296e305.

216. Liu H, Luo Y, Mao Z, Liao L, Xia Z. A novel single-composition trichromaticwhite-emitting Sr3.5Y6.5O2(PO4)1.5(SiO4)4.5:Ce

3þ/Tb3þ/Mn2þ phosphor: synthesis, lumi-nescent properties and applications for white LEDs. J Mater Chem C 2013;2:1619e27.

217. Li K, Shang M, Zhang Y, Fan J, Lian H, Lin J. Photoluminescence properties of single-component white-emitting Ca9Bi(PO4)7:Ce

3þ,Tb3þ,Mn2þ phosphors for UV LEDs.J Mater Chem C 2015;3:7096e104.

218. Geng D, Shang M, Zhang Y, Lian H, Cheng Z, Lin J. Tunable luminescence and energytransfer properties of Ca5(PO4)2SiO4:Ce

3þ/Tb3þ/Mn2þ phosphors. J Mater Chem C 2013;1:2345e53.

219. Zhang J, He Y, Qiu Z, Zhang W, Zhou W, Yu L, Lian S. Site-sensitive energy transfermodes in Ca3Al2O6: Ce

3þ/Tb3þ/Mn2þ phosphors. Dalton Trans 2014;43:18134.220. Zhang Y, Geng D, ShangM,Wu Y, Li X, Lian H, Cheng Z, Lin J. Eur., single-composition

trichromatic white-emitting Ca9MgNa(PO4)7:Ce3þ/Tb3þ/Mn2þ phosphorse Soft chemical

synthesis, luminescence, and energy-transfer properties. J Inorg Chem 2013;25:4389e97.221. Wang Z, Li P, Yang Z, Guo Q, Dong G. A white emitting phosphor BaMg2(PO4)2:Ce

3þ,Mn2þ, Tb3þ : luminescence and energy transfer. Ceram Int 2014;40:15283e92.

222. Wen D, Shi J. A novel narrow-line red emitting Na2Y2B2O7:Ce3þ,Tb3þ,Eu3þ phosphor

with high efficiency activated by terbium chain for near-UV white LEDs. Dalton Trans2013;42:16621e9.

223. Laulicht I, Meirman S. Direct evidence for excitation transfer from the 5D4, manifold ofTb3þ, to the 5D1, manifold of Eu3þ, in Tb0.66Eu0.33P5O14. J Lumin 1986;34:287e93.


224. Bettinelli M, Speghini A, Piccinelli F, Ueda J, Tanabe S. Energy transfer processes inSr3Tb0.90Eu0.10 (PO4)3. Opt Mater 2010;33:119e22.

225. Zhou J, Xia Z. Luminescence color tuning of Ce3þ, Tb3þ and Eu3þ codoped and tri-dopedBaY2Si3O10 phosphors via energy transfer. J Mater Chem C 2015;3:7552e60.

226. Zhang X, Zhou L, Pang Q, Shi J, Gong M. Tunable luminescence andCe3þ/Tb3þ/Eu3þ energy transfer of broadband-excited and narrow line red emittingY2SiO5:Ce

3þ, Tb3þ, Eu3þ phosphor. J Phys Chem C 2014;118:7591e8.227. Zhang X, Chen Y, Zhou L, Pang Q, Gong M. Synthesis of a broad-band excited and

multicolor tunable phosphor Gd2SiO5: Ce3þ, Tb3þ, Eu3þ for near-ultraviolet light-emitting diodes. Ind Eng Chem Res 2014;53:6694e8.

228. Zhang X, Zhou L, Pang Q, Gong M. A broadband-excited and narrow-line GdBO3: Ce3þ,

Tb3þ, Eu3þ red phosphor with efficient Ce3þ/(Tb3þ)n/ Eu3þ energy transfer for NUVLEDs. Opt Mater 2014;36:1112e8.

229. Zhang X, Zhang J, Gong M. Luminescence and energy transfer of La5Si2BO13: a (A¼Ce3þ/Tb3þ/Eu3þ/Sm3þ) phosphors under UV excitation. Mater Lett 2015;143:71e4.

230. Yu H, Yu X, Xu X, Zhou D, Qiu J. Realization of enhanced sensitization effect inCaYAl3O7:Ce

3þ, Sm3þ phosphors via Tb3þ ions. ECS J Solid State Sci Technol 2014;3(12):R245e50.

231. Liu L, Wang L, Zhang C, Cho Y, Dierre B, Hirosaki N, Sekiguchi T, Xie R-J. Strongenergy-transfer-induced enhancement of luminescence efficiency of Eu2þ-and Mn2þ-codoped gamma-AlON for near-UV-LED-pumped solid state lighting. Inorg Chem 2015;54:5556e65.

232. Li K, Geng D, Shang M, Zhang Y, Lian H, Lin J. Color-tunable luminescence and energytransfer properties of Ca9Mg(PO4)6F2:Eu

2þ, Mn2þ phosphors for UV-LEDs. J Phys ChemC 2014;118:11026e34.

233. Guo N, You H, Jia C, Ouyang R, Wu D. A Eu2þ and Mn2þ-coactivated fluoro-apatite-structure Ca6Y2Na2(PO4)6F2 as a standard white-emitting phosphor via energy transfer.Dalton Trans 2014;43:12373e9.

234. Feng Y, Huang J, Liu L, Liu J, Yu X. Enhancement of white-light-emission from single-phase Sr5(PO4)3F: Eu

(2þ),Mn(2þ) phosphors for near-UV white LEDs. Dalton Trans 2015;44:15006.

235. Geng D, Shang M, Zhang Y, Cheng Z, Lin J. Tunable and white-light emission fromsingle-phase Ca2YF4PO4:Eu

2þ, Mn2þ phosphors for application in W-LEDs. Eur J InorgChem 2013;16:2947e53.

236. Huang C-H, Chan T-S, Liu W-R, Wang D-Y, Chiu Y-C, Yeh Y-T, Chen T-M. Crystalstructure of blueewhiteeyellow color-tunable Ca4Si2O7F2: Eu

2þ, Mn2þ phosphor andinvestigation of color tunability through energy transfer for single-phase white-lightnear-ultraviolet LEDs. J Mater Chem 2012;22:20210e6.

237. Li P, Wang Z, Yang Z, Guo Q. A novel, warm, white light-emitting phosphor Ca2PO4Cl:Eu2þ, Mn2þ for white LEDs,. J Mater Chem C 2014;2:7823e9.

238. Choi S, Yun YJ, Jung H-K. Eu2þ, and Mn2þ, activated single phase white emittingphosphor Na(Sr,Ba)PO4 for phosphor converted-LEDs. Mater Lett 2012;75:186e8.

239. Guo M, Huang L, Zhao S, Deng D, Wang H, Hua Y, Jia G, Xu S. Fluorine effects onSr1.7Mg0.3SiO4:0.01Eu

2þ,0.08Mn2þ, phosphors for white light-emitting diodes. J Lumin2013;134:96e9.

240. Guo N, Huang Y, You H, Yang M, Song Y, Liu K, Zheng Y. Ca9Lu(PO4)7:Eu2þ,Mn2þ: a

potential single-phased white-light-emitting phosphor suitable for white-light-emittingdiodes. Inorg Chem 2010;49:10907.


241. Kwon KH, Im WB, Jang HS, Yoo HS, Jeon DY. Luminescence properties and energytransfer of site-sensitive Ca6�x�yMgx�z(PO4)4:Euy

2þ,Mnz2þ phosphors and their application

to near-UV LED-based white LEDs. Inorg Chem 2009;48:11525.242. Xie N, Liu J, Huang Y, Kim SI, Seo HJ. Tunable luminescence properties and efficient

energy transfer in Eu2þ, Mn2þ co-doped Ba2MgP4O13. Ceram Int 2012;38:1489e95.243. Liu W-R, Huang C-H, Yeh C-W, Tsai J-C, Chiu Y-C, Yeh Y-T, Liu R-S. A study on the

luminescence and energy transfer of single-phase and color-tunable KCaY(PO4)2:Eu2þ,Mn2þ phosphor for application in white-light LEDs,. Inorg Chem 2012;51:9636e41.

244. Xia Z, Zhang Y, Molokeev MS, Atuchin VV. Structural and luminescence properties ofyellow-emitting NaScSi2O6:Eu

2þ phosphors: Eu2þ site preference analysis and generationof red emission by codoping Mn2þ for white-light-emitting diode applications. J PhysChem C 2013;117:20847e54.

245. Guo N, Huang Y, Yang M, Song Y, Zheng Y, You H. Phys. A tunable single-componentwarm white-light Sr3Y(PO4)3: Eu

2þ, Mn2þ phosphor for white-light emitting diodes.Chem Chem Phys 2011;13:15077e82.

246. Guo N, Zheng Y, Jia Y, Qiao H, You H. Warm-white-emitting from Eu2þ/Mn2þ-codopedSr3Lu (PO4)3 phosphor with tunable color tone and correlated color temperature. J PhysChem C 2012;116:1329e34.

247. Guo N, Jia Y, L€u W, Lv W, Zhao Q, Jiao M, Shao B, You H. A direct warm-white-emitting Sr3Sc(PO4)3:Eu

2þ,Mn2þ phosphor with tunable photoluminescence via efficientenergy transfer. Dalton Trans 2013;42:5649e54.

248. Guo N, Zheng Y, Jia Y, Qiao H, You H. A tunable warm-white-light Sr3Gd (PO4) 3: Eu2þ,

Mn2þ phosphor system for LED-based solid-state lighting. New J Chem 2012;36:168e72.249. Huang CH, Wu PJ, Lee JF, Chen TM. (Ca,Mg,Sr)9Y(PO4)7:Eu

2þ, Mn2þ: phosphors forwhite-light near-UV LEDs through crystal field tuning and energy transfer. J Mater Chem2011;21:10489e95.

250. Huang CH, Chen TM. Ca9La(PO4)7:Eu2þ, Mn2þ: an emission-tunable phosphor through

efficient energy transfer for white light-emitting diodes. Opt Express 2010;18:5089e99.251. Guo N, You H, Song Y, Yang M, Liu K, Zheng Y, Huang Y, Zhang H. White-light

emission from a single-emitting-component Ca9Gd(PO4)7:Eu2þ,Mn2þ phosphor with

tunable luminescent properties for near-UV light-emitting diodes. J Mater Chem 2010;20:9061e7.

252. Xu Q, Han L, Di Q, Sun J. Tunable luminescence and efficient energy transfer ofNa2BaMgP2O8:Eu

2þ, Tb3þ phosphor for white light-emitting diodes. Ceram Int 2015;41:2699e705.

253. Li K, Zhang Y, Li X, Shang M, Lian H, Lin J. Tunable blue-green emission and energytransfer properties in a

ˇ

-Ca3(PO4)2: Eu2þ, Tb3þ phosphors with high quantum efficiencies

for UV-LEDs. Dalton Trans 2015;44:4683e92.254. Zhang M, Liang Y, Xu S, Zhu Y, Wu X, Liu S. Investigation of luminescence properties

and the energy transfer mechanism of tunable emitting Sr3Y2(Si3O9)2:Eu2þ,Tb3þ phos-

phors. CrystEngComm 2016;18:68e76.255. Lv W, Hao Z, Zhang X, Luo Y, Wang X, Zhang J. Tunable full-color emitting BaM-

g2Al6Si9O30:Eu2þ, Tb3þ, Mn2þ phosphors based on energy transfer. Inorg Chem 2011;50:

7846e51.256. Liu WR, Huang CH, Yeh CW, Chiu YC, Yeh YT, Liu RS. Single-phased white-light-

emitting KCaGd(PO4)2:Eu2þ, Tb3þ, Mn2þ phosphors for LED applications. RSC Adv

2013;3:9023e8.


257. Jiang L, Pang R, Li D, Sun W, Jia Y, Li H, Fu J, Li C, Zhang S. Tri-chromatic white-lightemission from a single-phase Ca9Sc(PO4)7:Eu

2þ, Tb3þ, Mn2þ phosphor for LED appli-cations. Dalton Trans 2015;44:17241e50.

258. Xu S, Li P, Wang Z, Li T, Bai Q, Sun J, Yang Z. Luminescence and energy transfer ofEu2þ/Tb3þ/Eu3þ in LiBaBO3 phosphors with tunable-color emission. J Mater Chem C2015;3:9112e21.

259. Jia Y, Lv W, Guo N, Lv W, Zhao Q, You H. Utilizing Tb3þ as an energy transfer bridge toconnect Eu2þeSm3þ luminescent centers: realization of efficient Sm3þ red emission undernear-UV excitation. Chem Commun 2013;49:2664e6.


Recent developmentof fabrication technologiesof nitride LEDs for performanceimprovement

6

Ray-Hua Horng 1, Dong-Sing Wuu1, Chia-Feng Lin 1, Chun-Feng Lai 21National Chung Hsing University, Taichung, Taiwan; 2Feng-Chia University,Taichung, TaiwanRevised by JianJang Huang

6.1 Introduction

Gallium nitride (GaN)-based materials have attracted a great deal of attention becausethe direct wide bandgap gives a range of optical emission wavelengths from red toultraviolet (UV) when alloyed with indium or aluminum. Therefore, their ternary andquaternary alloys have been used for optoelectronic devices over the past 10 years,1,2

such as light-emitting diodes (LEDs) and laser diodes (LDs). Nowadays, high-brightness (HB) GaN-based LEDs have attracted significant interest; this is a highlyenergy-efficient lighting technology3 with advantages such as small size and a longlifespan. Above all, GaN-based LEDs emit short wavelength light in the blue or UVregion, which can be used to excite a yellow phosphor. This produces white lightand the devices are used for solid-state lighting. The development of solid-state lightinghas unlocked a number of niche markets, for example, the automotive market (LEDs arenow integrated into the headlamps in a number of car models), backlighting of displayscreens (including TVs), display projectors, street signs, and, the important market ofcell phones (screens, keyboards, and camera flashes). It should be noted that generallighting is a major market: according to the United States Department of Energy(DOE), the total sales of lighting products are $60 billion each year worldwide. Besidesthe economic potential of this market, the opportunities are also huge in terms of energysavings. Indeed, in the United States lighting represents 8% of the total energyconsumption and 22% of the electrical energy consumption. Inefficient incandescentbulbs remain the most widespread sources: they consume 40% of the lighting energyto produce only 15% of the light output. A 50% penetration of LEDs into the generallighting market would yield considerable energy savings of more than 350 TWh.Currently, the LED market is worth $3.7 billion each year globally, of which 58% isthe cell phone market, while illumination only represents 5% of the market.

At present, the efficiency of GaN-based LEDs is still lower than that of fluorescentlamps in general lighting applications. Therefore, the optimization of all aspects of LED

Nitride Semiconductor Light-Emitting Diodes (LEDs). https://doi.org/10.1016/B978-0-08-101942-9.00006-XCopyright © 2018 Elsevier Ltd. All rights reserved.

https://doi.org/10.1016/B978-0-08-101942-9.00006-X

efficiency is necessary for solid-state lighting. In general, the performance of aGaN-based LED is characterized by its external quantum efficiency (EQE) orwall-plug efficiency (h), defined as the ratio of emitted optical power (Popt) to injectedelectrical power (Pel) so that h ¼ Popt/Pel. The wall-plug efficiency h of GaN-basedLEDs is limited by several factors,2 which may be broken down as follows:h ¼ hinj � hint � hf � hextract. The injection efficiency hinj is the fraction of electron-hole pairs injected into the LED that reaches the p-n junction. This value is, for instance,limited by leakage of the current in the LED and by Joule losses in the electricalcontacts of the LED. The internal quantum efficiency (IQE) hint is the fraction ofelectron-hole pairs reaching the p-n junction that recombines radiatively. It is limitedby the non-radiative recombination of electron-hole pairs. The feeding efficiency hfis the ratio of the energy of an emitted photon h�u to the energy of an electron-holepair injected into the LED: hf ¼h�u/qV, where V is the voltage drop across the LEDand q the elementary charge. In general, part of an electron’s energy can be lost tophonons and hf < 1. Finally, the light extraction efficiency hextract is the fraction ofthe photons emitted at the p-n junction that actually escape the LED. Its value is limitedbylightreflections,whichtrapphotonsinsidetheLEDwheretheyareeventuallyabsorbed.

One may think that the main limiting factor would be the internal light generationand the IQE. Nevertheless, this is not the case in a variety of materials where the con-version from carriers to photons reaches 50%e90% if the material’s quality is highenough. In GaN-based LEDs, the main limiting factor is the light extraction efficiencyhextract, i.e., the ability of photons generated inside the semiconductor material toescape into the air. Unfortunately, most of the light emitted inside an LED is trappedby total internal reflection (TIR) at the material’s interface with air as the refractiveindexes of GaN and air are nGaN ¼ 2.5 and nair ¼ 1.0, respectively. Thus, the criticalangle at which light generated in the InGaN/GaN active region can escape is approx-imately qc ¼ sin�1(nair/nGaN) w 23 degrees, which limits the EQE of conventionalGaN-based LEDs to only a few percent.4 Although many efficient light extraction stra-tegies have already been applied, they are mostly based on the principle of random-izing the paths followed by the light and this gives poor control of the far-fieldemission distribution in GaN-based LEDs. The next generation of applications ofGaN HB-LEDs in projectors, automobile headlights, and general lighting requiresfurther improvements in light extraction efficiency and the directional far-fieldemission distribution. In this chapter, the objective is to achieve high extractionefficiency and to control the directionality of the emitted light.

6.2 GaN-based flip-chip LEDs and flip-chip technology

6.2.1 Background of flip-chip LEDs

Nowadays, as GaN-based HB-LEDs have improved; applications such as trafficsignals, backlights for cell phones, and LCD televisions have become possible. How-ever, before conventional fluorescent lighting can be replaced with solid-state lighting,the light extraction efficiency and the IQE of LEDs must be significantly improved.


Conventional LEDs are inherently inefficient because photons are generated through aspontaneous emission process and emitted in all directions. A large fraction of the lightemitted downward and toward the substrate does not contribute to the useable lightoutput. In addition, there is an inherent problem associated with conventional nitrideLEDs, that is, the poor thermal conductivity of the sapphire substrate. Flip-chip(FC) techniques are an effective way to enhance light extraction and heat dissipation.5

Therefore, flip-chip LEDs (FCLEDs) have always been used in high current and highpower operations to alleviate the thermal budget problem. The FCLED configurationhas high extraction efficiency compared to a conventional LED due to the thicker lightextraction window layer and the smaller difference between the refractive index of thesapphire substrate (nsapphire ¼ 1.76) and air (nair ¼ 1.0). The critical angle of theoutput light is larger and TIR is reduced. Furthermore, a metal contact, includingthe n- and p-metal in FCLEDs, does not baffle the light output and can serve as a reflec-tive mirror to reflect the light back through the transparent sapphire substrate.6e8

However, in FCLEDs there is still the TIR effect between the sapphire substrate andair, which reduces the extraction efficiency of the transparent window layer. Thesurface roughness technique can enhance the light output; it works by scattering thephotons from the textured semiconductor surface and the probability of photonsescaping from the semiconductor can be increased.9e11 A combination of conductiveomnidirectional reflectors (ODRs) and a micro-pillar array (MPA) sapphire surface hasbeen developed.12 A conductive ODR13 serves as an ohmic contact layer and also ahighly reflective mirror. Highly reflective ODRs will reflect radiated light with anyincident angle to the top surface of the device.14e16 The formation of MPAs on thebottom side of the sapphire surface can increase the probability for photons to escapethrough the textured sapphire surface. FCLED performance for different micro-pillardepths and shapes will be discussed.

6.2.2 Flip-chip technology

FC technology has been developed by IBM to provide connections between thebonding pads of chips and the metallization on the substrates since 1960. The first tech-nique was called controlled collapse chip connection (C4), and it displaced wirebonding, giving increased input/output density at a lower cost.15 The C4 process startsby depositing under bump metallurgy (UBM) onto the bonding pads of the chips tosupply good adhesion between the bonding pads and the bumps. The UBM usuallyconsists of three layers: an adhesion and/or barrier layer, a wetting layer, and an oxida-tion barrier layer. After that, solder bumps are formed on the UBM and reflow tobecome a solder ball. The next step is to put down the top surface metallurgy(TSM) onto the substrate. The chips are aligned and joined to the substrate. Subse-quently, many methods of connecting the bonding pads of the chips to the metalliza-tion on the substrates have been developed, such as solder bump, tape-automatedbonding (TAB), conductive adhesives, anisotropic conductive adhesives, wirebonding, metal bump, polymer bump, and composite bump.

FC technology is a method for interconnecting semiconductor devices to externalcircuitry with flux-less solder bumps that have been deposited onto the chip pads.

Recent development of fabrication technologies of nitride LEDs for performance 211

The LED chips are mounted onto a substrate with interconnects produced usingvarious materials and methods. FC technology has been used for LEDs due to the bet-ter electrical and thermal dissipation. Fig. 6.1 shows the structure of a conventionalGaN-based LED. Sapphire is a common substrate used to grow a GaN film, but it isan insulator and a poor thermal material. Therefore, the n- and p-pads must be onthe same side and lead to bonding pads that baffle the light output and decrease lightextraction efficiency. Fig. 6.2 shows the structure of a GaN-based FCLED. The n- andp-pads of FCLEDs do not baffle the light output, and they can be used as a highlyreflective mirror on the sub-mount to direct downward traveling light back to the sap-phire substrate. The FC technique can enhance the light output power by 1.5e1.7times compared to conventional GaN-based LEDs.16,17 Finally, heat can be conductedaway by using a high thermal conductivity sub-mount bonded to metal, which givesbetter thermal dissipation. This is an important advantage for high-power LEDapplications.

TCLp-electrode

n-electrode

p-GaN

n-GaN

MQW

Sapphire substrate

Figure 6.1 Structure of a conventional GaN-based LED. MQW, multiple quantum well; TCL,transparent conductive layer.

Sapphire substrate

n-GaN

n-metalp-GaN

p-metal andreflector

MQWs

Solder or Au stud bumpSub-mount

Figure 6.2 Structure of a GaN-based FCLED.


6.3 GaN FCLEDs with textured micro-pillar arrays

The GaN LED structures with a dominant wavelength at 460 nm used in this studywere grown by metalorganic chemical vapor deposition (MOCVD) on c-plane sap-phire substrates. The LED structure consists of a 2-mm-thick undoped GaN layer, a2-mm-thick highly conductive n-type GaN layer, a 0.2-mm-thick InGaN/GaN multiplequantum well (MQW), a 0.2-mm-thick p-type GaN layer, and InGaN/GaN short-period superlattice (SPS) tunneling contact layers with indium-tin-oxide (ITO).Fig. 6.3 shows the fabrication steps of a GaN FCLED with MPAs. Top-emittingLEDs with a size of 1000 mm � 1000 mmwere fabricated using standard photolithog-raphy and BCl3/Cl2 inductively coupled plasma (ICP) etching for current isolationpurposes. The p-GaN and active layers were partially etched by an ICP etcher toexpose an n-GaN layer for an electrode. An ITO layer of 250 nm was depositedonto the p-GaN layer as a transparent conductive layer (TCL). The samples werethen annealed at 500�C for 10 min in air. A layer of Cr/Pt/Au metals (50/50/2500 nm) was deposited for the p- and n-contact pads. After completing the conven-tional face-up LED structure, the sapphire was ground down to a thickness of 100 mm.A 500-nm layer of nickel metal was deposited onto the bottom of the sapphiresubstrate as a dry-etching mask. The sample was then subjected to the ICP processusing Cl2/BCl3 (10/30 sccm) plasma with an ICP power of 850 W and radio-frequency power of 400 W to form MPAs for light extraction. The ICP etching ratefor the sapphire was approximately 800 Å/min. The processed LED wafer was brokeninto 1000 mm � 1000 mm chips using a laser scriber.

A layer of Ti/Al metals (500/2000 Å) was deposited onto the silicon sub-mount as amirror. A SiO2 film of 800 Å was deposited onto it for passivation. A 2-mm layer ofgold metal was deposited for the n- and p-bonding pads. The silicon sub-mount wasthen subjected to the stud bump process. Fig. 6.4 shows the top view of a siliconsub-mount before FC bonding. Finally, the chips were FC bonded to the siliconsub-mount using a Panasonic ultrasonic FC bonder for electrical and optical measure-ment, as shown in Fig. 6.5.

In an attempt to verify the effect of the MPA surface on light extraction efficiency,pillars with various depths and bevel angles were formed for comparison. The surfacemorphology of the GaN FCLEDs with MPA sapphire surfaces were examined byscanning electron microscope (SEM) as shown in Fig. 6.6. The periodic distance forthe pillar array was about 5.5 mm and the depth of the pillars was between 1.1 and3.2 mm. Furthermore, the bevel angle of the pillars changed from 8 to 35 degreeswith increasing dry-etching time. Fig. 6.6(a) and (b) show pillars with a smooth topsurface and sidewall. On increasing the dry-etching time, the surface of the micro-pillars becomes rougher, and there is a larger bevel angle as shown in Fig. 6.6(c).Fig. 6.6(d) shows pillars with a pineapple-like textured surface. The results may beascribed to the uniformity of the hard nickel mask, which results in partial over-etching and an uneven pillar surface.

The GaN FCLED chips were packaged into a TO can without epoxy resin for sub-sequent measurement. The light-current-voltage (LeIeV) characteristics were


measured using a high-current measurement instrument (Keithley-240). The lightoutput power of the GaN FCLEDs was measured using an integrated sphere with acalibrated power meter. The IeV characteristics of flat-surface FCLEDs andMPA-FCLEDs were also measured as shown in Fig. 6.7. The IeV curves of

Si sub-mount

Au-bump

100 μm

Figure 6.4 SEM image of silicon sub-mount before FC bonding.

p-GaN

p-GaN p-GaN

p-GaN

p-GaN

p-GaN

n-GaN

n-GaN n-GaN

n-GaN

n-GaNn-GaN

MQW

MQW MQW

MQW

MQW

MQWSapphire Sapphire

Sapphire (100 μm)

Sapphire (100 μm) Sapphire (100 μm)

Sapphire (100 μm)Ni mask (500 nm)

ITO

ITO

Cr/Pt/Au

Cr/Pt/Au

Cr/Pt/Au Cr/Pt/Au

Cr/Pt/Au Cr/Pt/Au

Cr/Pt/Au

Cr/Pt/Au

Cr/Pt/Au

Cr/Pt/Au

Au bump

Si sub-mount

(a)

(d) (e) (f)

(b) (c)

Figure 6.3 Fabrication steps for GaN FCLEDs with MPAs: (a) before process, (b) normalprocess, (c) polishing, (d) mask layer, (e) ICP etching, (f) flip-chip bonding.


50 μm

Figure 6.5 SEM image of a chip bonded to the silicon sub-mount.

1 μm

1 μm 1 μm

1 μm

(a)

(c) (d)

(b)

Figure 6.6 SEM images of the MPA surfaces of a sapphire backside with various depths andbevel angles: (a) 1.1 mm MPA, (b) 1.8 mmMPA, (c) 2.7 mmMPA and (d) 3.2 mm MPA.


MPA-FCLEDs show normal p-n diode behavior with a forward voltage of 3.4 V (at350 mA), indicating that there was no heating or charging damage during thefabrication of the MPAs by the ICP etching process. The light output power-current characteristics of the flat FCLEDs and MPA-FCLEDs are shown inFig. 6.8. We clearly observed that the output power of the MPA-FCLEDs is larger

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0

Volta

ge (V

)

0 50 100 150 200 250 300 350

Current (mA)

Flat surface FCLEDs1.1 μm MPA-FCLEDs1.8 μm MPA-FCLEDs2.7 μm MPA-FCLEDs3.2 μm MPA-FCLEDs

Figure 6.7 Current-voltage (IeV) characteristics of flat-surface FCLEDs and MPA-FCLEDs.

300

250

200

150

100

50

0

Out

put p

ower

(mW

)

0 50 100 150 200 250 300 350

Current (mA)

60

50

40

30

20

10

0

Wal

l-plu

g ef

ficie

ncy

(%)

Flat surface FCLEDs1.1 μm MPA-FCLEDs1.8 μm MPA-FCLEDs2.7 μm MPA-FCLEDs3.2 μm MPA-FCLEDs

Figure 6.8 Light output power-current (LeI) curves of flat-surface FCLEDs and MPA-FCLEDs.


than for the flat FCLEDs. At an injection current of 350 mA, it was found that theMQW emission peaks of those devices were located at about 460 nm, and the lightoutput power of the flat FCLEDs and MPA-FCLEDs with depths of the flat surface,1.1, 1.8, 2.7 and 3.2 mm were about 151, 165, 179, 227 and 252 mW, respectively.Fig. 6.9 shows that the enhancement of the light extraction efficiency forMPA-FCLEDs with various pillar depths ranges from 10% to 68% at 350 mA currentinjection compared to a conventional flat-surface FCLED. This indicates that thetextured sapphire surface reduces TIR and increases the probability that photonswill escape from the semiconductor to the air. Furthermore, with the increase of pillardepth (from 1.1 to 3.2 mm) and bevel angle (from 8 to 35 degrees), the light outputpower of the MPA-FCLED increased by 55% under 350 mA current injection. Theseresults can be attributed to the increase of the effective surface area on increasing thedepth and bevel angle of the micro-disk. Plate 2 (see color plate section) showsimages of a conventional flat-surface FCLED and an MPA-FCLED under 350 mAcurrent injection. The intensity distributions are also shown. The electroluminescence(EL) intensities for the MPA-FCLED clearly exceeded those from the conventionalflat FCLED under the same current injection, especially on the FCLED top surface.The improved light extraction efficiency was further supported by beam view analysisresults.

Obviously, the results indicate that a sapphire substrate with an MPA surface re-duces the internal light reflection and increases the light extraction efficiency. Theprobability of light escaping from the sapphire to air is increased due to the increaseof the escape cone by the MPA structure. This enhancement can be attributed to theroughness of the top surface and because photons are more likely to be emittedfrom a surface-roughened device, resulting in an increase of the light output powerof the MPA-FCLED.

100

90

80

70

60

50

40

30

20

10

0

Enh

ance

men

t (%

)

1.0 1.5 2.0 2.5 3.0 3.5MPA depth (μm)

Experiment (at 350 mA)

Figure 6.9 Light extraction enhancement versus MPA depth.


6.4 GaN FCLEDs with a geometric sapphire shapingstructure

There has been intensive research into improving the light extraction efficiency andenhancing the brightness of LEDs. The effect of a geometric chip structure on the lightextraction efficiency has been discussed in many papers.18e21 Krames et al. reportedthe enhancement of the extraction efficiency for truncated-inverted-pyramid AlGaInP-based LEDs.18 Eisert and Harle described their experimental and simulated results thatshowed enhanced light extraction efficiency of GaN-based LEDs chip with an under-cut SiC substrate.19 Chang et al. reported a 10% output power enhancement fromInGaN/GaN MQW LEDs with wave-like textured sidewalls.20 Kao et al. describeda light output enhancement for a nitride-based LED with 22 degrees undercut side-walls.21 All these methods have one thing in common, which is that photons generatedwithin the LEDs have multiple opportunities of finding the escape cone. As a result,the light extraction efficiency and the LED output intensity could both be enhancedsignificantly. A simple method of fabricating oblique sidewalls will help in increasingthe brightness of GaN-based LEDs. GaN-based FCLEDs with a geometric sapphireshaping (SS) structure have been developed.22 The formation of oblique sapphire side-walls on the bottom side of the sapphire surface could be a better way to improve theprobability that the photons will escape. The electrical and optical properties of thesapphire shaped FCLEDs were reported.23

The GaN LED structures used in this study were grown in the same way as thosewith textured MPAs (Section 6.3) and had the same dominant wavelength (460 nm)and LED structure. Fig. 6.10 shows the fabrication steps of GaN FCLEDs with a

GaN LED

GaN LED

GaN LEDGaN LED

GaN LED

Sapphire sub.

SapphireSapphire

Sapphire

Sapphire sub.

GaN LED

Sapphire

SiO2H2SO4:H3PO4=3:1

at 330°C

Pad TCL

Sub-mount

(a)

(d)

(b)

(e)

(c)

(f)

Figure 6.10 Fabrication steps for SS-FCLEDs: (a) MOCVD growth of GaN LED epilayer,(b) growth of SiO2 passivation layer, (c) wet etching of sapphire substrate, (d) normal process,(e) laser cutting and dicing, (f) flip-chip bonding.


geometric sapphire shaping structure (SS-FCLEDs). First, an SiO2 film of size of1000 mm � 1000 mm was deposited onto the backside of a sapphire substrate byplasma-enhanced chemical vapor deposition and defined using standard photolithog-raphy to serve as the wet-etching hard mask. To avoid damaging the epitaxial layer,an SiO2 film was also deposited onto the epitaxial layer as a sheathing. The sapphiresubstrate was then immersed in a H2SO4:H3PO4 (3:1) solution at an etching tempera-ture of 330�C for 70 min. The sapphire wet-etching rate was about 1.4 mm/min, whichis related to the H3PO4 composition and the etching temperature. After finishing the SSprocess, top-emitting LEDs with a size of 1000 mm � 1000 mm were fabricated usingthe standard photolithography process. They were aligned with the backside SS patternand were partially etched using an inductively coupled plasma etcher to expose an n-GaN layer for the electrode. An ITO layer (250 nm) was deposited onto the p-GaNlayer as a transparent conductive layer. The samples were then annealed at 500�Cfor 10 min in air. A layer of Cr/Pt/Au metals (50/50/2500 nm) was deposited for thep- and n-contact pads. After conventional LEDs processes, the processed LED waferwas broken into 1000 mm � 1000 mm chips using a laser scriber. A layer of Ti/Almetals (500/2000 Å) was deposited onto the silicon sub-mount as a mirror. A SiO2film of 800 Å was deposited onto it for passivation. A 2-mm layer of gold metalwas deposited for n- and p-bonding pads. The silicon sub-mount was then subjectedto stud bump process. Finally, the LED chips with an oblique SS sidewall were FCbonded to the silicon sub-mount using a Panasonic ultrasonic FC bonder for electricaland optical measurement. Fig. 6.11 is a schematic drawing of a GaN SS-FCLED and asketch indicating how light can exit from the oblique sapphire sidewall.

Fig. 6.12 shows SEM images of the sapphire shaping structure. The sapphire wasetched for 70 min with an etching rate of about 1.4 mm/min. The etching depth wasabout 100 mm, as shown in Fig. 6.12(b). The crystallography facets are the(1�102), (1�106), and (11�25) planes against the (0001) c-axis and their anglesagainst the (0001) c-axis are 60, 30, and 50 degrees, respectively, as shown inFig. 6.13. Furthermore, the etching structures are all V-grooves. The V-shaped struc-ture can be used to form a cleaving line to break the thick (w450 mm) sapphire sub-strate. SEM images of a conventional FCLED (C-FCLED) and a SS-FCLED areshown in Fig. 6.14. The SS area and the much thicker window layer are clearly visibleon the SS-FCLED structure compared with the C-FCLED. The oblique sapphire ge-ometry improves light extraction by reducing the number of totally internally reflectedphotons from the sidewall interfaces, and the photons can escape through the obliquesidewall. In addition, the thicker sapphire window layer has significant advantagesover a conventional thin sapphire window layer because it facilitates light emissionfrom the edges of the chip. These two processes significantly reduce the photonpath length for extraction in a SS-FCLED device compared to a conventional chip.The benefits are visible in the photomicrographs in Fig. 6.15. Note that light appearsto radiate evenly from the thicker window layer and oblique sidewall of the SS-FCLED compared with the C-FCLED, indicating that the light extraction efficiencywas improved due to the oblique sapphire geometry and thicker window layer.

The LED chips were packaged into a TO can without epoxy resin for the subse-quent measurement. The corresponding IeV characteristics of SS-FCLEDs and C-


100 μm

350 μm

Au bump

Sapphire

Sapphire

n-GaN

p-GaNMQW

Figure 6.11 Upper: GaN SS-FCLED. Lower: Light can exit from the oblique sapphire sidewall.

V-groove

Sapphire substrate

Sapphire substrate

100 μm

100 μm

∼100 μm

∼450 μm

(a)

(b)

Figure 6.12 SEM images of sapphire shaping structure: (a) top and (b) cross-sectional views.


FCLEDs were measured, as shown in Fig. 6.16. The IeV curve for SS-FCLEDsexhibits a normal p-n diode behavior with a forward voltage of 3.5 V (at 350 mA),indicating that the high-temperature sapphire wet-etching process does not appear toadversely affect the IeV characteristics of these devices. Fig. 6.17 shows the light

60°

30°

50°

Sapphire substrate

Sapphire substrate

Sapphire substrate

(a)

(b)

(c)

Figure 6.13 SEM images of the crystallography facets of (a) R-plane (60 degrees), (b) A-plane(30 degrees), and (c) M-plane (50 degrees) against (0001) c-axis.


output power and wall-plug efficiency as a function of injection current forlp w 460 nm SS-FCLEDs and C-FCLEDs. The light output power of the SS-FCLEDs is larger than for the C-FCLEDs. Under a current injection of 350 mA, thelight output power of the SS-FCLEDs compared to the C-FCLEDs was significantlyraised from 150 to 234 mW and the wall-plug efficiency was increased from12.26% to 18.98%. We note that bare SS-FCLEDs (without encapsulating an epoxylens) have a 55% enhancement of light extraction efficiency under a current injectionof 350 mA compared to the C-FCLEDs. This indicates that the geometric sapphiresidewall reduces TIR and increases the probability that photons will escape from thesemiconductor to the air. Furthermore, the thicker sapphire window layer has a signif-icant advantage over a conventional thin sapphire window layer because it facilitateslight emission from the edges of the chip.

Fig. 6.18 shows normalized far-field patterns for SS-FCLEDs and C-FCLEDs un-der a current injection of 20 mA. For a detailed comparison, the normalized far-fieldpatterns in two directionsdthe x axis ((1�106) plane to (1�102) plane) and the yaxis ((11�25) plane to (11�25) plane)dwere measured. The experiment resultsshow the electroluminescence (EL) intensities of C-FCLEDs are concentrated on thenear vertical direction. In contrast, EL intensities observed from the SS-FCLED are

GaN-LED

GaN-LED

Si sub-mount

Si sub-mount

Figure 6.14 SEM images of (a) C-FCLED and (b) SS-FCLED devices.


(a)

(b)

Figure 6.15 Photomicrographs of (a) C-FCLED and (b) SS-FCLED chips (40 � 40 mm2)operating at 20 mA (dc) with an emission wavelength of lp w 460 nm.

5

4

3

2

1

00 20 40 60 80 100

Current (mA)

Volta

ge (V

)

C-FCLEDsSS-FCLEDs

Figure 6.16 Current-voltage (IeV) characteristics of SS-FCLEDs and C-FCLEDs.


concentrated on the near horizontal direction (i.e., larger than 120 degrees). Plate 3(see color plate section) shows normalized 3D far-field patterns of SS-FCLEDs andC-FCLEDs, which confirm that the oblique sidewall changes the far-field patternand increases the viewing angle by 50%. This enhancement is attributed to the obliquesidewall and the thicker window layer so that the probability of photons being emittedfrom the device in the near horizontal directions has increased.

250

200

150

100

50

0

0 50 100 150 200 250 300 350 40010

15

20

25

30

35

40O

utpu

t pow

er (m

W)

Current (mA)

Wal

l-plu

g ef

ficie

ncy

(%)

C-FCLEDsSS-FCLEDs

Figure 6.17 Light output power and wall-plug efficiency as a function of injection current forlp w 460 nm SS-FCLEDs and C-FCLEDs.

SS-FCLED X axis (145.24°)SS-FCLED Y axis (142.20°)C-FCLED X axis (136.16°)C-FCLED Y axis (137.55°)

180

150

120

90

60

30

0

Figure 6.18 Normalized far-field patterns for SS-FCLEDs and C-FCLEDs for two directions.x axis: (1�106) plane to (1�102) plane; y axis: (11�25) plane to (11�25) plane. The intensityis shown with arbitrary units.


6.5 GaN thin-film photonic crystal (PC) LEDs

The GaN sapphire-based LED approach is limited by several factors. First, the sap-phire substrate causes substrate losses, which are difficult to avoid, other than by intro-ducing an efficient distributed Bragg reflector (DBR), ODR, etc. Second, because thickGaN epilayers have been grown, multiple guided modes propagate and some of thelower-order guided modes tend to ignore the photonic crystal (PC) interaction. A stron-ger photonic interaction is desirable but this can only be obtained if the GaN layer isthin enough. Third, the question of efficient p-contacts is still unsolved. One mayconsider using a semitransparent metallic contact over the PC or even using a trans-parent injector such as ITO, but both of these present drawbacks: there is still absorp-tion and ITO tends to lose its transparency over time. Finally, other crucial aspects foran efficient LED have been ignored, such as thermal management and mounting into apackage. Sapphire is a poor thermal conductor, which limits the maximum electricalpower in GaN LEDs with a sapphire substrate. Unfortunately, most of the work onGaN LEDs is directed at obtaining good high-power LEDs, which certainly requirehigh current densities.

For all these reasons, the solution is usually to use GaN epifilm-transferred technol-ogy to obtain high-powered LEDs. The metal substrate also provides enhanced thermaldissipation. Here, the sapphire substrate is removed and the GaN epilayer thinneddown. Thus, reducing the number of guided modes in a GaN waveguide LED byreducing the GaN thickness is of interest, to produce thin-film LEDs (TFLED). AGaN TFLED is actually a GaN microcavity LED composed of a GaN epifilm placedbetween the top air/GaN interface and the bottom metallic reflector mirror. A GaN-based TFLED combined with PC has been reported for the blue wavelengthrange.24e26 An AlInGaP TFLED combined with PC and a DBR was reported tohave enhanced light extraction efficiency and temperature stability in the red wave-length range.27 GaN PC TFLEDs emitting blue wavelengths have been fabricatedand the guided modes extraction behavior and the far-field emission distribution char-acteristics were studied in detail.26

6.5.1 Semiconductor wafer bonding

In the past, many approaches have been proposed and demonstrated for joining dissim-ilar systems onto one platform for monolithic integration. For example, in the early andmid-1980s, extensive work in integrating GaAs thin films with silicon substrates aimedto combine the optoelectronic functionality of GaAs with the processing power ofcomplementary metal-oxide-semiconductor silicon integrated circuit technology.28

The union of GaAs devices with silicon substrates would have created new possibil-ities for high-speed communications. However, the direct heteroepitaxial growth ofGaAs thin films on a silicon substrate with a low density of dislocations is difficult.Wafer-bonding technology is an alternative approach to this problem. In waferbonding, two highly polished, flat and clean wafers of almost any material, whenbrought into contact at room temperature, are locally attracted to each other by van


der Waals forces and adhere or bond to each other. Standard direct wafer bonding isattributed to relatively weak van der Waals forces and subsequent annealing at hightemperature is required to achieve a strong bond. This phenomenon has been knownfor a long time for optically polished pieces of material and was first investigatedfor polished pieces of quartz glass by Rayleigh in 1936.29 A variety of techniqueshave been employed to bond the two surfaces together, the methods being distin-guished by the characteristics of the bonded interface. Some of the common wafer-bonding techniques and their associated properties have reported.30e34

6.5.2 Epifilm-transferred technology

III-nitride semiconductors are promising materials for producing optoelectronicdevices working in the ultraviolet to visible spectrum. Due to the lack of a substratewith a good lattice match, the majority of III-nitride devices are grown onto a sapphiresubstrate, which provides a hexagonal template for growing wurtzite GaN. However,the poor electrical conductivity and low thermal conductivity of the sapphire substratescauses the poor characteristics of the GaN electronic devices. Now, GaN epifilm-transferred technology has been widely used to produce HB-LEDs, such asTFLEDs.35,36 Epifilm-transferred technology combined with wafer bonding and laserlift-off (LLO) techniques may be a direct approach for eliminating the sapphire growthsubstrate. In this approach, the GaN epifilm is transferred to a substrate with betterthermal and electrical conductivity to improve the light extraction efficiency anddrooping characteristics of GaN-based TFLED devices.

A GaN epifilm grown on a substrate can be separated from the substrate by a laserillumination process.37,38 LLO techniques39e41 use a KrF excimer pulsed laser. Sap-phire is transparent to the beam (lex ¼ 248 nm), which is absorbed by the interfacialGaN. The temperature increases at the interface, inducing the decomposition of theinterfacial GaN into gaseous nitrogen and gallium droplets, as shown in Fig. 6.19.

Ohmic-contact metal layer

Metal reflectorSiOx

Si substrate

Bonding metal layer

GaGaN epilayer thickness

Sapphire

Laser scanning

Figure 6.19 Laser lift-off process.


This process is termed laser-assisted film debonding and has considerable potential forfabricating devices when used in conjunction with wafer-bonding techniques. Howev-er, a GaN TFLED structure is actually a GaN microcavity LED composed of a GaNepifilm placed between the top air/GaN interface and the bottom metallic reflectormirror.

6.6 PC nanostructures and PC LEDs

A PC nanostructure is any structure with a periodic variation in its refractive index andthese structures provide exciting new ways to manipulate photons.42e44 The period-icity can be in one, two, or three spatial dimensions and can introduce a photonicbandgap (PBG) (a range of frequencies for which electromagnetic radiation is non-propagating) with the same dimensionality. A PBG arises due to Bragg’s reflectionand occurs when the spatial periodicity has a length approximately one-half that ofthe wavelength of the incident electromagnetic radiation. The same phenomenon givesrise to the electronic bandgap in semiconducting materials.45 Examples of PC struc-tures with periodicity in different spatial dimensions are shown in Fig. 6.20. Now,novel optical properties that are tunable by the period of the PC and the size of thenano-objects have been the subject of much research work. One-dimensional (1D)PCs are used as Bragg reflectors, which are part of the optical feedback mechanismin distributed feedback lasers46,47 and vertical cavity surface-emitting lasers.48 In addi-tion, 2D and 3D PCs have been the subject of intense research in areas related tosensing,49e51 telecommunications,52e54 slow light,55e58 and quantum optics.59e61

As mentioned above, PCs exhibit unique dispersion properties (e.g., PBG42) and canbe used to manipulate light emission. There are several schemes to obtain light extrac-tion through PC nano-structures on GaN-based LEDs, as shown in Fig. 6.21,62 such as

• Inhibition of emission of guided modes by PBG. Light can only escape using out-of-planeleakage modes. The emission region is etched with a deep pattern to forbid the propagationof guided modes, and thus force the emitted light to be redirected toward the outside, asshown in Fig. 6.21(a). This method needs high radiative efficiency. Etching through theactive layer can cause troublesome current injection and large non-radiative recombinationrates due to induced surface states.

(a) (b) (c)

Figure 6.20 Photonic crystals with periodicity in (a) one, (b) two, and (c) three dimensions. Thedifferent colors represent materials with different dielectric crystals.


• Enhancement of spontaneous emission in a small 3D PC cavity by the Purcell effect. Defectsin PCs behave as microcavities, as shown in Fig. 6.21(b), such that the Purcell effect can beused to enhance spontaneous emission.

• Emission extraction over the whole surface by leaky mode coupling. PCs can be used as 2Ddiffraction gratings in slabs or as waveguides to extract the guided modes into the air andredirect the emission, so that light can only escape through leaky mode couplings, as shownin Fig. 6.21(c). Because the extraction and generation regions are not separated, the activelayer is efficiently used as the total emitting surface.

p-GaN

n-GaNMQW

MirrorSubstrate

p-GaN

n-GaNMQW

MirrorSubstrate

p-GaN

n-GaNMQW

MirrorSubstrate

(a)

(b)

(c)

Figure 6.21 Various extraction methods using PCs: (a) PBG, (b) Purcell effect, (c) leaky modecoupling.


In this study, the extraction of waveguide light from GaN PC sapphire-based andmicrocavity LED structures was studied using 2D PCs as diffractive elements. Alight wave propagating in a GaN PC LED waveguide structure, with propagationpartially confined by TIR, can interact with the reciprocal lattice vectors of the 2DPC lattice to exhibit a variety of novel behaviors due to light localization.63 Onthe other hand, GaN PC LEDs can use Bragg diffraction to scatter the guided lightinto the escape cone to circumvent the deleterious effect of TIR, which traps mostof the light emitted in LED chips.64

Fig. 6.21(c) is a schematic of a surface grating device showing light extractionfrom a PC lattice, which will be described using the Ewald construction of Bragg’sdiffraction theorem. According to Bragg’s diffraction law: kgsinq1 þ mG ¼ k0sinq2.Phase-matching diagrams in the wave number space are shown in Fig. 6.22(a). Thetwo semicircles in Fig. 6.22(a) correspond to: (1) the waveguide mode semicirclewith radius kg ¼ 2np/l where n is the effective refractive index of the guided modeand (2) the air cone with radius k0 ¼ 2p/l. The light extraction from PC can be quan-titatively analyzed using the Ewald construction in the reciprocal space. The extrac-tion of waveguide light into air can be described by the relation jkg þ Gj < k0, whereG is the diffraction vector. This relation can be represented graphically with the Ewaldconstruction commonly used in X-ray crystallography. In the present case, forsimplicity, the PC is treated as 2D in an overall 3D structure as is commonly done.In this case, the reciprocal lattice of the 2D PC will be represented as rods protrudingperpendicular to the waveguide plane. Fig. 6.22(b) depicts an Ewald sphere for asquare PC lattice with the k vector of the incident light pointing directly at a reciprocallattice point. The center of the sphere is at the end of the vector and the radius is themagnitude of kg. The intersection points of the sphere with the protruding rods definethe extraction direction of the diffracted light. For simplicity, only in-plane propaga-tion needs to be considered and the projection onto the waveguide plane is sufficient.When the in-plane component of the resultant wave vector, after coupling to a recip-rocal lattice vector, falls inside the air circle, the diffracted light can escape into theair, as shown in Fig. 6.22(c).

For example, in an actual 2D square PC lattice used as a grating, the diffraction vec-tor exhibits anisotropy. Fig. 6.23 shows the diffraction vector for various values of thelattice constant a, and dispersion circles for the in-plane wave vector in air, k0, and inthe semiconductor material, kg. In the square lattice of the PC, GGX and GGM are 2p/aand 2

ffiffiffi2

pp�a, respectively. When GGX > k0 þ kg [a/l < 1/(n þ 1)], the zone-folded

curve does not enter the air curve, so diffraction does not occur, as shown inFig. 6.23(a). When a is larger than this value, some diffraction occurs, as shown inFig. 6.23(b). When a is large enough to satisfy GGM < k0 (a/l >

ffiffiffi2

p), the diffraction

vector is wholly included in the air curve, and this gives the maximum light diffractionefficiency. However, the diffraction efficiency cannot be unity for large a, since thelight not only enters the extraction light cone but also another solid angle not extractedby the diffraction. Even when diffracted into the extraction light cone, half of the lightwill go downward.


Air

PCExtracted light

Diffraction factork0θ

θ

2

1

kg

z

xk-spaceSemiconductor

material

G

Reciprocal lattice rodsAir cone

GaN material hemisphere

GΓX

(a)

(b)

(c)

Figure 6.22 (a) 2D PC structure with Bragg diffraction phase-matching diagrams. (b) Ewaldconstruction for a square PC lattice. (c) Projection of the Ewald sphere construction onto thewaveguide plane. The thick circle is the air cone and the dashed circle is the waveguide modecone.


6.7 Light emission characteristics of GaN PC TFLEDs

The blue GaN LED structure used consists of a 30-nm-thick GaN nucleation layer, a4-mm-thick undoped GaN buffer layer, a 3-mm-thick silicon-doped n-GaN layer,which consists of a 150-nm In0.05Ga0.95N/GaN superlattice (SL) with 10 periodsand a 120-nm In0.15Ga0.85N/GaN MQW active region (dominant wavelengthl0 ¼ 470 nm) with eight periods, a 20-nm-thick Mg-doped p-AlGaN electron blockinglayer and a 300-nm-thick Mg-doped p-GaN contact layer. The detailed wafer process-ing for the GaN TFLEDs with PC used the LLO technique to remove the growth sub-strate as described above and as shown in Fig. 6.24. The resulting structure was thenthinned by chemical-mechanical polishing (CMP) to obtain a GaN cavity thicknessT w 1500 nm. The PC with a square lattice of circular holes was then defined byholographic lithography. Holes were etched into the top n-GaN surface to a deptht ¼ 150 nm. Fig. 6.25(a) is a transmission electron micrograph (TEM) showing thecross-section of the GaN PC TFLED structure. The lattice constant, a, of the PCused for the far-field study was 290, 350, and 400 nm, and the hole diameter d wasfixed with the ratio d/a ¼ 0.7. A SEM image of a square PC lattice structure is shownin Fig. 6.25(b). Finally, a patterned Cr/Pt/Au (30/70/2000 nm) electrode was depositedonto the n-GaN as the n-type contact layer and Cr/Au metal (30/2000 nm) was depos-ited onto the backside of the silicon substrate.

A prepared sample structure is shown in Fig. 6.26. The light extraction character-istics of GaN PC TFLEDs with various values of the PC lattice constant a weremeasured and compared. The light-current-voltage (LeIeV) characteristics weremeasured using an integration sphere with a silicon photodiode. The turn-on voltagewas about 2.7 V. The light output power of the GaN PC TFLEDs with PC lattice con-stant a values of 290, 350 and 400 nm at a driving current of 200 mA is shown in

Small a Large a

Not diffracted Partly diffracted Diffracted

Lattice constant (a)

a/ = 1/(n + 1) a/ = √2λ λ(a) (b) (c)

GΓM

GΓX

Figure 6.23 Brillouin zones for 2D square PC lattices showing the dispersion curves for k0(center thick circle) and kg (dashed circles).


Fig. 6.27. The output power was enhanced by 45%, 68%, and 77%, respectively,compared to a GaN TFLED without PC. At 200 mA driving current, the forward volt-ages of the GaN PC TFLEDs with PC lattice constant a values of 290, 350, and400 nm were 6.2, 6.4, and 6.5 V, respectively. The high forward voltages could beattributed to the high series resistance in this thin PC device.

Step 1. Wafer bonding

Step 3. CMP polish

Step 5. PC pattern transferred to GaN

Step 7. SiOx passivation Step 8. n-GaN contact

Step 6. LED mesa define

Step 4. PC pattern define

Step 2. Laser lift-off

Laser scanning

SapphireSapphire

GaGaN epi thickness

GaN epi thickness

GaN epi thickness

GaN epi thickness

GaN epi thickness

Bonding metal layerBonding metal layer

Bonding metal layer

Bonding metal layer

Bonding metal layer Bonding metal layer

Bonding metal layer

Bonding metal layer

Si substrateSi substrate

Si substrate

Si substrate

Si substrate Si substrate

Si substrate

Si substrate

Ohmic-contactmetal layerMetal reflectorSiOx

VelocityPolishing pad

PC

PC

PRSiN etch mask

SiOx SiOx SiOx

Figure 6.24 Fabrication steps for GaN-based TFLEDs with PC lattice structures. PR,photoresist.


Angular-resolved spectra were measured under electrical current injection for thefar-field distribution. A continuous current of 20 mA was injected into aTO-mounted device at room temperature. Plate 4(aec) (see color plate section) showsspectra taken along the GX and GM directions for PC lattice constant a ¼ 290, 350,and 400 nm. The light emission spectra were taken for every zenithal angle with0.2 degrees resolution from�90 to 90 degrees. The plate shows the spectra on a wave-length versus angular plot with the color of a pixel representing the intensity of the

t

T

a = 350 nm

150 nm

(a)

(b)

Figure 6.25 (a) TEM showing a cross-section of a GaN PC TFLED structure with etch deptht ¼ 150 nm and GaN cavity thickness T ¼ 1500 nm. (b) SEM image of top of PCs on a TFLEDwith lattice constant a ¼ 350 nm and the diameter of air holes d ¼ 200 nm, fabricated usingholographic lithography.

(a) (b)n-contacts PC

n-GaNp-GaN

Bond metal

Si substrate

MQW

Ag mirror

500

μm

850 μm

Figure 6.26 (a) GaN TFLED structure with PC. (b) Optical micrograph showing the blue lightdistribution across the die operated at a low injection current of 5 mA.


light. Using this angular-resolved spectral technique, most of the states can be inves-tigated efficiently. In a GaN PC TFLED, the far-field distribution will be significantlymodified by PC lattice diffraction. The waveguide light traveling in the plane will bediffracted by the reciprocal wave vectors associated with the PC. Plate 4(d) shows the2D free-photon band structure for the transverse electric (TE) modes with averagerefractive index n ¼ 2.42. The three different PC lattice constant values created a rangeof a/l ¼ 0.52 to 0.91 for experimental investigation as enclosed in the boxes shown inPlate 4(d). These data were transformed into guided mode dispersion curves, whichcan be compared with the calculated band structures shown in the insets in Plate4(d). This serves as a useful framework for understanding the experimental results.Plate 4(d) shows the normalized dispersion curves for each mode line in the GXand GM directions. Above the air lines, the band structure has an abundance of reso-nant states that are involved in the PC-assisted light extraction. Due to the shallowetching of the samples, which resulted in a negligible narrow bandgap, the observeddispersion in this study was in good agreement with the calculated 2D free-photonband structure. Only the guided modes of effective refractive index neff ¼ 2.414 to2.15 from our sample are visible. We can accurately fit the lowest-order mode withthe free-photon band structure. The other modes are shifted higher than our red linebecause each mode has a different PC-induced effective index. Additionally, theextracted guided mode corresponds with the high symmetry point along the G axisof G1 and G2, which shows the light collimation profile. A detailed analysis of theangular-resolved emission patterns can determine the PC dispersion curves, as wellas the extraction efficiency for the various waveguide modes.

180

160

140

120

100

80

60

40

20

00 100 150 200 250 30050

Out

put p

ower

(mW

)

Current (mA)

GaN-based TFLEDsWithout PCWith PC a = 290 nmWith PC a = 350 nmWith PC a = 400 nm

Figure 6.27 Light output power-current (LeI) curve characteristic of the GaN TFLED with PCand without PC.


Furthermore, due to the discrete nature of the guided modes, this diffracted lightwill exhibit anisotropy in the far-field pattern in both in the zenith direction and theazimuthal direction.65 The far-field emission patterns in the zenith direction weremeasured at a driving current of 50 mA, as shown in Fig. 6.28. The samples with aPC lattice constant a of 290 and 400 nm have collimated far-field patterns that peaknear the normal to the GaN TFLED surface and have a small far-field angle at halfintensity of �31.7 (�41.05 degrees) and �42.45 (�49.7 degrees) in the GX (GM)orientation of the PC lattice, respectively, which are much smaller than for a typicalLambertian cone, �60 degrees. The measured far-field emission pattern of the GaNnon-PC TFLED was nearly Lambertian. In addition, the sample with PC lattice con-stant a ¼ 350 nm has lobes at around �17 (�15, �30 degrees) in the GX (GM)

–90

–90

–60

–60

–30

–30

0

0

30

30

60

60

90

90

1

1

3

3

2

2

4

4

Nor

mal

ized

inte

nsity

(a.u

.)N

orm

aliz

ed in

tens

ity (a

.u.)

(1) Non-Pc

(2) a = 290 nM(3) a = 350 nM

(4) a = 400 nM

(a)

(b)

Figure 6.28 Far-field pattern normalized with the peak intensity in (a) the GX direction and(b) the GM direction for a PC at a driving current of 50 mA. The far-field pattern shows thedifferent direction (a) GX and (b) GMwith PhC at driving current of 50 mA (solid line for non-PC, dash line for a ¼ 290 nm, dot line for a ¼ 350 nm, and dash dot line for a ¼ 400 nm).


orientation. Therefore, for a GaN PC TFLED, the far-field emission distribution issignificantly modified by the PC structure, that is, the lattice constant a. Furthermore,the far-field emission pattern remained unchanged when the current was varied from20 to 200 mA. This invariance of the far-field pattern indicates the temperature stabil-ity of the device since the junction temperature of the device can vary significantlyover this current range. We found that an encapsulated PC TFLED had similarfar-field characteristics. Hence, GaN PC TFLEDs can be encapsulated to increase lightenhancement while retaining the directional patterns.

In addition, the azimuthal anisotropy of the far-field distribution was measured as afunction of the azimuthal angles using the angular-resolved spectra.66 Plate 5(aec)plots the far-field distributions monochromatically in the azimuthal direction at a fixedwavelength of l0 ¼ 470 nm with PC lattice constant a values of 290, 350 and 400 nm.Different guided modes with different indexes will trace out an arc with the radiuscorresponding to the respective waveguide circle, which is well fitted by Ewald’sconstruction of Bragg’s diffraction theory.67 Several lower guided modes are extractedby the PC lattice, as shown in Plate 5 (see color plate section). Additionally, wemeasured the top view of the 3D far-field patterns for the three different PC lattice con-stant values, which reveal PC diffraction patterns with fourfold symmetry due to thesquare lattice, as shown in color Plate 5(def).

The light enhancement of the GaN PC TFLEDs compared to the GaN non-PCTFLED at a driving current of 50 mA is shown in Fig. 6.29. The light enhancementis defined as the ratio of the light output of a GaN PC TFLED divided by that ofthe GaN non-PC TFLED, and the power was collected from angles �0 to �90degrees. The light enhancement by collection angle strongly depends on the

2.7

2.4

2.1

1.8

1.50 ±10 ±20 ±30 ±40 ±50 ±60 ±70 ±80 ±90

Collection angle (degree)

Ligh

t enh

ance

men

t

GaN-based TFLEDsa = 290 nma = 350 nma = 400 nm

Figure 6.29 Light enhancement recorded at various output collection angles for GaN PCTFLEDs with three different values of a.


far-field patterns of the GaN PC TFLEDs. For the collimated GaN PC TFLEDs, thelight enhancement was approximately 2.4 for a�20 degrees collection cone. For colli-mated patterns, the light enhancement increases with smaller collection angles. Thediffering profile of the PC with a ¼ 350 nm had only slight light enhancement at smallcollection angles. Therefore, the collimation profile of the far-field pattern couldcontribute to strengthening the directional light enhancement in many applications,especially for etendue-limited applications. However, the extraction enhancement isnot only a function of the PC parameters, but also other variables such as the GaNthickness and MQW placement.68

6.8 Conclusion

In this chapter we studied the FC- and TF-types of GaN-based LEDs, including theirfabrication and electrical and optical characteristics. First, two types of GaN-basedFCLEDs with an MPA structure and an oblique geometric SS structure were demon-strated. For the MPA-FCLEDs, we used ICP to etch an MPA on the backside surfaceof the sapphire substrate to enhance light extraction. The light output power of theMPA-FCLEDs increases by 68% for a 3.2-mm textured MPA on the bottom side ofthe sapphire substrate. This enhancement can be attributed to the top surface rough-ness and the fact that photons are more likely to be emitted from the surface-roughened device, resulting in an increase of the light output power of theMPA-FCLED. For the SS-FCLEDs, FCLEDs with an oblique sapphire geometricstructure and much thicker sapphire window layer were fabricated. A H2SO4:H3PO4 (3:1) solution was used to etch the backside surface of the sapphire substrate.The enhancement of the light extraction efficiency for 100 mm SS-FCLEDs was 55%under a current injection of 350 mA compared to C-FCLEDs. The novel FCLEDstructure not only reduced the TIR effect but facilitated light emission from the edgesof the thicker sapphire window layer, resulting in an increase in the light extractionefficiency of the FCLEDs.

In the second part of the chapter we studied the combination of TF-types of GaN-based LEDs with PC nanostructures, including their fabrication and optical and elec-tronic phenomena. The enhancement of the directional light extraction of the GaN PCTFLEDs with blue light emission was experimentally investigated. Angular-resolvedspectra measurements revealed the directional profile and azimuthal anisotropy in thefar-field distribution, with the guided modes extraction due to Bragg diffraction. Dueto the shallow etching, the dispersion curve of the mode lines observed in the angular-resolved spectra closely resembled the band structures of the 2D free-photon band withthe index corresponding to the average index for the slab waveguides. The extractedguided mode corresponds with the high symmetry point along the G axis of G1 andG2, which shows the light collimation profile. The enhancement of the light fromthe GaN PC TFLEDs within the collection cone angle depended on the measured3D far-field patterns. In a �20 degrees collection cone, the collected light wasenhanced by a factor of approximately 2.4 for a collimated PC TFLED compared tothe GaN non-PC TFLED. This anisotropy and collimation of the PC slabs could be


used for the light propagation extraction of etendue-limited applications. This researchprovides important information for designing LEDs for projection displays that takefull advantage of what a PC can offer.

The fabrication development for the improvement of LED performance is notlimited to the technologies mentioned above. Ultraviolet LEDs require redesign ofepi-structure, as well as process steps, so that internal quantum efficiency, light extrac-tion efficiency, and device resistance can be optimized. As for micro-LED displays, thechip size is shrunk to be smaller than 10 � 10 mm2, by which a new process flow isimposed for the new products.

References

1. Schubert EF, Kim JK. Solid-state light sources getting smart. Science 2005;308:1274e8.2. Schubert EF. Light-emitting diodes. Cambridge, UK: Cambridge University Press; 2003.3. Krames MR, Shchekin OB, Mueller-Mach R, Mueller GO, Zhou L, et al. Status and future

of high-power light-emitting diodes for solid-state lighting. J Disp Technol 2007;3:160.4. Huh C, Lee KS, Kang EJ, Park SJ. Improved light-output and electrical performance of

InGaN-based light emitting diode by microroughening of the p-GaN surface. J Appl Phys2003;93:9383e5.

5. Chang SJ, Chang CS, Su YK, Lee CT, Chen WS, et al. Nitride-based flip-chip ITO LEDs.IEEE Trans Adv Packag 2005;28(2):273.

6. Hibbard DL, Jung SP, Wang C, Ullery D, Zhao YS, et al. Low resistance high reflectancecontacts to p-GaN using oxidized Ni/Au and Al or Ag. Appl Phys Lett 2003;83:311e3.

7. Hibbard DL, Jung SP, Wang C, Ullery D, Zhao YS, et al. Al-based Ohmic reflectors withlow leakage currents and high reflectance for p-GaN flip-chip processes. Appl Phys Lett2003;83:311.

8. Kondoh Y, Wantanabe S, Kaneko Y, Nakagawa S, Yamada N. Nitride semiconductor lightemitting device having a silver p-contact; 2001. US Patent 6,194,743.

9. Fujii T, Gao Y, Sharma R, Hu EL, DenBaars SP, et al. Increase in the extraction efficiencyof GaN-based light-emitting diodes via surface roughening. Appl Phys Lett 2004;84e86:855.

10. Windisch R, Rooman C, Meinlschmidt S, Kiesel P, Zipperer D, et al. Impact of texture-enhanced transmission on high-efficiency surface-textured light-emitting diodes. ApplPhys Lett 2001;79:2315.

11. Huang HW, Kao CC, Chu JT, Kuo HC, Wang SC, Yu CC. Improvement of InGaN-GaNlight-emitting diode performance with a nano-roughened p-GaN surface. IEEE PhotonTechnol Lett 2005;17(5):983.

12. Lee C-E, Lee Y-J, Kuo H-C, Tsai M-R, Cheng BS, et al. Enhancement of flip-chip light-emitting diodes with omni-directional reflector and textured micropillar arrays. IEEEPhoton Technol Lett 2007;19:1200.

13. Kim JK, Gessmann T, Schubert EF, Xi JQ, Luo H, et al. GaInN light-emitting diode withconductive omnidirectional reflector having a low-refractive-index indium-tin oxide layer.Appl Phys Lett 2006;88:013501.

14. Gessmann T, Luo H, Xi J-Q, Streubel KP, Schubert EF. Light-emitting diodes withintegrated omnidirectionally reflective contacts. Proc SPIE 2004;5366:53.


15. Lee YJ, Kuo HC, Lu TC, Wang SC. High light-extraction GaN-based vertical LEDs withdouble diffuse surfaces. IEEE J Quantum Electron 2006;42(12):1196.

16. Lin CH, Lai CF, Huang HW, Kuo HC, Hung YY, et al. Enhancement of InGaNeGaNindiumetineoxide flip-chip light-emitting diodes with TiO2eSiO2 multilayer stackomnidirectional reflector. IEEE Photon Tech Lett 2006;18:2050e2.

17. Lau JH. Flip chip technologies. McGraw-Hill; 1995.18. Krames MR, et al. High-power truncated inverted pyramid (AlxGa1-x)0.5In0.5P/GaP light-

emitting diodes exhibiting >50% external quantum efficiency. Appl Phys Lett 1999;75:2365.

19. Eisert D, Harle V. International Conference on Numerical Simulation of SemiconductorOptoelectronic Devices, Session 3: Photonic Devices, invited paper, 2002.

20. Chang CS, Chang SJ, Su YK, Lee CT, Lin YC, et al. Nitride-based LEDs with textured sidewalls. IEEE Photon Technol Lett 2005;16:750.

21. Kao CC, Kuo HC, Huang HW, Chu JT, Peng YC, et al. Light-output enhancement in anitride based light-emitting diode with 22 undercut side walls. IEEE Photon Technol Lett2005;17:19.

22. Lee C-E, Lee Y-J, Kuo H-C, Tsai M-R, Lu T-C, et al. Luminance enhancement of flip-chiplight-emitting diodes by geometric sapphire shaping structure. IEEE Photon Technol Lett2008;20:184.

23. Lee CE, Lee YC, Kuo HC, Tsai MR, Lu TC, et al. High brightness GaN-based flip-chiplight-emitting diodes by adopting geometric sapphire shaping structure. Semicond SciTechnol 2008;23:1.

24. David A, Fujii T, Moran B, Nakamura S, DenBaars SP, et al. Photonic crystal laser lift-offGaN light-emitting diodes. Appl Phys Lett 2006;88:133514.

25. Cho HK, Kim SK, Bae DK, Kang BC, Lee JS, et al. Laser liftoff GaN thin-film photoniccrystal GaN-based light-emitting diodes. IEEE Photon Technol Lett 2008;20:2096.

26. Lai CF, Chao CH, Kuo HC, Lee CE, Yeh WY. Directional light extraction enhancementfrom GaN-based film-transferred photonic crystal light-emitting diodes. Appl Phys Lett2009;94:123106.

27. Bergenek K, Wiesmann Ch, Zull H, Wirth R, Sundgren P, et al. Directional light extractionfrom thin-film resonant cavity light-emitting diodes with a photonic crystal. Appl Phys Lett2008;93:231109.

28. Tong QY, Gosele U. Semiconductor wafer bonding science and technology. New York:John Wiley & Sons, Inc; 1999.

29. Raleigh L. A study of glass surface in optical contact,. Proc Phys Soc A 1936;156:326e49.30. Christiansen SH, Singh R, Gosele U. Wafer direct bonding: from advanced substrate en-

gineering to future applications in micro/nanoelectronics. Proc IEEE 2006;94:2060e106.31. Christensen C, Bouwstra S. Eutectic bonds on wafer scale by thin film multilayers. Proc

SPIE, Int Soc Opt Eng 1996;2879:288.32. den Besten C, van Hal REG, Munoz J, Bergveld P. Polymer bonding of micro-machined

silicon structures. MEMS 1992;92:104.33. Wallis G, Pomerantz DI. Field assisted glass-metal sealing. J Appl Phys 1969;40:3946e9.34. Stengl R, Ahn KY, Gosele U. Bubble-free silicon wafer bonding in a non-cleanroom

environment. Jpn J Appl Phys 1988;27:L2364e6.35. Sparks D, Queen G, Weston R, Woodward G, Putty M, et al. Wafer-to-wafer bonding of

nonplanarized MEMS surfaces using solder. J Micromech Microeng 2001;11:630.36. Niklaus F, Enoksson P, Kalvesten E, Stemme G. Low-temperature full wafer adhesive

bonding. J Micromech Microeng 2001;11:100e7.


37. Kelly MK, Ambacher O, Dahlheimer B, Groos G, Dimitrov R, et al. Optical patterning ofGaN films. Appl Phys Lett 1996;69:1749e51.

38. Kelly MK, Ambacher O, Dimitrov R, Handschuh R, Stutzmann M. Optical process forliftoff of Group III-nitride films. Phys Status Solidi A 1997;159:R3.

39. Wong WS, Sands T, Cheung NW. Damage-free separation of GaN thin films from sapphiresubstrates. Appl Phys Lett 1998;72:599e601.

40. Yablonovitch E, Gmitter T, Harbison JP, Bhat R. Extreme selectivity in the lift-off ofepitaxial GaAs films. Appl Phys Lett 1987;51:2222.

41. Yablonovitch E, Sands T, Hwang DM, Schnitzer I, Gmitter TJ, et al. Van der Waalsbonding of GaAs on Pd leads to a permanent, solid-phase-topotaxial, metallurgical bond.Appl Phys Lett 1991;59:3159e61.

42. John S. Strong localization of photons in certain disordered dielectric superlattices. PhysRev Lett 1987;58:2486e9.

43. Yablonovitch E, Gmitter TJ, Leung KM. Photonic band structure: the face-centered-cubiccase employing nonspherical atoms. Phys Rev Lett 1991;67:2295e8.

44. Joannopoulos JD, Meade RD, Winn JN. Photonic crystals: molding the flow of light.Princeton. 1995.

45. Hamaguchi C. Basic semiconductor physics. Springer; 2001.46. Kogelnik H, Shank CV. Stimulated emission in a periodic structure. Appl Phys Lett 1971;

18:152e4.47. Nakamura M, Yariv A, Yen HW, Somekh S, Garvin HL. Optically pumped GaAs surface

laser with corrugation feedback. Appl Phys Lett 1973;22:515e7.48. Soda H, Iga K, Kitahara C, Suematsu Y. GaInAsP/InP surface emitting injection lasers. Jap

J Appl Phys 1979;18:2329e30.49. Loncar M, Scherer A, Qiu Y. Photonic crystal laser sources for chemical detection. Appl

Phys Lett 2003;82:4648e50.50. Chow E, Grot A, Mirkarimi LW, Sigalas M, Girolami G. Ultracompact biochemical sensor

built with two-dimensional photonic crystal microcavity. Opt Lett 2004;29:1093e5.51. Smith CLC, Wu DKC, Lee MW, Monat C, Hanic ST, et al. Microfluidic photonic crystal

double heterostructures. Appl Phys Lett 2007;91:121103.52. Noda S, Tomoda K, Yamamoto N, Chutinan A. Full three-dimensional photonic bandgap

crystals at near-infrared wavelengths. Science 2000;289:604e6.53. Jiang Y, Jiang W, Gu L, Chen X, Chen RT. 80-micron interaction length silicon photonic

crystal waveguide modulator. Appl Phys Lett 2005;87:221105.54. Aoki K, Guimard D, Nishioka M, Nomura M, Iwamoto S, et al. Coupling of quantum-dot

light emission with a three-dimensional photonic-crystal nanocavity. Nat Photon 2008;2:688e92.

55. Vlasov YA, O’Boyle M, Hamann HF, McNab SJ. Electromagnetically induced trans-parency with tunable single-photon pulses. Nature 2005;438:65.

56. Krauss TF. Slow light in photonic crystal waveguides. J Phys D 2007;40:2666.57. Baba T. Slow light in photonic crystals. Nat Photon 2008;2:465e73.58. Baba T, Mori D. Slow light engineering in photonic crystals. J Phys D 2007;40:2659.59. Yoshie T, Scherer A, Hendrickson J, Khitrova G, Gibbs HM, et al. Vacuum Rabi splitting

with a single quantum dot in a photonic crystal nanocavity. Nature 2004;432:200e3.60. Lodahl P, Driel AF, Nikolaev IS, Irman A, Overgaag K, et al. Controlling the dynamics of

spontaneous emission from quantum dots by photonic crystals. Nature 2004;430:654e7.61. Englund D, Fattal D, Waks E, Solomon G, Zhang B, et al. Controlling the spontaneous

emission rate of single quantum dots in a two-dimensional photonic crystal. Phys Rev Lett2005;95:013904.


62. Rattier M, Benisty H, Stanley RP, Carlin J-F, Houdre R, et al. Toward ultrahigh-efficiencyaluminum oxide microcavity light-emitting diodes: guided mode extraction by photoniccrystals. IEEE J Sel Top Quant Electron 2002;8:238.

63. Lai CF, Kuo HC, Chao CH, Hsueh HT, Wang J-FT, et al. Anisotropy of light extractionfrom two-dimensional photonic crystal light-emitting diodes. Appl Phys Lett 2007;91:123117.

64. Ichikawa H, Baba T. Efficiency enhancement in surface grating type two-dimensionalphotonic crystal light emitting diode. Appl Phys Lett 2004;84:457e9.

65. Lai CF, Chi JY, Yen HH, Kuo HC, Chao CH, et al. Polarized light emission from photoniccrystal light-emitting diodes. Appl Phys Lett 2008;92:243118.

66. Lai CF, Chi JY, Kuo HC, Chao CH, Hsueh HT, et al. Anisotropy of light extraction fromGaN two-dimensional photonic crystals. Opt Express 2008;16:7285.

67. David A, Benisty H, Weisbuch C. Optimization of light-diffracting photonic-crystals forhigh extraction efficiency LEDs. IEEE J Display Technol 2007;3:133e48.

68. Wierer JJ, David A, Megens MM. III-nitride photonic-crystal light-emitting diodes withhigh extraction efficiency. Nat Photon 2009;3:163.



Nanostructured LED 7Chien-Chung Lin 1, Ching-Hsueh Chiu 2, Da-Wei Lin 2, Zhen-Yu Li 2,Yu-Pin Lan 2, JianJang Huang 3, Hao-Chung Kuo 2

1National Chiao Tung University, Tainan, Taiwan; 2National Chiao Tung University,Hsinchu, Taiwan; 3National Taiwan University, Taipei, Taiwan

7.1 Introduction

To address the next generation applications of light emitting diodes (LEDs) in projec-tors, automobile headlights, and high-end general lightings, further improvements ontheir optical power and the external quantum efficiency (EQE) are required. Due to thedifficulties and high cost to grow native GaN substrates, sapphire substrates have beenplaying a major role for nitride-based LED production. The development ofGaN-based LEDs has shown significant progress over the past decade; in particular,the metal-organic chemical vapor deposition (MOCVD) growth of GaN on lattice-mismatched sapphire substrates.1,2 It has been shown that the epitaxial lateral over-growth (ELO) method with a microscale SiNx or SiOx patterned mask on as-grownGaN seed crystals can effectively reduce the threading dislocation density(TDD).3e5 However, the requirements of the two-step growth procedure and a suffi-cient thickness for GaN coalescence are costly and time-consuming. Meanwhile,high quality GaN-based LEDs have been demonstrated on a microscale patternedsapphire substrate (PSS) by wet etching,6 where the microscale patterns served as atemplate for the ELO of GaN and the scattering centers for the guided light. Boththe epitaxial crystal quality and the light extraction efficiency (LEE) were improvedby utilizing a micron-scale PSS. With the advances in fabrication processes, nanometerscale patterned substrates have become available and the improvements in LEDperformances are more obvious than the micron-scale ones. The benefits brought byreducing pattern sizes can be: (1) the nanoscale patterns can reduce the strain causedby lattice mismatch; (2) the nanoscale substrate can bend the threading dislocationmore efficiently and thus improve the crystal quality for epitaxial layers grown abovethis type of substrate; (3) the extra scattering caused by voids formed during the growthcan enhance the LEE which will in turn increase the power output.7 The LEDs grownon the nanoscale PSS showed more enhancement in the EQE than those grown on themicroscale PSS. However, the fabrication of nanoscale PSS generally requiredelectron-beam lithography8 or nanoimprinting techniques,9,10 making it unfavorablefor mass production. In this chapter, we review several techniques to fabricate aSiO2 nanorod-array PSS (NAPSS), serving as a template for the nanoscale ELO(NELO) of GaN byMOCVD to produce high efficiency GaN-based LEDs. The furtherdevelopment of these nanoscale substrates will be the nano-size LEDs (or core-shelltype) which can greatly reduce the volume of the device and explore the maximum


https://doi.org/10.1016/B978-0-08-101942-9.00007-1

of efficiency and flexibility. Some of the characterization methods for the nanostruc-tured LED in this chapter can be found in many other literatures. The strain relaxationof the subsequently grown GaN or InGaN layers can be analyzed by the peak positionof the Raman spectrum. The enhancement of scattering effect can either be calculatedby finite difference time domain (FDTD) simulation tool or measured by reflectivity inthe specialized template. The TDD reduction can be read from the TEM results andfinally the measured LIV curves can tell the actual improvement of the devices.

In the following sections, we will discuss two major methods of nanostructured/patterned LED: the first one is top-down technique and the second one is bottom-uptechnique. The top-down method is most widely used in the nanostructure formationbecause of the easy fabrication. The main focus will be the nano-patterning and how toavoid defect growth during coalescence. However, the dry-etch procedure whichusually occurs in this process can greatly damage the film quality if attention is notpaid. Several examples can be shown from our previous research for a successfultransfer of nano-patterning and etching into great LEDs.

The second method (bottom-up) is more difficult in terms of execution, mainly dueto the involved growth control. The nano-pattern is grown from the native substratewith or without pre-deposited patterns and the growth of LED structure is based onthis nanorod (NR) arrays. Many parameters, such as III/V ratio, chamber pressure,carrier gas composition, etc., need to be carefully adjusted in order to create suitablerod formation.10 Once the nanostructure can be established, the subsequent LEDgrowth is expected to be less defective and thus higher luminescent efficiency ispossible. A molecular beam epitaxy (MBE)-based technique will be discussed in thecontext and a 71% increase of output power is demonstrated.

7.2 Top-down technique for nanostructured LED

7.2.1 Nanoscale epitaxial lateral overgrowth of GaN-based lightemitting diodes on a SiO2 NAPSS

One of the simplest methods to carry out nano-patterned substrate is to use thenanoscaled SiO2 mask for lateral overgrowth. The previous researches focused onmicron-scale mask11,12 have produced a robust technology that are widely appliedin current LED industry. It is nature to shrink down the mask as the fabrication tech-nique progresses. To form the nano-patterns, self-assembled metal layer was consid-ered for mask material as it has been demonstrated numerously for nanotechnology.13

The GaN-based LEDs used in this study were grown on a 2 in. SiO2 NAPSS using alow-pressure MOCVD system (Aixtron 2400 G). The preparation of the SiO2 NAPSStemplate started with the deposition of a 200-nm thick SiO2 layer on a c-face (0001)sapphire substrate by plasma enhanced chemical vapor deposition (PECVD), followedby the evaporation of a 10-nm thick Ni layer, and the subsequent rapid thermal anneal-ing with a flowing nitrogen gas at 850�C for 1 min. The resulting self-assembled Niclusters then serve as the etch masks to form a SiO2 NR array using a reactive ionetch system for 3 min. Finally, the sample was dipped into a heated nitric acid solution


(HNO3) at 100�C for 5 min to remove the residual Ni masks. As shown in Fig. 7.1(a),the field-emission scanning electron micrograph (FESEM) indicated that the fabricatedSiO2 NRs were approximately 100e150 nm in diameter with a density of3 � 109 cm�2. The spacing between NRs was about 100e200 nm. Fig. 7.1(a) alsoshows that the exposed sapphire surface was flat enough for epitaxy. As the depositionprocess began, localized and hexagonal island-like GaN nuclei were first formed fromthe sapphire surface to initiate GaN overgrowth, as shown in Fig. 7.1(b). Fig. 7.1(c)shows the cross-sectional FESEM image of the GaN epilayer, where voids with asize varying from 150 to 200 nm were observed between the highlighted SiO2 NRs.The existing of the voids between NRs observed from the micrographs suggestedthat not all the exposed surface enjoyed the same growth rate. Hence, only the regionswith higher growth rates, which might be originated from larger exposed surface,could play the role of a seed layer, facilitating the lateral coalescence of GaN. Lastly,

Figure 7.1 FESEMs of (a) the fabricated SiO2 nanorod array, (b) GaN nuclei on the SiO2

NAPSS as growth seeds, (c) the GaN epilayer on an NAPSS in the cross-sectional view, and(d) the epitaxial pits on the p-GaN surface.

Nanostructured LED 245

the growth of a conventional LED structure, which consists of 10 periods of InGaN/GaN multiple quantum wells (MQWs) and a 100-nm thick p-GaN layer, wascompleted by MOCVD. The p-GaN layer of the NAPSS LED was grown at the rela-tively low temperature of 800�C, leading to the formation of hexagonal pits due toinsufficient migration length of Ga atoms.14 The FESEM image of the roughenedp-GaN surface with randomly distributed pits is shown in Fig. 7.1(d).

Transmission electron microscopy (TEM) was employed to investigate the crystal-line quality of GaN layers epitaxially grown on a planar sapphire substrate and on anNAPSS. As shown in Fig. 7.2(a), the TDD of GaN on the planar sapphire substrate washigher than 1010 cm�2 due to both the large lattice mismatch (13%) and the high ther-mal coefficient incompatibility (62%) between sapphire and GaN. On the other hand,the crystalline quality of GaN epilayer on an NAPSS was drastically improved fromthat grown on a planar sapphire substrate, as shown in Fig. 7.2(b). We found that anumber of stacking faults often occurred above the voids between SiO2 NRs, wherevisible TDs were rarely observed in the vicinities. It is believed that the presence ofstacking faults could block the propagation of TDs.15 Moreover, the TDs of theGaN layer on a NAPSS mainly originated from exposed sapphire surface, which couldbe bent due to the lateral growth of GaN. The inset of Fig. 7.2(b) shows the TEMimage of the dislocation bending with visible turning points. We summarized fourpotential mechanisms that were involved in the suppression of TDD, denoted as Types1e4 and illustrated in Fig. 7.3.

As shown in Fig. 7.3(a), the TDs originated from the sapphire surface during theinitial formation GaN growth seeds on an NAPSS. The presence of voids confirmedthe lateral coalescence of GaN, leading to the bending of dislocations near the edgeof SiO2 NRs. The bent TD eventually developed into stacking faults,6 as depictedby Type 1 in Fig. 7.3(b). Moreover, the coalescence fronts of GaN seeds provided a

0.5 µm 0.2µm

(a) (b)

Figure 7.2 The TEM images of the GaN/sapphire interface for the GaN epilayer grown on (a) aplanar sapphire substrate and (b) on an NAPSS. The inset of (b) shows the dislocation bendingphenomenon with visible turning points.


strain release layer where stacking faults could occur. These stacking faults were foundmostly above the voids or the small GaN seeds,16 blocking the TD propagation,denoted as Type 2. Occasionally, the blocked dislocation might also be bent to formstacking faults.17 If the growth rate was too slow to be a GaN seed, the dislocationcould be blocked by the formation of voids, as illustrated by Type 3. Finally, webelieved that the residual SiO2 between NRs could prohibit the GaN growth andfurther reduce the dislocation formation from sapphire surface, as depicted by Type4. It is also worth noting that the density of voids in the SiO2 NAPSS was higherthan that of a microscale PSS. Therefore, we believe that the formation of stackingfaults and voids were involved in the reduction and bending of dislocations.

The completed epitaxial structure then underwent a standard four-mask LED fabri-cation process with a chip size of 350 � 350 mm2 and packaged into transistor outline(TO)-18 with epoxy resin on top. The schematic of a fabricated NAPSS LED is shownin the inset of Fig. 7.4(a). The current-voltage (IeV) characteristics of an NAPSS-LEDand a conventional-LED (C-LED) with the same chip size were measured at room tem-perature, as shown in Fig. 7.4(a). The forward voltages at 20 mA were 3.27 V for theC-LED and 3.31 V for the NAPSS-LED. The nearly identical IeV curves indicate thatthe nanoscale roughness on the p-GaN surface had little impact on the IeV character-istics. Moreover, the NELO of GaN did not deteriorate the electrical properties.

Fig. 7.4(b) shows the measured light-output power versus the forward continuousDC current (LeI) for the NAPSS and conventional LED. At an injection current of20 mA, the light-output powers were approximately 22 and 14 mW for the NAPSS-and the C-LEDs, respectively. The output power of the NAPSS-LED was enhancedby a factor of 52% compared to that of the C-LED. The inset shows the normalizedelectroluminescence spectra for both devices at an injection current of 20 mA. A minorwavelength blueshift of approximately 2 nm was observed for the NAPSS-LED,attributed to the partial strain release by adopting the NELO scheme.18 The EQE of

GaN

Dislocations

Stacking faults

Type 1 Type 2 Type 3 Type 4

(a) (b)

Figure 7.3 The schematics of (a) the overgrowth process and the formation of dislocations,stacking faults, and voids at the initial stage of epitaxy, and (b) four potential mechanismsaccounted for the reduction of the TDD.


the NAPSS-LED was calculated to be approximately 40.2%, which is an increase of56% when compared to that of the C-LED, approximately 25.7%. We believe thatthe 56% enhancement in EQE originated from the improved internal quantum effi-ciency (IQE) and the enhanced extraction efficiency. The SiO2 NAPSS-assistedNELO method effectively suppressed the dislocation densities of GaN-based LEDs,which increased the IQE. Moreover, the embedded SiO2 NRs in the GaN epilayercontributed to light extraction due to scattering at the interfaces of different refractiveindices. Ueda et al.19 reported that the output power linearly increased with the surfacecoverage ratio of nanosilica spheres. Therefore, the extraction efficiency was enhancedby the SiO2 NR array.

7.2.2 High extraction efficiency GaN-based LED on embeddedSiO2 NR array and NPSS

Once the SiO2 nanomask can be built, further improvements can be added on thistemplate. One of them is to use the photonic quasi-crystal (PQC). We utilize anano-imprint technique to fabricate a NPSS and a SiO2 PQC on an n-GaN layer formass production. This idea is similar to previous Section 7.2.1, except the patternswere fabricated by a nanoscale master. Experimental results reveal that the light outputpower of LED with an NPSS and a SiO2 PQC pattern on an n-GaN layer is signifi-cantly greater than that of a C-LED.

Fig. 7.5 shows the schematic diagram of GaN-based LED with an NPSS and a SiO2

PQC structure on an n-GaN layer. Similar to the C-LED, the LED structure consists ofa Cr/Pt/Au p-electrode, an indium tin oxide (ITO) transparent layer, a LED epitaxiallayer, a smooth p-GaN surface, and a Cr/Pt/Au n-electrode on NPSS structure. Partic-ularly, the LED epitaxial structure has inset a SiO2 PQC pattern on an n-GaN layer bynanoimprint lithography (NIL) for comparison. In our study, three patterned LEDs arefabricated in order to investigate the influence of the NPSS and a SiO2 PQC on an

0 00

20

40

60

80

100

20 40 60 80 10020 40 60Current (mA) Current (mA)

Volta

ge (V

)

Ligh

t out

put p

ower

(mW

)

80 100

1

2

3

4

5

6Conventional LED on sapphireLED on NAPSS

Conventional LED on sapphireLED on NAPSS

0.0350 400 450

Wavelength (nm)

EL

inte

nsity

(a.u

.)

500 550

0.20.40.60.81.0

(a) (b)

Figure 7.4 Electrical and optical properties of an NAPSS- and a conventional-LED: (a) thecurrent-voltage (IeV) curves, where the inset shows a schematic of an NAPSS-LED, and(b) the current-output power (LeI) curves, where the inset shows the electroluminescencespectra for both devices at a driving current of 20 mA.


n-GaN layer on the LED light output power and beam profile performance. Forconvenience, LED with a SiO2 PQC, LED with an NPSS, and LED with an NPSSand a SiO2 PQC structure are denoted as LED A, LED B, and LED C, respectively.

The following details the process flow of NPSS by using NIL technique on a flatsapphire substrate. First, we spin-coat a 200-nm polymer layer on the sapphiresubstrate surface. Second, we place a patterned mold onto the dried polymer film.By applying a high pressure, we can heat the sapphire substrates to above the glasstransition temperature of the polymer. After that, the sapphire substrates and themold are cooled down to room temperature to release the mold. Finally, we use aninductively coupled plasma reactive ion etching (ICP-RIE) with BCl3 plasma to trans-fer the pattern onto sapphire substrate and remove the polymer layer with O2 plasmaetching gas in a reactive ion etching (RIE) system.

Fig. 7.6(a) shows the cross-section view of scanning electron microscope (SEM)image on a GaN epitaxial layer with an NPSS. The SEM image reveals that theNPSS exhibits a nanolens pattern. The lattice constant (a) of NPSS structure is750 nm and the nanolens diameter (d) is 455 nm. In addition, the etching depth ofNPSS is approximately 182 nm. Fig. 7.6(b) shows a cross-section SEM image on aGaN epitaxial layer with a SiO2 PQC pattern. The NIL process flow of inserting aSiO2 layer on an n-GaN layer is similar to the NPSS process described previously. Af-ter finishing molding, we use an RIE with CF4 plasma to transfer the pattern onto GaNsample and remove the polymer layer with O2 plasma etching gas in RIE system.

All LED samples are grown by MOCVD with a rotating-disk reactor (Veeco) on ac-axis sapphire (0001) substrate at a growth pressure of 200 mbar. The LED structureconsists of a 50-nm thick GaN nucleation layer grown at 500�C, a 3-mm thick undopedGaN buffer layer grown at 1050�C, a 2-mm thick Si-doped GaN layer grown at1050�C, an unintentionally doped InGaN/GaN MQWs active region grownat 770�C, a 50-nm thick Mg-doped p-AlGaN electron blocking layer grown at1050�C, and a 120-nm thick Mg-doped p-GaN contact layer grown at 1050�C. TheMQW active region consists of five periods of 3 nm/20 nm thick In0.18Ga0.82N/GaNquantum well layers and barrier layers.

Cr/Pt/Au

Cr/Pt/Au

N-GaN

U-GaN

SiO2 PQC

Sapphire

Figure 7.5 Schematic diagram of LED with an NPSS and a SiO2 PQC structure.


Different from the photonic crystals (PhCs) with high natural lattice symmetry,PQCs appear random at first glance; however, closer inspection reveals them topossess long-range order but short-range disorder. Fig. 7.6(c) shows a top-view imageof an atomic force microscopy (AFM) with 12-fold PQC pattern based on a square-triangular lattice. We choose the 12-fold PQC pattern due to the better enhancementof surface emission.20 This is obtained from the PhCs with a dodecagonal symmetricquasicrystal lattice, as opposed to regular PhCs with triangular lattice and eightfoldPQC.20 The recursive tiling of offspring dodecagons packed with random ensemblesof squares and triangles in dilated parent cells forms the lattice. The lattice constant androd diameters are 750 nm and 500 nm, respectively.

All LED samples are fabricated using the following standard processes with a mesaarea of 300 � 300 mm2. A SiO2 layer with thickness of 300 nm is deposited onto theLED sample surface by using PECVD. Photolithography is used to define the mesapattern after wet etchings of SiO2 by a buffered oxide etchant solution. The mesaetching is then performed with Cl2BCl3 Ar etching gas in an ICP-RIE system in orderto transfer the mesa pattern onto n-GaN layer. A 270-nm thick ITO layer is subse-quently evaporated onto the LED sample surface. The ITO layer has a high electrical

0 1 2 3(µm)

(µm

)

40

1

2

3

4

(a)

(c)

(b)

Figure 7.6 (a) Top view and (b) cross-section SEM images of sapphire surface with an NPSSand (c) top view AFM image of an n-GaN surface with a SiO2 PQC.


conductivity and a high transparency (>95% at 460 nm). Cr/Pt/Au contact is subse-quently deposited onto the exposed n- and p-type GaN layers to serve as the n- andp-type electrodes.

Fig. 7.7(a) shows the characteristics of a typical current voltage (IeV) measure-ment. It is found that the measured forward voltages under injection current 20 mAat room temperature for conventional LED, LED A, LED B, and LED C are 3.16,3.15, 3.15, and 3.23 V, respectively. In addition, the dynamic resistance of conven-tional LED, LED A, LED B, and LED C are about 14.7, 14.8, 15.3, and 15.4 U,

00

10

20

30

40

50

60

70

80Conventional LEDLED with a SiO2 PQC (LED A)LED with a NPSS (LED B)LED with a NPSS+a SiO2 PQC (LED C)

Conventional LED

LED with a SiO2 PQC (LED A)LED with a NPSS (LED B)

LED with a NPSS+a SiO2 PQC (LED C)

0

1

3

4

5

6

20 40 60Forward current (mA)

Ligh

t out

put p

ower

(mW

)Fo

rwar

d vo

ltage

(V)

80 100

0 20 40 60Forward current (mA)

80 100

(a)

(b)

Figure 7.7 (a) Current-voltage (IeV) and (b) intensity-current (LeI) characteristics ofconventional LED, LED with an NPSS, LED with a SiO2 PQC structure on an n-GaN layer,and LED with an NPSS and a SiO2 PQC structure on an n-GaN layer.


respectively. Therefore, in terms of dynamic resistance, there is no influence on thesetypes of devices by incorporating an NPSS and a SiO2 PQC structure by NIL process.The light output is detected by calibrating an integrating sphere with Si photodiode onthe device with TO-can package so that light emitted in all directions from the LEDcan be collected. The intensity-current (LeI) characteristics of conventional LED,LED A, LED B, and LED C are shown in Fig. 7.7(b). At an injection current of20 mA and peak wavelength of 460 nm, the light output powers of conventionalLED, LED A, LED B, and LED C with TO-can package are 12.8, 15.4, 17.3, and18.9 mW, respectively. Hence, the enhancement percentages of LED B, LED C,and LED D are 20%, 35%, and 48%, respectively, compared to that of C-LED.The enhancement of output power is due to the better crystal quality and more reflec-tion at the interface of GaN/sapphire. NPSS is regarded as an effective way to reduceTD between GaN and the underneath sapphire substrate, and allows more light toreflect from sapphire substrate onto the top direction. In addition, the use of a12-fold SiO2 PQC pattern also results in higher epitaxial crystal quality whichincreases more light output power.7,21,22

To confirm the speculations above, transmission electron microscopy (TEM) im-ages were employed to investigate the crystalline quality of GaN layers epitaxiallygrown on a flat sapphire substrate and an NPSS. As shown in Fig. 7.8(a) and (b), itis obvious that in Fig. 7.8(b) the TDD of GaN grown on an NPSS and a SiO2 PQCstructure was drastically reduced from that grown with flat sapphire substrate [asshown in Fig. 7.8(a)]. From Fig. 7.8(c), we found that a number of stacking faults oftenoccurred above the nanolens patterns, where visible TDs were rarely observed in thevicinities. It is believed that the presence of stacking faults could block the propagationof TDs.22 Moreover, the TDs of the GaN layer on an NPSS and a SiO2 PQC structuremainly originated from exposed sapphire surface, which could be bent due to thelateral growth of GaN. Fig. 7.12(c) shows the TEM image of LED with an NPSSand a SiO2 PQC structure is the dislocation bending with visible turning points.22

Accordingly, these observations support our assumption as well.

Figure 7.8 The TEM images of GaN/sapphire interface for the GaN epilayer grown on a (a) flatsapphire substrate and (b) an NPSS and a SiO2 PQC structure on an n-GaN layer. (c) Thedislocation bending phenomenon with visible turning points of an NPSS and a SiO2 PQCstructure on an n-GaN layer.


7.2.3 Highly efficient and bright LEDs overgrown on GaNnanopillar substrates

While nanoscale masks can improve the quality of the LEDs, a more elaboratedmethod based on this technology can be used, that is, the GaN nanopillar (NP) arraysubstrate. The formation of NP is one more step than the previous two methods: dryetch to the sapphire surface. The benefit of this method is to maximize the dislocationbending because of stronger lateral growth, and a larger air void array embedded in thesubstrate to scatter the light and increase the light-extraction efficiency.

The preparation procedures of the GaN epilayer with embedded micro air voids andSiO2 nanomask are as follows. First, we deposited a 200-nm thick SiO2 layer on a2-mm undoped GaN layer by PECVD, followed by the evaporation of a 10-nm thickNi layer, and the subsequent rapid thermal annealing (RTA) with a flowing nitrogengas at 850�C for 1 min. The resulting self-assembled Ni clusters are then served asthe etch masks to form a SiO2 NR array when we use a RIE system for 20 min to sap-phire surface. Finally, the sample was dipped into a heated nitric acid solution (HNO3)at 100�C for 5 min to remove the residual Ni masks. The 2-mm tall GaN NRs with SiO2

nanomask are produced as shown in Fig. 7.9(a). It can be seen that GaN NRs with anaverage diameter of about 250e500 nm were formed. It was also found that GaN NR’sdensity was around 3.3 � 108 cm�2 as shown in Fig. 7.9(b). Next, we deposited a

(a) (b)

(c) (d)

1 µµm 0.5 µm

0.5 µm 2 µm

Figure 7.9 (a) Cross-sectional; (b) tilted SEM image of GaN NRs template. (c and d) cross-sectional SEM image of GaN epilayer.


GaN-based LED structure on this GaN NRs template by a low pressure MOCVD(Veeco D75) system, denoted as NR-LEDs. At the same time, the identicalGaN-based LED structure was also grown on sapphire without GaN NRs for compari-son, denoted as C-LEDs. During the growth, trimethylgallium (TMGa), trimethylindium(TMIn), and ammonia (NH3) were used as gallium, indium, and nitrogen sources,respectively. Silane (SiH4) and biscyclopentadienyl magnesium (CP2Mg) were usedas the n-dopant and p-dopant source. The epitaxial structure of the GaN-based LEDovergrowth on GaN NRs template, consisting 3-mm n-doped GaN (n-GaN), 10-pairsInGaN/GaN MQWs, and 0.2-mm p-doped GaN (p-GaN) cap layer. Fig. 7.9(c) showsthe cross-sectional SEM image of GaN epitaxial layers with air voids and SiO2 nano-mask after all the growth is done. The estimated diameters of these air voids rangesfrom 0.5 to 1 mm from the SEM. These embedded air voids and SiO2 nanomask (asshows in Fig. 7.9(d)) shall be able to increase the LEE due to extra light scattering.23

Fig. 7.10(a)e(c) show the mechanisms of the air gap formation by nanoscaleepitaxial lateral overgrowth (NELOG) techniques on top of SiO2 nanomask. First,the GaN NRs with SiO2 nanomask were formed on sapphire substrate by top-downmethods in Fig. 7.10(a). As the GaN NRs grow upwards, there is also lateral growthon the sidewall of individual NRs. Such lateral growth eventually narrows the gap be-tween columns and forms holes with embedded air pockets as shown in Fig. 7.10(b).Our pervious study showed this growth process adding extra M-plane

�1010

�GaN on

the side walls of the etched pillars and inclined R-plane facets�1102

�close to the top

GaN nanorods

LT-GaN LT-GaN

u-GaN

VoidLT-GaN

Sapphire Sapphire Sapphire

M-plane

R-planeSiO2

(a)

(d) (e)

(b) (c)

1 µm 1 µm

Figure 7.10 (aec) The procedure of the air voids formation between GaN NRs and u-GaNepitaxial layer. (d) TEM image of GaN epilayer overgrowth on GaN NRs, (e) HRTEM imageof region I in (d). The diffraction condition is g ¼ 0002.


of nanopillars.24 Frajtag et al. have also reported that the semipolar planes coalesce firstdue to their higher growth rates relative to the growth rates on the nonpolar side-facesof nanowires.25 All of these growth mechanisms help the formation of air voids inbetween NRs. The final step consists of planar epitaxial GaN overgrowth, and air voidsand SiO2 nanomasks were encapsulated as show in Fig. 7.10(c). To analyze thedetailed epitaxial layer quality, we took the TEM pictures of GaN epilayer overgrowthon GaN NRs shown in Fig. 7.10(d), and calculated the dislocation densities. FewerTDDs are observable within the range in view. The dislocation density on the topof u-GaN is calculated to be around 5 � 107 cm�2. Meanwhile, we found TDDswere bent near SiO2 nanomask. The behavior is similar to those occurred in theNELOG method on a SiO2 NAPSS.26 The reduction of TDDs can be attributed tothe misfit (mainly perpendicular to the c-axis) and dislocation bending occurred justabove the voids, as shown in the inset of Fig. 7.10(e).

To investigate the optical properties of the microscale air voids and SiO2 nanomaskin GaN epilayer, the reflectance spectra were measured for both samples shown inFig. 7.11. The samples were grown with only GaN on the different substrates (planarand air void/nanomask) and coated with antireflection coating layer with SiO2. Thereflectance spectra bear interference fringes due to the substrate interfaces, and havean abrupt cutoff at the wavelength around the absorption edge of the GaN at379 nm. At the blue light emission wavelength, we found that the reflectance enhance-ment of the microscale air voids and SiO2 nanomask in GaN epilayer was 1.32 timeshigher than that of GaN epilayer on planar sapphire. This result indicates that thereflectance increased significantly due to the embedded microscale air voids andSiO2 nanomask, and it clearly demonstrates the light-scattering capability of the airvoid/nanomask design, and the extra reflected light from this layer can be harvestedin the front side more effectively.

5

10

15

20

25

30

35

0400 500 600

GaN on sapphireGaN on NRs

Wavelength (nm)

Ref

lect

ance

(%)

700 800

Figure 7.11 Diffuse reflectance spectra for both samples.


Fig. 7.12(a) displays the typical power-current-voltage (LeIeV) characteristics ofNR-LEDs and C-LEDs. With an injection current of 20 mA, the forward voltages are3.37 and 3.47 V, and the output powers are 21.6 and 13.1 mW, for NR-LEDs andC-LEDs, respectively. The light enhancement of LeIeV characteristics can be attrib-uted to the following factors: first, the TDDs reduction of epitaxial layers. This reduc-tion leads to much fewer non-radiative recombination events and increases the photongeneration efficiency. Second, more lights can be extracted from the LED because ofthe light scattering effect from the embedded micro/nanoscale air voids and SiO2. Inaddition, at the reverse bias, the leakage current of the NR-LEDs is smaller thanC-LEDs as shown in Fig. 7.12(b). Several types of dislocations can contribute tothe reverse-bias leakage current, and one of the dominant types is the screw disloca-tion.27 The reduction of screw dislocations can certainly help to reduce the reverse-bias current,27 and our measurement indicates a better crystal quality of NR-LEDs,which conforms with TEM results.

0

0.0

–0.2

–0.4

–0.6

–0.8

–1.0

0

–6 –5 –4 –3 –2 –1 0

20 40 60Current (mA)

Reverse voltage (V)

Rev

erse

cur

rent

(10–6

A)

Volta

ge (V

)

80 1000

C-LEDs

C-LEDs

NR-LEDs

NR-LEDs

20

40

60

80

100

4

6

2

(a)

(b)

Light output power (m

W)

Figure 7.12 (a) Forward IeV characteristics of all fabricated LEDs. (b) Reverse IeVcharacteristics of all fabricated LEDs.


On the other hand, we still need to quantify how much improvement of LeIeV iscoming from the better light extraction scheme due to air voids. A three-dimensionalfinite difference time domain (FDTD) simulation was applied to calculate the LEE ofthe LEDs using the FullWAVE program.28 The 5 � 5 random NRs with a heightof 2 mm were employed in the simulation structure. The unit cell has an area ofw7.56 mm2, corresponding to a density of w3 � 108 cm�2. The structural dimen-sions are extracted from the SEM images shown in Fig. 7.1. The light sources arecomposed of 15 dipole illuminators placed below 0.32 mm of the surface and thedetector at top of the simulated device. The calculated electric field distributionwith air void and SiO2 period of 0.5 mm is shown in Fig. 7.13(a). As it can be seenin the figure, the light intensity of NR-LEDs is higher than C-LEDs at the monitor.It indicates that the photons emitted from the MQWs escape out into the air mucheasier in NR-LEDs than in C-LEDs. The corresponding normalized light output asa function of the simulation time are calculated and plotted in Fig. 7.13(b), and theenhancement of extra light scattering is defined as the ratio of steady-state light outputof NR-LEDs to that of C-LEDs. From the simulated results, light output of NR-LEDsis around 1.447 times higher than that of the C-LEDs. Therefore, from the outputpower enhancement (65%) and LEE enhancement (44.7%), the IQE enhancementwould be 14% which is from the material quality improvement due to NELOGmethod on the SiO2 nanomask and NRs.

0

–1–0.2 0.20.0 –0.2 0.20.0

1

2

3

(µm

)

(µm) (µm)

4

5

6

7

8

9

0Sapphire

Air void

SiO2

GaN

Detector

–1

–4.00.0

0 100 200 300Simulation time (fs)

Nor

mal

ied

ligth

out

pot (

arb.

uni

ts)

400 500

NR-LEDsC-LEDs

0.4

0.8

1.2

1.6

2.0

0

4.0

1

2

3

4

5

6

7

8

9

C-LEDs NR-LEDs(a)

(b)

Figure 7.13 (a) Three-dimensional finite difference time domain of the calculated electric-fielddistribution of NR-LEDs and C-LEDs. (b) Normalized light output and power as functions ofthe simulation time for C-LEDs and NR-LEDs.


7.2.4 Freestanding high quality GaN substrate by associatedGaN nanorods self-separated hydride vapor-phaseepitaxy

In addition to the direct epitaxy growth on the native substrate, one can utilize this NRformation to fabricate a GaN substrate by hydride vapor-phase epitaxy (HVPE).29 Themost-used substrate for GaN-based optoelectronic devices in current days is sapphire,which is neither lattice matched nor themal-expansion matched to GaN material. Othermaterials frequently considered are silicon and silicon carbide. However, these twomaterials also have the same problems as sapphire, and silicon carbide is even moreexpensive than others. Native GaN is still the best candidate for device fabricationand previous researches also demonstrated encouraging results.30 However, diffi-culties in raw material growth hinder the progress in this type of material. Traditionalhigh pressure solution growth requires high temperature (>2800K) and high equilib-rium N2 pressure (>45 kbar), and these conditions are not very practical for industrialadaptation.31 The other methods such as Ammonothermal Growth were proposed andare under intense study nowadays. Meanwhile, HVPE is still one of the most practicaland favorable method to fabricate GaN substrates. A. Usui et al. proposed using ELOGand a thin MOCVD-grown GaN as seed layer and achieved dislocation density as lowas 6 � 107 cm�2 in the past.32 Here we proposed and demonstrated that a nanoscaleMOCVD grown GaN seed layer can be applied successfully in HVPE technology.29

Fig. 7.14(a) shows the proposed process of fabrication of the GaN substrate. In thebeginning, we grew an undoped 1.8-mm thick GaN layer on a 2-in. c-plane sapphiresubstrate by MOCVD. Then, a 500-nm thick SiO2 layer was deposited on theMOCVD-grown GaN layer by PECVD, followed by a 20-nm thick Ni layer depositedby an e-gun evaporator. The sample was then annealed at 850�C for 90 s in nitrogenambient to form self-assembled Ni nanoclusters on the SiO2 layer. The Ni nanoclustersacted as etching masks and subsequently, the RIE and inductively coupled plasma dryetching were performed in sequence to form GaN NR arrays.13 After etching, the GaNNR arrays were dipped into heated HNO3 and buffered oxide etchant to remove theresidual Ni and SiO2 from the top of the arrays. The diameter and the etched depthof the NRs were 200e500 nm and 1.8 mm, respectively. Afterward, we again depos-ited a 200-nm thick SiO2 layer on the NR template and then utilized RIE to remove theSiO2 layer on the top surfaces of the GaN NRs, as shown in Fig. 7.14(b). This figureclearly exhibits that the sidewalls of GaN NRs were surrounded by a thin SiO2 layerwith a thickness of approximately 40 nm. Finally, HVPE was used to regrow a 300-mmthick GaN layer; the details of the HVPE process are described elsewhere.33 During theHVPE cooling process, the 300-mm thick GaN substrate self-separates from the under-lying host sapphire substrate as a result of the release of thermal strain. For compari-son, a conventional as-grown GaN substrate was also prepared by HVPE on sapphiresubstrate. Fig. 7.14(c) shows the cross-sectional SEM image of the initial regrowthstage of HVPE. In this figure, the thickness of the GaN bulk is approximately3.3 mm and the surface is quite rough. The rough surface is believed to be associatedwith different growth rates of GaN seeds in the initial stage of HVPE. This problem issolved by increasing the growth time. Unlike in the growth of GaN on microscale


patterned substrates,34 no void was observed in this work because the GaN rod-to-rodintervals were small enough (200e400 nm).

Fig. 7.15(a) and (b) show the results of as-grown 300-mmGaN films separated fromGaN NR-array template and flat GaN template, respectively. In Fig. 7.15(a), a com-plete 2-in. self-separated freestanding GaN substrate was demonstrated. Accordingto Fig. 7.14(c), we can observe that the GaN regrowth layer was suspended on theNR-array template because of the SiO2 sidewall passivation. In general, duringHVPE cooling process, the large thermal stress will be induced by the quiet differentthermal expansion coefficient between GaN and Al2O3. In order to release the thermalstress during HVPE cooling process, the GaN NRs were broken and resulted in theself-separation of GaN from sapphire substrate. In contrast, as shown inFig. 7.15(b), the GaN grown on flat GaN smashed into several pieces because the ther-mal stress cannot be released in the HVPE cooling process.

Fig. 7.16(a) and (b) show the analysis of Nomarski images for the GaN substrateseparated from the NR-array template and flat GaN template, respectively. Comparingwith Fig. 7.16(a), a large amount of cracks was observed in Fig. 7.16(b). These crackswere motivated mainly by the influence of tensile stresses and thrived once the criticalthickness for GaN grown on the flat GaN substrate was reached.35 In other words, theabsence of cracks in the thick GaN film in Fig. 7.16(a) implies that the whole growth

Ni

SiOGaN nanorod Thick GaN

MOCVD GaNtemplate

Sapphire Sapphire Sapphire

(a)

(b) (c)VD 16 VD 17

Figure 7.14 (a) Schematic process flowchart for GaN nanorods (NRs), (b) SEM image of GaNNR arrays with SiO2 passivated sidewalls, and (c) cross-sectional SEM image for the initialstage of HVPE regrowth with a growth time of 2 min.


550 560 570 580 700

A1 (LO)

Regrowth u-GaN templateNanorodFlat

A1 (LO)

E2 (high)

E2 (high)

566.74 cm–1

567.15 cm–1

@ 300K

733.40 cm–1

735.01 cm–1

Raman shift (cm–1)

Ram

an in

tens

ity (a

rb. u

nits

)

720 740 760

(a) (b)(b)

(c)

Figure 7.16 Optical microscope images under the Nomarski illumination for GaN thick filmsobtained from (a) GaN nanorod (NR) arrays and (b) flat GaN surface. (c) Cross-sectionalRaman scattering analysis of GaN thick films obtained from GaN NR arrays and flat GaNsurface.

Figure 7.15 Results of GaN thick films obtained from (a) GaN nanorod arrays and (b) flat GaNsurface.


process is with low tensile stress. Some large, about 10e84 mm in diameter, and sparsehexagonal pits with six triangular {1011} facets were observed at the GaN surface inFig. 7.16(a). Basically we believe it comes from the extremely high growth rate about200 mm/h and a smoother surface could be achieved under a slower growth condition.

In order to further identify the difference in stress in these two samples, we per-formed the measurement of cross-sectional Raman scattering analysis. The E2-highphonon modes of GaN substrates obtained from the NR-array template and flat GaNtemplate were located at 567.15 and 566.74 cm�1, respectively, as shown inFig. 7.16(c). The E2-high peak for the substrate obtained from the NR-array templatewas very close to that of the stress-free GaN substrate, which is believed to be567.1 � 0.1 cm�1.36 Therefore, the residual stress in the substrate obtained from theNR-array template was negligible. The residual stress could be calculated by thefollowing equation:

Dug ¼ ug � u0 ¼ Kg$sxx; (7.1)

where ug and u0 represent the Raman peaks of GaN from flat surface and NR-arraytemplates, respectively. The estimated value of stress is about 0.160 GPa by adopting atheoretical Kg value of 2.56 cm�1/GPa reported byWagner and Bechstedt.37 The greatdifference in stress between these two samples is due to relaxation of the thermalstress, which is inherently accumulated during HVPE regrowth process.

Moreover, we examine the optical properties of these two samples. The 325 nmHeeCd laser was used to perform the photoluminescent (PL) measurement and laser’soutput power was set constantly to be 10 mW with a diameter of 100 mm. Fig. 7.17(a)shows the PL spectra (corresponding emission peak energy) at 20K for the GaN sub-strate obtained from the NR array and flat GaN template. Their emission peaks arelocated at 356.62 nm (3.477 eV) and 357.01 nm (3.473 eV), respectively. In compar-ison with the substrate obtained from the flat GaN template, the PL intensity for thesubstrate obtained from the NR-array template was enhanced by a factor of 1.66,and simultaneously the PL wavelength exhibited a blueshift of about 0.39 nm(3.8 meV). Also, its full width at half maximum (FWHM) is 4.717 meV, which isnearly half of that obtained from the flat GaN template (7.875 meV). The inset inFig. 7.17(a) shows the room temperature PL result. From the inset, the PL intensitywas enhanced by a factor of 1.72, and simultaneously the PL wavelength exhibiteda blueshift of about 2 nm as compared to that of the substrate obtained from the flatGaN template. Besides, a high resolution X-ray diffractometer (XRD) (Bede D1)with a Cu target was employed to investigate the crystalline quality of GaN substrates.The XRD peaks’ FWHM values are 196.3 in symmetric (002), and 152.9 arc sec inasymmetric (102) axis for NR-arrays template, and 350 and 227 arc sec in symmetricand asymmetric axis for flat template sample. We note the number of flat sample issimilar to previous flat results,38 and it is not surprising a low number is achievedfrom NR-array sample than flat one, hence the better crystal quality.

Regarding the emission wavelength, it is well-known that a residual strain in thesemiconductors would affect the energy band gap and then result in a shift in emission


wavelength. In general, the type of induced stress could be identified from the blueshiftor redshift in emission wavelength. For the GaN grown on the sapphire substrate, atensile strain was expected to be induced due to its relative small lattice constantthan that of sapphire underneath. In Fig. 7.17(a), therefore, we can consistentlyobserve a blueshift for the sample from NR arrays in both 20K and room temperaturePL measurement. In addition to the measurement of Raman scattering, it is anothersolid evidence that our proposed scheme can provide a GaN substrate which is freeof residual strain. On the other hand, the high TDD existing in GaN plays the rolesof non-radiative recombination centers to deteriorate the luminescence efficiency.

Fig. 7.17(b) and (c) show the typical bright field cross-sectional TEM images ofGaN substrate obtained from the NR-array and flat GaN template, respectively. Incontrast with Fig. 7.17(c), we found that a number of stacking faults often occurredabove the GaN NRs as shown in Fig. 7.17(b), where visible TDs were rarely observedin the vicinities. It is believed that the presence of stacking faults could block the prop-agation of TDs.15 From the TEM images, the TDD in the substrate obtained from NRarrays and flat GaN template is estimated to be approximately 107 and 5 � 109 cm�2,respectively.

350

PL temperature: 20KRegrowth u-GaN template

NanorodFlat

PL temperature: 300K

3400.0

0.2

0.6

0.4

0.8

1.0

360 380Wavelength (nm)

400 420

0.0

0.2

0.4

0.6

0.8

1.0

1.2

360Wavelength (nm)

Nor

mal

ized

inte

nsity

Nor

mal

ized

inte

nsity

370 380

(a)

(b) (c)

Figure 7.17 (a) Photoluminescent (PL) spectra of GaN substrates separated from GaN nanorods(NRs) and flat GaN surface at 20K. Inset shows the spectra at room temperature. Cross-sectional TEM images for the GaN substrates grown on (b) GaN NRs and (c) flat GaN surface.


7.3 Bottom-up technique for GaN nanopillar substratesprepared by molecular beam epitaxy

In Section 7.2, nanostructures were generally fabricated by top-down methods,39e41

such as etching process, in which the dry etching procedure normally generates defectstates on the column surfaces, causing reduction of IQE. In this section, we will discussa NELOG of high-quality GaN layer on bottom-up nanostructure [self-assembled GaNnanopillars (NPs)] grown MBE.42 Detailed analyses of the grown InGaN/GaN filmwill be demonstrated, and electro-optical properties of LEDs based on such GaNNPs template will also be discussed.

The epitaxial structure for GaN-based LED on sapphire with GaN NP was preparedas follows. First, the self-assembled GaN NP structure was grown on sapphire sub-strate by an RF plasma MBE system (ULVAC MBE), and the related processeshave been reported in our previous study.43 Fig. 7.18(a) shows SEM image of the

P-contact

1 mm

P-type GaN

N-contact

N-type GaN

AIN

Sapphire

ITO layer

InGaN/GaNMQWs (x10)

U-GaNMBE GaNnano-pillars

(a)

(b)

Figure 7.18 (a) Cross-sectional SEM image of GaN NPs template. The inset shows the funnel-like GaN NP. (b) Schematic of GaN-based LED structures on GaN NPs template.


grown GaN NPs. It can clearly be seen that the GaN NP is in a funnel-like form shownin the inset of Fig. 7.18, which might be beneficial for the following regrowth ofGaN-based LED structure. In addition, the density, the diameter, and the height areestimated to be around 1.15 � 1010 cm�2, 50 nm and 0.8 mm, respectively. Next,we deposited a GaN-based LED structure on this NP template by a low-pressureMOCVD (Veeco D75) system, denoted as NP-LEDs. In the meantime, the sameGaN-based LED structure was also grown on sapphire without GaN NP for compar-ison, denoted as C-LEDs. During the growth, trimethylgallium (TMGa), trimethylin-dium (TMIn), and ammonia (NH3) were used as gallium, indium, and nitrogensources, respectively. Silane (SiH4) and biscyclopentadienyl magnesium (CP2Mg)were used as the n-dopant and p-dopant sources. The epitaxial structure of theGaN-based LED overgrowth on NP is depicted in Fig. 7.18(b), consisting of 30-nmGaN nucleation layer (GaN NL), 1-mm undoped GaN (u-GaN), 3-mm n-doped GaN(n-GaN), 10 pairs InGaN/GaN MQWs, and 0.2-mm p-doped GaN (p-GaN) cap layer.

To find out what happened to these NPs after regrowth. Fig. 7.19 shows the pro-posed steps of the air voids formation during the entire material growth procedure.First, funnel-like GaN NPs were formed on a sapphire substrate by MBE at substratetemperature of 740�C shown in Fig. 7.19(a). As the NP grows upward, there is alsolateral growth on the sidewall of individual pillar. Such lateral growth eventually nar-rows the gap between columns and forms holes with 0.2e0.25 mm in size, which is

GaN nano-pillars

Sapphire (0001)

Sapphire (0001)

Sapphire (0001)

U-GaN

AIN

Air-void

1 µm

(a)

(d)

(b)

(c)

Figure 7.19 (aec) Procedure of the air voids formation between a GaN NPs and u-GaNepitaxial layer; (d) Cross-sectional SEM image. The inset shows air voids.


shown in Fig. 7.19(b). Next, the template was transferred to an MOCVD system to fin-ish the growth. The regrowth temperature of GaN film is about 1050�C. Under thishigh temperature, recrystallization of GaN is very possible and final coalescence ofu-GaN NPs template was performed and air voids were encapsulated, as shown inFig. 7.19(c) and (d). From the SEM pictures in Fig. 7.19(d), we can estimate theaverage diameter of these air voids to be about 100 nm. These embedded air voidsshall be able to increase the LEE due to extra light scattering from these air bubbles.44

The quality of the film can first be evaluated by its surface roughness. After the u-GaN layer was deposited, without growth of remaining LED layers, the surfacemorphology was measured by AFM, as shown in Fig. 7.20 The root mean square(rms) value of the surface roughness is about 1.4 nm, indicating high surface qualityand excellent coalescence overgrown on GaN NPs template. To analyze the detailedepitaxial layer quality, TEM was used to compare the cross-section between two typesof devices (NP-LEDs and C-LEDs). As in Fig. 7.21(a), in the case of the GaN epitaxiallayer grown on sapphire without GaN NPs, numbers of TD propagate vertically fromthe interface of GaN and sapphire, all the way to the top device layers. As a result, theTDs density in conventional GaN layer can be as high as 109 cm�2. Whereas, for theGaN epitaxial layer grown on sapphire with GaN NP [see Fig. 7.21(b)], it can beclearly found that the crystallography is drastically different from that of conventionalones. Fewer TDs are observable within the range in view. The dislocation density onthe top of n-GaN, MQWs is calculated to be around 7 � 107 cm�2. The reduction ofTDs density can be attributed to the misfit (mainly perpendicular to the c-axis) anddislocation bending occurred just above the voids, as shown in the inset ofFig. 7.21(b). Such behaviors are similar to those occurred in the NELOG method ona SiO2 NAPSS.

22

In addition to material defect density, another important feature to watch is the in-ternal stress of the epitaxial film since the nano-sized holes of template can potentiallyalleviate the built-in stress due to lattice mismatch. To analyze the residual strain in the

0 2.5µm

µm

5.0

0.0

15.0

30.0nm

0

2.5

5.0RMS=1.4 nm

Figure 7.20 Surface morphology of overgrown GaN NPs template scanned by AFM.


GaN films, Raman backscattering measurements were performed at room temperature.Fig. 7.22 shows the Raman spectrum for GaN epitaxial layer grown on sapphire withand without GaN NPs. The Raman shift peaks of E2 (high) mode for GaN epitaxiallayer grown on sapphire with and without GaN NPs are located at around 567.4 and569.3 cm�1, respectively. The in-plane compressive stress s for GaN epitaxial layeris estimated to decrease from 1.24 to 0.4 GPa with presence of GaN NP templates,by using the following equation27:

Du ¼ uE2 � u0 ¼ Cs; (7.2)

0.5 µm 0.5 µm 100nm

(a) (b) (c)

Figure 7.21 TEM image of (a) C-LEDs, (b) NP-LEDs, and (c) high-resolution TEM image ofregion I in (b). The diffraction condition is g ¼ 0002.

550

0.0

0.2

0.4

0.6

0.8

1.0

1.2

560

567.4cm–1 569.3cm–1GaN on NPGaN on sapphire

570Raman shift (cm–1)

Nor

mai

lze

inte

nsity

(arb

. uni

ts.)

580 590 600

Figure 7.22 Raman spectrum for GaN epilayer overgrown on GaN NPs template and sapphire.


where Du is the Raman shift peak difference between the strained GaN epitaxial layeruE2 and the unstrained GaN epitaxial layer u0 (566.5 cm�1), and C is the biaxial straincoefficient, which is 2.25 cm�1/GPa. Since the film on NP template bears less strain,consequently we can expect that the GaN-based LED grown on such template haveweaker quantum-confined Stark effect (QCSE).45

LED devices with a chip size of 350 � 350 mm2 were then fabricated from thecompleted epitaxial structures grown on sapphire with and without GaN NPs.Fig. 7.23(a) shows EL emission peak wavelength as a function of injection currentfor NP-LEDs and C-LEDs. The emission peak wavelength of NP-LED is slightlyred shifted (about 3.4 nm) from that of C-LED, and this is reasonable since lateralstrain relaxation favors higher indium incorporation.46e48 More importantly, as the in-jection current increases, the emission peak wavelength of NP-LEDs exhibits smallerblueshift (around 2.9 nm) compared with that of C-LEDs (around 5.6 nm). This resultindicates that the QCSE does become weaker due to the strain relaxation in epitaxiallayer overgrown on GaN NPs template, as expected. Fig. 7.23(b) displays the typicalpowerecurrentevoltage (LeIeV) characteristics of NP-LEDs and C-LEDs. With aninjection current of 20 mA, the forward voltages are 3.38 and 3.40 V, and the outputpowers are 25.3 and 14.8 mW, for NP-LEDs and C-LEDs, respectively. The lightenhancement of LeIeV characteristics can be attributed to the following factors: First,the TDD reduction of epitaxial layers. This reduction leads to much fewer non-radiative recombination events in the NP devices and increases the photon generationefficiency. Second, more lights can be extracted from the LED because of the light-scattering effect from the embedded nanoscale air voids.

In order to confirm the efficiency improvement of our NP-LED, the PL IQE mea-surement was performed. A general approach to evaluate the IQE of LEDs is tocompare the PL integrated intensity between low and room temperatures.49

Fig. 7.24 shows the measured IQE as a function of excitation power at 15 and300K for NP-LEDs and C-LEDs. The efficiency is defined as the collected photonnumbers divided by the injected photon numbers and normalized to the maximum

0 20 40 60 80 100 00

1

2

3

4

20 40 60 80 100465

470

475

480

485

Current (mA) Current (mA)

Out

put p

ower

(mW

)

EL

peak

wav

elen

gth

(nm

)

Volta

ge (V

)

0

20

40

60

80

100NP-LEDsC-LEDs

NP-LEDsC-LEDs

(a) (b)

Figure 7.23 (a) EL peak wavelength as a function of injection current of two fabricated LEDs.(b) LeIeV characteristics of the two fabricated LEDs.


efficiency at low temperature.50 At 20 mW of excitation power, it can be found that theIQE increases from 58% (C-LEDs) to 72% (NP-LEDs), which corresponds to 1.24times enhancement of efficiency. At this excitation level, we could calculate the cor-responding generated carrier density to be 2 � 1017 cm�3, approximately same levelof 20 mA at room temperature in our device. Thus, part of the efficiency improvementof GaN-NP-based LED can be linked directly to the improvement of IQE due to bettercrystal quality.

On the other hand, we still need to quantify how much improvement of LeIeV iscoming from the better light-extraction scheme due to air voids. A 2D FDTD simula-tion was applied to calculate the LEE of the LEDs using the FullWAVE program.28

The calculated electric field distribution with air void period of 0.25 mm is shown inFig. 7.25(a), where an array of air-filled rectangular holes represent air voids in ourdevices. The size of each rectangular hole is 0.2 mm � 0.1 mm. We set single dipole

1E–3

1E13 1E14 1E15 1E16 1E17 1E18

0.01 0.1 1 10 1000.0

0.2

0.4

0.6

0.8

1.0

Excitation power (mW)

Rel

ativ

e in

tern

al q

uant

um e

ffici

ency

(%) Carrier density (#/cm3)

NP-LEDs(15K)NP-LEDs(300K)

C-LEDs(300K)C-LEDs(15K)

Figure 7.24 Relative IQE as a function of excitation power for C-LEDs and NP-LEDs.

00.00

0.25

0.50

0.75

1.00

1.25

1.50

100 200 400 500300Nor

mal

ied

light

out

put (

arb.

uni

ts.)

–0.4 –0.2–0.2

0.2

0.2

0.40–0.4 –0.2 0.2 0.40

0

–1

0

1

2

3

4

5

6

7

8

(µm

)

(µm) (µm)

(µm

)

C-LEDs

C-LEDs

NP-LEDs

NP-LEDs

Simulation time (fs)

Detector

Light source

GaN

Air voids

Sapphire–1

0

1

2

3

4

5

6

7

8

(a) (b)

Figure 7.25 (a) Two-dimensional FDTD of the calculated electric field distribution of NP-LEDsand C-LEDs. (b) Normalized light output power as functions of the simulation time for C-LEDs and NP-LEDs.


illumination sources at 0.5 mm below the top of surface structures and the detectoraround the simulated device.51 As it can be seen in the figure, the electric field intensityof NP-LEDs is higher than C-LEDs at the monitor. It indicates that the photons emittedfrom the MQWs escape out into the air easier in NP-LEDs than in C-LEDs. The cor-responding normalized light output as functions of the simulation time are calculatedand plotted in Fig. 7.25(b), and the enhancement of extra light scattering is defined asthe ratio of steady-state light output of NP-LEDs to that of C-LEDs. From the simu-lated results, light output of NP-LEDs is around 1.48 times higher than that of theC-LEDs. Combining with previous PL IQE measurement, we can see a total enhance-ment of 82% (48% from LEE and 24% from IQE) when we compare the result toC-LED structure. The actual increase in power output of LED, which is 70%, is lowerthan prediction. This is possibly due to randomness of the air void formation, whichmakes our FDTD analysis overestimate the light-scattering effect.

7.4 Other nanostructures of interest for LEDs

Nowadays, there are more and more nanostructures employed for LEDs in the researcharea and applications. For example, nanostructure templates, such as AlN NRs, weregrown on sapphire substrate for the improvement of GaN-based epi-material qualityand device internal quantum efficiency.52 The AlN NRs were synthesized by vapor-liquid-solid mechanism. Alternatively, quantum dots were employed for enrichingcolor quality for white LEDs, as well as for flat panel displays. The II-VI or III-V quan-tum dots behave as the wavelength conversion material.53

7.5 Conclusion

In this chapter, we reviewed several technologies which can be applied onto nanoscalepatterned substrate for high-quality LEDs. The high-quality GaN-based LED structurewas successfully fabricated on NR template by using these nanotechnologies. Eithertop-down or bottom-up method possesses its own merit on device performance andthus attention shall be paid. The ease of fabrication and growth for top-down devicesshould be beneficial for initial commercialization. However, the bottom-up methodcan potentially reduce the dislocation and defects caused by in situ processes andthus lead to a more powerful nanoscale device.

References

1. Narukawa Y, Niki I, Izuno K, Yamada M, Murazki Y, Mukai T. Jpn J Appl Phys 2002;(Part2, No. 41):L371.

2. Schubert EF. Light emitting diodes. 1st ed. Cambridge, England: Cambridge UniversityPress; 2003.


3. Sakai A, Sunakawa H, Usui A. Appl Phys Lett 1997;71:2259.4. Zheleva TS, Nam OH, Bremser MD, Davis RF. Appl Phys Lett 1997;71:2472.5. Wuu DS, Wang WK, Wen KS, Huang SC, Lin SH, Huang SY, Lin CF, Horng RH. Appl

Phys Lett 2006;89:161105.6. Wuu DS, Wang WK, Wen KS, Huang SC, Lin SH, Horng RH, Yu YS, Pan MH.

J Electrochem Soc 2006;153:G765.7. Gao H, Yan F, Zhang Y, Li J, Zeng Y, Wang G. J Appl Phys 2008;103:014314.8. Xing A, Davanco M, Blumenthal DJ, Hu EL. J Vac Sci Technol B 2004;22:70.9. Huang HW, Lin CH, Yu CC, Lee BD, Chiu CH, Lai CF, Kuo HC, Leung KM, Lu TC,

Wang SC. Nanotechnology 2008;19:185301.10. Li S, Waag A. J Appl Phys 2012;111:071101.11. Zheleva TS, Nam Ok-H, Ashmawi WM, Grin JD, Davis RF. Journal of Cryst Growth 2001;

222:706e18.12. Usui A, Sunakawa H, Saki A, Atsushi Yamaguchi A. Jpn J Appl Phys 1997;36(Part 2, No

7B):L899e902.13. Chiu CH, Lo MH, Lu TC, Yu P, Huang HW, Kuo HC, Wang SC. IEEE J Lightwave

Technol 2008;26:1445.14. Lee YJ, Kuo HC, Lu TC, Wang SC. IEEE J Quan Electron 2006;42:1196.15. Feng ZH, Qi YD, Lu ZD, Lau KM. J Cryst Growth 2004;272:327.16. Nagai T, Kawashima T, Imura M, Iwaya M, Kamiyama S, Amano H, Akasaki I. J Cryst

Growth 2007;298:288.17. Cho HK, Lee JY, Kim KS, Yang GM, Song JH, Yu PW. J Appl Phys 2001;89:2617.18. Kusakabe K, Kikuchi A, Kishino K. Jpn J Appl Phys 2001;(Part 2, No. 40):L192.19. Ueda K, Tsuchida Y, Hagura N, Iskandar F, Okuyama K, Endo Y. Appl Phys Lett 2008;92:

101101.20. Zhang ZS, Zhang B, Xu J, Xu K, Yang ZJ, Qin ZX, Yu TJ, Yu DP. Appl Phys Lett 2006;88:

171103.21. Chen JJ, Su YK, Lin CL, Chen SM, Li WL, Kao CC. IEEE Photon Technol Lett 2008;20:

1193.22. Chiu CH, Yen HH, Chao CL, Li ZY, Yu PC, Kuo HC, Lu TC, Wang SC, Lau KM,

Cheng SJ. Appl Phys Lett 2008;93:081108.23. Park EH, Jang J, Gupta S, Ferguson I, Kim CH, Jeon SK, Park JS. Appl Phys Lett 2008;93:

191103.24. Lo MH, Cheng YJ, Kuo HC, Wang SC. Appl Phys Lett 2011;98:121101.25. Frajtag P, Samberg JP, El-Masry NA, Nepal N, Bedair S. M J Cryst Growth 2011;322:27.26. Chiu CH, Yen HH, Chao CL, Li ZY, Peichen Y, Kuo HC, Lu TC, Wang SC, Lau KM,

Cheng S. J Appl Phys Lett 2008;93:081108.27. Puech P, Demangeot F, Frandon J, Pinquier C, Kuball M, Domnich V, Gogotsi Y. J Appl

Phys 2004;96:2853.28. Tsai MA, Peichen Y, Chiu CH, Kuo HC, Lu TC, Lin SH. IEEE Photon Technol Lett 2010;

21:257.29. Chao CL, Chiu CH, Lee YJ, Kuo HC, Liu P-C. Appl Phys Lett 2009;95:051905.30. Wei TB, Wu K, Chen Y, Yu J, Yan Q, Zhang YY, Duan R, Wang J, Zeng Y, Li JM. IEEE

Electron Device Lett 2012;33(6):857.31. Paskova T, Evans KR. IEEE J Sel Top Quan Electron 2009;15(4):1041.32. Usui A, Sunakawa H, Sakai A, Yamaguchi A. Jpn J Appl Phys 1997;36:L899.33. Chao CL, Liu PC, Chi TW, Huang HH, Tsay JD. In: Proceedings of the International

Workshop on nitride semiconductors, Montreux, Switzerland, 6e10 October; 2008.


34. Lee YJ, Hwang JM, Hsu TC, Hsieh MH, Jou MJ, Lee BJ, Lu TC, Kuo HC, Wang SC. IEEEPhoton Technol Lett 2006;18:1152.

35. Etzkorn EV, Clarke DR. J Appl Phys 2001;89:1025.36. Goni AR, Siegle H, Syassen K, Thomsen C, Wagner JM. Phys Rev B 2001;64(035205).37. Wagner JM, Bechstedt F. Appl Phys Lett 2000;77:346.38. Gogova D, Hemmingsson C, Monemar B, Talik E, Kruczek M, Tuomisto F, Saarinen K.

J Phys D 2005;38:2332.39. Chen HS, Yeh DM, Lu YC, Chen CY, Huang CF, Tang TY, Yang CC, Wu CS, Chen CD.

Nanotechnology 2006;17(5):1454.40. Huang HW, Kao CC, Hsueh TH, Yu CC, Lin CF, Chu JT, Kuo HC, Wang SC. Mater Sci

Eng B 2004;113(2):125.41. Chiu CH, Lu TC, Huang HW, Lai CF, Kao CC, Chu JT, Yu CC, Kuo HC, Wang SC,

Lin CF, Hsueh TH. Nanotechnology 2007;18(44):445201.42. Kouno1 T, Kishino K, Yamano K, Kikuch A. Opt Expr 2008;17(22):0440.43. Yang TH, Ku JT, Chang J-R, Shen S-G, Chen Y-C, Wong YY, Chou WC, Chen C-Y,

Chang C-Y. J Cryst Growth 2009;311(7):1997.44. Park EH, Jang J, Gupta S, Ferguson I, Kim CH, Jeon SK, Park JS. Appl Phys Lett 2008;

93(19):191103.45. Chiu CH, Li ZY, Chao CL, Lo MH, Kuo HC, Yu PC, Lu TC, Wang SC, Lau KM,

Cheng SJ. J Cryst Growth 2008;310(23):5170.46. Zang KY, Wang YD, Liu HF, Chua SJ. Appl Phys Lett 2006;89(17):171921.47. Kikuchi A, Kawai M, Tada M, Kishino K. Jpn J Appl Phys 2004;43(12 A):L1524.48. Kikuchi A, Tada M, Miwa K, Kishino K. In: Proc. Int. Soc. Opt. Eng. (SPIE), vol. 6129;

2006. p. 36.49. Watanabe S, Yamada N, Nagashima M, Ueki Y, Sasaki C, Tamada Y, Taguchi T, Kudo H.

Appl Phys Lett 2003;83(24):4906.50. Lee YJ, Chiu CH, Ke CC, Lin PC, Lu TC, Kuo HC, Wang SC. IEEE J Sel Top Quan

Electron 2009;15(4):1137.51. Fullwave 6.1. Ossining (NY): RSoft Design Group Inc.; 2008.52. Tsai CH, Ma MH, Yin YF, Li H-W, Lai WC, Huang JJ. Nanoscale epitaxial lateral over-

growth of GaN-based light-emitting diodes on an AlN nanorod-array template. IEEE JQuan Electron May 2015;51(5):1e5.

53. Lin HY, Wang SW, Lin CC, Chen K-J, Han HV, Tu Z-Y, Tu HH, Chen TM, Shih MH,Lee PT, Chen HM, Kuo HC. Excellent color quality of white-light-emitting diodes byembedding quantum dots in polymers material. IEEE J Selected Top Quan ElectronJanuaryeFebruary 2016;22(1):35e41.



Nonpolar and semipolar LEDs 8Yuh-Renn Wu1, C.-Y. Huang 2, Yuji Zhao 3, James Speck 4

1National Taiwan University, Taipei, Taiwan; 2TSMC Solid State Lighting, Ltd, Hsinchu,Taiwan; 3Arizona State University, Tempe, AZ, United States; 4University of California,Santa Barbara, CA, United States

8.1 Motivation: limitations of conventional c-plane LEDs

8.1.1 Quantum-confined Stark effect (QCSE)

In the 1990s, breakthroughs in GaN epitaxial growth techniques1e3 and p-type GaNactivation4,5 led to the development of violet and blue light-emitting diodes (LEDs).High-quality nitride materials were grown with a GaN buffer layer on c-plane sapphiresubstrates along the [0001] direction of wurtzite unit cells, i.e., on the gallium polar(0001) surface.2,3 In 1994, Nakamura et al. produced the first candela-class blueLED with a zinc-doped InGaN/AlGaN active region.6 Today, high-brightness LEDswith efficacy greater than 100 lm/W are commercially available.

However, due to the absence of inversion symmetry in the wurtzite crystal structure,nitride-based materials are piezoelectric.7e9 The biaxial plane stress in (0001) InGaNquantum wells (QWs) results in internal electrical fields along the c-axis. The discon-tinuity of polarization between a QW and the barrier results in a sheet charge in theinterfaces. In the band profile for (0001)-oriented GaN/InGaN (QW)/GaN, thepolarization-induced charge separation results in a triangle-shaped potential profile,resulting in a reduced energy separation between the eigenstates in the conductionband and valence band. Therefore, the internal electric field causes a built-in red shiftof the emission wavelengths in c-plane LEDs. With increasing injection current, thesheet charges at the QW/barrier interfaces are screened by injected carriers, which flat-tens the QWs’ potential profile.10,11 As a result, conventional c-plane LEDs sufferfrom a blueshift of the emission wavelengths with increasing injection current. Thiswavelength shift with applied bias in a confined heterostructure is known as thequantum-confined Stark effect (QCSE).

Due to the triangular QW potential profile, the electron and hole wave functions inthe QWs are spatially separated, which reduces the radiative recombination rate and inturn likely reduces the internal quantum efficiency (IQE). Therefore, to maintain adecent wave function overlap and IQE, the QW thickness in conventional c-planeLEDs is usually between 2 and 3 nm. However, thin QWs are regarded as one ofthe major causes of carrier transport and droop issues of LEDs under a high injectioncurrent.

Nitride Semiconductor Light-Emitting Diodes (LEDs). https://doi.org/10.1016/B978-0-08-101942-9.00008-3© Woodhead Publishing Limited, 2014

https://doi.org/10.1016/B978-0-08-101942-9.00008-3

8.1.2 The green gap

In the AlGaInP system, high IQE LED can be realized for the red spectral region(l ¼ 630e650 nm). However, the peak IQE is significantly lower in the yellow-green spectral region due to the direct-indirect bandgap transition. InGaN materialshave a direct bandgap all over the visible spectral region. InGaN/GaNbased blueLEDs (l ¼ 440e460 nm) with high IQE have been demonstrated and are commer-cially available.12 However, the efficiency of nitride-based LEDs also drops signifi-cantly toward the green spectral region. The low efficiency of green to yellowLEDs, whatever the material system, is known as the “green gap” in the LEDcommunity.13

The low efficiency of nitride-based LEDs is attributed to two major causes:difficulties in epitaxial growth and strong QCSE. The lattice mismatch between InNand GaN is around 10%. To increase an LED’s emission wavelengths, the indium con-tent in the QWs has to be increased. For emissions in the green spectral region, theindium content in the QW has to be around 30%, which gives aw3% lattice mismatchbetween the QW and the unstrained GaN underlayer. With high strain in the active re-gions, the strain energy is prone to being relaxed by the generation of new defects, suchas pits, dislocations and stacking faults. Therefore, the epitaxial growth of nitride-based LEDs with high indium content is an area of significant interest.

Furthermore, QCSE also limits the performance of LEDs in the green spectralregion. Since the polarization discontinuity between InGaN and GaN is proportionalto the lattice mismatch strain, c-plane LEDs with a higher indium content in theQWs suffer more from QCSE. Therefore, green LEDs have a stronger wavelength shiftand lower IQE compared to blue LEDs. The challenges in epitaxial growth might bedealt with by strain management14 and further optimization of growth conditions.15e18

However, the limitation due to QCSE originates from the nature of the materials,which is difficult to solve or alleviate by engineering. Hence, there is significant moti-vation for growing devices along crystal orientations with no or reduced internalpolarization, to give further improvements in device performance in the green gap.

8.1.3 Carrier transport problems inmultiple-quantum-well LEDs

As mentioned above, the QW thickness of c-plane LEDs is kept thin to maintain adecent IQE. Therefore, increasing the number of QWs is a common method forincreasing the total active volume. However, the achievable effective active volumeof multiple-quantum-well (MQW) LEDs is limited by carrier transport betweenQWs. Electrons traveling in the conduction band have a smaller effective mass andhigher mobility than holes in the valence band, resulting in an unbalanced carrierdistribution among the QWs.18e22

For InGaN/GaN QWs, the band offset in the conduction band DEc is higher thanthat in the valence band DEv (DEc ¼ 0.6e0.7DEg, where DEg is the full bandgap dif-ference). Thus, the electrons should have a higher barrier for escaping the QW bythermionic emission. However, experimental data show that hole transport is the majorlimiting factor in c-plane MQW LEDs.23,24 Highly p-doped AlGaN layers are grown


as electron-blocking layers (EBLs) above the active region to alleviate electron leakageproblems.25,26 Schubert attributed the electron leakage problem to the short dwell timeof electrons after being injected into the QW.27 The deep conduction offset and smalleffective mass give a high initial velocity to electrons injected into QWs. In a thin QW,the injected electrons have a high probability of coherently traveling through the QWwithout being captured by phonon relaxation. Sizov et al. proposed that ballistic trans-port is the dominant transport mechanism in long wavelength LEDs and laser diodes(LDs) instead of the commonly used drift-diffusion transport models.28 Both of theabove studies suggested that the polarization-related electric fields in QWs favor bal-listic transport. The sheet charge on the QW/barrier interface accelerates the injectedelectrons and significantly reduces their dwell time. For holes injected into the QW,the initial velocity is much lower because of the smaller offset in the conductionband and their large effective mass. The probability of a hole experiencing ballistictransport through a QW is much lower. For those holes escaping from a QW via therm-ionic emission or via interacting with phonons, the strong polarization-related electricfields in the barrier forbid hole injection between adjacent QWs, causing most injectedholes to populate the QWs that are nearest to the p-side.

Many barrier structures have been proposed to improve the hole injection efficiencyin c-plane MQW LEDs, for example, the magnesium-doped barrier,29,30 InGaNbarrier,31,32 compositionally graded InGaN barriers,33,34 etc. However, introducingmagnesium into the barrier causes magnesium to diffuse into the QWs and lowersthe IQE of the devices, while using InGaN as a barrier causes strain managementissues. Since nonpolar and semipolar LEDs have no or reduced polarization-inducedelectric fields, the hole injection efficiencies are improved compared to c-planeLEDs.28,35

8.1.4 Efficiency droop

Although LEDs with a high peak IQE under a low injection current are commerciallyavailable, their efficiency drops quickly with increasing current injection. The phe-nomenon of IQE deterioration with increasing injection current density is known as“droop”. To increase the total radiation flux of commercial LED bulbs, the totalchip area was increased instead of driving up the current density. Therefore, theefficiency was retained though with increased fabrication costs. Hence, overcomingdroop in LEDs for general illumination became a major focus of research.

To date, the physical origin of droop is still inconclusive. Some droop mechanismshave been proposed with potential solutions. Lumileds suggested that Auger recombi-nation is the major cause of efficiency droop.36,37 Auger recombination is a nonradia-tive recombination process in which the rate is proportional to the cube of the carrierdensity n (wn3) in the materials. Lumileds observed an efficiency droop in the photo-luminescence (PL) of an InGaN layer with increasing excitation power. The experi-ment was done under zero bias to exclude the effects of carrier dynamics in theactive region. They concluded that droop is an inherent property in materials regardlessof the design of a device. Kioupakis et al. from UCSB supported this conclusion with afirst principles calculation of the Auger recombination coefficient (C) in InGaN.38 It

Nonpolar and semipolar LEDs 275

was argued that the phonon-assisted hole-hole-electron (h-h-e or Cp0 phonon) Augerrecombination process dominates under a high carrier density especially for longwavelength emitters. The solution proposed by Lumileds’ theory was to use a thickInGaN QW to reduce the carrier density in the active region.39 Li et al. at the NationalTaiwan University observed that the peak efficiency of c-plane LEDs was pushed from10 to 200 A/cm2 by increasing the QW thickness from 2.5 to 13 nm, which is in agree-ment with Lumileds’ argument.40

However, researchers at Rensselaer Polytechnic Institute have an alternativeview on the origin of droop. Instead of Auger recombination, they suggested thatelectron leakage under a high injection current is the dominant factor for efficiencydroop.41 It was proposed that the electrons are swept through the active region dueto their small effective mass and the existence of a sheet carrier at QW/barrier in-terfaces. Those electrons that are not captured by QWs eventually vanish outsidethe active region, resulting in the low efficiency under high injection. In this pointof view, droop can be alleviated by designing epilayer structures to balance theelectron and hole transport in the active region. Samsung used a quaternary (Al,In, Ga)N alloy as barriers to match the polarization between barriers and QWs.42

Using a quaternary alloy provides another degree of freedom when designing thepolarization and bandgap of nitride materials. With the same polarization in the bar-riers and QWs, the sheet charge at the interface can be eliminated. Researchers atVirginia Commonwealth University designed QWs with a stepped potential to miti-gate the effects of hot electrons. They observed an improvement in droop perfor-mance.26,43 Wang et al. from National Chiao-Tung University used a gradedInGaN barrier in MQW c-plane LEDs to improve hole transport. Droop is improvedwhen there is a more uniform carrier distribution among the QWs.33 Most recentlystudies using atom probe tomography (APT)44 show that fluctuations in indiumcomposition in a QW might be the reason for spectrum broadening. Later numericalstudies that included indium alloy fluctuations44 showed that the alloy fluctuationcauses a carrier to become localized in the QW, which will increase the peakIQE. However, the much higher local carrier density will enhance Auger recombi-nation. At the same time, the smaller active volume will also increase the chancesof an overflow. The unstable alloy fluctuation in a QW also explains the variation indevice characterization.

8.1.5 Advantages of nonpolar and semipolar LEDs

The QCSE imposes a limit on device performance for c-plane LEDs. To circumventthe detrimental effects of the internal polarization, growing devices along orientationsthat have zero or minimal polarization has been proposed as the solution for manyunsolved issues. When there are no polarization-induced electric fields in the QW,the IQE is enhanced and there is an improved overlap between carrier wave functions.The QWs can be grown thicker without much reduction of the IQE as long as thedefect density is not significantly increased. Therefore, the unbalanced carrier transportin MQW devices resulting from the over-thin QWs can be mitigated. Additionally,lower polarization reduces the energy barrier for hole injection into MQW devices.


Kawaguchi et al. observed superior hole injection efficiency in semipolar MQW LEDsin dichromatic LED experiments with a simple GaN barrier structure.35 Recently,improved droop was observed in nonpolar and semipolar LEDs with similar QW/bar-rier structures in c-plane LEDs.45e47 Pan et al. from UCSB produced semipolar (2021)single-quantum-well blue LEDs with a uniform planar emission and low droop.47

Further theoretical calculations and experimental data for nonpolar and semipolarLEDs will be presented in other sections of this chapter.

8.2 Introduction to selected nonpolar and semipolarplanes

8.2.1 Crystallography of wurtzite nitride

The Miller indices of nonpolar and semipolar planes are (hkil) or (hkl), where i ¼ hþk and either h or k has a nonzero value, for example,

�1010

�and

�1122

�. The plane of

hkl is perpendicular to the vector (h, k, l) in the reciprocal lattice of wurtzite nitride. w isdefined as the angle between the surface normal vectors of the (0001) plane and anarbitrary

�hkil�plane and is given as:

q ¼ arccos

ffiffiffi3

pal

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi4c2ðh2 þ k2 þ hkÞ þ 3a2l2

p

!

[8.1]

Planes with l ¼ 0 are called nonpolar planes and those with h ¼ k ¼ 0 are polarplanes or the c-plane. All other non-c-planes with a nonzero l are all called semipolarplanes. Selected planes are schematically illustrated in Table 8.1 and Fig. 8.1.

Table 8.1 Selected wurtzite crystal planes

Plane q (degrees) Polarity

(1013) 32 Semipolar

(1012)/r-plane 43 Semipolar

(1122) 58 Semipolar

(1011) 62 Semipolar

(2021) 75 Semipolar

(1120)/a-plane 90 Nonpolar

(1010)/m-plane 90 Nonpolar

(2021) 105 Semipolar


8.2.2 Changes in piezoelectric polarization charge withorientation

GaN has a wurtzite structure (space group P63mc and point group 6 mm) and as aresult it is polar and piezoelectric. The polarization in the unstrained state is referredto as spontaneous polarization and additional polarization is referred to as piezoelectricpolarization; the c-axis is the polar axis. For the c-plane, semipolar and nonpolar struc-tures, a crystal can be rotated through angles q and 4, which are the angles of rotationfrom the z-axis (c-axis) to the x-axis and from the x-axis to the y-axis, respectively(Fig. 8.2).

In a semipolar or nonpolar quantum well, the crystal growth direction is differentfrom the traditional c-axis. If an InGaN quantum well is grown in a different direction,the lateral strain and vertical strain are in the new growth direction. But all equationsuse the old coordinate system. Therefore, we need to rotate the coordinates back to theregular plane to get the strain components. The relation between the new coordinatesand the old coordinates is defined by the equations:

Px0 ¼ Ux0xPx þ Ux0yPy þ Ux0zPz

Py0 ¼ Uy0xPx þ Uy0yPy þ Uy0zPz

Pz0 ¼ Uz0xPx þ Uz0yPy þ Uz0zPz

c-plane

m-plane

r-plane

a-plane

(0001)

(1010)

[0001] [0001] [0001]

[0001] [0001] [0001]

–(1120)–

(1012)–

(1122)–

(3031)–(2021)–

(1011)–

(1011)— (2021)— (3031)—

a3

a2

a1

a3

a1

a2 a2

a3

a1

a2a2a2

a3a3a3

a1a1a1

Figure 8.1 Polar, nonpolar and semipolar planes in wurtzite crystal structures.


Px, Py and Pz are some physical property, such as strain or the elastic tensor. Ux0x is thecosine of the angle between the x0 and x axes. Uxz is the cosine of the angle between x0and z axes. Note thatUx0zsUz0x. If we rotate the coordinates by q and 4, we can deriveU as:

U ¼

0

BB@

cos q cos 4 cos q sin 4 �sin q

�sin 4 cos 4 0

sin q cos 4 sin q sin 4 cos q

1

CCA [8.2]

We rotate the crystal growth axis (defined as the z0-axis) from the (x, y, z) coordi-nates to the (x0, y0, z0) coordinates. The relations between the wave vectors, strain ten-sors and elastic stiffness constants in the original coordinate system and the rotatedcoordinate system are:

k0i ¼

X

a

Uiaka;

εij ¼X

ab

UiaUjbεab;

C0ijkl ¼

X

abgd

UiaUjbUkgUldCabgd:

[8.3]

From Eq. (8.3) we can obtain the rotated elastic stiffness constants, which aredefined through Hooke’s law. The parameters for the elastic stiffness constants weused are also listed in Huang and Wu.48

According to Romanov et al.,49 the lateral strain of a rotated semipolar plane InGaNlayer grown on a GaN substrate is:

˛m1 ¼ aS � aLaL

¼ ˛y0y0 [8.4]

zz

x

y

y

y

xx

z′z′

x′

y ′

y ′

y ′

x ′

x′

φφ

φ

φ

θθ

θ

(a) (b) (c)

Figure 8.2 New and old coordinates.


˛m2 ¼aScS �

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�ðaLcSÞ2 cos2 qþ ðaScLÞ2 sin2 q

�r

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðaLcSÞ2 cos2 qþ ðaTcLÞ2 sin2 q

q ¼ ˛x0x0 [8.5]

where (cS, aS) and (cL, aL) are the lattice constants for the substrate and InGaN layer,respectively. q is the rotation angle from the z-axis to the x-axis. The strain in theoriginal coordinates can be obtained using the rotation matrix U as shown in Eq. (8.3).To calculate the piezoelectric polarization Ppz, we use:

Ppz ¼

0

BB@

0 0 0 0 e15 0

0 0 0 e15 0 0

e31 e31 e33 0 0 0

1

CCA

0

BBBBBBBBBBBB@

˛xx

˛yy

˛zz

˛yz

˛zx

˛xy

1

CCCCCCCCCCCCA

¼

0

BB@

e15˛xz

e15˛yz

e31ð˛xx þ ˛yyÞ þ e33˛xx

1

CCA [8.6]

where eij is the piezoelectric tensor in Voigt notation.49 When the plane is rotated to the

nonpolar or the semipolar plane, the strength of polarization changes as well.According to Romanov et al.,49 the polarization strength of an InGaN semipolarquantum well becomes:

DP0z ¼ Ppz

Lz0 þ�PspL � Psp

S

�cos q [8.7]

where PspL and Psp

T are the spontaneous polarizations for the layer and the substrate,respectively. Psp

Lz0 can be calculated from:

PspLz0 ¼ Psp cos q.

With some mathematical algebra,49 the strain-induced piezoelectric polarization,PspLz0 , can be expressed as:

PspLz0 ¼e31 cos q˛x0 x0 þ

�e31 cos

3 qþ e33 � e152

sin q sin 2 q

�˛y0y0

þ�ðe31 þ e15Þ

2sin q sin 2 qþ e33 cos

3 q

�˛z0 z0

þ ½ðe31 � e33Þcos q sin 2 qþ e15 sin q cos 2 q�˛z0 z0 .


The simplified polarization strength of an InGaN quantum well under a c-planegrown on a GaN substrate can be found in Kawaguchi et al.35:

PpzlnGaN�GaNðxÞ ¼ ½0:148x� 0:0424xð1� xÞ�C m�2

InGaN �GaNThe polarization strength for different rotation angles can be plotted as shown in

Fig. 8.3.

(a)

(b)

0.064

0.048

0.032

0.016

0.000

–0.016

0.064

0.048

0.032

0.016

0.000

–0.016

0 20 40 60 80

0 20 40 60 80

θ °

θ °

In0.1Ga0.9N

In0.2Ga0.8N

In0.3Ga0.7N

In0.4Ga0.6N

In0.1Ga0.9N

In0.2Ga0.8N

In0.3Ga0.7N

In0.4Ga0.6N

Pz’

(C/m

2 )P

Lz’ (

C/m

2 )pz

Δ

Figure 8.3 Calculated polarization charge density as a function of inclination angle for anInGaN quantum well.48


8.2.3 Influence of polarization on band bending, QCSE andcarrier transport

As mentioned in the previous section, the polarization strength will change as thegrowth plane is rotated. When q is larger than 45 degrees, the polarization directionwill even become switched. This makes the transport behavior of electrons and holesrelatively complicated. For a nonpolar plane, such as the m-plane or a-plane, thepolarization strength is zero, and there are no polarization-related electric fields orpolarization-related QCSE. For a rotation angle between 0 and 45 degrees, the polar-ization direction is the same as for the c-plane with a gallium face but with a smallerelectric field. When q > 45 degrees, the band bending direction will change and this issimilar to the nitrogen faces, which might significantly affect carrier transport. Fig. 8.4shows the band bending profile for the c-plane, the nonpolar m-plane, (1013) and(2021). As shown in Fig. 8.3, the nonpolar plane (q ¼ 90 degrees) and the r-plane(1012) (q ¼ 45 degrees), where the polarization is close to zero, are like a normalQW. For the (1013) plane, the polarization direction is close to that of the c-planebut with less band bending. For (2021), the polarization has switched. This will causean additional potential barrier when the carrier wants to escape from the first quantumwell to the next quantum well, which has been observed experimentally by Kawaguchiet al.35 The result showed that the direction of the polarization field will affect carriertransport significantly, especially for an MQW. The extra potential barrier caused bythe polarization field in the case like a nitrogen face will increase the forward voltage

(a) (b)

(c) (d)

1

0

–1

–2

–3

–4

1

0

–1

–2

–3

–41.4 1.6 1.8 2 2.2 2.4 2.6

1

0

–1

–2

–3

–41.4 1.6 1.8 2 2.2 2.4 2.6

1

0

–1

–2

–3

–41.4 1.6 1.8 2 2.2 2.4 2.6

1.4 1.6 1.8 2 2.2 2.4 2.6

EcEvEfnEfp

EcEvEfnEfp

EcEvEfnEfp

EcEvEfnEfp

V (e

V)

V (e

V)

V (e

V)

V (e

V)

z (10–5 cm)

z (10–5 cm) z (10–5 cm)

z (10–5 cm)

Figure 8.4 Calculated band profiles of InGaN MQWs for different growth orientations:(a) c-plane, (b) m-plane, (c) (1013), (d) (2021). Ec is the conduction band potential, Ev thevalance band potential, Efn the electron quasi-Fermi level and Efp the hole quasi-Fermi level.


of an LED significantly. Hence, the choice of semipolar plane will affect the powerefficiency, which is a critical issue for energy applications.

8.2.4 Influence of anisotropic strain on E-k relaxation, bandstate mixing and optical polarization emission of light

To understand the light emission polarization of a semipolar quantum well system, weneed to address the following issues: (1) the valence band state mixing, (2) the energyseparation of the first valence band (CH1) and the second valence band (CH2) and (3)the effective mass ratio of the CH1 to CH2 bands, which will strongly affect thestrength of the quantum confinement. The 6 � 6 k/p method can be used to calculatethe influence of the valence band mixing and the strain. The 6 � 6 Hamiltonian can beexpressed as:

Hv ¼

0

BBBBBBBBBBBB@

F 0 �H� 0 K� 0

0 G D �H� 0 K��H D l 0 I� 0

0 �H 0 l D I�K 0 I D G 0

0 K 0 I 0 F

1

CCCCCCCCCCCCA

ju1iju2iju3iju4iju5iju6i

[8.8]

where

F ¼ D1 þ D2 þ lþ q

G ¼ D1 � D2 þ lþ q

l ¼�

Z2

2m0

�hA1k

2z þ A2

�k2x þ k2y

�iþ D1˛zz þ D2ð˛xx þ ˛yyÞ

q ¼ Z2

2m0

hA3k

2z þ A4

�k2x þ k2y

�iþ D3˛zz þ D4ð˛xx þ ˛yyÞ

K ¼ Z2

2m0A5ðkx þ ikyÞ2 þ D5ð˛xx � ˛yy þ 2i˛xyÞ

H ¼ Z2

2m0i

A6kzðkx þ ikyÞ þ A7ðkx þ ikyÞ þ i D6ð˛xz þ i˛yzÞ

I ¼ Z2

2m0i

A6kzðkx þ ikyÞ � A7ðkx þ ikyÞ þ i D6ð˛xz þ i˛yzÞ

D ¼ffiffiffi2

pD3


D1 to D6 are the deformation potentials and A1 and A7 are the fitting parameters tothe valence band structure. ki and ˛ij (i, j ¼ x, y, z) are the wave vector and the straintensor. D1 is the crystal-field energy. D2 and D3 are the spin-orbit energy parameters.The parameters can be found in Huang and Wu.48

For the Hamiltonian of the crystal growth orientation z0-axis, kz0 will be transformedinto the differential forms �iv/vz0. The bases of the Hamiltonian

u1itou6iare

1ffiffiffi2

pX þ iY ;[i; 1

ffiffiffi2

pX þ iY ;Yi; jZ[i; jZYi 1

ffiffiffi2

pX � iY ;[i;

and1ffiffiffi2

pX � iY ;Yi

The polarization of the emission light is strongly affected by these bases.Plate 6 shows the valence band E-k diagrams of an InGaN quantum well for the c-

plane, the m-plane and the semipolar (1122) plane. Red, green and blue represent thestates of jX0i, jY 0iand jZ 0i. If a state is mixed, it will have a mixed color as shown in thetriangle in Plate 6(a).

In a c-plane InGaN/GaN strained QW, the basis of the valence band is mainlydominated by the jX � iYi states. Therefore, for the interband transition, the emittingstrength for different polarized light is mainly given by the momentum matrixelement:

hfeðzÞjfhðzÞihSjja$

�� iZV

. �jX � iYij2 [8.9]

where hfeðzÞjfhðzÞidenotes the overlap between the z-dependent envelope function ofthe conduction and valence bands, which determines the dipole strength. a

ˇ

is thepolarization of the generated light or the incident light. Since the bases are mainlydominated by the jXi-like and jYi-like states for the c-plane case, the radiating dipolesare oriented along the plane of the QW. For a nonpolar plane, such as an m-planequantum well (grown along the y direction), the jXi-like (X0) (defined as the a-axishere) state will become the first subband due to the anisotropy. The jZi-like (Y0) statewill become the second or even third subband and the jYi-like (Z0) state will bethe final one due to the anisotropic strain and quantum-confined effect. Therefore,the radiating dipole will be oriented along the a-axis, which will be a linearly polarizedlight source along the plane of the QW. For the semipolar plane, the situation isrelatively more complicated. From the literature,48,50,51 due to the anisotropicstrain and shear strain in the semipolar QW, the valence band will be mixed with jXi-,jYi- and jZi-like states depending on the rotation angle and the strain (or indiumcomposition). For example, for the (1122) plane shown in Plate 6(c), which is rotatedby w58 degrees with respect to the (0001) plane, the first subband is mainly domi-nated by jY 0i-like states for a low indium composition (<30%). Note that here we userotated coordinate x0, y0 and z0 axis for ease of comparison. The second subband is


mixed with mjX0i þ njZ 0i like states. However, as the indium composition increases,the mjX0i þ njZ 0i subband will rise to the first subband due to the increase of theanisotropic strain. Therefore, the out-of-plane polarization will be observed for thehigher indium cases.52 From Huang and Wu,48 we know that if the indium compo-sition is less than 20%, the emitting dipole is mainly oriented along the in-plane of theQW for most rotating angles. When the indium composition continues to increase, theout-of-plane dipole will rise, especially for the angles around 20e60 degrees.

8.3 Challenges in nonpolar and semipolar epitaxialgrowth

8.3.1 Heteroepitaxy of nonpolar and semipolar planes

In the early stages when bulk GaN substrates were not available, researchers tried togrow nonpolar and semipolar GaN templates on foreign templates. As in c-planeGaN heteroepitaxy, sapphire and SiC substrates were chosen because of their accept-able lattice mismatch and availability. (1120) plane GaN, i.e., a-plane GaN, has beengrown on an r-plane sapphire substrate.53e57 Although a-plane GaN on r-plane sap-phire is cost effective, the template suffered from a high density of facetted surfacepits,48,58 basal plane stacking faults and associated terminating partial threading dislo-cations.55 GaN templates grown on m-plane sapphire have been reported for variousplanes with different pretreatments, such as the (1010), (1013) and (1122) planes.59e62

To produce an m-plane GaN template, g-LiAlO263,64 and m-SiC65e67 substrates are

better candidates because of their smaller lattice mismatch. However, LiAlO 2 sub-strates are chemically unstable and large m-SiC substrates are still expensive e and,in any case, m-plane GaN on either substrate had high densities of basal plane stackingfaults and partial threading dislocations. For semipolar GaN templates, a (1122) GaNtemplate has been reported on m-plane sapphire,53,54,68,69 patterned r-plane sap-phire70,71 and patterned Si substrates.72. (1011) or (1011) GaN templates have beenreported on (100) MgAl2O4 spinel73,74 and patterned (001) silicon substrates.75 A(2021) GaN template has been prepared on a patterned (2243) sapphire substrate.76

However, the crystal quality of nonpolar and semipolar GaN heteroepitaxy is stillunsatisfactory. There is a high density of extended effects such as basal plane stackingfaults (BPSFs) and pits. Although the defect density can be significantly reduced bycoalescence growth on a patterned substrate,70,71,75,76 high-quality large-area (e.g.,200) nonpolar and semipolar GaN templates are still unavailable. To date, there are stillno materials that can be used as substrates for GaN heteroepitaxy in an arbitraryorientation.

8.3.2 Homoepitaxy and the need for bulk GaN substrates

Instead of overcoming all the challenges of nonpolar and semipolar heteroepitaxy,homoepitaxy on free-standing bulk GaN substrates is widely regarded as the ultimate


solution for nonpolar and semipolar epitaxy. Strain-free homoepitaxy greatly reducesthe challenges in metal organic chemical vapor deposition (MOCVD) growth with alower defect density. Substrates with arbitrary orientation are also easily availableby cutting and polishing the free-standing bulk GaN. Therefore, the availability ofhigh-quality bulk GaN is the key issue in the development of nonpolar and semipolardevices.

Because nitride materials decompose before they melt under atmosphere pressure,conventional liquid-phase bulk crystal growth techniques used in silicon and conven-tional III-V arsenide and phosphide systems are nearly impossible for nitrides. Free-standing bulk GaN can be grown by hydride vapor-phase epitaxy (HVPE). TheGaN is synthesized from gas-phase gallium chloride (GaCl) and ammonia (NH3) ona heated foreign substrate such as sapphire or SiC under atmosphere pressure.77

Free-standing GaN is then obtained by a subsequent lift-off process from thesubstrate.78e80 Smooth surfaces are obtained after several lapping and polishing steps.

To fabricate bulk nonpolar and semipolar substrates, two approaches are possible:(1) directly grow GaN along a nonpolar or semipolar direction via HVPE81e83 or (2)slice the bulk GaN grown along the c-direction to expose the desired nonpolar or semi-polar surface.84 Issues similar to those for nonpolar and semipolar heteroepitaxy byMOCVD have also been encountered in HVPE growth. The high extended defect den-sity and limited selection of planes impose strong challenges on nonpolar and semipo-lar HVPE growth. In addition, HVPE GaN grown along the c-direction has muchbetter crystal quality.77,84 High-quality substrates with an arbitrary surface can beobtained by cutting HVPE GaN grown along the c-direction.

The scalability of the HVPE technique is limited by undesired gas-phase parasiticreactions.85 Alternatively, the ammonothermal method has attracted significant atten-tion because of its high scalability for mass production. The ammonothermal methoduses supercritical ammonia to dissolve poly-crystalline GaN with mineralizers (forexample, KNH2 and NH4X) under high pressure and high temperature.86e88 Thesupercritical ammonia carries the dissolved GaN to seeds, which are at a lower temper-ature. The dissolved GaN crystallizes onto the seeds because of the lower solubility.The ammonothermal method can be scaled up by simply increasing the autoclavesize and the number of seeds. The challenges for the ammonothermal method comefrom the high pressure (>100 MPa) and corrosive environment. Safety issues wouldraise the cost, despite its potential for high scalability. Growing Bulk GaN withhigh quality and low cost is still under development.89

8.3.3 Indium incorporation in nonpolar and semipolar planes

The InxGa1�xN alloy plays an important role in nitride-based optoelectronics becauseits bandgap spans from UV to IR. However, the growth of the InGaN alloy is generallyregarded as difficult even on free-standing GaN. Besides the large lattice mismatch be-tween InN and GaN, the volatile indium adatoms impose challenges on InGaNepitaxial growth. To attain a high indium composition in MOCVD growth, the temper-ature has to be low to reduce indium desorption. However, poor adatom diffusion atlow temperatures causes difficulties in the growth of high-quality InGaN with a


high indium composition. The incorporation of indium is influenced by growth condi-tions such as chamber pressure and growth rates, but mostly it is determined by theatomic configuration at the surface. Therefore, a plane surface that favors indiumincorporation is highly desirable.

The incorporation of indium varies for the different surfaces in a wurtzite nitridestructure. For example, the m-plane showed relatively low indium incorporationcompared to semipolar planes.90 A theoretical calculation attributed the low indiumincorporation to repulsive interactions between indium adatoms.91,92 Durnev et al.suggested that semipolar planes with a 60 degrees inclination angle to the c-plane(w w 60 degrees) have the highest indium incorporation due to the minimal strain ef-fects, which is consistent with data from co-loaded growth experiments.92 The (1122)plane (w w 58 degrees) had the highest incorporation of indium among variousnonpolar and semipolar planes with q ranging from 58 to 90 degrees.93 Besides thestrain-induced repulsive effects, the surface polarity also has a strong influence onindium incorporation. Zhao et al. showed that the (2021) surface has a higher indiumincorporation than the (2021) surface.93 These two planes have the same strain schemewith reversed polarities. A similar phenomenon was also observed between gallium-polar and nitrogen-polar surfaces.94 A first principles calculation suggested that ahigher binding energy between indium and the nitrogen-polar surface reduces theprobability of indium desorption on the surface.95 Therefore, a surface with anitrogen-polar surface component has higher indium incorporation.

High-efficiency green LEDs have been produced on the (2021) and (1122)planes.96,97 The high quality of the InGaN QWs was attributed to the high indiumincorporation on the these surfaces. However, green (2021) LEDs have not yet beenproduced due to other challenges in MOCVD growth.

8.3.4 Morphology of nonpolar and semipolar epitaxy

The morphology of nonpolar and semipolar epitaxy varies because of the uniqueatomic structure near the surface. A rough surface morphology causes a deteriorationin the quality of epilayers and limits device performance. For nonpolar planes, striatedmorphologies have been observed for (Al, In, Ga)N epitaxy.98e101 Atomic force mi-croscope images showed that the striations on the a-plane are along the c-axis andthose on the m-plane are along the a-axis. A first principles calculation suggestedthat the anisotropic adatom diffusion causes striation morphologies along a directionwith a lower diffusion kinetic barrier.102 For semipolar planes, striated morphologieswere seen to be similar to their vicinal nonpolar planes. For example, striations wereparallel to the c-axis projection for (1122) GaN epitaxy101,103 and striations were alsoobserved along the a-axis on (2021) and (2021) epilayers.104,105

Threading dislocations (TDs) also influence the surface morphology for nonpolarand semipolar epitaxy. TDs with a screw dislocation component enhance the spiralgrowth centered at the TDs. The spiral growth creates pyramidal hillocks several mi-crons high on the surface.101,106e109 Hillock morphologies cause a non-uniformgrowth rate and indium incorporation across the surface. Lin et al. observed non-uniform photoluminescence in an m-plane laser diode structure with hillock


morphologies.109 To suppress the spiral growth centered at the TDs, an intentionalmiscut was introduced onto nonpolar and semipolar substrates. The right miscut ona substrate enhances terrace growth by introducing free steps on the surface, whichinterrupt the spiral growth by an interaction with the terrace growth. With a 1 degreemiscut toward the nitrogen-polar surface, a planar surface morphology was achievedon m-plane LEDs and LDs.109,110

8.3.5 Strain-induced defect generation in nonpolar andsemipolar epitaxy

The large lattice misfit between InN and GaN causes the growth of QWs with a highindium content challenging. The misfit stresses in the QWs are likely to be reduced byextended defect generation. Basal plane stacking faults (BPSFs) are commonlyobserved in nonpolar and some semipolar green emitters.90,111 Wu et al. found thatthe BPSFs in m-plane InGaN QWs (indium w 26%) were bounded by a half loopof partial dislocations extending to the surface.90 Similar defect structures were alsofound in (2021) and (2021) green QWs. Black triangles observed using a fluorescencemicroscope in semipolar green emitters were also expected to be related to the forma-tion of BPSFs.105,109 The radiative efficiency of long wavelength nonpolar and semi-polar LEDs is therefore limited due to the high defect densities.

Beside the formation of BPSFs, strain relaxation also generates misfit dislocations(MDs) in semipolar and nonpolar epitaxy. MDs along the m-axis can be formed by abasal plane slip in semipolar (1122) InGaN/GaN heteroepitaxy.112e115 MDs along thea-direction have also been observed in (2021) InGaN/GaN heteroepitaxy and wereproduced by a similar mechanism.116 For the semipolar (3031) plane and the m-plane,the relaxation is initiated through a prismatic m-plane slip instead of a basal plane slipdue to the high shear stress.114e116 The formation of MDs could introduce surfacemorphologies and facilitate the nucleation of new defects. Romanov et al. calculatedthe critical thicknesses for the formation of MDs in nonpolar and semipolar heteroepi-taxy with the Matthewe Blakeslee equilibrium model.117 The critical thickness calcu-lation can be used a reference in epilayer structure design to avoid the generation ofdefects from strain relaxation.117e119

8.4 Light extraction for nonpolar and semipolar LEDs

8.4.1 Light extraction efficiency as a limiting factor

Despite the many advantages of nonpolar and semipolar LEDs (e.g., reducedpolarization-related effect, higher optical gain, etc.),120e127 the output power and effi-ciency of current semipolar and nonpolar LEDs are still lower than those of the bestreported c-plane devices, mainly due to the poor light extraction efficiency (hextr)compared with their polar counterparts.128e132 Light extraction efficiency has becomethe most important limiting factor for the efficiency of LEDs, since the IQE of nitride-based LEDs has been greatly improved (by more than 80%)133 by the availability of


low-dislocation GaN substrates and advances in MOCVD techniques. The low hextr isprimarily caused by the low critical angle (23 degrees) of the light escape cone, due tothe large difference between the refractive indices of GaN (n w 2.5) and air (n ¼ 1).134

For c-plane devices, both hextr and the corresponding output power have been greatlyimproved by surface roughening methods such as patterned sapphire substrate (PSS)and photoelectrochemical (PEC) etching techniques.134e138 Semipolar and nonpolardevices, however, still suffer from low light extraction efficiency due to the lack ofproper roughening techniques, which have hindered their performance. In this section,we review several popular approaches in processing techniques, device structures andpacking methods to increase the hextr of InGaN-based LEDs.

8.4.2 Increasing hextr via surface roughening

There have been many reports of advanced surface roughening techniques through theperiodic patterning of the semiconductor material or an overlying dielectric layer, withmicro or nano features to increase the hextr of LEDs. For c-plane devices, PSS and PECetching techniques have been widely implemented for commercial devices.134e138 Forsemipolar and nonpolar devices, however, there are very few reports. Recently, Zhonget al. and Zhao et al. at UCSB demonstrated that surface patterning with conical fea-tures by inductively coupled plasma (ICP) etching could have the potential to achieverelatively high power and high efficiency semipolar devices.139,140 The backsides ofthe LED devices were polished and patterned with conical features by conventionalcontact lithography followed by ICP etching. It was discovered that the hextr of theLEDs increased dramatically as the density of the patterns increased, compared witha reference device with a smooth surface. Fig. 8.5 demonstrates the dependence oflight output power (corresponding to hextr) on the density of the conical features.The experimental finding was also consistent with theoretical Monte Carlo ray tracingsimulations on c-plane devices. This patterning technique resulted in semipolar blueLEDs with over 50% external quantum efficiency (EQE), which is comparable tothe best c-plane devices. It is also noteworthy that nonpolar and semipolar LEDsare usually characterized by a high degree of polarized emission, which is of signifi-cant interest for laser and display applications.141,142 Novel surface patterning design,such as with photonic crystal structures, to control or manage this optical polarizationis a very interesting topic of research.143

8.4.3 Thin-film flip-chip LEDs

One of the most popular chip designs adopted by the LED industry to increase hextr isthe thin-film flip-chip (TFFC) LED.144 The fabrication process for a TFFC LED typi-cally includes a substrate removal step (sapphire or SiC substrate for a c-plane LED).After that, the LED chips are flipped and mounted. The thinness of the LEDs greatlyreduces the modes that are trapped in the LED chip due to total internal reflection.Furthermore, the n-GaN surfaces are roughened using the process techniquesdescribed in Section 8.4.2 to further increase hextr.

145 To date, Lumileds and Osramhave successfully produced TFFC or thin-GaN device structures on InGaN-based


LEDs.146,147 A high light extraction efficiency of 80% has been reported for theseTFFC LEDs.148 However, up to now, most of the TFFC LEDs have been fabricatedon c-plane devices. There are very few reports of TFFC LEDs on nonpolar or semipo-lar devices.

8.4.4 High light extraction packaging

Current conventional LED packaging structures suffer significantly from the backcoat-ing effect, where a fraction of the light in an LED die is absorbed by the active layer,

(a)

(b) (c)

Smooth

Smooth

5.8 x 105 cm–2

9.0 x 105 cm–2

1.6 x 106 cm–2

1.6 x 106 cm–2

0

2

4

6

8

10

12

0 5 10 15 20 25 30 35 40Current (mA)

Ligh

t out

put p

ower

(mW

)

10μm 10μm

Figure 8.5 (a) Dependence of light output power on the density of conical features patterned onthe backside of the devices. Scanning electron microscopy (SEM) images of the backside ofthe device with (b) a smooth surface and (c) a high density (r ¼ 1.6 � 106 cm�2) of conicalfeatures (upper diameter Ø ¼ 3 mm).140


semi-transparent contacts (e.g., indium-tin-oxide or ITO) or reflective surfaces in thepackage (e.g., silver header coatings) due to the lead frames or the reflectors on theback surface of the LED dies. As a result, only a fraction of light is able to escapeto the free space and the overall efficiency of the LEDs is reduced. To circumventthis problem, advanced LED package architectures, such as the suspended LED pack-age, in which an LED die is suspended from gold wire-bonding leads, have been pro-posed.149,150 However, this type of packaging has caused several heat dissipationproblems, which greatly reduced the light output power of the LEDs at high currentinjections. More recently, Pan et al. at UCSB used a novel vertical-stand LED archi-tecture based on a transparent submount (ZnO).151 This package structure allows lightto be extracted from all faces of an LED die, in particular from the backside surface,which is enhanced by the high refractive index and roughened surface of the verticalZnO stand. In addition, using a ZnO submount helped to improve the heat dissipationfrom the LEDs. This packaging architecture has been successfully demonstrated forsemipolar blue LEDs, which had a high EQE of over 50%.46 The estimated hextr isaround 75%, which is much higher than for the conventional packaging.152

References

1. Amano H, Sawaki N, Akasaki I, Toyoda Y. Appl Phys Lett 1986;48:353.2. Nakamura S. Jpn J Appl Phys 1991;30:L1e705.3. Nakamura S, Senoh M, Mukai T. Jpn J Appl Phys 1991;30:L1e708.4. Amano H, Kito M, Hiramatsu K, Akasaki I. Jpn J Appl Phys 1989;28:L2e112.5. Nakamura S, Mukai T, Senoh M, Iwasa N. Jpn J Appl Phys 1992;31:L1e39.6. Nakamura S, Mukai T, Senoh M. Appl Phys Lett 1994;64:1687.7. Bernardini F, Fiorentini V. Phys Stat Sol (B) 1999;216:391.8. Takeuchi T, Wetzel C, Yamaguchi S, Sakai H, Amano H, et al. Appl Phys Lett 1998;73:

1691.9. Takeuchi T, Sota S, Katsuragawa M, Komori M, Takeuchi H, et al. Jpn J Appl Phys 1997;

36:L382.10. Kuroda T, Tackeuchi A. J Appl Phys 1999;92:3071.11. Chichibu S, Azuhata T, Sota T, Nakamura S. Appl Phys Lett 1996;69:4188.12. Narukawa Y, Ichikawa M, Sanga D, Sano M, Mukai T. J Phys D Appl Phys 2010;43:

354002.13. Krames MR, Shchekin OB, Mueller-Mach R, Mueller GO, Zhou L, et al. J Display

Technol 2007;3:160.14. Houghton DC, Davies M, Dion M. Appl Phys Lett 1994;64:505.15. Oh MS, Kwon MK, Park IK, Baek SH, Park SJ, et al. J Cryst Growth 2006;289:107.16. Wetzel C, Li P, Detchprohm T, Nelson JS. Phys Stat Sol (C) 2005;2:2871.17. Detchprohm T, Zhu M, Zhao W, Wang Y, Li Y, et al. Phys Stat Sol (C) 2009;6:840.18. Shioda T, Yoshida H, Tachibana K, Sugiyama N, Nunoue S. Phys Stat Sol (A) 2012;209:

473.19. David A, Grundmann MJ, Kaeding JF, Gardner NF, Mihopoulos TG, et al. Appl Phys Lett

2008;92:053502.20. Liu JP, Ryou JH, Dupuis RD, Han J, Shen GD, et al. Appl Phys Lett 2008;93:021102.


21. Li YL, Gessmann T, Schubert EF, Sheu JK. J Appl Phys 2003;94:2167.22. Peter M, Laubsch A, Stauss P, Walter A, Baur J, et al. Phys Stat Sol (C) 2008;5:2050.23. Galler B, Laubsch A, Wojcik A, Lugauer H, Gomez-Iglesias A, et al. Phys Stat Sol (C)

2011;8:2372.24. Zhu JH, Zhang SM, Wang H, Zhao DG, Zhu JJ, et al. Appl Phys Lett 2011;109:093117.25. Han SH, Lee DY, Lee SJ, Cho CY, Kwon MK, et al. Appl Phys Lett 2009;94:231123.26. Ni X, Li X, Lee J, Liu S, Avrutin V, et al. J Appl Phys 2010;108(033112).27. Schubert MF, Schubert EF. Appl Phys Lett 2010;96:131102.28. Sizov DS, Bhat R, Zakharian A, Song K, Allen DE, et al. IEEE J Sel Top Quan Electron

2011;17:1390.29. Xie J, Ni X, Fan Q, Shimada R, €Ozg€ur €U, et al. Appl Phys Lett 2008;93:121107.30. Han SH, Cho CY, Lee SJ, Park TY, Kim TH, et al. Appl Phys Lett 2010;96:051113.31. Kuo Y-K, Chang J-Y, Tsai M-C, Yen S-H. Appl Phys Lett 2009;95:011116.32. Lin RM, Yu SF, Chang SJ, Chiang TH, Chang SP, et al. Appl Phys Lett 2012;101:081120.33. Wang CH, Chang SP, Ku PH, Li JC, Lan YP, et al. Appl Phys Lett 2011;99:171106.34. Xia CS, Simon Li ZM, Lu W, Zhang ZH, Sheng Y, et al. Appl Phys Lett 2011;99:233501.35. Kawaguchi Y, Huang CY, Wu YR, Yan Q, Pan CC, et al. Appl Phys Lett 2012;100:

231110.36. Shen YC, Mueller GO,Watanabe S, Gardner NF, Munkholm A, et al. Appl Phys Lett 2007;

91:141101.37. David A, Grundmann MJ. Appl Phys Lett 2010;96:103504.38. Kioupakis E, Rinke P, Delaney KT, de Walle CGV. Appl Phys Lett 2011;98:161107.39. Gardner NF, M€uller GO, Shen YC, Chen G, Watanabe S, et al. Appl Phys Lett 2007;91:

243506.40. Li Y-L, Huang YR, Lai YH. Appl Phys Lett 2007;91:181113.41. Kim MH, Schubert MF, Dai Q, Kim JK, Schubert EF, et al. Appl Phys Lett 2011;91:

183507.42. Schubert MF, Xu J, Kim JK, Schubert EF, KimMH, et al. Appl Phys Lett 2008;93:041102.43. Ni X, Li X, Lee J, Liu S, Avrutin V, et al. Appl Phys Lett 2010;97:031110.44. Wu Y-R, Shivaraman R, Wang K-C, Speck JS. Appl Phys Lett 2012;101:083505.45. Ling SC, Lu TC, Chang SP, Chen JR, Kuo HC, et al. Appl Phys Lett 2010;96:231101.46. Zhao Y, Tanaka S, Pan CC, Fujito K, Feezell D, et al. Appl Phys Express 2011;4:082104.47. Pan CC, Tanaka S, Wu F, Zhao Y, Speck JS, et al. Appl Phys Express 2012;5:062103.48. Huang H-H, Wu Y-R. J Appl Phys 2010;107:053112.49. Romanov AE, Baker TJ, Nakamura S, Speck JS. J Appl Phys 2006;100:023522.50. Park SH, Ahn D, Chuang SL. IEEE J Quan Electron 2007;43:1175.51. Yamaguchi AA. Jap J Appl Phys 2007;46:L789.52. Koslow IL, Hardy MT, Hsu PS, Dang PY, Wu F, et al. Appl Phys Lett 2012;101:121106.53. Craven MD, Lim SH, Wu F, Speck JS, DenBaars SP. Appl Phys Lett 2002;81:469.54. Hollander JL, Kappers MJ, McAleese C, Humphreys CJ. Appl Phys Lett 2008;92:101104.55. Sun Q, Kong BH, Yerino CD, Ko TS, Leung B, et al. J Appl Phys 2009;106:123519.56. Ni X, Fu Y, Moon YT, Biyikli N, Morko H. J Cryst Growth 2006;290:166.57. Ko TS, Wang TC, Gao RC, Chen HG, Huang GS, et al. J Cryst Growth 2007;300:308.58. Ambacher O, Majewski J, Miskys C, Link A, M Hermann, et al. J Phys Condens Matter

2002;14:3399.59. Sharma R, Pattison PM, Masui H, Farrell RM, Baker TJ, et al. Appl Phys Lett 2005;87:

231110.60. Baker TJ, Haskell BA, Wu F, Speck JS, Nakamura S. Jpn J Appl Phys 2006;45:L154.61. Masui H, Baker TJ, Iza M, Zhong H, Nakamura S, et al. J Appl Phys 2006;100:113109.


62. Jonen H, Bremers H, Langer T, Rossow U, Hangleiter A. Appl Phys Lett 2012;100:151905.

63. Waltereit P, Brandt O, Trampert A, Grahn HT, Menniger J, et al. Nature 2000;406:865.64. Ghosh S, Waltereit P, Brandt O, Grahn HT, Ploog KH. Phys Rev B 2002;65:075202.65. Sun Q, Kwon SY, Ren ZY, Han J, Onuma T, et al. Appl Phys Lett 2008;92:051112.66. Cho YS, Sun Q, Lee IH, Ko TS, Yerino CD, et al. Appl Phys Lett 2008;93:111904.67. Sun Q, Yerino CD, Zhang Y, Cho YS, Kwon SY, et al. J Cryst Growth 2009;311:3824.68. Sun Q, Leung B, Yerino CD, Zhang Y, Han J. Appl Phys Lett 2009;95:231904.69. Vennegues P, Zhu T, Martin D, Grandjean N. J Appl Phys 2010;108:113521.70. Okada N, Kurisu A, Murakami K, Tadatomo K. Appl Phys Express 2009;2:091001.71. Mierry P, Kriouche N, Nemoz M, Chenot S, Nataf G. Appl Phys Lett 2010;96:231918.72. Hikosaka T, Tanikawa T, Honda Y, Yamaguchi M, Sawaki N. Phys Stat Sol (C) 2008;5:

2234.73. Kaeding JF, Iza M, Sato H, DenBaars SP, Speck JS, et al. Jpn J Appl Phys 2006;45:L536.74. Baker TJ, Haskell BA, Wu F, Fini PT, Speck JS, et al. Jpn J Appl Phys 2005;44:L920.75. Chen GT, Chang SP, Chyi JI, Chang MN. Appl Phys Lett 2008;92:241904.76. Okada N, Oshita H, Yamane K, Tadatomo K. Appl Phys Lett 2011;99:242103.77. Fujito K, Kubo S, Fujimura I. MRS Bull 2009;34:313.78. Park SS, Park I, Choh SH. Jpn J Appl Phys 2000;39:L1141.79. Williams AD, Moustakas TD. J Cryst Growth 2007;300:37.80. Ashraf H, Kudrawiec R, Weyher JL, Serafi nczuk J, Misiewicz J, et al. J Cryst Growth

2010;312:2398.81. Haskell BA, Wu F, Matsuda S, Craven MD, Fini PT, et al. Appl Phys Lett 2003;83:1554.82. Haskell BA, Chakaraborty A, Wu F, Sasano H, Fini PT, et al. J Electron Mater 2005;34:

357.83. Haskell BA, Wu F, Craven MD, Matsuda S, Fini PT, et al. Appl Phys Lett 2003;83:644.84. Ashraf H, Weyher JL, van Dreumel GWG, Gzregorzyck A, Hageman PR. J Cryst Growth

2008;310:3957.85. Lucznik B, Pastuszka B, Grzegory I, Bo�ckowski M, Kamler G, et al. J Cryst Growth 2005;

281:38.86. Hashimoto T, Wu F, Speck JS, Nakamura S. Nat Mater 2007;6:568.87. Dwilinski R, Doradzinski R, Garczynski J, Sierzputowski LP, Puchalski A, et al. J Cryst

Growth 2009;311:3015.88. Ehrentraut D, Fukuda T. Proc IEEE 2010;98:7.89. Avrutin V, Silversmith DJ, Mori Y, Kawamura F, Kitaoka Y, et al. Proc IEEE 2010;98:

1302.90. Wu F, Lin YD, Chakraborty A, Ohta H, DenBaars SP, et al. Appl Phys Lett 2010;96:

231912.91. Northrup JE. Appl Phys Lett 2009;95:133107.92. Durnev MV, Omelchenko AV, Yakovlev EV, Evstratov IY, Karpov SY. App Phys Lett

2010;97:051904.93. Zhao Y, Yan Q, Huang CY, Huang SC, Hsu PS, et al. Appl Phys Lett 2012;100:201108.94. Keller S, Fichtenbaum NA, Furukawa M, Speck JS, DenBaars SP, et al. Appl Phys Lett

2007;90:191908.95. Zywietz T, Neugebauer J, Scheffler M. Appl Phys Lett 1998;73:487.96. Sato H, Hirasawa H, Asamizu H, Fellows N, Tyagi A, et al. J Light Vis Environ 2008;32:

107.97. Yamamoto S, Zhao Y, Pan CC, Chung RB, Fujito K, et al. Appl Phys Express 2010;3:

122102.


98. Chen CQ, Gaevski ME, Sun WH, Kuokstis E, Zhang JP, et al. Appl Phys Lett 2002;81:3194.

99. Wang H, Chen CQ, Gong Z, Zhang J, Gaevski M, et al. Appl Phys Lett 2004;84:499.100. Li DS, Chen H, Yu HB, Zheng XH, Huang Q, et al. J Cryst Growth 2004;265:107.101. Wernicke T, Ploch S, Hoffmann V, Knauer A, Weyers M, et al. Phys Stat Sol (B) 2011;

248:574.102. Lymperakis L, Neugebauer J. Phys Rev B 2009;79:241308.103. Kappers MJ, Hollander JL, McAleese C, Johnston CF, Broom RF, et al. J Cryst Growth

2007;300:155.104. Ploch S, Wernicke T, Thalmair J, Lohr M, Pristovsek M, et al. J Cryst Growth 2012;356:

70.105. Zhao Y, Huang CY, Wu F, Fujito K, DenBaars SP, et al. Appl Phys Lett 2013;102:091905.106. Hirai A, Jia Z, Schmidt MC, Farrell RM, DenBaars SP, et al. Appl Phys Lett 2007;91:

191906.107. Yamada H, Iso K, Masui H, Saito M, Fujito K, et al. J Cryst Growth 2008;310:4968.108. Farrell RM, Haeger DA, Chen X, Gallinat CS, Davis RW, et al. Appl Phys Lett 2010;96:

231907.109. Lin YD, Hardy MT, Hsu PS, Kelchner KM, Huang CY, et al. Appl Phys Express 2009;2:

082102.110. Kelchner KM, Lin YD, Hardy MT, Huang CY, Hsu PS, et al. Appl Phys Express 2009;2:

071003.111. Fischer AM, Wu Z, Sun K, Wei Q, Huang Y, et al. Appl Phys Express 2009;2:041002.112. Tyagi A, Wu F, Young EC, Chakraborty A, Ohta H, et al. Appl Phys Lett 2009;95:251905.113. Wu F, Tyagi A, Young EC, Romanov AE, Fujito K, et al. J Appl Phys 2011;109:033505.114. Young EC, Wu F, Romanov AE, Tyagi A, Gallinat CS, et al. Appl Phys Express 2010;3:

011004.115. Hsu PS, Young EC, Romanov AE, Fujito K, DenBaars SP, et al. Appl Phys Lett 2011;99:

081912.116. Hardy MT, Hsu PS, Wu F, Koslow IL, Young EC, et al. Appl Phys Lett 2012;100:202103.117. Romanov AE, Young EC, Wu F, Tyagi A, Gallinat CS, et al. J Appl Phys 2011;109:

103522.118. Wu F, Young EC, Koslow I, Hardy MT, Hsu PS, et al. Appl Phys Lett 2011;99:251909.119. Hsu PS, Hardy MT, Young EC, Romanov AE, DenBaars SP, et al. Appl Phys Lett 2012;

100:171917.120. Hangleiter A, Im JS, Kollmer H, Heppel S, Off J, et al. MRS Internet J Nitride Semicond

Res 1998;3:15.121. Chitnis A, Chen C, Adivarahan V, Shatalov M, Kuokstis E, et al. Appl Phys Lett 2004;84:

3663.122. Chakraborty A, Haskell BA, Keller S, Speck JS, DenBaars SP, et al. Appl Phys Lett 2004;

85:5143.123. Chakraborty A, Haskell BA, Masui H, Keller S, Speck JS, et al. Jpn J Appl Phys 2006;45:

739.124. Schmidt M, Kim K, Sato H, Fellows N, Masui H, et al. Jpn J Appl Phys 2007;46:L126.125. Masui H, Baker TJ, Sharma R, Pattison PM, Iza M, et al. Jpn J Appl Phys 2006;45:L904.126. Okamoto K, Ohta H, Nakagawa D, Sonobe M, Ichihara J, et al. Jpn J Appl Phys 2006;45:

L1197.127. Park SH, Ahn D. Appl Phys Lett 2007;90:013505.128. Sato H, Chung RB, Hirasawa H, Fellows N, Masui H, et al. Appl Phys Lett 2008;92:

221110.


129. Zhong H, Tyagi A, Fellows N, Wu F, Chung RB, et al. Appl Phys Lett 2007;90:233504.130. Funato M, Ueda M, Kawakami Y, Narukawa Y, Kosugi T, et al. Jpn J Appl Phys 2006;45:

L659.131. Zhao Y, Sonoda J, Koslow I, Ha JS, Ohta H, et al. Jpn J Appl Phys 2010;49:070206.132. Koslow IL, Sonoda J, Chung RB, Pan CC, Brinkley S, et al. Jpn J Appl Phys 2010;49:

080203.133. Nishida T, Saito H, Kobayashi N. Appl Phys Lett 2001;79:711.134. Fuji T, Gao Y, Sharma R, Hu EL, DenBaars SP, et al. Appl Phys Lett 2004;84:855.135. Yamada M, Mitani T, Narukawa Y, Shijoji S, Niki I, et al. Jpn J Appl Phys 2002;41:

L1431.136. Morita D, Sano M, Yamamoto M, Nonaka M, Yasutomo K, et al. Phys Stat Sol (A) 2003;

200:114.137. Huh C, Lee K, Kang E, Park S. J Appl Phys 2003;93:9383.138. Kasugai H, Miyake T, Honshio A, Mishima S, Kawashima T, et al. Jpn J Appl Phys 2005;

44:7414.139. Zhong H, Tyagi A, Pfaff N, Saito M, Fujito K, et al. Jpn J Appl Phys 2009;48:030201.140. Zhao Y, Sonoda J, Pan CC, Brinkley S, Koslow I, et al. Appl Phys Express 2010;3:102101.141. Zhao Y, Tanaka S, Yan Q, Huang CY, Chung RB, et al. Appl Phys Lett 2011;99:051109.142. Zhao Y, Yan Q, Feezell D, Fujito K, Van de Walle CG, et al. Opt Express 2013;21:A53.143. Matioli E, Brinkley S, Kelchner KM, Nakamura S, DenBaars SP, et al. Appl Phys Lett

2011;98:251112.144. Crawford MH. IEEE J Sel Top Quan Electron 2009;15:1028.145. Schnitzner I, Yablonovitch E, Caneau C, Gmitter TJ, Scherer A. Appl Phys Lett 1993;63:

2174.146. Shchekin OB, Epler JE, Trottier TA, Margalith T, Steigerwald DA, et al. Appl Phys Lett

2006;89:071109.147. Haerle V, Hahn B, Kaiser S, Weimar A, Bader S, et al. Phys Stat Sol (A) 2004;201:2736.148. Krames MR, Shchekin OB, Mueller-Mach R, Mueller GO, Zhou L, et al. J Display

Technol 2007;3:160.149. Masui H, Fellows NN, Sato H, Asamizu H, Nakamura S, et al. Appl Opt 2007;46:5974.150. Fellows N, Masui H, Sato H, Asamizu H, Iza M, et al. Phys Stat Sol (C) 2008;5:2216.151. Pan CC, Koslow I, Sonoda J, Ohta H, Ha JS, et al. Jpn J Appl Phys 2010;49:080210.152. Pan CC, [PhD Dissertation] UCSB; 2012.

Further reading

1. Wu F, et al. (to be submitted).2. Yoshida S, Yokogawa T, Imai Y, Kimura S, Sakata O. Appl Phys Lett 2011;99:131909.



Part Two

Performance of nitride LEDs


Efficiency droop in GaInN/GaNLEDs 9Houqiang Fu, Yuji ZhaoArizona State University, Tempe, AZ, United States

9.1 Introduction

III-nitride semiconductors, that is, wurtzite (In, Ga, Al)N, have been the workhorse ofblue LEDs and laser diodes (LDs), which lay the foundations for modern solid-statelighting, full-color displays, traffic signals, visible light communications, and soon.1,2 The successful demonstration of the first blue GaInN/GaN LEDs on sapphirein the early 1990s sparked worldwide research on III-nitride optoelectronics.3 Currentcommercially available GaInN/GaN LEDs grown on conventional c-plane substrates,however, suffer from reduced efficiency with increasing current density, a notoriousphenomenon known as “efficiency droop” or “current droop.”4 In addition, withincreasing demand for high power and bright GaInN/GaN LEDs, thermal managementalso becomes important. It’s also been reported that increasing operating temperaturereduces the efficiency of GaInN/GaN LEDs, a phenomenon termed “thermal droop.”5

In this chapter, we will first introduce the efficiency and efficiency droop of GaInN/GaN LEDs; then, some popular mechanisms explaining the origin of efficiency droopwill be presented including Auger recombination, carrier leakage, and so on; then,from both material and devices perspectives, we talk about progress on low-droopblue GaInN/GaN LEDs based on nonpolar and semipolar GaN substrates; finally, lat-est work on thermal droop of GaInN/GaN LEDs is discussed.

9.1.1 GaInN/GaN LED efficiency

The light output power (LOP) of GaInN/GaN LEDs is defined as the number ofphotons generated per unit time per unit volume and determines how bright theLED appears to us. The LOP of LEDs is proportional to their external quantumefficiency (EQE), which is the ratio of the number of photons emitted from the deviceto the number of injected electrons. The EQE can be further decomposed into thefollowing three components:

EQE ¼ IQE� LEE� hINJ (9.1)

where IQE is the internal quantum efficiency (IQE), LEE is the light extraction effi-ciency (LEE), and hINJ is the injection efficiency. The LEE is the ratio of the number ofphotons emitted from the LED to the number of photons generated inside the LED.


http://dx.doi.org/10.1016/B978-0-08-101942-9.00009-5

The LEE is related to the device geometries and generally doesn’t play a role in theefficiency droop of GaInN/GaN LEDs.6 Injection efficiency is the ratio of the numberof electrons injected into the LED to the number of electrons supplied by the powersources. Usually the injection efficiency is assumed to be 100% and is not a factor inthe efficiency droop of GaInN/GaN LEDs. The IQE is the ratio of the number ofphotons generated inside the LED to the number of electrons injected into the LED. Ina certain device where LEE and injection efficiency are fixed, EQE is proportional toIQE. Therefore, IQE plays a vital role in analyzing and improving the efficiency droopof GaInN/GaN LEDs. The IQE of LEDs is generally characterized by the carrier rateequation model with ABC coefficients (ABC model)7:

IQE ¼ Bn2��

Anþ Bn2 þ Cn3�

(9.2)

where n is the carrier density inside the GaInN/GaN LEDs, and A, B, and C are theShockley-Read-Hall (SRH), radiative, and Auger coefficients, respectively. Specif-ically, all the injected electron-hole pairs have two recombination pathways: radiativerecombination that generates photons and non-radiative recombination that dissipatesenergy without releasing photons. The radiative recombination rate RR is proportionalto Bn2, while the non-radiative recombination rate RNR is the sum of SRH and Augerrecombination rate (An þ Cn3). Therefore, IQE is expressed as:

IQE ¼ RR=ðRR þ RNRÞ ¼ Bn2�

Anþ Bn2 þ Cn3� �

(9.3)

It’s worth mentioning that the ABC model is a simplified model without accountingfor other mechanisms such as carrier leakage4 and the phase-space filling (PSF)effect.7,8 The popularity of the ABC model in analyzing GaInN/GaN LEDs stemsfrom its simplicity and flexibility. The A, B, and C coefficients can be extracted byfitting experimental data with the ABC model. However, the ABC model has its limita-tions and sometimes shows a large discrepancy between simulation and experimentalresults. Therefore, some modified ABC models were proposed. For example, Linet al. modified the ABC equation where a drift-induced leakage (CDL) term was incor-porated into the C coefficient along with the Auger (CAuger) term, to include the impactof carrier leakage.9 David et al. substituted B(1 þ n/n0) for the radiative recombinationcoefficient B to account for the PSF effect.7 We will talk about PSF effect in detail later.

9.1.2 Efficiency droop in GaInN/GaN LEDs

Efficiency droop in GaInN/GaN LEDs refers to the reduction of efficiency (EQE orIQE) with increasing current densities or operating temperatures where the former iscalled “current droop” and the latter is called “thermal droop.” In literature, peopleoften mean current droop when talking about efficiency droop of GaInN/GaNLEDs. These two terms are interchangeable unless stated otherwise. The onset ofcurrent droop is usually at very low current densities, 1e10 A/cm2. The current droopof GaInN/GaN LEDs is defined as:

Droop ¼ ðIQEMax � IQEJÞ=IQEMax � 100%; (9.4)


where the IQEMax and IQEJ represent the IQE maximum and the IQE at a given currentdensity.10 Fig. 9.1 shows the efficiency and efficiency droop of a typical c-planeGaInN/GaN LED. Current droop is the most studied, as well as the most controversial,topic in GaInN/GaN LEDs. Many different mechanisms have been proposed toexplain this phenomenon including Auger recombination,11,12 carrier leakage,4 PSFeffect,7,8 carrier delocalization,13,14 quantum confined Stark effect (QCSE),15

defects,16 and so on. The physical origin, however, is still being debated. In thefollowing sections, we will discuss some of the major mechanisms causing the currentdroop of GaInN/GaN LEDs. The study of thermal droop of GaInN/GaN LEDs is in itsinitial stage and is becoming an increasingly important topic as GaInN/GaN LEDs areused more and more in high power and high temperature applications. Fig. 9.2 shows

60

50

40

30

20

10

00 100 200

Current density (A/cm2)

Droop

Dro

op (%

)

Onset of droop1 ~ 10 A/cm2

300 400 500

IQE

(%)

60

50

40

30

20

10

0

Figure 9.1 The internal quantum efficiency (IQE) and efficiency droop as a function of currentdensity for a typical c-plane GaInN/GaN LED.

100

80

60

40 Ambienttemperature

Junctiontemperature

20

020 40 60

Temperature (°C)

Droop

IQE

(%)

80 100

Figure 9.2 IQE curve of an LED showing a decreased efficiency with increasing temperature,that is, thermal droop.

Efficiency droop in GaInN/GaN LEDs 301

the IQE as a function of temperature for a typical LED. The performance of the LEDpeaks at low temperature (20�C); when temperature increases, the efficiency of theLED drops dramatically. At high temperature (100�C), the reduction in LED efficiencyis more than 50%. Currently, thermal droop in LEDs is mitigated by packaging designssuch as high thermal conductivity submounts and the use of large heat sinks. Theresults of these packaging methods are less satisfactory; they not only increase the sizeof the LED devices, but also increase the manufacturing cost.

9.2 Physical mechanisms of current droopin GaInN/GaN LEDs

Over many years of investigation, lots of mechanisms have been proposed to explain theorigin of current droop in GaInN/GaN LEDs; of these, Auger recombination and carrierleakage have been the most debated and popular ones. The Auger theory is mainlybacked up by Lumileds,11 as well as Nakamura et al.12 (UCSB group), while the carrierleakage theory is mainly supported by Schubert et al.4 (RPI group). For the former case,experimental Auger coefficients were determined to be large enough to cause currentdroop; the Auger electrons were directly observed and strongly linked to the currentdroop of GaInN/GaN LEDs. For the latter case, carrier leakage was experimentallyobserved and exhibited a close relation with the current droop in GaInN/GaN LEDs.We’ll discuss these and other mechanisms, such as the PSF effect, carrier delocalization,effective active region volume, and indirect Auger recombination.

9.2.1 Auger recombination

Auger recombination is a non-radiative process where the excess energy from theelectron-hole recombination is transferred to electrons or holes that are subsequentlyexcited to higher energy states within the same band instead of giving off photons(the radiative process). Another way to look at it is: two electrons collide in the vicinityof a hole which leads to a non-radiative e-h recombination event. The energy andmomentum is absorbed by the second electron (eeh process). The hhe process in similarto eeh process except that it involves two holes and one electron. As the ABC modelstates, the Auger recombination rate is proportional to the cube of carrier density, n3.This means Auger recombination may play an important role at high carrier/currentdensity, leading to current droop in GaInN/GaN LEDs. Piprek showed that the Augercoefficient has to be larger than 10�31 cm6/s to reasonably contribute to efficiencydroop.6 Therefore, it’s of vital importance to properly characterize and extract the Augercoefficient.

Shen et al. first experimentally determined the Auger coefficient of quasi-bulkGaInN (0001) (In composition 9%e15%) in the range of 1.5e2.0 � 10�30 cm6/s bya photoluminescence lifetime measurement based on resonant optical excitation.11

This method provided a powerful tool to investigate Auger coefficients since it hassimpler carrier dynamics than electroluminescence measurements. Piprek summarizedother Auger coefficients (>10�31 cm6/s) from other groups using various device


structures.6 For example, Meneghini et al. characterized a packaged single quantumwell (QW) GaInN/GaN with peak wavelength of 450 nm using an electro-opticalmethod and obtained an Auger coefficient of 10�30 cm6/s17 Laubsch et al. got thesame Auger coefficient of 3.5 � 10�31 cm6/s on 523 nm single QW GaInN/GaNLEDs at both 4 and 300K.18 Please note that all the Auger coefficients were extractedbased on the ABC model without considering carrier leakage or the PSF effect.Furthermore, Iveland et al. directly observed the generation of Auger electrons fromGaInN/GaN LEDs under electrical injection by electron emission spectroscopy.12,19

Fig. 9.3 shows how the Auger electrons are detected. There are various types of elec-trons generated inside the active region including the energetic Auger electrons thatlead to high-energy peaks in the energy distribution of electrons in vacuum. As longas these Auger electrons reach the surface and retain enough energy, they can be iden-tified by measuring their energies. In addition, Iveland et al. also proved the relationbetween efficiency droop and Auger recombination. In Fig. 9.4(a), the droop currentwas obtained from the dependence of EQE on current with/without efficiency droop.They also observed a linear correlation between Auger current and the droop current,which substantiates the idea that these two phenomena are the same. It also rules out

Auger(non-radiative)

Semiconductor Vacuum

Vacuumlevel

BBR

p-GaN

QW

n-GaN

VB

CB

Ene

rgy

EBL

Cesiatedp-GaN

L-valley

Γ-valley

Leakage/tunneling(non-radiative)

Direct recombination(radiative)

Figure 9.3 Schematic view of bandstructure of GaInN/GaN LED under positive bias. Electronswith different energies are emitted from the active region to vacuum. The cesium is depositedon the p-GaN surface to obtain negative electron affinity so that electrons with energy belowthe conduction band (CB) minimum can be detected with the assistance of energy relaxation inthe band-bending region (BBR). VB is valence band and EBL is the electron blocking layer.After Iveland J, Martinelli L, Peretti J, Speck JS, Weisbuch C. Direct measurement of Augerelectrons emitted from a semiconductor light-emitting diode under electricali: identification ofthe dominant mechanism for efficiency droop. Phys Rev Lett 2013;110:177406; Weisbuch C,Piccardo M, Martinelli L, Iveland J, Peretti J, Speck JS. The efficiency challenge of nitridelight-emitting diodes for lighting. Phys Status Solidi A 2015;212(5):899e913.


other droop mechanisms as droop current is proportional to the cube of injected currentdensity according to the ABC model. These results provide strong support for theAuger recombination mechanism. An Auger coefficient one magnitude larger than10�31 cm6/s, combined with high carrier densities, may explain the onset of currentdroop at very low current densities. Therefore, Auger recombination may be a domi-nant non-radiative mechanism and the main contributor to the current droop.

However, one of the major drawbacks for the Auger recombination theory is thatthe theoretical Auger coefficient C obtained by the direct intraband Auger recombina-tion process is too low to account for the observed experimental results.20,21 Fig. 9.5shows four different eeh Auger processes in III-nitride materials including direct/indi-rect and intraband/interband. Hader et al. report a very small direct band-to-bandAuger coefficient of 3.5 � 10�34 cm6/s using a microscopic many-body model andan 8 � 8 kp bandstructure model.20 Bertazzi et al. investigated direct interband andintraband Auger recombination by first-order perturbation theory and obtained amaximum Auger coefficient of approximately10�32 cm6/s at a bandgap of 2.9 eV.21

In order to overcome this discrepancy in the ABC model, many efforts have beendeveloped from different perspectives. For example, Kioupakis et al. studied an indi-rect intraband Auger recombination process mediated by electron-phonon couplingand alloy scattering using atomistic first-principle calculations and obtained a largerC coefficient,22 supporting the idea that Auger recombination is the origin of current

1.0

0.8

0.6

0.4

0.2

0.00 0 20 40 60 80 100 12050 100

100

100

80

60

40

20

10

50

0150

Droop current

Optical power without droopOptical power with droopHigh energy auger peak

Injected current (mA) Droop current (mA)In

tegr

ated

cur

rent

of p

eak

(nA

)In

tegr

ated

cur

rent

of p

eak

(nA

)

Ligh

t int

ensi

ty (a

.u.)

200 250

(a) (b)

Figure 9.4 (a) Integrated current of high energy Auger peaks (solid circle) and light intensity ofthe GaInN/GaN LED without efficiency droop (solid line) and with efficiency droop (dash-dotline). The solid line without efficiency droop is calculated based on a linear extrapolation fromthe maximum IQE at low currents. The droop current is the current difference between thedash-dot line and solid line, that is, between the actual injected current and current necessary toobtain the same light intensity when without efficiency droop. (b) Integrated high energyAuger peak versus the droop current.After Iveland J, Martinelli L, Peretti J, Speck JS, Weisbuch C. Direct measurement of Augerelectrons emitted from a semiconductor light-emitting diode under electricali: identification ofthe dominant mechanism for efficiency droop. Phys Rev Lett 2013;110:177406; Weisbuch C,Piccardo M, Martinelli L, Iveland J, Peretti J, Speck JS. The efficiency challenge of nitride light-emitting diodes for lighting. Phys Status Solidi A 2015;212(5):899e913.


droop in GaInN/GaN LEDs. Delaney et al. investigated interband Auger recombina-tion in GaInN using first principle density-functional and many-body-perturbation the-ory and reported a peak Auger coefficient of 2 � 10�30 cm6/s at bandgap of 2.5 eV.23

In an analytic model, Lin et al. modified the ABC equation where a drift-inducedleakage (CDL) term was incorporated into the C coefficient along with the Auger(CAuger) term as discussed in the following section.9

9.2.2 Carrier leakage

A standard GaInN/GaN LED structure contains an AlGaN electron blocking layer(EBL) at the p-side of the active region in order to suppress carrier (electron) leakage.Carrier leakage refers to electrons escaping from the active region and recombiningwith holes in the p-GaN or at the p-contact. Carrier leakage is limited by the conduc-tion band offset DEC between the quantum barrier and QW in the active region, andactive region and AlGaN EBL. The free carriers in the QW follow a Fermi-Diracdistribution. As a result, a certain fraction of the carriers will have an energy that ishigher than the energy of the barrier and thus escape from the QW. Such an escapingmechanism is illustrated in Fig. 9.6. Carrier leakage is suggested to be one of the majormechanisms causing current droop partly because the quantum barrier and EBL cannotcompletely block electron leakage. Fig. 9.7 illustrates all the non-radiative recombina-tion mechanisms (including carrier leakage) and the radiative recombination process.Vampola et al. provided experimental proof of the presence of electron leakage overthe EBL using test structures comprising of a standard LED with an extra low-indium QW (probe QW) on the p-side of the EBL.25 If there is carrier leakage,short-wavelength light will be emitted due to the probe QW. They observed that theshort-wavelength emission occurs at the onset of current droop and is enhancedwith increasing current density and decreasing EQE, thus indicating strong correlationbetween carrier leakage and current droop.

CB(a) (b) (c) (d)

VB VB VB VB

ElectronHole

CB CB CB

Figure 9.5 Schematics of different Auger recombination process: (a) direct intraband,(b) indirect intraband, (c) direct interband, and (d) indirect interband. The solid circlesrepresent electrons and empty circles are holes. The upward bands are conduction band (CB)while the downward bands are valence band (VB). The solid arrows are momentumtransferred from electron-hole recombination while the dashed arrows in indirect Augerprocesses represent a scattering mechanism, such as alloy scatttering, Coulombic scattering, orelectron-phonon coupling.22


The basis for analyzing the effect of carrier leakage on current droop is understand-ing electron-hole imbalance in the active region. Usually, there are more electrons thanholes in the active region. The excess electrons can flow through the active regionwithout radiative recombination due to the lack of holes, thus leading to current droop.This imbalance can be caused by either electrons escaping from the active region orpoor hole injection from p-side of devices. For the first mechanism, the electrons inthe active region cannot be completely confined by the QW and EBL; for the secondmechanism, the combination of the low hole concentration of p-GaN and the low holemobility and potential barrier created by the EBL can result in low hole injection. Therole of EBL can be controversial because it can block electrons from the active region,as well as serve as a barrier for hole injection. Some researchers reported GaInN/GaNLEDs with EBL having higher peak efficiency and better droop performance whileother works have shown opposite result due to worse hole injection.24 In addition,Yen et al. proposed n-type AlGaN EBL instead of the normally used p-type AlGaNEBL to reduce current droop because the former results in a more uniform distributionof holes in the QW, as well as sufficient confinement of electrons.26

n-GaN MQW EBL p-GaN

Carrierleakage

AugerEc

Ev

If qV

SRHRadiativehν

If qV

Figure 9.7 Schematic views of all the recombination processes in the GaInN/GaN LEDsincluding radiative recombination, SRH recombination, Auger recombination, and carrier leakage.After Cho J, Schubert EF, Kim JK. Efficiency droop in light-emitting diodes: challenges andcountermeasures. Laser Photon Rev 2013;7(3):408e21.

GaN barrier GaN barrier

High energy carriersescape from QW

ΔEc

ΔEc

ΔEv

Carrierdistribution

Carriers are injectedinto QW

GaInN QW EBL

Figure 9.6 Carrier escape in a QW structure. Also shown is the carrier distribution.


For conventional polar c-plane GaInN/GaN LEDs, additional complexities areadded to the carrier leakage due to the polarization charges at the hetero-interfaces.The first interface worth considering is the interface between the last QW and quantumbarrier at the p-side of the devices. This interface features a positive sheet charge thatlowers the barrier for electrons to escape from the QW. The second interface is the inter-face between the last quantum barrier and AlGaN EBL where positive sheet chargesexist. This lowers the conduction band edge of EBL and thus reduces the barrier heightthat is supposed to block electrons from leaving the active region. More detailed discus-sion can be found in Ref. 24. In order to reduce current droop by decreasing the polar-ization effects on c-plane GaInN/GaN LEDs, several methods were proposed. One suchmethod involves using polarization-matched AlInGaN quantum barriers and EBL27;another method involves growing devices on nonpolar and semipolar GaN substrates.The latter method is becoming a hot topic and will be discussed in detail later.

To account for carrier leakage, Dai et al.28 proposed the “ABC þ f(n) model,”wherethe total recombination rate Rtot is expressed as:

Rtot ¼ Anþ Bn2 þ Cn3 þ f ðnÞ; (9.5)

where f(n) represents the contribution of carrier leakage. On the one hand, f(n) can addanother third-order recombination term to the Auger recombination term, explaining themeasured large “Auger coefficients” that are actually the sum of the Auger coefficientand third-order coefficient from f(n). On the other hand, f(n) may include otherrecombination terms that are higher than third order. In order to enter the region ofefficiency droop, the LEDmust be in high level injection where drift current also comesinto play. With increasing total current Jtot, the drift-induced leakage current Jdrift playsa more and more important role in causing efficiency droop as it is proportional to Jtot:

JdriftfnJtot (9.6)

At medium current density where radiative recombination dominates and the LEDapproaches the onset of efficiency droop, the total current density becomes

Jtot ¼ qdRtot z qdBn2 (9.7)

Plugging Eq. 9.7 into Eq. 9.6, the Jdrift is rewritten as:

Jdrift ¼ qdCDLn3; (9.8)

where CDL is the third-order radiative recombination coefficient. Using some commonvalues of GaN material properties, CDL on the order of 10�29 cm6/s can be obtained.

At high current density in the efficiency droop region, third-order non-radiativerecombination becomes significant and Jtot is dependent on n3. This gives us a newrelationship between Jdrift and carrier density:

Jdrift ¼ qdCDLn4; (9.9)


where DDL is the fourth-order non-radiative recombination coefficient.By including the drift-induced leakage term, the onset of the droop current can be

calculated as:

Jonset ¼ dqBA=CDL (9.10)

Using some typical values such as d ¼ 3 nm, B ¼ 10�10 cm3/s, A ¼ 107 s�1, andCDL ¼ 2.4 � 10�29 cm6/s, we can obtain an onset droop current of 2 A/cm2, which isconsistent with the experimental onset droop current of 1e10 A/cm2.27

Combining all the mechanisms, the total recombination rate can be described as:

Rtot ¼ Anþ Bn2 þ CAugern3 þ CDLn

3 þ DDLn4; (9.11)

where the first three terms are from the conventional ABCmodel and the last two termsaccount for drift-induced leakage. It should be noted that electron leakage in the p-GaNlayer will recombine with injected holes before reaching the active region, reducing theinjection efficiency and thus leading to efficiency droop. Dai et al.28 used theABC þ f(n) model to fit the EQE cures of two blue c-plane GaInN/GaN LEDs andobtained a better fit than ABC model, especially at high current densities as shown inFig. 9.8. It’s worth mentioning that the fourth-order recombination term contributes14% to the total recombination of the two LEDs.

4

5

3

2

1

00 20 40 60 80 100

EQ

E (a

.u.)

Current (mA)

Peak wavelength at 25 mALED1: 444 nmLED2: 465 nm

ABC model for LED1ABC+f(n) model for LED1

ABC+f(n) model for LED2

Experimental data of LED1

Experimental data of LED2

ABC model for LED2

Figure 9.8 Experimental EQE as a function of driving current for two blue (dark gray in printversions) c-plane GaInN/GaN LEDs. These two LEDs have peak wavelength of 444 and465 nm, respectively. Simulations using ABC model and ABC þ f(n) model are also present.After Dai Q, Shan Q, Wang J, Chhajed S, Cho J, Schubert EF, Crawford MH, Koleske DD, KimM-H, Park Y. Carrier recombination mechanisms and efficiency droop in GaInN/GaN light-emitting diodes. Appl Phys Lett 2010;97:133507.


9.2.3 Other mechanisms

Though Auger recombination and carrier leakage are the two most promising mech-anisms for explaining efficiency droop of GaInN/GaN LEDs, some other mechanismsare also proposed and investigated. David et al.7 proposed the PSF effect as an alter-native explanation. At low carrier densities, the radiative recombination is expressedas Bn2 as shown in the ABC model. However, at high carrier densities, this doesn’thold and the radiative recombination rate is reduced and now proportional to n.They did differential carrier lifetime measurements and successfully fitted the droopcharacteristics of a c-plane blue GaInN/GaN LED with the PSF effect incorporated.The modified ABC model can be expressed as7,29:

J ¼ qd

0

B@Anþ Bn2

1þ n

n0

þ Cn3

1þ n

n0

1

CA; (9.12)

IQE ¼ Bn2

1þ n

n0

,0

B@Anþ Bn2

1þ n

n0

þ Cn3

1þ n

n0

1

CA ; (9.13)

n ¼ Dnþ N0 z Dn; (9.14)

where n is the carrier density, N0 is the doping density, and n0 is the PSF coefficient.Since N0 is much smaller than Dn, n is roughly equal to Dn. A smaller n0 indicatesa stronger PSF effect, and vice versa. The PSF effect comes from the utilization of aFermi-Dirac distribution at high carrier density instead of Boltzmann distribution. As aresult, the radiative recombination is proportional to n instead of n2 at high carrierdensity. This effect is accounted for by replacing coefficient B by B/(1 þ n/n0). Also,the Auger recombination is also assumed to become subcubic. Fig. 9.9(a) shows IQEcurves versus current density for various n0. Changing n0 can dramatically modify theefficiency curves and related droop performance. In Fig. 9.9(b), the droop ratios arecalculated for the IQE curves in Fig. 9.9(a). We see that smaller n0 (stronger PSFeffect) shows a larger efficiency droop. Using this model, the droop characteristics ofboth c-plane and nonpolar/semipolar GaInN/GaN LEDs can be simulated, as discussedin detail later. Ryu et al. showed that the combination of indium composition fluctu-ation, internal polarization, and inhomogeneous carrier distribution lead to reducedactive region thickness or volume, which affects the ABC model and therefore impactsthe droop properties.30 Because of poor hole injection due to low hole concentrationand mobility in p-GaN, most of the holes are distributed in a few QWs that are close tothe p-GaN, significantly reducing the volume of active region. Besides this, both theindium fluctuation and polarization limit carriers to certain regions of the QWs.Therefore, the actual active volume responsible for the light emission is smaller thanthe physical volume of the QWs. The reduced active region volume will increase the


carrier density, resulting in larger non-radiative Auger recombination and thusefficiency droop. Ryu et al. modified the relation between current and carrier density asthe following30:

I ¼ qVeff

0

B@Anþ Bn2

1þ n

n0

þ Cn3

1þ n

n0

1

CA; (9.15)

where the Veff is the effective volume of the active region where all recombinationprocesses occur. Current I can be converted to current density J, and Veff to theeffective thickness deff of the active region when divided by the area of the activeregion. The IQE is the same as Eq. (9.13). Ryu et al. obtained very good fitting of thedroop characteristics of a commercial c-plane GaInN/GaN LED using a smalleffective volume of active region with large Auger coefficient. Chichibu et al.reported that the spontaneous emission from GaInN/GaN LEDs is mainly due tolocalized exciton recombination at the potential minima in the density of states.13,31

Wang et al.14 compared the EQE curves of two GaInN/GaN LED structures withInGaN (LED3) and GaN (LED4) underlying layers under the active region,respectively, as shown in Fig. 9.10. The higher efficiency peak of LED3 indicatesthat LED3 has a higher degree of carrier localization than LED4. Carrier delocal-ization has been attributed to increasing efficiency droop, where more carriersparticipate in non-radiative recombination processes. LED3 also shows larger effi-ciency droop because the delocalization effect is stronger for LED3 with its higherlocalization degree. This phenomenon is unique and important in GaInN/GaN LEDsbecause, as confirmed by Chichibu et al.31 there are very large amount of localizedstates in GaInN materials.

60

50

40

30

20

10

0

60

50

40

30

20

10

01 10 100 1000

IQE

(%)

Dro

op ra

tio (%

)Current density (A/cm2) Current density (A/cm2)

n0 decreases

From left to right:n0 = 1020, 1019, 5 x 1018, 2 x 1018, 1018 cm–3

100 250 400 550 700 850 1000

n0 = 1020 cm–3

n0 = 1019cm–3

n0 = 5 × 1018cm–3

n0 = 2 × 1018cm–3

n0 = 1018cm–3

n0 decreases

(a) (b)

Figure 9.9 Calculated IQE curves as a function of current densities using the ABC model withdifferent n0 coefficients; (b) Calculated droop ratio of different IQE curves in (a).


9.3 Progress of low-droop GaInN/GaN LEDs

Due to the low cost, high thermal stability and commercial availability of large sap-phire wafers, GaInN/GaN blue LEDs grown on c-plane substrates have been industrialstandards. However, the conventional c-plane GaInN/GaN blue LEDs suffer fromlarge efficiency droop due to the presence of strong polarization field inside the QWwhich reduces the electron-hole wavefunction overlap. Over nearly a decade, therehas been tremendous effort dedicated to solving this issue.

From the perspective of Auger recombination, not much can be done to alter theAuger coefficient, but the carrier density n inside the QW can be decreased to reducethe Auger recombination rate. In devices, low carrier density can be realized byincreasing QW thickness or chip size. C-plane GaInN/GaN LEDs usually have a quan-tum QW thickness of 3 nm in order to mitigate the separation of electrons and holes.From the perspective of carrier leakage, researchers tried to engineer the QW barrierlayer and EBL to achieve better carrier confinement and more uniform carrier distribu-tion at high current densities. However, it is rather difficult to directly modify the QWprofile due to the polarization effect of c-plane GaInN/GaN LEDs. And some strategiesmentioned above make the growth process more complicated and degrade the materialquality. Nowadays, growing GaInN/GaN LEDs on novel nonpolar and semipolar GaNsubstrates has been proposed as a possible solution to suppress efficiency droop.32e34

These planes have eliminated or reduced the QCSE effect and allow for the growth ofthick QWs (up to 12 nm).

EQ

E (a

.u.)

0 20 40 60 80Current (mA)

LED3LED4

Figure 9.10 EQE as a function of current for LED3 with InGaN underlying layer and LED4with GaN underlying layer.After Wang J, Wang L, Zhao W, Hao Z, Luo Y. Understanding efficiency droop effect inInGaN/GaN multiple-quantum-well blue light-emitting diodes with different degree of carrierlocalization. Appl Phys Lett 2010;97:201112.


9.3.1 Polar c-plane, nonpolar/semipolar GaN

Wurtzite III-nitride materials have a strong spontaneous polarization effect along thec-axis ([0001] direction) due to the lack of inversion symmetry. When a GaInN QWis coherently grown on a GaN substrate, the lattice mismatch-induced strain will resultin piezoelectric polarization. The difference of the two kinds of polarizations at theGaInN/GaN interface induces a strong electric field inside the GaInN QW, whichresults in a tilted QW profile. This reduces the electron-hole wavefunction overlap andthus decreases the radiative recombination rate. In order to mitigate the polarization-related effect, GaInN/GaN LEDs grown on novel nonpolar and semipolar planes arecurrently being intensively explored. Fig. 9.11 shows the polar c-plane, nonpolarm-plane,and other widely used semipolar planes for III-nitride wurtzite crystal structure.

The polarizations of c-plane and nonpolar/semipolar planes can be calculated with anew method proposed by Romanov et al.35 where an analytical formalism in linearelasticity is incorporated. In the calculations, two coordinate systems are defined,whose details can be found in Ref. The primed coordinate z0 is along the growth direc-tion and x0 and y0 are in the substrate surface plane. For a plane tilted from c-plane byan angle of q, the total polarization difference along z0 at GaInN/GaN interface can beexpressed as a function of q by:

DPtot ¼ PGaInNpz þ�PGaInNsp � PGaNsp

�cosq; (9.16)

Polar c-plane(0001) (0º)

Semipolar(1122) (58º)

Nonpolar m-plane(1010) (90º)




Figure 9.11 Schematic views of polar (c-plane), semipolar (�1122

�,�1011

�,�2021

�, and�

2021�) and nonpolar planes (m-plane) of the III-nitride wurtzite crystal structure. The degree

indicates the angle by which the plane inclines from the c-plane.


where DPtot is the total polarization difference between InGaN layer and GaNsubstrate, and PGaInNsp and PGaNsp are the spontaneous polarizations of GaInN layer and

GaN substrate, respectively. PGaInNpz is the strain-induced piezoelectric polarization in

GaInN layer, which is expressed as:

PGaInNpz ¼ e31 cosqεx0x0 þ�e31 cos

3 qþ e33 � e152

sinqsin 2q

�εy0y0

þ�ðe31 þ e15Þ

2sinqsin 2qþ e33 cos

3 q

�εz0z0

(9.17)

where elements εk0m’ are the strain tensor components and elements eij are the com-ponents of the piezoelectric tensor in Voigt notation. Fig. 9.12 presents the PGaInNpz and

DPtot of GaInN/GaN with different indium compositions as a function of q. Since thePGaInNpz is dominant for GaInN/GaN heterostructures, DPtot shows minimum change

with the addition of spontaneous polarization. DPtot becomes zero at q ¼ 45 degrees(semipolar plane) and q ¼ 90 degrees (nonpolar plane). The two crossovers are onlyweakly impacted by the indium composition.

The bandstructure and electron-hole wavefuction of QWs on different planes can becalculated using SiLENSe, a commercial software package developed by STRGroup,37 where the one-dimensional Schr€odinger-Poisson equation is solved self-consistently with drift-diffusion models included. The software calculates spontaneousand piezoelectric polarizations for arbitrary crystal orientations of III-nitride materials,which are critical factors in device performance. Fig. 9.13 presents the results of

0.064

0.048

0.032

0.016

0.000

–0.016

00 20 40 60 80 20 40 60 80θ (º) θ (º)

P p

zIn

GaN

(C/m

2 )

In0.4Ga0.6NIn0.3Ga0.7NIn0.2Ga0.8NIn0.1Ga0.9N

In0.4Ga0.6NIn0.3Ga0.7NIn0.2Ga0.8NIn0.1Ga0.9N

ΔPto

t (C

/m2 )

0.064

0.048

0.032

0.016

0.000

–0.016

(a) (b)

Figure 9.12 Calculated (a) piezoelectric polarization PInGaNpz and (b) total polarization differenceDPtot as a function of semipolar plane orientation q for InGaN/GaN heterostructure with Incomposition of 10%, 20%, 30%, and 40%.After Zhao Y, Farrell RM,Wu YR, Speck JS. Valence band states and polarized optical emissionfrom nonpolar and semipolar IIIenitride quantum well optoelectronic devices. Jpn J Appl Phys2014;53:100206.


Ga0.8In0.2N (3 nm)/GaN (15 nm) single quantum well (SQW) LEDs at 100 A/cm2. Forc-plane GaInN/GaN LED, Ebi and Epz are antiparallel and Epz is much larger than Ebi,which results in the QW profile tilted downward along the growth direction. The Epz ofsemipolar

�1011

�and

�2021

�GaInN/GaN LEDs are also in the opposition direction of

1

0

–1

–2

–3

–4

1

0

–1

–2

–3

–4

1

0

–1

–2

–3

–4

603 608 613 618 623 628

603 608 613 618 623 628 603 608 613 618 623 628

603 608 613 618 623 628Position (nm)

Position (nm)

603 608 613 618 623 628Position (nm)

603 608 613 618 623 628Position (nm)

Position (nm)

Position (nm)

Ebi

Epz

Ebi

Epz

Ebi

EbiEbi

Epz

Ebi

Epz

Epz

Ene

rgy

(eV

)E

nerg

y (e

V)

1

0

–1

–2

–3

–4

Ene

rgy

(eV

)

Ene

rgy

(eV

)E

nerg

y (e

V)

1

0

–1

–2

–3

–4

Ene

rgy

(eV

)

1

0

–1

–2

–3

–4

Epz=0

(a) (b)

(c) (d)

(e) (f)

Figure 9.13 Simulated band diagram for In0.2Ga0.8N (3 nm)/GaN (15 nm) QW grown on(a) polar c-plane, semipolar (b)

�1122

�, (c)

�1011

�, (d)

�2021

�, (e)

�2021

�and (f) nonpolar

m-plane at a current density of 100 A/cm2. Moving left to right on the x-axis, it is n-GaN/i-InGaN/p-GaN. The directions of the junction built-in electric field Ebi and piezoelectricelectric field Epz are indicated by arrows whose thickness represents the relative magnitude ofEbi and Epz.


Ebi but smaller, which results in less tilting compared with c-plane QWs. For them-plane LED, Epz is zero and only Ebi exists inside the InGaN QW. In terms ofelectron and hole wavefuction overlap, the wavefuction of the c-plane GaInN/GaNLED is more spatially separated than in nonpolar and semipolar GaInN/GaN LEDs.

9.3.2 Low-droop nonpolar/semipolar GaInN/GaN LEDs

Nakamura et al.3 first demonstrated c-plane blue GaInN/GaN LEDs grown on sap-phire. Disadvantages such as high density of thread dislocations (TDs), poor thermalperformance, and different thermal expansion coefficients severely limited the deviceperformance of sapphire c-plane GaInN/GaN LEDs. Later, patterned sapphire sub-strates (PSS) or bulk GaN substrates were employed to enhance device performance.The highest EQE reported for c-plane blue LEDs is 81% from Nichia.38 However,c-plane LEDs still suffer from large efficiency droop. To overcome this issue, nonpolarand semipolar GaInN/GaN LEDs grown on bulk GaN substrates have been proposedas a solution. Fig. 9.14 shows a typical EQE curve of c-plane and semipolar InGaNblue LEDs. It is evident that semipolar blue GaInN/GaNLEDs show higher efficiencyat high current density due to lower efficiency droop. Table 9.1 summarizes recentlyreported polar, nonpolar and semipolar blue GaInN/GaN LEDs.

9.3.2.1 m-plane, ð1122Þ and ð1011 �GaInN/GaN LEDs

Before 2010, most of the studies focused on nonpolar planes (mostly m-plane) andsemipolar planes oriented near 45 degrees from c-plane such as

�1122

�and

�1011

�

planes. Chakraborty et al.39 first demonstrated m-plane blue LEDs with an outputpower <1 mW and EQE <1%. The low performance could result from high basalplane stacking faults (BPSFs) and TD density. Okamoto et al.40 optimized growthconditions and obtained dislocation-free m-plane MQW LEDs with output power of

70

60

50

40

30

20

10

00 50 100 150 200 250 300

EQ

E (%

)

c-plane LEDsSemipolar LEDs

@ 445 nm, chip size 0.1 mm2

Current density A/cm2

Figure 9.14 Comparison of EQE as a function of current density for typical c-plane andsemipolar blue (dark gray in print versions) GaInN/GaN LEDs.


Table 9.1 Device structure and performance of blue GaInN/GaN LEDs

Plane StructureChip size(mm2) Method

Wavelength(nm)

LOP(mW)

EQE(%)

Droop(%) Year Institution

�1010

�3 nm 5 QWs 0.09 DC 435 1.8 3.1 23.1 2006 Okomato

et al.40

�1122

�3 nm SQW 0.10 DC 426 1.8 3.0 17.5 2006 Funato et al.42

�1122

�3 nm 6 QWs 0.11 Pulsed 480 9.0 17.4 23.4 2007 Zhong et al.43

�1011

�3 nm 6 QWs 0.11 Pulsed 444 16.2 29.0 16.8 2007 Zhong et al.44

(0001) 2.5 nm 6QWs

0.11 Pulsed 444 30.4 55.7 48.0 2009 Vampolaret al.47

�1010

�2 nm 3 QWs 0.10 Pulsed 460 5.5 10.5 24.5 2009 Lin et al.41

(0001) 5 QWs 1 Pulsed 440 e e 1.6 2009 Chung et al.51

(0001) 5 QWs 1 Pulsed 440 e e 2.4 2009 Xu et al.52

(0001) 3 nm 6 QWs 1.00 Pulsed 460 37.0 80.8 18.3 2010 Narukawaet al.38

(0001) 3 nm 5 QWs 0.09 Pulsed 444 e e 4.5 2010 Dai et al.28

�2021

�3 nm 3 QWs 0.10 Pulsed 423 30.6 52.2 8.5 2011 Zhao et al.32

�2021

�12 nm SQW 0.10 Pulsed 446 28.2 52.7 4.7 2012 Pan et al.33

(0001) 4 nm 3QWs 0.09 - 445 e e 0.8 2013 Lin et al.53

(0001) 3 nm 5 QWs 0.36 Pulsed - e e 5.6 2013 Park et al.54

�3031

�15 nm SQW 0.10 Pulsed 413 29.8 49.5 9.5 2014 Becerra et al.34

Nonpolar and semipolar LEDs are grown on bulk GaN substrates; Ref. 47 is on (0001) bulk GaN substrate, and the rest of (0001) LEDs are on sapphire substrate. Wavelength is the peak ELwavelength at 20 mA. LOP and EQE are also measured at 20 mA. Droop of most devices is calculated at 100 mA, except for Ref. 47 at 75 mA, Ref. 42 at 200 mA, Ref. [28] at 200 mA, andRefs. 51 and 52 at 350 mA.

316Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

1.79 mW and EQE 3.1% at 20 mA. Lin et al.41 investigated the growth and propertiesof high indium composition (26%) m-plane InGaN LEDs showing EL peak wave-length of 460 nm, output power of 5.5 mW, and EQE of 10.5% at 20 mA.

Despite the encouraging improvement of m-plane blue LEDs, growth challengessuch as BPSFs and alloy inhomogeneity still make it very difficult to fabricate highquality m-plane QWs. The advantage of semipolar QWs is much improved materialquality while having weak polarization properties. Semipolar planes oriented near45 degrees from c-plane such as

�1122

�and

�1011

�have been extensively studied

since there is no polarization at 45 degrees. In 2006, Funato et al.42 demonstratedblue

�1122

�single QW (SQW) LED with output power of 1.76 mW and EQE of

3% at 20 mA, which were comparable to the performance of Okamoto et al.’s m-planeblue LEDs.40 Zhong et al.43 optimized the device structure and realized an output po-wer of 9 mW and EQE of 17.4% at 20 mA. Though

�1122

�LEDs have comparable

performance with respect to m-plane LEDs, they still cannot outperform c-planeblue LEDs. Semipolar

�1011

�LEDs show promising capability to achieve high effi-

ciency. Zhong et al.44 reported semipolar�1011

�LEDs on low defect density bulk

GaN substrate with an output power of 16.2 mW and EQE of 29% at 20 mA. Zhaoet al.45 performed systematic optimization of device structure by MOCVD, obtainingan output power of 22.75 mW and EQE of 39.5% at 20 mA. Furthermore, Zhao et al.46

incorporated backside roughening and transparent stand package techniques into�1011

�LEDs. At 20 mA under pulsed conditions, the LED exhibited an output power

of 31.1 mW and EQE of 54.7%, which are comparable to commercial c-plane LEDs.This is the first semipolar LED with over 50% EQE and low efficiency droop up to300 mA. This work indicates semipolar blue LEDs on bulk GaN substrates cancompete with commercial c-plane blue LEDs in terms of both material propertiesand chip processing techniques.

9.3.2.2�2021 Þ and ð3031Þ GaInN/GaN LEDs

Semipolar planes with high inclination angle such as�2021

�and

�3031

�have attracted

considerable interests for blue LEDs due to their unique material properties and deviceperformance such as high critical thickness, increased IQE, low efficiency droop, andlow thermal droop. Thick SQW structures of

�2021

�and

�3031

�planes can reduce

both efficiency droop and thermal droop. Zhao et al.32 first demonstrated the�2021

�

LEDs with output power of 30.6 mW and EQE of 52%, which are comparable tothe best device performance ever reported for nonpolar and semipolar devices. Thisdevice showed low droop ratios of 0.7% at 35 A/cm2, 4.3% at 50 A/cm2, 8.5% at100 A/cm2, and 14.3% at 200 A/cm2. In order to correlate the carrier dynamics withthe low-droop performance of

�2021

�LEDs, Pan et al.33 fabricated a small-area

(0.1 mm2) LED chip with a QW thickness of 12 nm. Fig. 9.15(a) shows the lightoutput power (LOP) and EQE of the SQW blue LED. This LED had an output powerof 30 mW and EQE of 51% at 20 mA, with low droop ratios of 4.7% at 100 A/cm2,14.2% at 200 A/cm2, and 22.2% at 400 A/cm2. The scanning transmission electron


microscope (STEM) image of the active region in Fig. 9.15(b) indicates uniform QWthickness and smooth interfaces. The authors postulated that, the combination ofreduced electric field, enhanced carrier uniformity (SQW structure) and smaller poten-tial fluctuation (thick high-quality InGaN layer) increased the effective active regionvolume, which lowers the average carrier density. Therefore, the Auger recombinationand carrier leakage can be mitigated and efficiency droop can be reduced.

Calculations show that the critical thickness of the semipolar�3031

�plane is larger

than the�2021

�plane due to reduced resolved shear stress on the basal plane.34 Thus

thicker active regions can be grown on semipolar�3031

�GaN substrates, which can

serve to decrease efficiency droop. Fig. 9.16(a) presents LOP and EQE for�3031

�

SQW blue-violet LED. This device has shown output power of more than 1W at a cur-rent density of 1 kA/cm2 with a droop ratio of only 33%, which is ideal for high powerapplications. In addition, results also show that increasing QW thickness lowers theefficiency droop. Fig. 9.16(b) shows the peak EL wavelength and full-width at half-maximum (FWHM) at different current densities. The

�3031

�LED exhibited only a

7-nm wavelength shift and around 15-nm FWHM up to 1 kA/cm2, which indicate amuch reduced electric field and high material quality. These results indicate that a thickactive region design for semipolar LEDs is favorable for reducing efficiency droop.

9.3.3 Modified ABC model for c-plane and semipolarGaInN/GaN LEDs

Despite the tremendous progress on low-droop blue semipolar GaInN/GaN LEDs, thephysical origin of the low droop performance is elusive and still being debated.

500

400

300

200

100

00 100 200 300 400

Ligh

t out

put p

ower

(mW

)

Pulsed conditionwith 2% duty cycle

60

50

40

20

EQ

E (%

)

10

0

p-GaN

AlGaN EBL

12 nm InGaN QW

InGaN/GaN SL

n-GaN

50 nm


30

(a) (b)

×

Figure 9.15 (a) Light output power and EQE versus current densities for GaN (10 nm)/InGaN(12 nm)/GaN (15 nm) SQW

�2021

�blue (dark gray in print versions) LED under pulsed

conditions. The inset is the schematic view of the device structure. (b) STEM image of activeregion of the LED.After Pan CC, et al. High-power, low-efficiency-droop semipolar single-quantum-well bluelight-emitting diodes. Appl Phys Express 2012;5:062103.


Although the IQE curve of c-plane GaInN/GaN LEDs was very well fitted with theABC model, similar A, B, and C coefficients were not able to model the semipolarGaInN/GaN LEDs. This indicates a different ABC model has to be used for semipolarLEDs where the different physical properties and resulted carrier dynamics must betaken in to account.

Fu et al.29 performed a systematic comparison of carrier lifetime between c-planeand semipolar GaInN/GaN LEDs. It was found out that semipolar GaInN/GaNLEDs have much smaller carrier lifetime s compared to c-plane devices, possiblydue to the reduced QCSE and material qualities such as high growth temperature,lower defects, and smaller indium fluctuation. Furthermore, we successfully linkedthe carrier lifetime to the droop performance of GaInN/GaN LEDs by the current den-sity equation and modified ABC model in Eqs. (9.12)e(9.14). The current density Jcan be expressed as a function of s: J ¼ qdDn/s, where q is the carrier charge, d isthe active region thickness, and Dn is the excess carrier density. Therefore, at thesame injected current density J, semipolar GaInN/GaNLEDs will have less excesscarrier density Dn than c-plane samples due to the smaller carrier lifetime s. Thismay explain the observed low droop performance on semipolar LEDs. In order toconfirm this assumption, we used a modified ABC model with PSF effect to simulatethe droop characteristics of the semipolar GaInN/GaNLEDs.

According to the discussions above, semipolar GaInN/GaNLEDs should have aweaker PSF effect due to lower carrier density and a larger n0 should be utilized inthe simulation. Fig. 9.17 presents the simulation results of semipolar and c-planeLEDs. The assumptions about the LEE are reasonable, given the current technologystatus. The injection efficiency is assumed to be 100% for all three LEDs. A verygood agreement between experimental data and the theoretical modeling was obtainedfor the semipolar LED using a weak PSF effect (n0 ¼ 5.0 � 1019 cm�3) and the two

1.2

1.0

0.8

0.6

0.4

0.2

0.0

Ligh

t out

put p

ower

(W)

0 400200 600 800 1000

Pulsed operation1% duty cycle

60

50

40

30

20

10

0

EQ

E (%

)

FWH

M (n

m)

Wav

elen

gth

(nm

) 430

440

420

410

4000 200 400 600 800 1000

Current density (A/cm2)Current density (A/cm2)

20

15

10

5

0

(a) (b)

Figure 9.16 (a) Light output power and EQE as a function of current density for GaN (15 nm)/InGaN (15 nm)/GaN (10 nm) SQW

�3031

�LED under pulsed conditions with 1% duty cycle.

The inset shows the device under current injection. (b) Peak wavelength and full-width at half-maximum (FWHM) versus current densities for the device.After Becerra DL, et al. High-power low-droop violet semipolar InGaN/GaN light-emittingdiodes with thick active layer design. Appl Phys Lett 2014;105:171106.


c-plane LEDs using a strong PSF effect (n0 ¼ 1.0 � 1018 or 6.0 � 1018 cm�3). TheIQE curve of the UCSB semipolar LED was obtained by using an LEE of 65%.The fitting parameters n0, A, B, and C are 5.0 � 1019 cm�3, 6.0 � 106 s�1,5.0 � 10�11 cm3/s, and 4.5 � 10�30 cm6/s, respectively. The IQE curve of theUCSB c-plane LED was obtained by using an LEE of 80%. The dash dot line showsthe simulated IQE curve using n0 ¼ 1.0 � 1018 cm�3, A ¼ 1.6 � 106 s�1,B ¼ 2.8 � 10�11 cm3/s, and C ¼ 4.8 � 10�30 cm6/s. The IQE curve of the c-planeblue LED from Nichia was fitted well when n0, A, B, and C are 6.0 � 1018 cm�3,8.0 � 105 s�1, 3.3 � 10�11 cm3/s, and 2.4 � 10�30 cm6/s, respectively. Furthermore,the Nichia c-plane blue LED38 has a larger chip size than the UCSB c-plane LED47 andshows smaller efficiency droop, which also confirms the correlation between excesscarrier density and efficiency droop. Though the UCSB semipolar plane33 has onemagnitude smaller chip size, the efficiency droop is much smaller than the Nichiac-plane LEDs with comparable IQE. These results all indicate the advantages ofdeveloping high-power, low-droop, small-area semipolar blue LEDs.

9.4 Thermal droop in GaInN/GaN LEDs

Thermal performance of GaInN/GaN LED is becoming an increasingly importantissue especially in high power and high temperature applications (such as automotivehead lights). A GaInN/GaN LED usually runs at junction temperatures of 80e100�C,which will significantly degrade the device performance.5 In order to avoid self-

100

90

80

70

60

50

40

30Empty shape: experimental dataLine: simulated data bymodified ABC model

Figure 9.17 Simulated IQE curves as a function of current densities for semipolar�2021

�LEDs

with weak (solid line) PSF effect, Nicha c-plane LEDs with strong PSF effect (dot line), andUCSB c-plane LEDs with strong PSF effect (dash dot line). Reported experimental data arealso plotted for UCSB semipolar

�2021

�LED (circle),33 Nichia c-plane38 (triangle), and

UCSB c-plane LED (diamond).47


heating effects, the thermal performance of GaInN/GaN LED is usually measuredunder pulse conditions. C-plane blue GaInN/GaN LEDs have not only huge efficiencydroop but also significant thermal droop: a 20% thermal droop, calculated usingEq. (9.18), is observed at 100 A/cm2 when the temperature is increased from roomtemperature to 100�C. Thermal droop is defined as:

Thermal droop %ð Þ ¼ EQE Jð Þ20�C � EQE Jð Þ100�C� �

=IQE Jð Þ20�C � 100%

(9.18)

where EQE(J)100�C and EQE(J)20�C are the EQEs of LEDs at the junction temperatureof 100�C and 20�C at current density J, respectively. Another factor can also be used:

hot=cold factor ¼ EQEðJÞ100�C�EQEðJÞ20�C (9.19)

Meyaard et al.48 calculated the percentage contribution of SRH recombination andhigh-order recombination (Auger þ f(n)) as shown in Fig. 9.18. At low currentdensity such as 10 mA, SRH recombination contributes 10.0% of total recombinationat 300K and 47.3% at elevated temperature (450K). This accounts for the thermaldroop of GaInN/GaN LEDs at low current density. At 2 A, SRH accounts for only0.6% of total recombination at 300K and only 3.1% at 450K. Therefore, SRH is notresponsible for thermal droop at high current densities. However, the high-orderrecombination such as Auger recombination and drift-induced carrier leakage playan important role at high current densities. At 2 A, they account for 39.3% of totalrecombination at 300K and nearly 60% at 450K. The contribution increases with

60

40

20

0

% o

f SR

H re

com

bina

tion

320 340 360 380 400 420 440 460 300 320 340 360 380 400 420 440 460Temperature (K)

100 mA10 mA

2 A100 mA10 mA

2 A

70

60

50

40

30

20

10

0

Temperature (K)

% o

f (A

uger

+f(n

)) re

com

bina

tion

(a) (b)

Figure 9.18 Percentage of (a) SRH recombination and (b) high-order recombination(Auger þ f(n)) to total recombination as a function a temperature at different current levels.Plot (a) is calculated based on An/Rtot and plot (b) is calculated from (Cn3þf(n))/Rtot. Themeasured GaInN/GaN LEDs have five pairs of QWs with a peak wavelength of 460 nm.After Meyaard DS, et al. On the temperature dependence of electron leakage from the activeregion of GaInN/GaN light-emitting diodes. Appl Phys Lett 2011;99:041112.


increasing temperature, which may be a result of more energetic electrons flying overthe active region and EBL or greater defect-assisted tunneling through the EBL. Inaddition, Meyaard et al.49 observed increased thermal droop with larger LED chipsize (i.e., lower current density), which is attributed to SRH recombination since SRHrecombination dominates at low current densities. Therefore, increasing chip size, atechnique for reducing current droop, will have an adverse effect on the thermalperformance of GaInN/GaN LEDs.

Semipolar LEDs with thick QW structure show a much improved thermal perfor-mance. Fig. 9.19(a) shows thermal droop of 12 nm SQW

�2021

�blue LEDs as a func-

tion of temperature.50 At 100 A/cm2, the thermal droop is only 9% at 100�C with hot/cold factor of 0.9 while c-plane blue LEDs show thermal droop >20% and hot/coldfactor <0.8. From 1 to 40 A/cm2, thermal droop is decreased, and a slight increasein thermal droop is measured at 40 A/cm2. This trend can be explained by the carrierrate equation model where Shockley-Read-Hall (SRH), radiative, and Auger recombi-nation are considered. SRH recombination dominates at low current densities. Radia-tive recombination becomes important with increasing current density, resulting inreduction of thermal droop from 1 to 40 A/cm2. When the current density is toohigh, Auger recombination becomes dominant and increases the thermal droop.Fig. 9.19(b) shows the EQE curve of 20 nm SQW

�3031

�LEDs at different temper-

atures. It exhibited improved thermal performance with thermal droop <10% andhot/cold factor >0.9 at 100�C.

20

25

15

10

5

004020 60 80 100 40 80 120 160 200

0

5

10

15

20

25

30

Ther

mal

dro

op (%

)

Temperature (°C)

EQ

E (%

)


1 A/cm2

2 A/cm2

5 A/cm2

10 A/cm2

40 A/cm2

100 A/cm2

(2021) 12 nm SQW–– (3031) 20 nm SQW

––

Thermal droop10%20°C

40°C60°C80°C100°C120°C

(a) (b)

Figure 9.19 (a) Thermal droop versus temperature at different current densities for 12 nm SQW�2021

�LEDs. (b) EQE as a function of current density with temperature 20e120�C for 20 nm

SQW�3031

�LEDs. This LED has only 10% EQE droop at 120�C.

(a) After Pan CC, et al. Reduction in thermal droop using thick single-quantum-well structure insemipolar (20-2-1) blue light-emitting diodes. Appl Phys Express 2012;5:102103. (b) AfterBecerra DL, et al. High-power low-droop violet semipolar InGaN/GaN light-emitting diodeswith thick active layer design. Appl Phys Lett 2014;105:171106.


References

1. Nakamura S, Fasol G, Pearton SJ. The blue laser diode: the complete story. 2nd ed.Springer; 2000.

2. Nakamura S. Background story of the invention of efficient blue InGaN light emittingdiodes (Nobel lecture). Ann Phys 2015;527(5e6):334e49.

3. Nakamura S, Mukai T, Senoh M. Candela-class high-brightness InGaN/AlGaN doubleheterostructure blue light emitting diodes. Appl Phys Lett 1994;64(13):1687.

4. Kim MH, Schubert MF, Dai Q, Kim JK, Schubert EF, Piprek J, Park Y. Origin of efficiencydroop in GaN-based light-emitting diodes. Appl Phys Lett 2007;91:183507.

5. Feezell DF, Speck JS, DenBaars SP, Nakamura S. Semipolar (20-2-1) InGaN/GaN lightemitting diodes for high efficiency solid-state lighting. J Disp Tech 2013;9(4):190e8.

6. Piprek J. Efficiency droop in nitride-based light-emitting diodes. Phys Status Solidi A 2010;207(10):2217e25.

7. David A, Grundmann MJ. Droop in InGaN light-emitting diodes: a differential carrierlifetime analysis. Appl Phys Lett 2010;96:103504.

8. Hader J, Moloney JV, Koch SW. Beyond the ABC: carrier recombination in semiconductorlasers. Proc SPIE 2006;6115:61151T.

9. Lin GB, Meyaard D, Cho J, Schubert EF, Shim H, Sone C. Analytic model for the efficiencydroop in semiconductors with asymmetric carrier-transport properties based on drift-induced reduction of injection efficiency. Appl Phys Lett 2012;100:161106.

10. Fu H, Lu Z, Zhao Y. Analysis of low efficiency droop of semipolar InGaN quantum welllight-emitting diodes by modified rate equation with weak phase-space filling effect. AIPAdv 2016;6:065013.

11. Shen YC, M€uller GO, Watanabe S, Gardner NF, Munkholm A, Krames MR. Augerrecombination in InGaN measured by photoluminescence. Appl Phys Lett 2007;91:141101.

12. Iveland J, Martinelli L, Peretti J, Speck JS, Weisbuch C. Direct measurement of Augerelectrons emitted from a semiconductor light-emitting diode under electricali: identificationof the dominant mechanism for efficiency droop. Phys Rev Lett 2013;110:177406.

13. Chichibu SF, Azuhata T, Sugiyama M, Kitamura ITY, Okumura H, Nakanishi H, Sota T,Mukai T. Optical and structural studies in InGaN quantum well structure laser diodes. J VacSci Technol B 2001;19:2177.

14. Wang J, Wang L, Zhao W, Hao Z, Luo Y. Understanding efficiency droop effect in InGaN/GaN multiple-quantum-well blue light-emitting diodes with different degree of carrierlocalization. Appl Phys Lett 2010;97:201112.

15. Fiorentini V, Bernardini F, Della Sala F, Di Carlo A, Lugli P. Effects of macroscopicpolarization in III-V nitride multiple quantum wells. Phys Rev B 1999;60:8849.

16. Monemar B, Sernelius EB. Defect related issues in the ‘current roll-off’ in InGaN basedlight emitting diodes. Appl Phys Lett 2007;91:181103.

17. Meneghini M, Trivellin N, Meneghesso G, Zanoni E. A combined electro-optical methodfor the determination of the recombination parameters in InGaN-based light-emittingdiodes. J Appl Phys 2009;106:114508.

18. Laubsch A, Sabathil M, Baur J, Peter M, Hahn B. High-power and high-efficiencyInGaN-based light emitters. IEEE Trans Electron Devices 2010;57(1):79e87.

19. Weisbuch C, Piccardo M, Martinelli L, Iveland J, Peretti J, Speck JS. The efficiencychallenge of nitride light-emitting diodes for lighting. Phys Status Solidi A 2015;212(5):899e913.


20. Hader J, Moloney JV, Pasenow B, Koch SW, Sabathil M, Linder N, Lutgen S. On theimportance of radiative and Auger losses in GaN-based quantum wells. Appl Phys Lett2008;92:261103.

21. Bertazzi F, Goano M, Bellotti E. A numerical study of Auger recombination in bulk InGaN.Appl Phys Lett 2010;97:231118.

22. Kioupakis E, Rinke P, Delaney KT, Van de Walle CG. Indirect Auger recombination as acause of efficiency droop in nitride light-emitting diodes. Appl Phys Lett 2011;98:161107.

23. Delaney KT, Rinke P, Van de Walle CG. Auger recombination rates in nitrides from firstprinciples. Appl Phys Lett 2009;94:191109.

24. Cho J, Schubert EF, Kim JK. Efficiency droop in light-emitting diodes: challenges andcountermeasures. Laser Photon Rev 2013;7(3):408e21.

25. Vampola KJ, Iza M, Keller S, DenBaars SP, Nakamura S. Measurement of electronoverflow in 450 nm InGaN light-emitting diode structures. Appl Phys Lett 2009;94:06116.

26. Yen SH, Tsai MC, Tsai ML, Shen YJ, Hsu TC, Kuo YK. Effect of N-type AlGaN layer oncarrier transportation and efficiency droop of blue InGaN light-emitting diodes. IEEEPhoton Tech Lett 2009;21(14):975e7.

27. Schubert MF, Xu J, Kim JK, Schubert EF, Kim MH, Yoon S, Lee SM, Sone C, Sakong T,Park Y. Polarization-matched GaInN/AlGaInN multi-quantum-well light-emitting diodeswith reduced efficiency droop. Appl Phys Lett 2008;93:041102.

28. Dai Q, Shan Q, J.Wang, Chhajed S, Cho J, Schubert EF, Crawford MH, Koleske DD,Kim M-H, Park Y. Carrier recombination mechanisms and efficiency droop in GaInN/GaNlight-emitting diodes. Appl Phys Lett 2010;97:133507.

29. Fu H, Lu Z, Zhao X, Zhang YH, DenBaars SP, Nakamura S, Zhao Y. Study of lowefficiency droop in semipolar (20-2-1) InGaN light-emitting diodes by time-resolvedphotoluminescence. J Display Technol 2016;12:736.

30. Ryu HY, Shin DS, Shim JI. Analysis of efficiency droop in nitride light-emitting diodes bythe reduced effective volume of InGaN active material. Appl Phys Lett 2012;100:131109.

31. Chichibu SF, et al. Origin of defect-insensitive emission probability in In-containing (Al, In,Ga)N alloy semiconductors. Nat Mat 2006;5:810.

32. Zhao Y, Tanaka S, Pan CC, Fujito K, Feezell D, Speck JS, DenBaars SP, Nakamura S.High-power blue-violet semipolar InGaN/GaN light-emitting diodes with low efficiencydroop up at 200A/cm2. Appl Phys Express 2011;4:082104.

33. Pan CC, et al. High-power, low-efficiency-droop semipolar single-quantum-well bluelight-emitting diodes. Appl Phys Express 2012;5:062103.

34. Becerra DL, et al. High-power low-droop violet semipolar InGaN/GaN light-emittingdiodes with thick active layer design. Appl Phys Lett 2014;105:171106.

35. Romanov A, Baker T, Nakamura S, Speck J. Strain-induced polarization in wurtziteIII-nitride semipolar layers. J Appl Phys 2006;100:023522.

36. Zhao Y, Farrell RM, Wu YR, Speck JS. Valence band states and polarized optical emissionfrom nonpolar and semipolar IIIenitride quantum well optoelectronic devices. Jpn J ApplPhys 2014;53:100206.

37. http://www.str-soft.com/products/SiLENSe/.38. Narukawa Y, Ichikawa M, Sanga D, Sano M, Mukai T. Whit light emitting diodes with

super-high luminous efficacy. J Phys D Appl Phys 2010;43:354002.39. Chakraborty A, et al. Demonstration of nonpolar m-plane InGaN/GaN light-emitting diodes

on free-standing m-plane GaN substrates. Jpn J Appl Phys 2005;44(5):L173e5.40. Okamoto K, et al. Dislocation-free m-plane InGaN/GaN light-emitting diodes on m-plane

GaN single crystals. Jpn J Appl Phys 2006;45(45):L1197e9.


http://www.str-soft.com/products/SiLENSe/

41. Lin YD, et al. Characterization of blue-green m-plane InGaN light emitting diodes. ApplPhys Lett 2009;94:261108.

42. Funato M, et al. Blue, gree, and amber InGaN/GaN light emitting diodes on semipolar{11-22} GaN bulk substrates. Jpn J Appl Phys 2006;45(26):L659e62.

43. Zhong H, et al. Demonstration of high power blue-green light emitting diode on semipolar(112-2) bulk GaN substrate. Electron Lett 2007;43(15).

44. Zhong H, et al. High power and high efficiency blue light emitting diode on freestandingsemipolar (10-1-1) bulk GaN substrate. Appl Phys Lett 2007;90:233504.

45. Zhao Y, et al. Optimization of device structures for bright blue semipolar (10-1-1) lightemitting diodes via metalorganic chemical vapor deposition. Jpn J Appl Phys 2010;49:070206.

46. Zhao Y, et al. 30-mW-class high-power and high-efficiency blue semipolar (10-1-1) InGaN/GaN light-emitting diodes obtained by backside roughening technique. Appl Phys Express2010;3:102101.

47. Vampola KJ, Fellows NN, Masui H, Brinkley SE, Furukawa M, Chung RB, Sato H,Sonoda J, Hirasawa H, Iza M, DenBaars SP, Nakamura S. Highly efficient broad-area blueand white light-emitting diodes on bulk GaN substrates. Phys Stat Sol A 2009;206:200.

48. Meyaard DS, et al. On the temperature dependence of electron leakage from the activeregion of GaInN/GaN light-emitting diodes. Appl Phys Lett 2011;99:041112.

49. Meyaard DS, et al. Temperature dependent efficiency droop in GaInN light-emitting diodeswith different current densities. Appl Phys Lett 2012;100:081106.

50. Pan CC, et al. Reduction in thermal droop using thick single-quantum-well structure insemipolar (20-2-1) blue light-emitting diodes. Appl Phys Express 2012;5:102103.

51. Chung HJ, et al. Improved performance of GaN-based blue light emitting diodes withInGaN/GaN multilayer barriers. Appl Phys Lett 2009;95:241109.

52. Xu J, et al. Reduction in efficiency droop, forward voltage, ideality factor, and wavelengthshift in polarization-matched GaInN/GaInN multi-quantum-well light-emitting diodes. ApplPhys Lett 2009;94:011113.

53. Lin GB, et al. Effect of quantum barrier thickness in the multiple quantum well active regonof GaInN/GaN light-emitting diodes. IEEE Photon J 2013;5(4):1600207.

54. Park JH, et al. Enhanced overall efficiency of GaInN-based light-emitting diodes withreduced efficiency droop by Al-composition-graded AlGaN/GaN superlattice electronblocking layer. Appl Phys Lett 2013;103:061104.



Photonic crystal nitride LEDs 10Martin D.B. CharltonUniversity of Southampton, Southampton, United Kingdom

10.1 Introduction

In view of raised global awareness of the depletion of non-renewable energy sourcesand the associated impact on the environment, there is intense interest and highlyfocused efforts to develop green, renewable energy technologies and to increase thewall-plug efficiency of existing technologies where possible. Approximately 19% ofthe world’s energy consumption is currently attributed to lighting. Present lightingtechnologies include incandescent lamps, fluorescent lamps, light-emitting diodes(LEDs) and organic light-emitting diodes (OLEDs). Various factors affect wall-plugefficiency.

10.1.1 Epitaxial materials

LEDs are based on a semiconductor heterostructure grown on a planar substrate.On application of a voltage across the heterostructure, electron-hole pairs (exci-tons) are generated in the junction regions of the epitaxial heterostructure, whichthen recombine after a short time delay (known as the relaxation time) generatinga photon with energy equal to the electronic band gap energy associated with theheterostructure.

LEDs can be fabricated from a number of different semiconductor materials. Due tothe dependence upon the intrinsic band gap energy, the wavelength of emission isprimarily related to the choice of this material. Common epitaxial materials for con-ventional LEDs include: InGaN/GaN for blue, GaIn/GaN for green, AlGa/InP forred and InGaAsP/InP for infrared LEDs. To some degree, the wavelength of emissioncan be adjusted by inducing and controlling the strain in the epitaxial heterostructure,since this induces a change in relative energy levels of the band structure either side ofthe junction region. This technique has, for example, been employed in GaN-basedLEDs to obtain green emission whereas the natural band gap energy would lie inthe blue part of the spectrum.

10.1.1.1 Factors affecting internal quantum efficiency

In a conventional LED, radiative emission resulting from recombination of excitonpairs in the multiple quantum well (MQW) layers must be maximized. This is achievedby optimizing the internal quantum efficiency (IQE) of the epitaxy structure. IQE istherefore a measure of the efficiency of conversion of electrons to photons. To obtain


https://doi.org/10.1016/B978-0-08-101942-9.00010-1

high IQE, the epitaxial heterostructure must be substantially monocrystalline and havea low epitaxial defect density. Other factors affecting electrical efficiency include thelayout and resistivity of the electrical contacts, and any change in strain due to local-ized heating of the layers, which in turn is dependent upon methods of heat extraction.

10.1.1.2 Base substrate material

In order to reduce the defect density the epitaxial layers must have a lattice constantthat is well matched to that of the underlying substrate material. This determines thechoice of underlying substrate. Sapphire is very well matched to GaN and so is thesubstrate of choice. However, sapphire is electrically insulating, is not a good heatconductor and is expensive to produce. Requirements for substrate materials placeconstraints on LED design and cost. Considerable efforts have been made to relievesubstrate-dependent growth issues resulting in a variety of LED epitaxialconfigurations.

10.1.2 Types of LED

LEDs have evolved over the years into a variety of types, which can be groupedaccording to the epitaxial substrate structure.

10.1.2.1 P-side up lateral current spreading LEDs

The most conventional LED design, which is often used for blue GaN LEDs, is theP-side up lateral current spreading design (Fig. 10.1). This comprises a GaN layergrown directly onto a sapphire substrate with the N-doped layer at the bottom of thestack and P-doped layer at the top. Since sapphire is electrically insulating, the back-side (N) contact must be off to the side of the LED, meaning that the injected currentmust spread to the sides of the contact (via the N-doped layer), hence the namedlateralcurrent spreading design.

A thick undoped seed layer is first deposited onto the substrate to improve latticematching to the sapphire. After some distance, the growth conditions are changed tocreate successive N-and P-doped layers with multiple quantum wells sandwichedin-between to form the junction regions. The N-doped layer is usually thick due tothe need for efficient lateral current spreading from the side contact.

N-contact

P-contact

N-GaN

P-GaN

MQW

Sapphire substrate

Current blocking layer

Transparent contact layer

Figure 10.1 P-side up lateral current spreading LED.


As the conductivity of the top P-side layer is relatively poor, an additional trans-parent conductive layer (TCL) is deposited on top of the device to improve electricalconductivity and allow uniform charge injection to the top P-GaN layer. The TCLmaterial of choice is indium tin oxide (ITO); however, ITO is far from perfectly trans-parent, hence the TCL introduces optical and electrical losses through light-trappingand increased resistivity. ITO also degrades through absorption of moisture, so theoverall device must be encapsulated in silicon dioxide with windows for the electricalcontacts.

Non-transparent metal contact stacks are then deposited onto the TCL. The currentspreads laterally via the contacts and TCL to the P-type layer. The top contact arrange-ment for P-side up LEDs is non-ideal since the nontransparent contacts mask part ofthe light-emitting area reducing the efficiency of the device.

Sapphire (which is well lattice matched to GaN) is a poor heat conductor. P-side updevices suffer reduced lifetimes if driven at the very high current densities required forgeneral lighting applications. The need for a thin optically transparent top currentspreading layer also places constraints on the achievable contact resistance, and there-fore on electrical efficiency of the device.

Finally, the underlying sapphire substrate is often thinned by chemical mechanicalpolishing (CMP) and a reflective metal layer applied to the underside to reflect more ofthe light out of the top of the LED.

10.1.2.2 N-side up LEDs

A much more sophisticated configuration, which is used for high-power blue LEDs, isknown as the vertical current spreading design. This overcomes lattice mismatchissues and provides improved heat extraction and reduced electrical contact resistance,but at the expense of complex fabrication. Fig. 10.2 shows an N-side up verticalcurrent spreading LED.

In this case the epitaxial layers are initially grown on a sacrificial sapphire substrate(similar to a conventional P-side up LED), but once grown, the P-side of the epistruc-ture is bonded onto a highly conductive silicon or metallic (usually copper) carriersubstrate. It is then released from the sacrificial growth substrate by laser lift-off.Finally, CMP is used to thin down the top surface of the N-GaN layer.

In this configuration, electrical injection to the MQWs is via the silicon or metalsubstrate and the underlying P-GaN layer, across the MQWs to the highly conductiveN-GaN layer. In this case, as the entire substrate acts as the back contact, current is

Eutectic interface layer

N-contactN-GaN

P-GaNMQW

Conductive siliconsubstrate (P-contact)

Figure 10.2 N-side up vertical current spreading LED.

Photonic crystal nitride LEDs 329

spread uniformly across the entire area of the P-GaN and travels vertically through theepitaxial layer stack, hence the configuration is known as vertical current spreading.

A key advantage of this configuration is that the epilayers can be grown much morethinly, which is important for light extraction. As the silicon or metal substrate has highthermal conductivity, heat extraction is much more efficient. In addition the metalinterface layer between the substrate and GaN provides a highly reflective mirror,greatly improving light extraction efficiency by radiating light initially emitted towardsthe substrate (which would otherwise be absorbed by the substrate in a P-side up GaNon sapphire LED), back towards the surface. The best (vertical) GaN-based LEDsachieve an internal quantum efficiency of 80%.1

10.1.2.3 Patterned substrate LEDs

Another method for improving light extraction efficiency is to grow the epitaxial layersover patterned substrates. In this case, the underlying sapphire growth substrate ispre-patterned with arrays of truncated cones, straight-sided pillars or curved structuresby photolithography and chemical etching. The initial epitaxial growth position ispre-seeded by the pattern arrangement. Epitaxial strain can be relaxed at thesubstrate/semiconductor interface. The main benefit of the underlying patterning isto scatter light, which would otherwise be absorbed by the substrate, back in thevertical direction.

10.1.3 Light-trapping in LEDs

One of the limiting factors for LED efficiency is the issue of light being trapped insidethe epitaxial heterostructure. At a smooth GaN/air interface the critical angle for totalinternal reflection (TIR), ac, is given by:

ac ¼ sin�1�

n1nGaN

�[10.1]

Hence, ac is just 24.1 degree. Although light is initially emitted in all directions bydipoles in the quantum wells (e.g., over a spherical solid angle of 4p steradians) onlylight emitted within a reduced solid angle given by a cone with an apex angle of24.1 degree has any possibility of escaping (Fig. 10.3).

By calculating the ratio of the solid angle of the escape cone to the solid angle ofemission given by:

ð1� cosðacÞÞ=2 [10.2]

we can calculate the percentage of light that has any chance of escaping. We find that ata GaN/air boundary only about 4.4% of the total light emitted has any possibility ofescaping directly from the LED surface. The remaining light is reflected back andrecycled within the LED heterostructure and substrate layers.


Light emitted within the light extraction cone is subject to Fresnel reflection loss atthe GaN/air interface. Fresnel loss is highly dependent on angle of incidence (a) to theboundary and the polarisation state with respect to the plane of incidence to the bound-ary. Amplitude reflection coefficients for the S and P polarisation states are given byEq. (10.3) below. The intensity reflection coefficients are given simply by the squaresof these equations.

r ¼cos a�

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�n2n1

�2

� sin2 a

s

cos aþffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�n2n1

�2

� sin2 a

s r ¼


�2

� sin2 a

s

��n2n1

�2

cos a


�2

� sin2 a

s

þ�n2n1

�2

cos a

[10.3]

Substituting refractive index values into Eq. (10.3), we find that Fresnel loss in-creases from a minimum value of 18% at normal incidence to the interface to 100%at perpendicular incidence (along the boundary).

Taking all of this into account and assuming a random polarisation of emitted light,overall only around 3.4% of light emitted from the quantum wells is emitted immedi-ately from the surface of a smooth unencapsulated LED. We must also remember thathalf the light emitted from the quantum wells radiates downwards towards thesubstrate. At the GaN/sapphire substrate boundary the critical angle is 46.6 degreesand about 14.4% of emitted light becomes coupled into the substrate and is lost.The remainder of the light (82%) remains trapped inside the GaN layer (Fig. 10.4).

Of course, some of the trapped light reflected back into the epitaxial layers is scat-tered from epitaxial defect sites to an angle lying within the allowed extraction cone.Some of the recycled light is also emitted from the edges of the LED chip so the actuallight extraction figure will be slightly higher.

Air

GaNn = 2.3

n = 1

αc αc

Figure 10.3 Light extraction cone. Dotted lines indicate the critical angle for total internalreflection and define the angular extent of the light extraction cone. The thick black dashed line isan example of a ray emitted within the extraction cone. It is refracted at the top surface andradiated at a glancing angle from the surface. The dash-dot line indicates a ray emitted at anangle above the critical angle. It is reflected at the top interface and trapped in the device.


We can now appreciate that due to the relatively high refractive index of LED semi-conductor materials combined with the fact that light is generated deep within theepitaxial structure, Fresnel reflection at the semiconductor/air interface causes confine-ment of a large proportion of the radiated energy. Hence, most of the light generated inthe structure gets recycled internally. This factor severely limits the wall-plugefficiency for conventional LEDs. We can see from Fig. 10.4 that a small increasein critical angle for TIR can greatly increase the amount of light radiated from theLED. To improve wall-plug efficiency we need to defeat two key light-trappingprocesses: light recycling due to the small critical angle for total internal reflectionand the Fresnel loss.

10.1.3.1 Radiometry and discussion of solid angle

We now take a closer look at the light recycling problem in relation to the direction oflight emission from the quantum wells. Consider a radiative dipole positioned on oneof the quantum wells inside an LED. If for the moment we take a simple 2D cross sec-tion through an LED structure then this dipole radiates over a flat 360 degrees disc ofangles. Light radiated at angles up to the critical angle for total internal reflection isradiated from the top surface; however, it is also refracted at the GaN/air interface.Due to refraction, light emitted at the critical angle for total internal reflection is actu-ally emitted from the top of the LED in a direction parallel to the interface (e.g., at aglancing angle along the surface). Only light emitted at angles up to the critical anglefor TIR actually contributes to the radiated power. For a simple unroughened smooth-surface LED, there is a direct relation between the direction of light emission withinthe LED and the angle at which it emerges from the top surface. Ignoring Fresnelreflection for the moment, in this simplified case there is an equal contribution to radi-ated power for all angles of emission from the radiative dipole up to the critical anglefor total internal reflection.

However, if we now consider that in reality light is actually emitted from the dipoleinto a 3D solid angle, the contribution to total radiative power is no longer equal for allangles of emission up to the critical angle. Intuitively you might expect that most of thepower radiated from the LED would have originated from light initially emitted at (or

Air

Substrate

GaN

Air

14.34%

82.26%

3.4%

Figure 10.4 Example ray paths inside an LED. The dotted line is for a ray radiating incident tothe boundaries at angles above the critical angles. The dashed line shows a ray trapped in thesubstrate layer. The solid line shows a ray trapped in the GaN layer. The percentages are for theoptical power trapped in the GaN and substrate layers, and radiated to air.


close to) normal incidence to the boundary, and you might expect the contribution toemitted power from rays emitted at the critical angle (which are radiated from thesurface at a glancing angle nearly parallel to the boundary) to be far less than thecontribution from those rays emitted at normal incidence.

However, in reality the situation is exactly the opposite way around. Fig. 10.5 plotsthe (relative) solid angle (equivalent to radiative surface area) subtended by a 1 degreewide annular extraction ring (the shaded annular surface area A in Fig. 10.6) as a func-tion of launch angle (b) measured with respect to the surface normal. This is equivalentto the total power radiated from the dipole within a 1 degree angle, at a predeterminedazimuth angle. We see from Fig. 10.5 that as the launch angle (b) increases fromnormal incidence (0 degree) to perpendicular to the interface (90 degrees), the relativesolid area subtended increases from 0 to 100. As a Lambertian source radiates equalpower in all directions (i.e., uniform illumination per solid angle), this means thatmore optical power is actually emitted within a 1 degree annular ring at glancingangles (b ¼ 90 degrees) than within a 1 degree ring close to normal incidence(b ¼ 0 degrees), as the solid angle (corresponding to the relative area of a sphere)subtended is not actually the same in both cases, but in fact is very much larger atglancing angles.

This illustrates that in order to improve light extraction efficiency it is more usefulto try to redirect rays of light initially emitted from the quantum wells at large angles tothe normal that radiate nearly parallel to the interface (dotted ray in Fig. 10.7) ratherthan rays emitted at shallow angles to the normal that radiate nearly perpendicularto the boundary (dash-dot ray in Fig. 10.7).

0 10 20 30 40 50 60 70 80 900

10

20

30

40

50

60

70

80

90

100

Pow

er (n

orm

aliz

ed to

100

)

Cone apex half-angle (b) (degrees)

Figure 10.5 Relative power emitted into a 1 degree annular emission cone as a function of coneapex (polar) half-angle. 0 degree corresponds to emission in a vertical direction perpendicular tothe top surface of the LED in which case there is no radial (sideways) emission. 90 degreescorresponds to emission parallel to the surface of the LED (sideways). For non-zero polar coneangles, the LED emits power uniformly in all radial (azimuth) directions.


Alternatively, increasing the critical angle for total internal reflection by even asmall amount will increase the total emitted power from the LED by a large amount.Hence strategies that slightly shift the critical angle for total internal reflection effec-tively can have a big impact on overall light extraction efficiency. This is a very impor-tant point, which we will return to in discussions later about the effect of surfacepatterning. We will now look in turn at several different methods for doing this.

10.1.4 Methods of improving light extraction from LEDs

Several methods can be employed to improve light extraction from conventionalLEDs. All attempt to overcome Fresnel reflection loss at the surface/air interface.The most common technique is to utilize random surface texturing to provide randomscattering centers for trapped light such that photons eventually become directed into

Extractionangle β

Emission axis

Annular ring(area A)

Figure 10.6 Annular emission cone and polar angle (b).

Air

GaN

Substrate

Quantum well

Figure 10.7 Examples of emitted ray paths. The dashed line is for a ray emitted at an angleabove the critical angle; it does not escape and no power is transferred from the LED. The dottedline is for a ray emitted at an angle just below the critical angle; it radiates along the surface of thedevice. When summed over all azimuth angles, these rays actually transmit most of the radiatedpower. The dash-dot line is for a ray emitted at angle of emission close to the normal. Whensummed over all azimuth angles these rays actually transmit very little power.


rays propagating within the normal extraction cone of the structure. In effect, rough-ening introduces sets of angled facets. This changes the conditions for total internalreflection in a highly localized way close to the angled facet, such that light emitted(or recycled) at an angle below the critical angle for TIR will be radiated from thesurface.

Random texturing can be induced either during epitaxial growth or by subsequentelectro-chemical etching of the surface layer. In both cases the positions of the randomstructures are seeded by epitaxial defect sites, and hence this technique is not compat-ible with or beneficial to very high-quality low defect density epitaxy. In addition,random roughening is very difficult to control, giving rise to large chip-to-chip andwafer-to-wafer performance variability, and the consequent need for performancebinning. Overall, random surface texturing can improve light extraction efficiencyfor a P-side up LED by around 30% to 35%. However, that is the maximum perfor-mance benefit. The vast majority of devices within a wafer will be improved by asignificantly smaller factor. Surface roughening for P-side up LED devices requiresa very lengthy epitaxial overgrowth step, which is highly undesirable in terms ofthroughput of wafers through an LED production line.

Patterning the underlying substrate with large, widely spaced, curved (dome-shaped) structures (as is the case for a patterned sapphire substrate (PSS) device) worksin a similar way e by redirecting recycled rays to angles lying within the normalextraction cone, but has the added benefit of changing the conditions for TIR at theGaN/substrate boundary, thereby reducing substrate absorption as well.

An alternative method (which is the real subject of this chapter) is to use small-scaleperiodic pattering in the form of a photonic crystal lattice etched some way into theheterostructure. In this case the photonic crystal provides a far more precisely engi-neered leakage mechanism for confined photons (which still reside in the high indexmaterial surrounding the holes) by redirecting confined modes to leaky radiatingmodes and actually reshaping the extraction light cone, for example by changingthe Fresnel reflection conditions at the top surface in a complex way.

For surface patterning of any kind, it is extremely important that the etched struc-tures do not puncture the quantum wells otherwise an electrical conduction path will becreated. This can cause short-circuiting (shunting) of the device upon application of thecurrent spreading layers. Alternatively, the leakage current is increased, which isundesirable since it reduces IQE.

10.2 Photonic crystal technology

Photonic crystals (PCs) are structured materials formed by etching 2D or 3D arrays ofshaped structures into a dielectric medium. When fabricated with submicron dimen-sions, the arrays form the optical equivalent of a semiconductor material. That is,photons can only travel across the material if they are localized into distinct energystates and obey strict rules relating to direction of travel, polarization state and energy(wavelength). An energy range can also exist for which there are no allowed states for


propagation. This energy range is known as a photonic band gap. Before looking atphotonic crystals in more detail, we will first expand on this concept and put it intoclearer perspective using an analogy familiar to solid-state physicists.

10.2.1 The workings of PCs

Consider the motion of electrons through silicon. Single-crystal silicon is made ofatoms arranged in a diamond lattice. Electrons moving through the silicon lattice expe-rience the presence of the silicon nuclei through Coulomb interactions. As an electrontravels through a silicon crystal, its interactions with the periodic potential of theatomic nuclei results in the formation of allowed and forbidden energy states. An en-ergy range can exist for which electrons cannot travel through the lattice. This is calledthe electronic band gap. At other energies, electrons can travel across the siliconcrystal, but only if they follow very special rules and conditions. For example, an elec-tron must travel in a particular direction and must have a very specific energy state.

A photonic crystal is very similar. However, in this case, instead of electrons trav-elling in a silicon crystal, we now consider photons travelling in a block of dielectricmaterial. For a photonic crystal, instead of having a periodic arrangement of atomicnuclei, we instead have arrays of tiny air holes placed carefully in a lattice arrange-ment. If we consider a truly 3D photonic crystal analogous to silicon, then the latticewould have a diamond (hexagonal close-packed) arrangement, and the air holes wouldbe spherical. But a photonic crystal can be simplified to a 2D arrangement, in whichcase it could take a different shapedsuch as triangular or square, and the air holescould be cylindrical air tubes (Fig 10.8). A photon passing through this patterneddielectric will see regions of high refractive index (the dielectric material itself)

Figure 10.8 Photonic crystal with triangular symmetry etched into the surface of a dielectricslab waveguide structure. Light couples into (and out from) the device from the slab waveguideseither side of the porous region. The lattice is designed and fabricated such that it is slightlyangled with respect to this facet to allow propagation along a certain direction within the crystal.


interspersed with regions of low refractive index (the air holes). To a photon thiscontrast in refractive index is equivalent to the periodic potential an electron experi-ences in travelling through single-crystal silicon, and we find that there are allowedand forbidden photon energies.

As photon energy is inversely proportional to photon wavelength, this means thepatterned dielectric block has certain allowed and forbidden wavelength states forphotons. As illustrated in Fig. 10.9, the patterned dielectric will not allow passageof photons whose wavelength or energy lies in the photonic band gap, but allowsfree passage of photons with allowed wavelengths, but only provided they obey strictrules in terms of direction of propagation, polarisation state and energy. Of course thedielectric material must also be inherently transparent to the operating wavelengthband (otherwise the light would simply be absorbed by the material).

The permitted states of transmission of an optical device (such as a waveguide, forexample) are known as modes. For photonic crystals the transmitted states are periodicin nature and are known as Bloch modes. Ranges of energy (or wavelength) with nopermitted transmission states are known as photonic band gaps.

This is a highly simplified description of how a photonic crystal works. In realitythe useful properties of a photonic crystal are due to partial reflection of the travellingwaves at the interface of each and every etched hole, giving rise to a huge number ofplane waves within the photonic crystal. The plane waves interfere with one another,either constructively or destructively depending on the direction of travel and wave-length. Overall the spatial modulation of the refractive index causes the localisationof photons into distinct energy states.

10.2.2 Classes of PC device

Photonic crystal devices can be grouped into two generic classes: those which makeuse of photonic band gaps and those which make use of dispersion properties. The firstclass of device includes spectral filters, microscale optical interconnects2 and micro-cavity laser devices.3 All of these applications make use of the reflective propertiesof a photonic band gap either to confine light within a microcavity region or a line-defect waveguide, or to provide back reflection over a certain wavelength range.

The second set of generic applications includes spatial beam-steering, polarisationcontrol and light extraction from LEDs. These applications do not require the existenceof a photonic band gap but instead require transmission of light directly across thelattice. In these applications light must couple from the external environment (suchas an input waveguide for an integrated optical device or quantum well for an LED)into guided modes, which exist and are allowed to travel across the photonic crystal.

For a given wavelength of light incident to the lattice, there is either partial trans-mission plus partial reflection or perfect reflection. Consider the first scenario (partialtransmission plus partial reflection). Permitted states of transmission across a photoniccrystal can be represented by an energy versus k-vector diagram e as for electricalsemiconductors (Plate 7, see color plate section).

Lines on the band diagram (known as dispersion relations to the solid-state physi-cist) provide vital information about exactly how photons of a given wavelength travel


Wavelengths outsideband gap are transmitted

Wavelengths withinband gap are reflected

Λ

Magnitude of kchanging

In-plane k vector

Direction of

k changing

X T J X

TM bandsTE bands

Λ/λ

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

φ

Filling of air rods:silicone, polymer, SiO2

Triangularsquarerectangularquasi-crystal

Engineering ‘parameters’:SymmetryΔn, Λ/φ

100 nm to ∼ 1000 nm

(a)

(b)

(c)

Figure 10.9 (a) Simple illustration of a photonic crystal. (b) Key parameters that determine itsoperation. (c) Example band diagram showing dispersion curves and band gaps (explained indetail later). Wavelengths within the band gap (shaded area) are reflected (there is nodispersion curve for these wavelengths on the band diagram), whereas all others are partiallytransmitted. The engineering parameters determine what happens to the light as it travelsthrough the device. TE, transverse electric polarization; TM, transverse magnetic polarization.


within the lattice. As travelling waves in a photonic crystal are known as Bloch modes,the lines on the band diagram show the properties of the guided Bloch modes. Theposition and shape of the dispersion curves encodes considerable useful informationabout how light travels through the lattice.

Most importantly the y-axis (scaled as frequency) tells us the wavelength (colour) ofthe light that can travel across the PC lattice. The x-axis (k-vector) tells us the directionin which light of that wavelength must travel in order to go across the PC lattice. Forphotonic crystals, direction is defined with respect to irreducible lattice symmetrydirections as shown on the top right of Plate 7 (see color plate section). We seefrom the diagram that light of a given wavelength can often travel in multiple direc-tions across the PC (for example, the T-J and T-X directions). The gradient of the linesis the group velocity of the light as it travels across the photonic crystal and so curva-ture of the lines (the second derivative) relates to changes in group velocity (groupvelocity dispersion). Strangely enough, it is possible for photonic crystals effectivelyto slow down and trap light.

The relation between the x-axis and y-axis of the band diagram (showing the posi-tion of the bands) gives the connection between the wavelength of permitted transmis-sion modes in free space, and the effective wavelength as they travel within the lattice(we will discuss this in more detail in the next section). It also gives information aboutthe effective refractive index that a wave sees as it travels across the lattice. The lowerright of Plate 7 (see color plate section) plots the effective index of the correspondingmodes (this example is for the case of air holes on a triangular lattice in a glass materialwith refractive index 1.55; the actual geometry and relative size of the holes is shownon the top right). As we shall see later, all of these factors are very relevant to theproblem of light extraction from LEDs.

Consider the second scenario for a photonic crystal e perfect reflection. In this caselight within the lattice is effectively forbidden because of the existence of a photonicband gap. The existence of photonic band gaps is predominantly determined by thecontrast in the index of refraction between the materials used in constructing the physicallattice. Band gaps can be non-polarisation-dependent and non-directional (known ascomplete and absolute band gaps), polarisation-dependent and non-directional (shadedorange in Plate 7 (see color plate section) and known as polarisation-dependent bandgaps) or directional and polarisation-dependent (shaded green in Plate 7 (see color platesection) and known as partial directional polarisation-dependent band gaps).

The transmission of light through a photonic crystal is illustrated more clearly inPlate 8 (see color plate section). In this figure, an example band diagram for a trian-gular lattice photonic crystal etched into a silicon nitride slab is shown, and the trans-mission properties of the crystal (calculated using the finite difference time domainmethod) are superimposed. The dotted black lines are the photon dispersion relations(which as we have just seen illustrates allowed modes of propagation for TE polarisedwaves as a function of frequency (y-axis) and k-vector (x-axis)). The red trace showsthe corresponding transmission spectrum as a function of frequency (y-axis) andamplitude (x-axis). We clearly see that at frequencies where allowed Bloch modes(indicated by the dotted black curves) exist, transmission occurs. The solid grey curveshows reflection from the front of the crystal as a function of frequency and amplitude.


At frequencies where Bloch modes do not exist (the gaps between the dotted blackdispersion curves) there is strong reflection from the crystal. It is clear that the redand grey traces are exactly complementary over the normalized frequency range0e0.5. At higher frequencies (shorter wavelengths) above 0.5 we find that light is stilltransmitted where allowed Bloch modes exist, but the transmitted light is subject toother optical effects such as diffraction (shown by the green curve in Plate 8 (see colorplate section)), hence the red and grey curves (transmission and reflection) are nolonger exactly complementary above that value. In practice, this means that incidentlight will split into multiple beams at frequencies lying above the diffraction limit,and the light will be partially transmitted and diffracted at the same time.

10.2.3 Regular PCs versus photonic quasi-crystals (PQCs): effectof lattice symmetry

Photonic crystals have traditionally emulated natural atomic lattice structures, and themost popular lattice shapes are 1D gratings (this is the simplest form of a photoniccrystal) and 2D square or triangular lattices (these are the most common forms of pho-tonic crystals). These lattices have, respectively, twofold, fourfold or sixfold symme-try. As the level of symmetry increases, the properties of the photonic crystals becomeless directional. The highest degree of symmetry found in nature is for the triangularlattice, which has sixfold symmetry.

For laser or LED applications the choice of lattice shape is very important becausethis in turn affects the far-field properties of the emitted light. (We will come back tothis and look at it in much more detail later in this chapter.) Since photonic crystals areman-made structures defined by photolithographic techniques, lattice patterns do nothave to be constrained to the limited set of arrangements commonly found in nature.More complex geometrical arrangements can be used instead.

A completely different class of artificial lattice shapes are known as photonic quasi-crystals (PQCs). Some example PQC structures are shown in Plate 9 (see color platesection).On first sight PQCs appear random; however, on closer inspection PQCs theycan be seen to possess long-range order but short-range disorder.4

PQCs are generated by simple geometrical algorithms. For example, the square-triangular lattice of Plate 9(c) (see color plate section) was grown using the randomStampfli inflation method. The lattice was generated by recursive shrinking and tilingof a parent cell. The parent cell is shown as the red dashed lines in Plate 9(c) (see colorplate section). The lines show that the parent cell is composed of an arrangement ofsquares and triangles in a dodecagon shape. The corner points of intersection of thelines then become the positions for the holes. This class of PQC has orders of symme-try relating to the original parent cell shape. The sunflower lattice of Plate 9(b) (seecolor plate section) was constructed using a different algorithm based on the Fibonacciseries. This lattice has the unusual property of providing maximal packing of holesaround a central region of space.

Analysis of the far-field diffraction pattern of a lattice reveals the true nature of itssymmetry. The far-field diffraction pattern can be obtained simply by applying a


Fourier transform (FT) to the arrangement of holes. Regular lattices and those withfinite levels of symmetry, such as square-triangular, Penrose and Archimedean tilings,5

reveal their symmetry by way of a few distinctly defined Bragg spots. There is a directrelationship between the number of Bragg spots, the number of lattice symmetryplanes and the corresponding number of directions for light propagation within aPC lattice. For an LED it is desirable for there to be very many possible directionsof propagation for light within the PC lattice as this results in a more uniform far-field illumination pattern (we will come back to this shortly).

Plate 9(d and e) (see color plate section) shows the Fourier transform correspondingto some example quasi-crystal lattices. The far-field transform of the Fibonacci lattice(Plate 9(e) (see color plate section)) has a well-defined circular ring with no sharplydefined artefacts in the space immediately surrounding the central Bragg spot (e.g.,there is a clear dark disc surrounding the central spot). Hence, the Fourier transformshows near perfect circular symmetry. Devices with a Fibonacci lattice are expectedto have extremely isotropic dispersion properties, meaning that light can travel equallywell in any direction.

The far-field pattern of an LED will show structural artefacts similar to those of thefar-field diffraction pattern (the Fourier transform of the lattice shape). SimulatedPC-LED beam patterns are shown in Plate 9(f and g) (see color plate section). Thefar-field beam profile for the triangular lattice clearly shows triangular artefacts,whereas the Fibonacci lattice shows a circularly symmetric far-field beam profile.

Fig. 10.10 shows actual measured far-field emission patterns for commercial PC-enhanced LEDs (the contrast range has been greatly enhanced in these images to bringout the features). Triangular artefacts are clearly visible for the blue LED based on atriangular lattice, whereas the red LED based on a quasi-crystal has a much more uni-form, circularly symmetric far-field pattern.

Artefacts such as these will affect the efficiency of phosphor colour-conversioncoatings applied to the surface of an LED. For example, since there is a triangulararrangement of periodic dark and bright spots in the far-field illumination of the trian-gular lattice shape, the phosphor coating will not be uniformly illuminated, slightlyreducing the overall wall-plug efficiency of the device.

10.3 Improving LED extraction efficiency throughPC surface patterning

Moving back to the main topic of this bookdLED light extractiondtypical P-side upsurface-emitting LEDs consist of a GaN slab layer several microns thick with a highrefractive index and embedded MQWs positioned a small distance from the surface.6

In a conventional unpatterned surface-emitting LED, the majority of the light emittedfrom the quantum wells becomes trapped in the high index GaN slab layer through TIRat the GaN/air and GaN/substrate boundaries. Since only a small fraction of emittedlight radiates away from the top surface to free space, unroughened LEDs sufferfrom poor wall-plug efficiency.7


(a) (b)

(c) (d)

(e)

(f)

Eightfold symmetric PQCPitch = 570 nmHole size = 210 nm

Triangular PCPitch = 430 nmHole size = 200 nm

Figure 10.10 Contrast enhanced images of far-field beam intensity for (a) a blue LED with atriangular PC lattice and (d) a red LED with an eightfold symmetric PQC lattice. Angularchanges in intensity relate to the underlying PC lattice shape. A PQC lattice (d) gives acircularly symmetric beam profile with respect to azimuth angle whereas a triangular lattice(a) has triangular artefacts. (b, c) Scanning electron micrographs (SEMs) of the surface of thetriangular lattice. (e) Fourier transform of the PQC lattice arrangement reveals eightfoldsymmetry. (f) SEM of the surface of the PQC lattice.


In the first section of this chapter we used a very simple model to estimate light-trapping, taking into account the solid angle of emission and the solid angle subtendedby the extraction cone, at the critical angle for total internal reflection. This gives aballpark figure for how much light becomes trapped inside the LED epitaxial layers.We now look in more detail at exactly how the back-reflected light becomes arrangedinto sets of trapped modes within a GaN epitaxy structure, and how surface patterningenables interaction with these modes.

In effect the GaN epistructure and substrate form a very thick, highly multi-modeslab waveguide. Light emitted at angles above the critical angle for TIR is localizedinto a multitude of distinct slab waveguide modes. By introducing periodic patterning,light trapped in some of these modes will become disrupted by scattering from thesidewalls of the patterned holes. In order to improve light extraction, we ideallywant the patterning (of a photonic crystal) to interact with as many of the confinedmodes as possible. To extract all of the trapped light, the photonic crystal wouldneed to interact (at least partially) with all of the trapped modes. (In practice this isvery difficult for a normal P-side up LED, which has very thick epitaxial layers andmay support 50 or more trapped modes.) Even better, we would ideally like to redirectlight confined in trapped modes to specific emission angles inside the extraction cone,depending on the application for the LED.

Extending the analogy between photonic crystals and semiconductor materialsfurther, we can borrow a concept generally used in solid-state semiconductor theoryto illustrate the behaviour of electrons inside a semiconductor material. The physicalprocess of light confinement in an LED can be explained (as before) diagrammaticallyby the energy versus k-vector (E-K) diagram. Fig. 10.11 shows a much simplified illus-trative example of a photonic band diagram for a thick slab waveguide with(Fig. 10.11(b)) and without (Fig. 10.11(a)) surface patterning. Fig. 10.11 is a bitdifferent to Plate 7 (see color plate section) shown earlier, which we recall showedthe full band structure corresponding to several directions of travel across a thin pho-tonic crystal slab supporting just one set of waveguide modes. In the previous example(Plate 7 (see color plate section)) we considered the case where light travels directlyacross the photonic crystal, exactly perpendicular to the sidewalls of the etched holes(e.g., from side to side across the PC lattice).

In this more realistic example (Fig. 10.11) we consider light travelling within athick LED, which has very many trapped modes. The light travels at a set of distinctangles with respect to the rods and vertical direction. To simplify the discussion weonly consider one direction of travel across the device (e.g., one azimuthal directiondefined with respect to the symmetry directions of the photonic crystal patterning),and we look in detail at what would correspond to just one section of the first disper-sion curve in Plate 7 (see color plate section) (equivalent to the G-J direction in Plate 7(see color plate section), for example). Hence in the following illustrative example, wecompletely disregard the higher dispersion bands and other directions of travel (G-X).

As before, the y-axis (energy) of Fig. 10.11 is related to the free space wavelengthof the emitted light and the x-axis (k-vector) is related to the effective wavelength of theemitted light as it travels inside the patterned LED. The wavelength of the light will besignificantly smaller when it travels inside the LED than the wavelength actually


c/sin(θ)

c/nc/n

c/sin(θ)c c

Light line Light line

Wavelengthsample line

Wavelengthsample line

k k

Ene

rgy

Ene

rgy

π/Λ π/Λ

Radiative modes(above light line)

Radiating PC bloch modes(above light line)

Confined slab modes(below light line)

Confined slab modes(below light line)

(a) (c)

(d)(b)

(e)

y

z

xxy plane

θφ

Figure 10.11 (a) Energy versus k-vector diagram for an unpatterned LED device. Solid curved lines: confined slab modes, shaded area: extent of lightextraction cone, vertical dashed line: Brillouin zone boundary. (b) Trapped and radiating modes for an unpatterned LED. (c) Energy versus k-vectordiagram showing modes in a PC-LED. Dashed lines: band-folded PC Bloch modes. (d) Modes for a PC-patterned LED. (e) Definition of angles: q isthe azimuth angle and f the polar angle.

344Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

radiated from the top to the free space because it becomes reduced in proportion to therefractive index of the GaN material. In fact, for a patterned (photonic crystal) LED thesituation is more complicated because the light sees an effective refractive index ratherthan the actual material refractive index. The effective refractive index is less than thematerial refractive index because some of the light will travel through the region withthe air holes (an example of this was shown earlier at the top right of Plate 7 (see colorplate section)). Hence the effective refractive index for a mode of the photonic crystalcould be anything between 1 (air) and 2.45 (GaN). Luckily the k-vector (x-axis of theband diagram) takes this into account.

We saw earlier that the k-vector is also related to the direction of propagation of aradiated wave. For an LED, the direction relates to both the azimuth angle of emissionfrom the quantum well (q as illustrated in Fig. 10.11(e)), which is defined with respectto the symmetry directions of the surface patterning (we saw earlier that directions oftravel across the photonic crystal are denoted by symmetry directions such as G-X andG-J, as indicated on the top right Plate 7 (see color plate section)) and the polar emis-sion angle (j as illustrated in Fig. 10.11(e)). In this case we can consider that thephotonic crystal behaves equivalently to a simple slab waveguide for some of theBloch modes (those which are trapped), and for those modes that are trapped, the polarangle j is equivalent to the slab waveguide mode angle. For modes that are not trapped(which leak out of the top of the device), the polar angle j is related to the radiationdirection (it is not actually the radiation direction as refraction will occur at thesurface).

For an unpatterned device the azimuth angle of emission is essentially irrelevant asall properties should be circularly symmetrical about any chosen point. This is notnecessarily the case for a patterned device, where we need to consider both theazimuthal direction of emission with respect to the symmetry directions of the surfacepatterning as well as the transverse angle of emission (waveguide mode angle).

In Fig. 10.11, the curved solid lines (dispersion curves) represent the confinedmodes of the LED and indicate the permitted energy states for emitted light(y-axis). There should be one dispersion curve associated with each and every confinedwaveguide mode of the LED epistructure. For a typical 4.5-mm-thick LED emitting at450 nm, we would expect around 50 trapped modes in A GaN layer (note Fig. 10.11(aand c) are just representative and only a few modes are shown). Each of the trappedmodes has an associated angle of propagation (j as illustrated in Fig. 10.11(e)). Lightemitted from a radiative dipole positioned on the quantum well at an angle above thecritical angle for TIR will couple into the mode with the closest matching mode angle(j) at the wavelength of emission (in effect the k-vector of the emitted dipoles willcouple to the most closely matched k-vector of the available confined modes). Thedashed line running at an angle (called the light line) represents the condition for totalinternal reflection. Modes lying diagonally below the light line will be confined withinthe LED GaN layer by TIR, whereas those lying in the shaded region diagonally abovethe light line are free to radiate to free space.

In reality LEDs emit over a narrow range of wavelengths (about 25 nm). OnFig. 10.11, this is approximated by a broad horizontal (dashed) sample wavelengthline (dashed blue line). Points of intersection between the sample wavelength line


and the dispersion curves are the permitted modes of emission for the LED structure.From Fig. 10.11(a) we see that for an unpatterned LED, the majority of the modes liebelow the light line and so are strongly trapped within the GaN slab.

Fig. 10.11(c) shows the equivalent situation for the case where periodic patterningis applied to the top surface of the LED. This causes the dispersion curves to becomefolded at the Brillouin zone boundary (represented by the vertical dashed line inFig. 10.11(c)). In this case the sample wavelength line (dashed horizontal blue line)then intersects not only a number of trapped modes lying below the light line, butalso a set of new band-folded photonic crystal Bloch modes lying above the lightline. In this case scattering events within the structure (originating from multiple reflec-tions of photons from the dielectric interfaces of the holes etched into the surface)allow emitted photons to switch freely between confined slab modes and radiatingleaky photonic crystal Bloch modes. Hence by etching a photonic crystal somedistance into the top surface of a conventional LED structure, an engineered leakagemechanism can be introduced into the LED, which can significantly improve extrac-tion efficiency.

For a surface-patterned LED, there is no longer any direct relation between the orig-inal direction of photon emission from a dipole on a quantum well, and the final direc-tion of radiation from the top surface of the LED, whereas for an unroughened (orunpatterned) smooth-surface LED the original direction of emission from the dipolecan be deduced directly from the angle of emission from the surface.

Geometric factors such as lattice shape, hole size and the dielectric constants of thematerials used play an important role in determining the efficiency and directionality ofthe resultant leaky modes. Other key factors include etch depth and lattice symmetry.

10.3.1 Effect of etch depth

Etch depth plays an important role in determining the efficiency of cross-couplingbetween trapped waveguide modes associated with a thick LED slab structure andleaky Bloch modes associated with the 2D surface patterning. Typical P-side upLEDs support over 50 trapped slab waveguide modes. For maximum extraction effi-ciency, there must be strong coupling between trapped slab waveguide modes andleaky PC Bloch modes. Shallow etched PC structures only allow cross-couplingbetween a few higher-order trapped slab modes whereas PCs etched within close prox-imity to the MQWs allow extraction of more trapped slab modes. For maximum lightextraction, the photonic crystal should ideally penetrate through the quantum wells andthrough the entire thickness of the epitaxial structure. However, this is not possible inpractice since holes that puncture the quantum wells create electrical short circuits orcurrent leakage paths. For P-side up LEDs, the top P-layer is usually very thin (just afew hundred nanometres), which means that the patterned holes can only go a smallpercentage distance through the thickness of the epitaxial layers. Top side patterningcan only interact strongly with a few trapped modes, and weakly with a small numberof others. Simple top side patterning (or roughening) for P-side up LEDs yields limitedgains in terms of improvement in overall light extraction efficiency.


The situation is better for N-side up LEDs, as the epitaxial layers are usually muchthinner thus supporting fewer modes to start with. Furthermore, the top N-GaN layer isrelatively thick meaning that the quantum wells are positioned further below thesurface of the LED. The PC patterning can penetrate a larger percentage of the distanceto the MQWs, in which case it interacts with a larger number of the trapped modes.

10.4 PC-enhanced light extraction in P-side up LEDs

10.4.1 Fabrication of P-side up LEDs

One of the key advantages of photonic crystal light extraction technology is that it canbe applied to a standard LED chip fabrication process. That is, photonic crystals can beapplied to most types of commercial LED epitaxy to provide the benefits of improvedlight extraction efficiency and production yield, with small changes in productionprocedure.

Key requirements for PC-LEDs are that the epitaxy should produce a smooth sur-face and a low defect density. A high defect density is undesirable as it reduces internalquantum efficiency (IQE); however, it is actually beneficial in standard LED produc-tion where epitaxial or chemical surface roughening is used to improve light extractionsince epitaxial defects seed surface roughness features. The additional process stepsrequired for PC-LED fabrication are epitaxial process tuning to obtain a smoothsurface and low defect density, surface patterning and etching.

Starting from pre-grown epitaxial GaN on a sapphire substrate with a smooth(rather than rough) surface, a thin (50e100 nm) SiO2 layer is deposited by plasma-enhanced chemical vapour deposition (PECVD). This is later used as a hard maskfor PC etching.

10.4.1.1 PC patterning options

For R&D prototyping and design development, a photonic crystal can be patterned bydirect-write electron-beam lithography. In production, patterning is performed bynano-imprint lithography.

10.4.1.2 Ultraviolet (UV) lithography

In principle, deep UV (DUV) lithography could be an alternative patterning process;however, it is problematic in practice as even the highest quality GaN substrates havesignificant surface roughness due to drop-down particulates becoming embedded inthe layers during epitaxial growth and have significant wafer bow due to the straininduced during growth. Both of these factors pose a serious problem for DUV lithog-raphy as these systems have very limited depth of focus at the required dimensions,making it extremely difficult to get high yield across a full 200 (or larger) wafer. In addi-tion DUV lithography is primarily intended for silicon-patterning and requires a care-fully designed anti-reflective coating to achieve maximum resolution. Silicon is highly


reflective to DUV light and has a very smooth surface. Hence a simple anti-reflectivecoating consisting of a quarter-wave stack is sufficient. GaN and sapphire, on the otherhand, are both quite transparent to UV and so exposure light interacts strongly with theunderlying substrate and epitaxial layers creating multiple back-reflections, blurringthe edges of the features. This greatly reduces the fidelity of the surface-patterned fea-tures. This problem could be reduced by use of an anti-reflective coating but this mayrequire multiple layers to make any significant difference and so is far harder to design.Furthermore, accurately applying a removable multi-layer coating to a rough andbowed GaN surface is very difficult.

10.4.1.3 Nano-imprint lithography (UV-NIL)

Currently for production, nano-imprint lithography (NIL) is the preferred solution. Asthere are many other texts dedicated to this subject, NIL will not be described in fulldetail here, but a brief description follows. In NIL, patterned features are transferredfrom a pre-fabricated stamp by a physical print process. The pattern is transferred toa thin polymer resist-like layer, which is applied to the top surface of the wafer. Asthis is a direct transfer method, the stamp is required to have features produced on a1:1 basis with the final required design. The two primary methods of NIL are thermaland UV NIL.

10.4.1.4 Thermal nano-imprint lithography

In thermal imprinting, the substrate is heated to the glass transition temperature of thepolymer layer whilst applying high pressure to a hard imprint stamp. The combinationof heat, pressure and capillary action force the polymer into the features of the stamp.The substrate then cools down and the pressure is released solidifying the polymerlayer. Due to the thermal mismatch of the substrate, stamp and polymer materialsthe final imprint may not have the same dimensions as the original stamp, and stamplife expectancy is low due to the stresses involved in releasing the stamp from themould. Thermal imprinting is very problematic for photonic crystal structures due tothe small dimensions of the structures and the high physical forces and temperaturesrequired. Also the yield is extremely poor as an inflexible hard stamp must be used,which does not work well in combination with the rough surface topology due toembedded drop-down surface particulates typical for GaN wafers.

10.4.1.5 UV-nano-imprint lithography (UV-NIL)

UV-NIL imprints into a semi-liquid polymer layer, which flows easily around the fea-tures of the stamp by capillary action. The polymer is instantly hardened by applyingUV light after application of the stamp to the surface under reduced pressure comparedto thermal imprinting. The stamp is then withdrawn leaving a patterned surface.UV-NIL requires either a transparent stamp or an intermediate hard-mask layer. Anintermediate hard mask is preferred since it can be a flexible membrane, and theprocess is more immune to wafer bow or surface particulates, increasing the yieldper wafer to acceptable levels for production.


10.4.1.6 Electron-beam lithography

Electron-beam lithography (EBL) is the preferred patterning method for product devel-opment and is also the preferred method for producing the stamps used for nano-imprint lithography. In EBL, a resist layer is directly patterned by scanning with anelectron beam electronically. Modern EBL systems have very good depth of focus(several hundred nanometres) and are able to correct for large-scale height variationsof the wafer (of several hundred microns), and so are able to cope well with the roughsurface topology of typical GaN wafers and associated wafer bow. EBL also has theadvantage of allowing multiple designs to be fabricated together on one wafer. EBLis, however, a slow and expensive process, which is not practical for production. Sub-strate charging and proximity error effects must be taken into account to get good qual-ity devices. Charging effects can be overcome by application of a sub-nanoscaleremovable conductive layer on top of the resist. Proximity error correction effectsare overcome using specialised design correction software.

10.4.1.7 Photonic crystal etching

After pattering, the top SiO2 hard mask is etched using inductively couple plasma(ICP) reactive ion etching (RIE) using a CHF3/Ar gas mixture. The GaN layers arethen etched to the required depthdbeing very careful not to penetrate through thequantum wellsdusing ICP-RIE with a Cl2/O2 gas mixture. The SiO 2 hard mask layeris removed by a simple wet chemical etch in a buffered oxide etchant (dilute hydroflu-oric acid). Following removal of the sacrificial hard mask layer, the process resumes asper the regular production cycle.

10.4.1.8 Current spreading layer

Standard P-side up GaN on sapphire LED devices require efficient current spreading asthe top P-doped GaN has low conductivity. This is usually an ITO coating with atypical thickness of 200e300 nm. The ITO current spreading layer (CSL) is depositedon top of the patterned LED either by reactive sputtering or electron gun evaporation.During deposition, oxygen is introduced into the chamber and the substrate is heated toaround 150e200�C. Alternatively (mainly for R&D) a thin (10 nm thick) nickel-goldbi-layer can be used instead of the ITO. This has the advantage that under highly direc-tional deposition conditions the metal does not completely fill up the holes of the pho-tonic crystal or coat the internal sidewalls of the holes and so preserves the highrefractive index contrast required for efficient operation of a photonic crystal. Afterdeposition of the current spreading layer, the devices are annealed in oxygen to in-crease the transparency of the current spreading layer further.

10.4.1.9 MESAs and contacts

The remainder of the chip process proceeds as for a normal P-side up LED. To isolateneighbouring chips electrically and to expose the underlying N-GaN layer in the regionof the contact pad, MESA trenches are defined by optical contact lithography and


ICP-RIE. The devices are then encapsulated in a thin SiO2 overcoat layer using a com-bination of PECVD, optical contact lithography and wet chemical etching to definewindows for the N- and P-side contacts. The SiO2 encapsulation layer is mainlyrequired to protect the ITO current spreading layer from the ingress of air as it degradesin moisture and air.

N and P contact stacks are then deposited by a metal lift-off process consisting ofbi-layer resist patterning followed by directional e-gun evaporation and metal lift-off.The entire wafer is then backside polished by CMP and a metal mirror may be appliedto the rear surface of the sapphire layer by e-gun evaporation. Finally the chip dice aredefined by laser dicing.

Fig. 10.12 shows an unencapsulated P-side up photonic quasi-crystal LED atvarious stages of fabrication. Fig. 10.12(cee) shows the top view of the device atdifferent magnifications both before and after deposition of contact stacks and wirebonding. Fig. 10.12(f) is a cross section through an etched photonic crystal LED. Itshows relatively vertical sidewalls with an angle of 5e10 degrees.

10.5 Modelling PC-LEDs

As LEDs are highly multi-mode incoherent optical systems, accurate modelling oflight emission from photonic crystal LEDs is extremely tricky. Most conventionalmodelling tools for analysis of photonic crystals (such as plane wave and conventionalfinite difference time domain (FDTD) methods) are not easily able to take either ofthese factors into account. FDTD methods can, however, be adapted to investigatethe emission properties of PC-LEDs. Properly adapted FDTD simulations can returnuseful and accurate information about the far-field beam profile, angular power distri-bution, and internal beam-steering efficiency of a photonic crystal. It is well beyond thescope of this chapter to give a full review of modelling methods or how they work (thisis a very broad and complex subject), but in the next few pages I aim to give a veryrough overview discussing some of the key problems involved.

10.5.1 Finite difference time domain (FDTD)simulation methods

FDTD is a very powerful simulation tool but requires detailed understanding of how itactually works for it to be used effectively and to obtain meaningful results. In theFDTD simulation method, an initial electro-magnetic field is stepwise propagatedacross a simulation space, which has been divided into a grid. The grid is initiallyset up to include a spatial representation of the dielectric function of the componentsto be analysed. Maxwell’s equations for the E and H fields are discretised across thegrid and the boundary conditions are solved at the interfaces between each grid cellon each iteration of the simulation. Each iteration of the simulation corresponds to asmall time step, and so initial waves gradually propagate across the simulation spaceas the simulation progresses.


(a)

(b)

(f)

(c)

(d)

(e)

P-GaN

N-GaNQWs

Sapphire

PQC patternMetal contact

Figure 10.12 Fabricated PQC LED. (a) Diagram of the PQC LED, (b) photograph of a fully packaged LED, (c) SEM of a packaged LED, (d) SEM ofundiced LEDs before mesa-etching and contact deposition, (e) high magnification SEMs of the PQC pattern, (f) cross section of the PQC pattern.

Photonic

crystalnitride

LEDs

351

FDTD calculations require both spatial and time-dependent derivatives of the fieldfunctions at the grid cell boundaries. Spatial derivatives are easily computed by sam-pling the field across two or more consecutive grid points and applying simple differ-encing equations. To calculate the temporal derivatives, snapshots of the fielddistribution across the entire simulation grid are kept from one time step to the nextto allow time-dependent information to be passed forward through the calculation.Better accuracy can be obtained by using higher-order differencing schemes, whichrequire even more grid snapshots to be kept. Whilst it is relatively easy to implementhigh-order spatial derivative calculation schemes, there is a large computational over-head in terms of the size of the data-set that must be kept to implement a high-ordertemporal derivative calculation, as many snapshots of the computational grid mustbe kept in memory. High-order temporal differencing schemes also require complexleapfrogging differencing schemes, which are again tricky to implement.

Boundary conditions must be applied at the edges of the simulation space as appro-priate for the type of simulation and position within the device. For example, perfectlymatched layer (PML) absorbing boundary conditions are needed for parts of the devicewhere light radiates to free space (for example, at the top surface of an LED or the exitfacet of a waveguide). Periodic boundary conditions are required at the sides if thestructure is periodic. Provided the structure is symmetric, reflective boundary condi-tions can be used in conjunction with periodic or PML boundary conditions to reducethe overall simulation space.

Virtual sampling planes are placed within the simulation space. A set of consecutivegrid cells (arranged in a line) are sampled periodically (every few simulation timesteps). By applying a computational near-to-far-field transform to the field sampledacross the sample plane, the angle of propagation and intensity of a wavefront canbe derived. In fact, scattered beams propagating at several different angles can beresolved simultaneously.

10.5.2 Limitations of FDTD for modelling PC/PQC LEDs

10.5.2.1 Light launch considerations

Conventional FDTD simulation tools are designed to analyse coherent optical systemssuch as laser and optical waveguide devices, which support very few modes and areexcited by coherent monochromatic light sources. Even if these devices have multi-modes, the number of modes is usually very small (<10) and the correct light launchconditions can be accurately represented by carefully positioning a single dipoleemitter or a single plane-wave launch field. This rapidly couples coherently to theavailable modes allowing interference between them. For lasers and optical wave-guides this is representative of the real situation.

LEDs on the other hand are very highly multi-mode systems (often supporting over50 modes). The light launch is far more complicated as emission occurs simulta-neously across the entire area of several different quantum wells. The emission pro-cesses in an LED are also non-coherent, and so light launch conditions cannot berepresented by a single dipole or plane-wave emitter. To model LEDs correctly, a


vast number (ideally an infinite number) of dipole emitters must be incorporated in thecalculation. These should be distributed randomly across a number of spatial planeswithin the structure corresponding to the positions of each quantum well, and thedipoles should emit randomly in time.

10.5.2.2 Boundary condition considerations

Conventional FDTD software is normally used to model devices that naturally confinethe light to a region of space in very close proximity to the boundaries of the device.Consider a simple rib waveguide (Fig. 10.13). An evanescent wave propagates exactlyperpendicular to the waveguide surface decaying exponentially with distance from thewaveguide walls. Light that escapes from the sidewalls of the waveguide has a verywell controlled and narrow range of angles (exactly 0 degree). These are ideal condi-tions for the correct operation of the perfectly matched boundary conditions at the sidesof the simulation space.

In fact, virtually no field energy actually impinges upon the simulation space sideboundaries at wavelengths of 400 nm and 700 nm, so the type of boundary conditionis almost irrelevant in this case up to the position of the exit from the waveguide. Inother words, the side boundaries actually do virtually no work at all. At the end of thewaveguide, things become more complicated as waves become diffracted and sodiverge over a range of angles. They subsequently impinge upon the top boundaryover a limited range of angles (as shown in the far-field angular plot in Fig. 10.13).Very little field energy impinges on the corners of the simulation space and the topboundary does not need to be particularly well tuned or non-directional for reasonableresults to be obtained over the narrow angular range of projection from the guidefacet. Usually when modelling conventional integrated optical devices, the collectionplane is placed somewhere within the confined region of the waveguide, in which casethe top boundary does not need to be very good at absorbing waves incident atshallow angles.

The situation is very different for LEDs. In this case the aim is to examine theangular scattering properties of the surface patterning to free space accurately. In orderto model the angular properties of light emission from an LED we must rely onperfectly matched boundary layers to absorb the radiated light. In this case, thePML boundary conditions need to be extremely well tuned to ensure they do not reflectany power back into the simulation space at any angle, and they must work well up tolarge glancing angles. PML boundary conditions normally work very well for near-normal wave incidence. However, even well-tuned PMLs become imperfect absorbersfor waves incident at large glancing angles (e.g., travelling nearly parallel to thesurface). In particular problems arise at the corners where the PML side and top bound-aries meet.

10.5.2.3 Considerations for collection plane placement

For accurate analysis of far-field angular emission profiles, great care must be takenin positioning and calibrating the sampling (collection) planes to ensure that


–1

1

2

3

4

5

6

7

8

–1

1

2

3

4

5

6

7

8

1

0

2

3

4

5

6

7

8

–6 –4 –2 0 2 4 6 –1.0

–6 –4 –2 0 2 4 6 –1.0

–6 –4 –2 0 2 4 6 –1.0

1.0 1.0

1.0

Wavelength 400 nm Wavelength 700 nm

Wavelength 1500 nmx (μm)

z (μ

m)

z (μ

m)

x (μm)

X (μm)

400 nm

700 nm

1550 nm

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Inte

nsity

(a.u

.)

–80 –60 –40 –20 0 20 40 60 80Angle (degrees)

Figure 10.13 FDTD simulation for a rib waveguide with PML boundaries at all edges of the simulation space for wavelengths of 400 nm, 700 nm and1500 nm. The graph shows the corresponding far-field beam scattering angles.

354Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

spurious errors are not introduced due to the imperfect behaviour of the boundaryconditions. As even small amplitude back-reflected waves can have an adverseeffect on the simulation results (particularly for a far-field profile) it is extremelyimportant that the simulation space is very well set up and the boundary conditionsfinely tuned.

For example, there will be a partially bound evanescent wave for light incidentclose to the critical angle for total internal reflection to a high/low index dielectricinterface. If the sampling plane is placed too close to the top surface of the LED, itwill continuously capture part of this bound wave giving spurious results. On the otherhand, if the sampling plane is placed too far away from the surface of the LED then itbecomes incapable of receiving light travelling at shallow glancing angles and inca-pable of resolving the large angular spread of light at steep angles giving false results.Going back to the examples shown earlier in Fig. 10.13, narrow sampling planes areplaced close to the exit of (or even within) the waveguide, in which case they are wellaway from the corners of the simulation space and are unlikely to sample back-reflected waves from the corners between boundaries.

10.5.2.4 Requirements for the extent of a cross-sectional profile

To take into account resonant interactions, which occur via light recycling within anLED, a full cross-sectional profile of the LED heterostructure must be incorporatedinto the simulation space. Hence the full epitaxial layer structure (including the under-lying substrate) should form the basis for the simulation. The FDTD simulation mustincorporate a broad emission bandwidth and emission from multiple quantum wells.The latter requires either very long computational times or distributed or clustercomputing techniques.

10.5.2.5 Comparative normalisation of results

Most of the problems associated with the fine calibration of the simulation space andboundaries can be overcome by making comparisons (or normalisations) between sim-ulations for a completely empty simulation space, an unpatterned smooth-surface LEDand a surface-patterned (PC) LED. This provides a good sanity test by confirming thatthe simulation set-up is good, and allows small errors to be deconvolved from the finalsimulation results.

10.5.2.6 Modeling tools

As no suitable commercial modeling tools existed when the field of PC/PQC LED lightextraction emerged, we (the authors of this chapter) developed our own FDTD soft-ware for modeling LEDs. This software takes all of the factors above into accountand the simulations make a realistic representation of LEDs. The software runs on amulti-processor system in a Microsoft high-performance computing cluster (MSHPC) environment. All results presented in the following sections of this chapter arisefrom these calculations.


10.5.3 Unpatterned P-side up LED simulation example

By way of a worked example we now look in detail at a simple sanity test simulationfor an unpackaged, smooth-surface P-side up LED. Fig. 10.14 is a schematic cross sec-tion through the device (not to scale) and Fig. 10.15 shows the corresponding simula-tion results. As discussed in the previous section, to derive the angular far-field profileaccurately and deconvolve any errors or inefficiencies associated with the boundaryconditions and corner points of the simulation space, comparisons are made betweensimulations for an unpatterned LED and simulations for a blank (air-filled) simulationspace, with the collection (sampling) plane placed in the same position for both. In thiscase, the empty air-filled simulation provides a reference.

10.5.3.1 Simulation set-up

We recall (from the previous section) that the collection (observation) plane must beplaced a small distance above the top surface of the LED, but ideally should not inad-vertently oversample evanescent waves associated with partly trapped modes. Weneed to be even more careful about the position of the light injection (launch) planefor the two simulations as the correct relative positions are not actually the same.For the LED there are multiple launch (light injection) planes corresponding to the po-sition of each quantum well layer positioned just below the surface of the LED. In thissimulation, dipoles are placed along the positions of the individual quantum wells. Thecollection plane, however, observes what goes on at the top surface of the LED since itacts as the radiant surface as far as the external observer is concerned (in this case thecollection plane) and not the quantum wells. Hence, the launch plane for the emptyair-filled reference simulation should be at a position equivalent to the top surfaceof the LED. The dipoles are therefore placed along the position of the surface of theLED, not the positions of the multiple quantum wells.

Although this is a small and subtle point it is very important for accurate calibrationof the simulations for the following reason. If for the reference simulation the launchplane were placed at the positions of the quantum wells, the distance between the radi-ative surface (the launch plane for an empty simulation space) and the collection plane

Air

P-GaN (140 nm) n = 2.45

N-GaN (2000 nm) n = 2.44

Sapphire substrate (150 nm) n = 1.8

CBL (30 nm) n = 2.45

MQW (150 nm) n = 2.46

Figure 10.14 Cross section through a P-side up LED showing physical dimensions andrefractive index values. CB, current blocking layer.


would be unrealistically large. The reference simulation would then not mimic over-sampling of waves emitted at glancing angles for the case when the collection planeis very close to the LED surface, whereas with the launch plane at the LED surfaceit would and this allows these effects to be deconvolved from the results. Overall,this strategy simulates the far-field angular profile accurately up to a much larger anglethan would otherwise be the case, since the collection plane can be placed closer to theLED surface (actually within the distance of near-field effects).

Finally by normalizing the results of the LED simulation with the results from thereference simulation, the inherent response of the simulation space (e.g., the angularperformance of the boundary conditions) can be decoupled from the measurement data.

10.5.3.2 Simulation results

The dot-dash curve in Fig. 10.15(a) plots the intensity of the extracted (radiated) lightfrom the top surface of the unpatterned LED as a function of the original dipole launch

20 40 60 80 100 120 140 160

20 40 60 80 100 120 140 160

Angle (degrees)

Launch angle (degrees)

0

2

4

6

8

10

12

Inte

nsity

(a.u

.)To

tal c

olle

cted

pow

er (a

.u.)

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

(a)

(b)

Angular emission (x-axis = collection angle)Angular extraction (x-axis = launch angle)Lambertian reference (x-axis = collection angle)

Figure 10.15 FDTD simulation for an unpatterned, P-side up LED device. (a) Angular in-tensity. (b) Total emitted power as a function of dipole emission angle.


angle within the quantum wells of the LED. Note in this and the following plots, anglesare measured with respect to the interface (rather than the normal to the interface as isusually the case). Hence 90 degrees corresponds to emission normal to the surface ofthe LED. The solid curve in Fig. 10.15(a) shows the corresponding angular intensity oflight actually radiated from the top surface. The results show that the angular intensityfollows a Lambertian profile (dotted curve) as would be expected for an unpatternedLED. Comparing the angular intensity extraction curve (dot-dash curve) over theangular range above the critical angle for total internal reflection (e.g., 65e90 degrees)to the angular far-field intensity curve (over the full angular illumination range(0e90 degrees), we see that the shape and fine features are the same for both. Howev-er, the angular range is different.

As discussed earlier, the FDTD simulations presented in Fig. 10.15(a) confirm thatthere is a direct relation between the original angle of emission from dipoles in thequantum wells and the final angle of emission from the top surface of the LED, cor-responding to Snell’s law of refraction. For an unpatterned smooth-surface LED,Snell’s law of refraction simply applies at the interface between the top surface andair for angles of emission above the critical angle for TIR. In other words, we cansee from the simulation results that there is no diffuse light scattering at the surface,as would be the case for a randomly roughened LED.

Fig. 10.15(b) shows the corresponding total collected power as observed by thecollection plane, but integrated across all collection angles as a function of originalangle of emission (launch angle) from within the quantum wells. The power extractioncurve gives a clear picture of how much light remains trapped inside the LED, and pro-vides an insight into the light extraction mechanisms at work. From the power extrac-tion plot we can clearly see that there is no light extracted for launch angles lyingbelow the critical angle for total internal reflection (0e65 degrees). For launch anglesabove the critical angle of TIR (65e90 degrees) we see a near continuous amount oftotal collected powerdas would be expected from a Lambertian source.

In the final sections of this chapter we shall see that the situation is much morecomplicated for a photonic crystal LED, where there is a large amount of diffuse lightscattering at the surface. The key difference between photonic crystal patterning andrandom surface roughening is the ability of the photonic crystal to provide directionalrather than diffuse scattering of trapped light. This mechanism provides a means toredirect trapped beams to specific useful angles of emission.

10.5.4 P-side up PC-LED performance

So far we have looked at the physical principles utilised by PC patterning to improvelight extraction from LEDs, and touched on some of the pitfalls involved in simulatingreal devices, and performed a sanity test simulation for a conventional LED.

We now take a look at what happens if we introduce surface patterning to a P-sideup LED. Fig. 10.16 shows a cross section through a device (this diagram does notshow all of the epitaxial layers, but is simplified to the most important ones). Theonly difference to the previous example is the introduction of a set of holes into thetop surface penetrating some distance through the top P-GaN layer but stopping short


of the quantum wells and current blocking layer (CBL). (Fabrication of this type ofdevice was described earlier and images are shown in Fig. 10.12.) Fig. 10.17(aec)shows the calculated (solid traces) and experimentally measured (dot-dash traces)angular far-field intensity profiles of radiated light as polar plots. (The same methodsas described above are used for the simulation.)

In this case the surface patterning consists of a 12-fold symmetry photonic quasi-crystal lattice, based on square triangular tiling. Results are shown for three differentlattice constants (450, 550 and 750 nm), but the same etch depth and air-filling fractionof 0.22. This means that in each case the holes have the same diameter relative to thelattice pitch, and so results for different lattice pitch are directly comparable to eachother in these plots.

The oscillations on the simulation curves are due to the limited number of dipolelaunch angles used for the simulation (5 degrees angular steps were used in thiscase). These oscillations would diminish if more dipole launch angles were used butat the expense of longer calculation time. It is the average envelope of the simulationcurves which is of interest.

The legends on each sub-figure of Fig. 15.17(aec) show the expected improvementin total extracted power compared to a smooth-surface LED. Firstly there is very goodagreement between the measured and simulated directional far-field emission profiles.Comparing plots for the different lattice pitches, we see that there is a small but definitechange in angular emission behavior. The 450-nm pitch gives a slightly narrower beamthan is the case for the 750-nm, pitch which has a divergent beam. More importantlythere is a large improvement in predicted total power extraction for each of these de-signs compared to a reference unroughened LED. The extraction improvement variesfrom þ18% for the 750-nm pitch device to þ60% for the 450-nm pitch device, asshown by the legends.

Fig. 10.17(d) shows the actual power extraction enhancement for a set of real LEDdevices measured experimentally using an integrating sphere set-up. The experimentaldata is in line with the theoretical predictions. From the experimental data we see thatextraction enhancement of up to 50% is easily possible for a P-side up PC patternedLED. An equivalent conventionally roughened P-side up LED, taken from the top

Air

P-GaN (140 nm) n = 2.45

Air holes (n = 1)

N-GaN (2000 nm) n = 2.44

Sapphire substrate (150 nm) n = 1.8

CBL (30 nm) n = 2.45

MQW (150 nm) n = 2.46

Figure 10.16 Cross section through a surface-patterned P-side up LED showing physical di-mensions and refractive index values.


0.2

0.40.60.8

1

180

150

12090

60

30

0

Simulation (extraction improvement = 60%)Experiment



330

300270

240

210

0.2

0.40.60.8

1

180

150

12090

60

30

0

330

300270

240

210

0.2

0.40.60.8

1

180

150

12090

60

30

0

300270

240

450 nm

550 nm 750 nm

0

10

20

30

40

50

0.15 0.20 0.25 0.30Air fraction

Tota

l ext

ract

ion

enha

ncem

ent (

%)

(a)

(c)

(b)

d)

Figure 10.17 Comparison between experimentally measured and simulated angular far-field projection for PQC LED devices: (a) 750 nm pitch, (b)450 nm pitch, (c) 550 nm pitch (130 nm etch depth for all). (d) Experimentally measured power extraction improvement values for surface-patternedP-side up LED.

360Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

power bin, would be expected to show about 30% improvement in power extractionefficiency compared to an unroughened LED (remember this is from the top powerbin). So photonic crystal patterning provides a definite improvement (20%) in powerextraction compared to top bin conventional surface-roughened devices. More impor-tantly, perhaps, there is far more control of PC surface roughening across the full areaof the epitaxial wafer and so all PC devices should fall into the top power bin giving abenefit in terms of improved product yield per wafer.

10.5.5 Effect of etch depth on extracted power

Fig. 10.18(a) shows the simulated far-field emission intensity for P-side up 300-nmand 400-nm pitch 12-fold symmetric PQC LEDs with etch depths of 110 and130 nm. We see that there is very little change in far-field directionality with etch depth(the dashed and solid traces overlap) but there is a definite increase in total extractedpower for both 300- and 400-nm lattice pitch devices (5% and 10% respectively) forjust a 20-nm increase in etch depth (as shown by the legend).

10.5.6 Beam-steering effects

Fig. 10.18(b) shows the corresponding angular power extraction as a function of orig-inal emission direction from the quantum wells. Power extraction from an unpatternedreference device is also shown (dot-dash grey trace). As before the critical angle forTIR is clearly visible for the reference device at 65 degrees.

By analysing the angular power extraction behaviour of emitting dipoles within thequantum wells of an LED (as shown in Fig. 10.18(b)), we can gain a clearer under-standing of how the photonic crystal patterning works in terms of the underlying phys-ical mechanisms. There are clear differences in angular power extraction between thePC devices and the unpatterned reference device. We see that in this particular case thePQC surface patterning actually reduces power extraction over the angular range 65e90 degrees, which is the normal light extraction range for an unpatterned LED, provingthat the surface patterning interacts with light radiated within the normal extractionconedin this case in a detrimental way. However, the PC also scatters power radiatedat angles far below the critical angle (30e40 degrees) in a beneficial way to radiativemodes, so overall we observe a net increase in total extracted power from the LED.

Looking more closely at the plots (Fig. 10.18(b)) for the 300-nm lattice pitch (110and 130 nm etch depth), we see that like the unpatterned reference device there is asharp drop in extracted power for dipole emission at the critical angle, but in thiscase instead of dropping straight to zero (as is the case for the unpatterned LED) powerextraction drops to about 30%, and continues to reduce linearly to around 30 degrees(note the gradual slope of the emission power curve). Hence, a small amount of addi-tional power is extracted at angles far below the critical angle. This tells us that thesurface patterning interacts with a very large number of higher-order trapped modes(as shown earlier in Fig. 10.11), but not very efficiently, so overall we obtain a smallimprovement in total power extraction efficiency (10% to 15%). Looking back at theangular intensity profile (Fig. 10.18(a)), the emission profile is Lambertian.


The 400-nm pitch device on the other hand extracts significantly more power overthe range 40e60 degrees (Fig. 10.18(b)) peaking at 100% extraction efficiency for the130-nm etch depth at a dipole emission angle of 55 degrees. If the etch depth is reducedto 110-nm the height of this new power extraction peak decreases, but the peak wave-length remains the same. This tells us that the extra power extracted by making a small(20 nm) increase in etch depth originates from a stronger interaction between the sur-face patterning and the same (previously trapped) mode. If the etch depth were

0.2

0.4

0.6

0.8

1

180

150

12090

60

30

0

400 nm pitch, 130 nm depth, improvement = 57%








Lambertian reference

Unpatterned LED (reference)

20 40 60 80 100 120 140 160Launch angle

0

0.2

0.4

0.6

0.8

1

Tota

l em

itted

pow

er (a

.u.)

(a)

(b)

Figure 10.18 (a) Angular emission profile of radiated beam, (b) total emitted power as afunction of dipole emission angle for 300 nm and 400 nm pitch 12-fold symmetric PQC P-sideup LEDs with etch depths of 110 nm and 130 nm (as shown in the legend).


increased further (which would in this case penetrate our quantum wells and so is notsensible in practice), then the surface patterning may interact with an additional higher-order trapped mode, but we would expect the shape and/or wavelength of the peak tochange. Looking at the angular intensity profile (Fig. 10.18(a)) we also find that thedevice has a substantially reduced angular emission cone. Overall, although lightextraction is greatly improved by the surface patterning compared to an unpatterneddevice, PC surface patterning is still very inefficient at extracting light radiated fromthe quantum wells at very shallow angles (0e30 degrees).

Going back to the highly multi-mode waveguide analogy made earlier (presented inFig. 10.11), this shows that the shallow surface-patterned PC only interacts with a fewtrapped modes, and in fact the surface patterning interacts most strongly with thehigher-order modes. In an ideal case, the surface pattering would scatter light equallyefficiently over the entire angular emission range (0e90 degrees). In practice this isvery difficult to achieve.

An analysis of the dipole angle power extraction behavior therefore allows us togain a strong physical insight into the mechanisms altering the far-field profile andimproving power extraction, and so provides a powerful tool when analyzingsurface-patterned LEDs.

10.5.7 P-side up PQC LED electro-optical performance

Fig. 10.19 shows actual measured electro-optical performance data comparing anunpatterned smooth-surface P-side up LED to a 12-fold symmetric PQC patternedP- side up LED. Both devices were fabricated from neighbouring areas of the samewafer to eliminate inaccuracies due to variation in IQE (which can change dramaticallyacross the area of 200 of GaN on a sapphire wafer).

Fig. 10.19(a) is a schematic diagram of the experimental set-up used to measure thefar-field angular beam profiles. Fig. 10.19(b) shows the measured far-field profiles forthe PQC patterned LED and the expected profile for a Lambertian emitter. The fullwidth at half maximum of the unpatterned device is 120 degrees compared to 60 de-grees for the PQC LED, demonstrating the beam-shaping capability of the technology.Fig. 10.19(c) compares the emission spectra of the patterned and unpatterned devices.Both show a centre wavelength of 450 nm and very similar spectrum confirming thatthe PQC does not degrade the LED’s spectral performance. Fig. 10.19(d) compares theon-axis power versus electrical input power for the patterned and unpatterned LEDs.The PQC patterning increases the on-axis extraction power by over 20%.

Fig. 10.19(e) shows the IeV characteristic. The LED was pulsed to avoid heating,which would otherwise distort the results. There is a relatively high forward resistancefor both the unpatterned LED and the PQC LED due to the thin nickel-gold currentspreading layer applied to the surface of these test devices (ITO was not used as thetransparent current spreading layer to preserve the refractive index contrast betweenthe GaN and the air inside the holes). Nonetheless the IeV curve confirms that theinclusion of the PQC into the GaN layer does not adversely degrade the electricalcharacteristics.


Unpatterned LEDPQC LED

Unpatterned LEDPQC LED measurementPQC LED theory



0

5

10

15

20

2 4 6 8Vf (V)

I f (m

A)

0 50 1000

0.2

0.4

0.6

0.8

Inte

nsity

(a.u

.)

Nor

mal

ized

inte

nsity

(a.u

.)

Wavelength (nm)300 350 400 450 500 550 600

0

0.2

0.4

0.6

0.8

1

1.2

Input power (mW)

0.2

0.4

0.6

0.8

1

180

150

12090

60

30

0LED

Rotating arm

Siliconphotodiode

(c) (d)

(e)

(a) (b)

Figure 10.19 Electro-optical performance data comparing an unpatterned smooth-surface P-side up LED to a 12-fold symmetric PQC patterned P- sideup LED. (a) Far-field measurement set-up. (b) Measured and simulated angular beam profiles. (c) Emission spectra. (d) On-axis intensity as a functionof electrical input power (PQC patterning provides an increase in on-axis extraction power of over 20%). (e) IeV characteristic (high forwardresistance is due to a nickel-gold current spreading layer applied to the surface of these test devices). Inset is a zoomed in photograph of the LED inoperation. The glowing LED chip is mounted on a metal can and bond wires can be seen.

364Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

10.5.8 Effect of transparent current spreading and passivationlayers on PQC performance

The final ITO transparent current spreading layer (TCSL) and oxide passivation layershave a dramatic effect on the performance of a PC/PQC LED. For a conventional(unpatterned) LED, the transparency of the transparent contact layer can be improvedby optimising the optical thicknesses of the layers so that they are a quarter wavelengthat the emission wavelength. In this case the TCL forms an anti-reflective coating andreduces the angle-dependent loss due to Fresnel reflections at the top surface of anLED over a wide range of angles within the normal extraction cone.

Applying ITO CSL layers to a PC-LED can, however, reduce the performance ofthe device by around 15%. This is due to the fact that the ITO will fill up the holesof the photonic crystal during deposition, greatly reducing the refractive index contrastbetween the holes and the GaN, which in turn reduces the efficiency of the PQC.

A NieAu CSL on the other hand has virtually no effect on the efficiency of thePQC, even if it penetrates the holes, since its thickness is a small fraction of the wave-length (just a few nanometres, typically <10 nm total). However, a NieAu currentspreading layer has the disadvantage of poor optical transparency. The typical opticaltransmission of NieAu is in the region of 65% at a wavelength of 450 nm while thetransmission of a thick ITO layer can reach 90% at a wavelength of 450 nm.

Assuming the conductivity of ITO is similar to that of NieAu, an ITO CSL canincrease the efficiency by 38% over a NieAu CSL as the slight loss of efficiency inthe PQC due to the ITO is compensated by reduced light absorption. The use ofITO as a current spreading layer is also highly preferable for production.

There are also possibilities for using more exotic materials as the TCL for PC/PQCLEDs, such as carbon nanotubes and graphene, both of which have very high conduc-tivity, but very small layer thickness and low optical absorption (for the required layerthickness) in the visible region.

10.6 PC-enhanced light extraction in N-side up LEDs

10.6.1 Advantages of N-side up over P-side up deviceconfiguration for light extraction

Photonic crystals can also be etched into the top surface of N-side up vertical currentspreading LEDs (power chips). As we shall see, in this case they can improve lightextraction efficiency far more than for a P-side up LED. Fig. 10.20 shows a schematiccross section through an N-side up PC-LED. There are several very important differ-ences in comparison to a conventional P-side up device. There is an underlying metalmirror layer at the interface with the substrate and the quantum wells are placed just acouple of 100 nm from it. For a vertical LED the N-GaN layer can be polished back tocreate a thin GaN device, so that the overall thickness of the epistructure is usually verymuch thinner than is the case for a P-side up LED. The reduced overall device thick-ness greatly reduces the number of trapped optical modes within the epistructure,making the job of extracting light from the remaining trapped modes far easier.


Even after CMP, the top N-GaN layer is still significantly thicker (w2 mm) than thetop P-GaN layer (130 nm) for P-side up LEDs. Consequently the quantum wells areburied further below the top surface of the device. PC holes are etched into the topsurface of the N-GaN layer of an N-side up LED and it is possible to etch thesurface-patterned PC a more significant distance (depth) into the epistructure (82%instead of just 6% of total device thickness). As discussed earlier, this has a big impacton the way in which the surface patterning interacts with trapped modes, and has a bigimpact in increasing the efficiency of light extraction.

10.6.2 Utilization of microcavity effect in N-sideup LEDs to improve power extraction

Another key difference between N- and P-side up LED configurations is the underly-ing metal mirror close to the quantum wells in N-side up devices. This creates a weakmicrocavity effect for light emitted in the downwards direction transverse to the quan-tum well. The thickness of the epistructure and the position of the quantum wells rela-tive to the mirror can be optimised such that the radiative rate of emission from thequantum wells is slightly enhanced by optical feedback between the backside mirrorand the quantum wells.8 In addition, the microcavity effect changes the angular far-field intensity profile of the LED. When properly optimised, the microcavity effectcan direct more energy into the forward emission direction. Fig. 10.21 illustrates sche-matically how this works. Microcavity LEDs can be described by a simple analyticalmodel, which has been validated against experimental measurements on fabricateddevices. Fig. 10.22 shows extraction efficiency as a function of distance betweenthe mirror and the quantum wells, calculated using both rigorous FDTD and simpleanalytical methods (d is the distance between the mirror and the most central quantumwell). In Fig. 10.22, a value of two corresponds to the benchmark figure for a simplethick LED with a backside mirror placed a large distance from the quantum wells (inwhich case microcavity effects do not contribute); a value of one corresponds to thebenchmark figure for an LED without a backside mirror (such as a simple P-side upLED). Both methods show that if properly tuned, a 75% improvement in light extrac-tion can be obtained in comparison to a standard LED with a perfect lossless mirrorplaced a large distance behind the quantum wells.

N-GaN (<2000 nm) n = 2.44

MQW (150 nm) n = 2.46

P-GaN (140 nm) n = 2.45

CBL (30 nm) n = 2.45

Air holes (n = 1)

Mirror Sapphire substrate (150 nm) n = 1.8

Figure 10.20 Cross section through a surface-patterned N-side up LED showing typicalphysical dimensions and refractive index values.


10.6.3 N-side up PC-LED performance

We now take a look at what happens if surface patterning is introduced to an N-side upLED. To determine the performance benefit of PQC patterning, 400 of N-side up GaNon a silicon wafer (provided by Luxtaltek Corp, Taiwan) was patterned with a numberof different 12-fold symmetric PQC designs by direct-write electron-beam lithography(EBL) and reactive ion etching (Southampton University, UK). The lattice pitch and

Mirror

P-GaN

MQW

N-GaN

Forward emitted wave inphase with reflected wave

Forward emitted wave

Backward emitted wave

Wave reflected forwardsfrom rear mirror

Dipole emitter

Constructive interference between forwardtravelling waves increases power output

Figure 10.21 Microcavity effect in an N-side up vertical LED.

Model

FDTD

0.5 0.7 0.9 1.1 1.3 1.5d/λn

0

0.5

1

1.5

2

2.5

3

3.5

4

Inte

nsity

(nor

mal

ized

to L

ED

with

out m

irror

)

Figure 10.22 Comparison of FDTD and analytical models for cavity-enhanced thin GaN LEDdesigns in comparison to conventional thick LEDs.


air-filling fraction were varied across the devices on the wafer. Plate 10(a) (see colorplate section) shows a photograph of the fully patterned 400flip-chip bonded GaN on asilicon wafer after etching to a depth of 600 nm (note there are no metal contacts orMESA isolation trenches at this stage). Under white light illumination the PQC areasshow a variety of colours due to the different lattice pitch and air-filling fraction. Eachsquare patterned area is 3 � 3 mm2 in size. Blank areas are deliberately left betweenthe PQC patterned areas to provide control (reference) chips. It is important that thereference chips are positioned close to the PQC device areas as the IQE can changesignificantly across a wafer. Devices placed in close proximity to each other on thewafer will have very similar IQE and so are more comparable. In the photographwe also see a superimposed square grid pattern (1 mm spacing). This is an artefactof the laser lift-off process used earlier in the fabrication of the devices. Many otherdefects can be seen across the waferdthis is due to imperfections in the laserlift-off and the wafer bonding process. Plate 10(b) (see color plate section) showsSEM images of one of the surface-patterned areas showing the high uniformity ofthe pattern (note that the holes are not intended to be completely circular).

To determine optical light extraction performance, the photoluminescence (PL) wasmeasured. This enables a relative comparison to be made between different devices, andhas the advantage that it can be used at wafer level on devices that are only part fabri-cated and do not have electrical contacts. Wafer-level measurement has a further advan-tage as light emitted from the side edges of the chips is not collected and the effects ofthe PQC can easily be derived by comparison with unpatterned wafer areas.

Light at a shorter wavelength than the LED emission is shone onto the device. Thequantum wells absorb the energy, and then give out light (luminescence) at a newwavelength corresponding to the band gap energy of the quantum wells. The experi-mental configuration used for PL measurement is shown in Fig. 10.23. A collimated405.5-nm laser diode was focused to a wide (250 mm) spot on the LED surface. Inci-dent light was filtered with a bandpass filter with transmission peaks centred at400 nm. Reflected light was filtered with a bandpass filter otherwise the photodetectorbecomes saturated by the very strong pump reflection, and is unable to discriminate theluminescence from the excitation power. The GaN wafer had a peak emission wave-length of 456 nm. Measurements were recorded using a luminance camera placed0.52 m away from the sample giving a collection angle of 0.5 degrees. The camerarotated about the focus to record angular emission spectra.

10.6.3.1 Demonstration of microcavity effect in practicefor an unpatterned N-side up LED

Fig. 10.24 shows FDTD simulations and experimental measurements of the angularemission profile and angular dipole extraction profiles for an unpatterned verticalLED with N-GaN layer thickness of 2 mm. The solid black line shows the simulatedfar-field angular intensity profile into air (in this case the x-axis corresponds to thecollection angle) and the solid grey line shows power extracted to air as a functionof internal launch angle for the device (in this case the x-axis corresponds to the dipoleemission angle within the GaN). The solid grey line shows a clear cut-off in extraction


Sample

∼15 degrees

400 nm, 10 nmbandpass filter

450 nm, 12 nmbandpass filter

Laser diode and collimator,distance to sample 22 cm

Camera – 0.5 degrees collection angle,distance to sample 52 cm,rotation about sample perpendicularto the plane of the figure

Beam block

Figure 10.23 Measurement system for recording photoluminescence from PQC samples.

Angular intensity (experiment PL)Angular intensity (experiment EL)Angular intensity (model)

Power extraction cone (model)Power extraction scaled by refraction at GaN/air interface (model)

–80 –30 20 70Angle (degrees)

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Inte

nsity

(a.u

.)

Figure 10.24 Measured and simulated angular emission (intensity) profiles of the radiatedbeam, and corresponding (computed) dipole extraction angles for an unpatterned vertical LEDwith N-GaN layer thickness of 2 mm. Overlap of rescaled power extraction curve with actualemitted angular intensity curves proves that simple refraction occurs at the surface of thisdevice and there is little diffuse scattering.


for launch angles below �26 degrees or above þ26 degrees corresponding to thecritical angle for total internal reflection and so shows the limit of the extractioncone for an unpatterned device.

Comparing features between the two traces, the angular intensity profile (solidblack line) has two prominent peaks at �66 degrees (�5 degrees) and �36 degrees(�5 degrees) with smaller modulations between �30 degrees. Features in the angularextraction curve (solid grey line) follow exactly the same profile, but over a restrictedangular range of �26 degrees, with prominent peaks at �21.5 degrees (�1 degree)and �13 degrees (�1 degree). This data indicates a direct correspondence betweenthe internal angle of emission and far-field extraction angle, and so emission followsconventional refraction at the air/GaN interface (recall from the previous section thatthis was also the case for an unpatterned P-side up LED). To confirm this, the angularrange for power extraction (solid grey line) was rescaled by the refraction condition atan air/GaN boundary according to Snell’s law and a new trace (dashed black line) issuperimposed onto the plot. This shows the expected angular power emission into air.There is excellent agreement between the solid black line and the dashed black lineconfirming that this is the case. The expected corresponding peak intensity angleswould be at �59 degrees (�5 degrees) and �36 degrees (�5 degrees), which is ingood agreement with the simulated angular intensity profile.

Modulations in the angular intensity (emission) profile for an unpatterned LED arepurely due to weak microcavity resonance effects occurring in the back-plane of theLED between the quantum wells and the rear contact/mirror (as discussed in the pre-vious section). Hence it is possible to modify the emission profile to some degreethrough microcavity effects and careful tuning of the epitaxial layer thickness.

Fig. 10.24 also shows the measured photoluminescence (thin dotted line) and electro-luminescence (EL) (thin dash-dot line) angular far-field profiles for a fabricated referencedevice with N-GaN thickness in the range 1.9e2.6 mm. The experimental photolumines-cence has two smooth peaks at�62 degrees and �28 degrees, whereas the electrolumi-nescence has two smooth peaks slightly shifted to shallower angles of �66 degrees and�32 degrees. These values are in excellent agreementwith the peak angles predicted fromthe angular extraction curve (�66 degrees and�36 degrees). The offset between the ELand PL values can be explained by the large variation in thickness of the N-GaN layeracross the area of the wafer (the CMP process used was not very precise). PL and ELmeasurements were made on different positions across the wafer.

These results confirm that rigorous FDTD simulation correctly predicts: angular in-tensity profile, refractive confinement effects within the underlying epitaxy and powerextraction enhancement for an unpatterned microcavity LED.

10.6.4 Performance improvement for N-side up PQCpatterned LEDs

10.6.4.1 Far-field beam-shaping

Fig. 10.25(a) shows the angle-dependent PL measured under the same excitationconditions for three different devices: an unpatterned control (reference) device and


two 12-fold symmetric PQC patterns with pitch of 300 and 400 nm. There is a very sig-nificant (8�) improvement in on-axis intensity for both 300- and 400-nm pitch PQC de-signs compared to the unpatterned reference device, demonstrating the ability of thePQC to interact with and scatter power efficiently from the trapped modes of the device.

Fig. 10.25(b) shows the PL normalised to the maximum value and compared with aLambertian emission profile. The 300-nm pitch device shows broadening of theangular emission profile, whereas the 400-nm pitch shows narrowing of the emissionprofile. We also see that the unpatterned control device has a far from Lambertianemission profile due to the microcavity effect discussed earlier.

–75 –60 –45 –30 –15 0 15 30 45 60 75

Angle (degrees)

–75 –60 –45 –30 –15 0 15 30 45 60 75

Angle (degrees)

Unpatterned LED300 nm pitch400 nm pitch

Unpatterned LED300 nm pitch400 nm pitchLambertian reference

Inte

nsity

(nor

mal

ized

)In

tens

ity (a

.u.)

(a)

(b)

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Figure 10.25 Angle-dependent photoluminescence for N-side up PQC patterned microcavityLEDs. (a) Raw PL measurement data showing 8� increase in power extraction over a 40degrees angular range for a 300-nm pitch PQC patterned device compared to an unpatternedreference device. (b) Normalised PL measurement data showing changes in angular beamprofile due to PQC patterning. The unpatterned control device has a non-Lambertian emissionprofile due to weak microcavity effects between the quantum wells and rear mirror.


10.6.4.2 Enhanced power extraction

Fig. 10.26 shows the simulated angular dipole extraction profile for a PQC patternedLED device. The dotted trace shows the predicted angular extraction cone for anunpatterned device revealing strong emission over the central angular range abovethe angle for total internal reflection (64 degrees). The other lines show the angularextraction profile for various PQC patterned devices.

In stark contrast to the extraction profile for the unpatterned device the patterneddevice radiates light into air at emission angles well below the critical angle forTIR. In particular there is a new angular range of emission between 30 and 50 degrees,and also for glancing angles below 20 degrees.

As explained in previous sections of this chapter, the angular dipole emission plot(Fig. 10.26) gives a physical insight into the mechanisms that give rise to increasedextraction efficiency. Specifically, it demonstrates that the patterning causes scatteringbetween strongly confined guided modes and radiative modes. For this to occur thephotonic crystal must interact with a number of lower-order trapped modes insidethe LED heterostructure.

This is very different to an average index effect, which could alternatively beinduced by the surface patterning. A reduction in average index of the surface layerwould be expected simply to shift the critical angle for total internal reflection to asmaller angle, thereby increasing extraction efficiency. From Fig. 10.26 we see thatthis is not the case. Looking closely at the angular range between 50 and 66 degreeswe see that extraction is increased a little and over this range the cut-off condition

020

50

100

150

200

250

300

Inte

nsity

(a.u

.)

40 60 80 100 120 140 160Launch angle (degrees)

300 nm depth, improvement = 44%400 nm depth, improvement = 54%500 nm depth, improvement = 95%600 nm depth, improvement = 72%Unpatterned LED (reference)

Figure 10.26 Simulated angular extraction profile for a PQC patterned device with 300 nmpitch for various etch depths.


for TIR is less sharp (now sloped) indicating that there is a partial contribution toimproved extraction efficiency, which can be attributed to a reduction in effectiverefractive index of the surface layers.

We also see that large angular ranges remain where no PC scattering occurs (20e30degrees and 45e50 degrees), hence, for this design a large proportion of emitted radi-ation still remains trapped in the heterostructure. The theoretical modelling indicates amaximum improvement in total power extraction efficiency of 72% (relative to anunpatterned device with a smooth-surface condition and no encapsulation) could beachieved for this structure, assuming that introduction of the PQC does not compro-mise electrical performance, induce further damage to the quantum wells or introducenon-radiative surface states.

10.7 Summary

In this chapter we reviewed factors affecting wall-plug efficiency for LEDs, focusingon issues that affect light extraction from the epitaxial layers. We discussed the possi-bility of using photonic crystals to help improve light extraction and looked in detail atthe physical means by which this can be done. The first half of the chapter wasintended to give an understanding of the physical principles utilized by PC-LEDs.The second half of the chapter discussed the practical implementation of PC-LEDsand levels of performance enhancement attainable for P-and N-side up LEDs.

In Section 10.1, we briefly reviewed the most common configurations for LEDs anddiscussed factors that affect wall-plug efficiency for each type. We used simple radi-ometry to calculate ballpark figures for light-trapping in specific layers of a smooth-surface GaN on sapphire LED, and discussed general methods for improving lightextraction.

In Section 10.2 a general overview of how photonic crystals work was presented.Photonic crystals are very complicated devices to understand. The properties of pho-tonic crystals are normally represented or analyzed using photonic band diagrams.These can be confusing and difficult to understand by the non-physicist. I tried toexplain clearly why these tools are useful and how they can be interpreted and usedeffectively. A very simplistic PC device that supports only one trapped mode wasused as an example to introduce the basic concepts. Some strange properties of photoniccrystals that are utilized in other fields (for example, silicon photonics), such as slowlight and multi-directional diffraction, were mentioned. Photonic quasi-crystals werealso introduced and connections between LED far-field beam profile, lattice patterningand beam symmetry and choice of patterned lattice shape were discussed.

In Section 10.3 we discussed the physical mechanism for enhancing light extractioninvoked by PC patterning, using a more realistic example of a thick P-side up LEDsupporting many trapped modes. Comparisons were made between a thick slab wave-guide and an LED. We then discussed the effect of etch depth and epitaxial layer thick-ness on expected light extraction efficiency.

In Section 10.4 we looked at general issues affecting fabrication and mass produc-tion of PC and PQC LEDs. A step-by-step overview as to how PCs can be incorporated


into P-side up LEDs was given and potential methods of patterning the photonic crys-tal that are scalable for mass production were discussed.

In Section 10.5 we discussed issues in modelling PC-LEDs using the FDTDmethod, limitations of conventional FDTD software and ways of improving simulationaccuracy. In particular LEDs require very complicated light launch conditions, whichare difficult to implement using conventional FDTD tools. Careful fine tuning of PMLboundary conditions and calibration of the overall simulation space response alongwith careful placement of collection planes are absolutely critical for obtaining mean-ingful results. To obtain high accuracy, the full cross-sectional profile of the device,including the substrate, should be included in the simulation space, and comparisonsmade to unpatterned devices to ensure the simulation set-up is good. An examplesimulation for an unpatterned P-side up LED was used to describe more clearly subtleissues in setting up the simulations and to discuss simple internal beam-steeringeffects.

There are already a number of very different device configurations for commercialLEDs. In particular the two main configurations are P-side up lateral current spreadingwith and without a patterned substrate, and N-side up vertical current spreading. Notall configurations for LEDs lend themselves well to incorporation of PCs and perfor-mance benefits can vary greatly dependent on the type of LED.

In Section 10.6, example simulation and experimental measurements for PQCpatterned P-side up LEDs were presented. We showed that an improvement in lightextraction up to around 50% in comparison to a smooth-surface LED with an identicaldesign is easily possible, but it is very difficult to change the far-field LED beam profilesignificantly. We discussed the origin of light extraction improvement in terms of trap-ped modes. Despite the surface patterning, the level of improvement provided by thePC/PQC patterning is relatively small, and a large amount of light still remains trappedwithin a P-side up device.

In Section 10.7, example simulation and experimental measurements for PQCpatterned N-side up LEDs were presented. We showed that improvements in lightextraction well beyond 100% are easily possible, and that there is scope to dramaticallychange the far-field beam profile as well.

The key difference in performance benefits between N-and P-side up patterned de-vices lies in the thickness of the underlying epitaxial layers and the relative percentagedepth to which the patterned holes can be etched.

10.8 Conclusions

We have seen that photonic crystal surface patterning can provide significant improve-ment for light extraction for both N- and P-side up LEDs, with improvements of 50%compared to a smooth-surface P-side up LED easily possible. Conventional LEDs aresurface roughened and so there is considerable device-to-device performance


variation. This introduces the requirement to sort finished devices into power bins. Thevery best P-side up devices taken from the top power bin would have an efficiencyimprovement of up to 35% compared to smooth-surface LEDs. Overall, PCs providean improvement of around 15% over top bin devices. Apart from improved overalllight extraction there are secondary gains from the PC technology when applied toP-side up LEDs. As the nano-imprint process is more reproducible than random rough-ening, PC surface patterning can greatly improve the overall production yield, as alldevices should effectively fall into the top power bin. In addition, fabrication becomesslightly easier as there is no longer a requirement for the final epitaxial overgrowth pro-cess used to create the surface roughening. The overgrowth step takes a considerablelength of time and so reduces overall production throughput for the reactors. Applica-tion of PC patterning to patterned substrate P-side up LEDs provides smaller gains asPSS LEDs already have greatly improved performance over conventional surface-roughened LEDs.

For N-side up LEDs, the performance gains are far greater and improvements inextraction efficiency greater than 100% in comparison to a smooth-surface LED areeasily possible. It is also more practical to change the shape of the emitted beam tosuit a given application. However, the overall fabrication process for the final powerchip is still very complex and costly compared to the simple P-side up configuration.For all types of LEDs improvement in device performance is still relative to the qualityof the initial epitaxial material. PC patterning does nothing to improve the IQE of apoor quality epitaxial wafer.

Although this chapter has specifically discussed photonic crystal surface patterningfor improving light extraction and beam-shaping for LEDs, all the techniques andbackground theory described are equally applicable to solar cells, photodetectorsand OLEDs. In the case of solar cells, PCs can improve overall efficiency byimproving light-trapping. Application of PCs to the surface of photodetectors canimprove sensitivity and directionality of detection. OLEDs have an inherently verythin device layer construction and so are ideal for incorporation of photonic crystals.However, the refractive index of the polymer layers is low compared to GaN and theamount of light trapped within these layers is relatively small, and so the overall gain interms of improvement in light extraction by utilizing a PC is smaller than is the case forconventional LEDs but nonetheless not insignificant.

Photonic crystal light extraction methods have only recently been applied to com-mercial LEDs and currently there are very few commercial companies that have adop-ted the technologydmost notable are Luminus Devices (US) and Luxtaltek Corp(Taiwan). PC-enhanced LEDs are finding their way already into mainstream consumerdevices such as flat-screen televisions. The slow take-up of the technology is in partdue to the need to scale up the patterning technology to mass production, which hasinvolved a huge amount of effort and technical development in nano-imprint lithog-raphy. These processes are now becoming more mainstream. Over the next few yearswe can expect to see greater market penetration for devices utilising photonic crystalsurface-patterning technology.


References

1. Nishida T, Saito H, Kobayashi N. Efficient and high-power AlGaN-based ultraviolet light-emitting diode grown on bulk GaN. Appl Phys Lett 2001;79(6):711e2.

2. Joannopoulos JD, Villleneuve PR, Fan S. Photonic crystals: putting a new twist on light.Nature 1997;386:143.

3. Painter OJ, Husain A, Scherer A, O’Brien JD, Kim I, et al. Room temperature photoniccrystal defect lasers at near-infrared wavelengths in InGaAsP. J Lightwave Technol 1999;17:2082e8.

4. Zoorob ME, Charlton MDB, Parker GJ, Baumberg JJ, Netti MC. Complete photonicbandgaps in 12-fold symmetric quasicrystals. Nature 2000;404:740e3.

5. Rattier M, Benisty H, Schwoob E, Weisbuch C, Krauss TF, et al. Omnidirectional andcompact guided light extraction from Archimedean photonic lattices. App Phys Lett 2003;83:1283.

6. Shubert EF. Light emitting diodes. Cambridge University Press; 2003.7. Schnitzer I, Yablonovitch E, Caneau C, Scherer A. 30% external quantum efficiency from

surface textured, thin-film light-emitting diodes. Appl Phys Lett 1993;63:2174.8. Shen YC, Wierer JJ, Krames MR, Ludowise MJ, Misra MS, et al. Optical cavity effects in

InGaN/GaN quantum-well-heterostructure flip-chip light-emitting diodes. Appl Phys Lett2003;82:14.


Nitride LEDs based on quantumwells and quantum dots 11J. Verma 1,2, S.M. Islam 3, A. Verma 3,4, V. Protasenko 3, D. Jena 3

1University of Notre Dame, Notre Dame, IN, United States; 2Intel Corporation, Hillsboro, OR,United States; 3Cornell University, Ithaca, NY, United States; 4Indian Institute ofTechnology, Kanpur, India

11.1 Light emitting diodes

Light emitting diodes (LEDs) are used in a multitude of applications. Large area dis-plays and solid-state lighting stand to immensely benefit from high efficiency lightemitters. The availability of efficient solid-state light sources over the entire visiblespectrum is also enabling a number of new applications that were previously notpossible. Unlike the inefficient heat-induced radiation in incandescent light bulbs,solid-state light sources are designed to be efficient sources of photons at thequantum-mechanical level. Semiconductors form the basic fabric of LEDs. The oper-ating mechanism of semiconductor LEDs involves the transition of electrons in a con-duction band into an empty state in the valence band. If the process produces photons,it is called radiative recombination and results in useful light output. If the transitionprocess is mediated by emission of phonons, it is nonradiative and results in loss ofuseful energy. The essence of LED design is to maximize the radiative transitionsand reduce the nonradiative ones under the constraints of the semiconductor materialsystem used to realize them. We start the discussion with an introduction to the physicsof electron motion in a p-n junction.

11.1.1 p-n junction diodes

Light emission was first observed from a semiconductor device in 1907.1 Unlike cur-rent LEDs that use p-n junctions, the first semiconductor LEDs were metalesemiconductor Schottky diodes.2 Later p-n homojunction LEDs were found to bemore efficient light emitters than Schottky diodes. A p-n junction is comprised ofan electron-rich n-type layer and an electron-deficient p-type layer in metallurgicalcontact. Chemical doping is typically achieved either during crystal growth or subse-quently by ion implantation or thermally driven diffusion. An n-type semiconductorlayer is formed when a donor atom provides an extra electron to the conductionband. This electron can move around freely through the crystal in response to electricfields or concentration gradients. Similarly, a p-type layer is formed when an acceptoratom captures an electron from the valence band creating a vacancy (hole), which can


http://dx.doi.org/10.1016/B978-0-08-101942-9.00011-3

then move inside the crystal. From basic semiconductor physics the electron and holeconcentrations in a semiconductor in equilibrium are given by

n ¼ NceðEf �EcÞ

kT ; [11.1]

and

p ¼ NveðEv�Ef Þ

kT ; [11.2]

where n and p are the volume concentrations (cm�3) of electrons and holes, respec-tively, Nc is the conduction band edge density of states, Nv is the valence band-edgedensity of states, Ef is the Fermi energy, Ec is the conduction band energy mini-mum, Ev is the valence band energy maximum, k is the Boltzmann constant, and T isthe ambient temperature. When the n- and p-regions are in contact, the free carriers inthe p-n junction diffuse into regions of lower concentration. Holes diffuse from the p-layer into the n-layer and electrons diffuse into the p-type layer. Thus, a depletionregion is created that contains ionized acceptor and donor atoms and is devoid of freecarriers. The carrier diffusion is contained by the electric field that is produced due tothe ionized charge dipole in the depletion region. This field creates a built-in potentialthat presents an energy barrier for carrier motion. Using the above equations alongwith Poisson equation and assuming complete ionization of donor and acceptor atoms(n ¼ ND, p ¼ NA) applying charge neutrality, the built-in potential is obtained as

Vbi ¼ q

2εs

NAND

NA þ NDW2; [11.3]

or

Vbi ¼ kT

qlnðNANDÞn2i

; [11.4]

where Vbi is the built in potential, W is the depletion region width, NA is the acceptorconcentration, ND is the donor concentration, ni is the intrinsic carrier concentration, qis the electron charge, and εs is the semiconductor dielectric constant. Under equi-librium conditions, the electron and hole Fermi levels are same throughout the junc-tion. Application of an external voltage across the junction or optical excitation createsa nonequilibrium situation. Under nonequilibrium conditions, the electron distributionand the hole distributions are considered to be in “equilibrium amongst themselves”,but not in equilibrium with each other. This nonequilibrium feature is quantitativelycaptured by separate quasi-Fermi levels for electrons (Efn) and holes (Efp). The dif-ference of the quasi-Fermi levels is a measure of how far from equilibrium the electron/hole system has been driven, with Efn � Efp ¼ 0 signifying equilibrium. The concept


of quasi-Fermi levels allows the analysis of current transport and the resulting opticalphenomena in p-n junction LEDs.

The electron and hole current components in a p-n junction diode are given by thesum of the drift and diffusion currents:

Jn ¼ qnmnF þ qDndn

dx; [11.5]

Jp ¼ qnmpF � qDpdp

dx; [11.6]

where Jn and Jp represent electron and hole currents, respectively, F is the electric field,Dn is the electron diffusion constant, Dp is the hole diffusion constant, and q is the unitelectron charge. On applying a voltage bias, the minority carrier concentrations at thedepletion region boundaries are changed from their equilibrium values; the newconcentration is given by the Shockley boundary condition:

pnðxnÞ ¼ pnoeqVakT ; [11.7]

and

npðxpÞ ¼ npoeqVakT [11.8]

where pn(xn) is the hole concentration at the depletion region edge xn in the n-typelayer, np(xp) is the hole concentration at the depletion region edge xp in the p-typelayer, pno is the equilibrium hole concentration in n-type region, npo is the equi-librium electron concentration in p-type region, and Va is the applied bias. Theminority carrier concentration near the depletion edge changes exponentially withthe applied bias. The minority carrier concentration decreases exponentially awayfrom the depletion region due to recombination of excess minority carriers withmajority carriers. The minority carrier concentration thus drops exponentially fromthe depletion edge till it reaches the equilibrium concentration as shown inFig. 11.1. The spatial variation of minority carriers in the quasi-neutral regions isobtained by solving the continuity equation in the region outside the depletionregion:

pnðxÞ ¼ pno þ ½pnðxnÞ � pno�e�xLp ; [11.9]

and

npðxÞ ¼ npo þ ½npðxpÞ � npo�e�xLn [11.10]

where Ln is the electron diffusion length and Lp is the hole diffusion length.

Nitride LEDs based on quantum wells and quantum dots 379

In the depletion region, the current is carried primarily through diffusion becausethe Fermi levels are approximately constant. The total current is the sum of electrondiffusion current in the p-depletion edge and the hole diffusion current at the n-depletion edge:

Jtotal ¼ q

�Dn

Lnnpo þ Dp

Lppno

��eqVakT � 1

�; [11.11]

which can be written as

Jtotal ¼ Js�eqVakT � 1

�. [11.12]

This is the celebrated diode equation. The current increases exponentially withbias. In a p-n junction LED, electrons are injected into the depletion region fromthe n-side and holes are injected from the p-side. Thus, the depletion region thatwas almost devoid of mobile carriers under equilibrium (Va ¼ 0) is flooded withexcess electrons and holes under the application of a forward bias. If the situationin the depletion region is conducive to radiative recombination, the electronsrecombine with holes to produce light. If not, then they either recombine nonradia-tively producing heat or overshoot to the other side contributing to recombinationin the quasi-neutral n- and p-regions. The recombination in the quasi-neutralregion again may be radiative or nonradiative depending upon which processdominates.

Quasi-neutralregionEc

Efp

Ln Lp

npo

Ev

npopno

EfnqVa

p-n junction

Forward bias Quasi-neutralregion

kTkT

p(x)~exp(–x/Lp)

n(x)~exp(–x/Ln)

pnoExcess minority

carriers

exp( exp(

)) qVa

qVa

Figure 11.1 Band diagram and minority carrier distribution in a forward biased p-n junction.


11.1.2 Recombination mechanisms

The carriers injected in a diode undergo competitive radiative and nonradiative recom-bination processes. The generation rate is equal to the recombination rate at equilib-rium. The photon absorption rate per unit volume, Ro, from the vanRoosbroeckeShockley model can be termed as the equilibrium recombination rateand is given by

Ro ¼ 8pcn02ao�kT

ch

�3ffiffiffiffiffiffiffiffiffiffiffiffiffi�kT

Eg

�s Z N

xg

x2ffiffiffiffiffiffiffiffiffiffiffiffiffix� xg

pex � 1

dx; [11.13]

where c is the speed of light, n0 is the refractive index of the material, k is theBoltzmann constant, T is the ambient temperature, ao is the absorption constant atE ¼ 2Eg (Eg is the band gap), x ¼ E/kT and xg ¼ Eg/kT. The radiative recombinationrate in a semiconductor is given by

R� ¼ Bnp; [11.14]

where B is the bimolecular recombination coefficient, n is the electron concentration,and p is the hole concentration. At equilibrium the rate is

Ro ¼ Bnopo ¼ Bn2i . [11.15]

Using Eqs. (11.13) and (11.15) we can calculate B corresponding to differentmaterials.

In a crystal, the carriers can undergo nonradiative recombination processes throughdefects, traps or by many-particle processes such as Auger recombination. Shockley,Read, and Hall developed the theory for nonradiative recombination of carriers due totraps that act as deep energy levels in the band gap. The net recombination rate fromthe SRH theory is given by

RSRH ¼ np� n2ispðnþ ntÞ þ snðpþ ptÞ ; [11.16]

where n is the electron concentration, p is the hole concentration, ni is the intrinsiccarrier concentration, sn is the electron lifetime, sp is the hole lifetime, nt is the electrondensity, and pt is the hole density if the Fermi level is at the trap energy level. Thus,

nt ¼ nieEt�EikT ; [11.17]

and

pt ¼ nieEi�EtkT ; [11.18]


where Et is the trap energy level and Ei is the intrinsic Fermi energy level. Atequilibrium, we have np ¼ n2i and thus there is no net recombination andgenerationdmeaning the two processes still occur, but exactly cancel each other.On the other hand, np < n2i results in a eve recombination, which is the same asgeneration and np > n2i results in recombination. The first case is achieved in a p-njunction under reverse bias conditions and is useful for photovoltaic effect, whichcapitalizes on generation processes. The second case np > n2i occurs under forwardbias conditions: this nonradiative process competes for carriers with radiativeprocesses.

In a defect-free intrinsic semiconductor, the nonradiative lifetime decreases with in-crease in temperature, and it is minimum when Et ¼ Ei. Therefore, at high tempera-tures the radiative efficiency becomes low due to high nonradiative recombinationrate. Though here we have considered nonradiative recombination due to traps, insome cases deep level traps can lead to radiative recombination. An example is N-doped GaP, where N forms spatially localized states that are extended in momentumspace allowing direct optical transitions in GaP, which is actually an indirect bandgapsemiconductor.

The other major nonradiative process is Auger recombination. In this process, anelectron scatters from the conduction band to a hole state in the valence band. Butinstead of releasing the energy radiatively to a photon, it is spent in exciting anotherconduction band electron to a higher energy level in the conduction band (or a holeto the relevant higher energy level in the valence band). The excited electron (orhole) then relax back to the band minima and release the excess energy in theform of phonons, thus making Auger processes nonradiative. Because this processrequires the presence of three carrier particles, the Auger recombination rate isgiven by

RAug ¼ Cnn2p; [11.19]

for n-type semiconductors and

RAug ¼ Cpp2n; [11.20]

for p-type semiconductors, where n is the electron concentration, p is the hole con-centration, and Cn and Cp are the Auger recombination coefficients. Under high-levelinjection conditions, the Auger recombination rate becomes proportional to the thirdpower of the carrier density,

RAug ¼ Cn3. [11.21]

For IIIeV (V: As or P) semiconductors, C has been reported to be 10�29 cm6/s andfor III-nitrides the value has been found to be close to 10�30 cm6/s.3 Thus, due to thelow value of C, Auger recombination becomes dominant only at high injectioncurrents.


The various recombination mechanisms that occur in p-n junction LEDs describedearlier are shown schematically in Fig. 11.2. The simplified total recombination rate ina semiconductor is given by

R ¼ Anþ Bn2 þ Cn3; [11.22]

where A is the monomolecular nonradiative recombination coefficient, B is thebimolecular radiative recombination coefficient, C is the Auger recombination coef-ficient, and n is the electron carrier concentration.

Although calculating the diode current equation, recombination processes in thequasi-neutral regions were considered, but those occurring inside the depletion regionwere neglected. If we consider this component, an additional recombination currentterm adds to the ideal diode current equation when the diode is operated in forward bias:

Jnet ¼ Js�eqVakT � 1

�þ Jrec [11.23]

where

Jrec ¼ JSRH þ Jrad þ JAug ¼ q

ZRdx [11.24]

where x is the distance along the length, R is the net recombination rate given by Eq.(11.22), and the integral is taken over the active region.

11.1.3 LED efficiency

The efficiency of a p-n junction LED is the watts of light produced divided by the wattsspent in injecting current into the device. The overall conversion efficiency of LED isgiven by

RadiativeRecombination

Trap-assistedNon-radiativeRecombination

RecombinationAuger

EC

ET

EV

Figure 11.2 Band diagram showing different recombination processes.


htot ¼ hinj � hrad � hextr; [11.25]

where htot is the total efficiency, hinj is the injection efficiency, hrad is the radiativeefficiency, and hextr is the extraction efficiency. The injection efficiency is the ratio ofthe carriers injected into the active region to the carriers injected into the LED structurefrom the contacts. The radiative efficiency is the fraction of the carriers injected intothe active region that recombines radiatively. The extraction efficiency is the ratio ofthe photons that make it out of the LED to the photons generated in the active region.The radiative recombination efficiency is given by

hrad ¼ Bn2

Anþ Bn2 þ Cn3; [11.26]

which can be written in terms of the carrier lifetime as

hrad ¼1

srad1

sradþ 1snr

; [11.27]

where srad is the radiative recombination lifetime and snr is the nonradiative recom-bination lifetime. Identifying and minimizing the nonradiative loss processes andmaximizing the radiative processes boosts the efficiency of LEDs.

11.1.4 Homojunctions and heterojunctions, quantum wells anddots

The active region of a LED can be

• A depletion region of a p-n junction structure (a homojunction)• A quantum well (QW) or quantum dot (QD) region embedded in the depletion region sur-

rounded by barriers (a heterojunction)• a small band gap depletion region in a double heterostructure p-n junction (a heterojunction)

To achieve higher radiative recombination rate, the excess injected carrier concen-trations should be high in the layers where the optical recombination is desired. Forthis reason, a double heterostructure design was adopted in which a small band gapmaterial is sandwiched between an n-type and p-type layer of a large band gap mate-rial. Thus, the electron and holes would be trapped in the small band gap layer ratherthan being distributed along the whole diffusion length (Fig. 11.3). This results in anincreased carrier concentration in the active region leading to higher radiative recom-bination efficiency.

Another way to solve the problem of carrier diffusion is to utilize QW or QD het-erostructures. These are equivalent to a double heterostructure except that the width ofthe low bandgap region is much smaller (of the order of few nanometers). One can use


either a single QW or multiple QWs (MQW). A benefit of a smaller bandgap activeregion is that the quasi-Fermi levels can be pushed into the bands at small voltages,implying higher carrier injection for the same voltage. This is essential in lasers, butalso useful in LEDs.

The carriers in an energy band in all semiconductor structures have a FermieDirac dis-tribution, due towhich the carrier concentration has an exponential tail extending over theFermi level at finite temperature. Due to this feature, a fraction of excess carriers injectedinto the active region leak out of the active region into the contact layers by surmountingany barriers in the way. This leakage current reduces hinj. To reduce carrier leakage,

Electrons diffuse

p-n junction

Holes diffuse

Holes confined

Holes confined

Electrons confined

Electrons confined

Doubleheterostructure

Efp

Efp Wqw

Efn

Efp

Ec

Ec

Ec

Wdh

Efn

Efn

Ev

Ev

Ev

Quantum wellheterostructure

(a)

(b)

(c)

Figure 11.3 Conduction band diagrams showing carrier diffusion and confinement in (a) a p-njunction, (b) a double heterostructure (heterojunction), and (c) a quantum well (QW) hetero-structure (heterojunction). As Wqw < Wdh, the carrier confinement is higher in a QW than adouble heterostructure resulting in higher carrier concentration in QW.


electron blocking layers (EBLs) are used in LEDs.2 A higher bandgap layer is insertedbetween the p-type layer and the active region. The layer is heavily p-type doped in orderto obtain a flat valence band profile to allow efficient hole injection. The band gap differ-ence would then be completely reflected in the conduction band as shown in Fig. 11.4.Recently, graded electron blocking layers (GEBL) have been employed to obtain thesame effect.4 The electron blocking layer (EBL) blocks the electrons from leaking intothe contact layer and thus increases recombination in the quantum well.

It took a long time for the mechanism of photon emission from semiconductors andthe working of LEDs to be understood after the first experimental report. The firstreport of light emission from a semiconductor was in 1907 by Henry Joseph Round,a British engineer.1 He observed yellow light emission from SiC by applying a poten-tial across two points of the crystal and reported the observation in “Electrical World”.In the latter half of the 20th century, the growth methods for semiconductors maderapid progress and liquid phase epitaxy, vapor phase epitaxy (VPE), and molecularbeam epitaxy (MBE) were increasingly used for compound semiconductor growth.

The wavelength (or energy) of light emitted from semiconductor active regions isdesigned around its energy bandgap. The bandgaps of III-nitrides/arsenides/phos-phides alloy system and their lattice constants are shown in Plate 11 (see color platesection). These alloys are predominantly used for fabricating LEDs covering the elec-tromagnetic spectrum from the ultraviolet to infrared.

11.1.5 LED development

The first LED was a Schottky diode, but subsequently p-n junctions were fabricatedusing GaAs, which emitted coherent infrared light at 842 nm.5 Later Holonyak and

Electrons blocked

Single quantum well junction

Single quantum well junctionwith EBL

No electrons blocked

Efn

Efn

Ev

Ev

EB

L

Ec

Ec

Efp

Efp

Figure 11.4 Band diagram of single quantum well (QW) heterostructure, at forward bias, withand without electron blocking layer (EBL). The EBL blocks the electrons from leaking into thecontact layer and thus increases recombination in the QW.


Bevacqua6 demonstrated visible light emission at 710 nm from a GaAsP p-n junction.The realization of visible light emission was ground breaking for solid-state lighting.LEDs became a component of displays. The GaAsP alloy system doped with N wasused to fabricate green, yellow, and amber LEDs.7 But the efficiency of such LEDswas low due to the indirect band gap of GaAsP alloys. The direct-gap AlGaInP alloysystem was explored to produce high brightness visible LEDs ranging from 560 nm to650 nm.8e13 Thus, yellow, orange, red, and green color emission was obtained by us-ing arsenide- and phosphide-based IIIeV semiconductors. But blue light emission wasneeded to complete the color spectrum in order to produce white light for lighting anddisplays.

Pankove et al.14 demonstrated the first GaN LED at RCA Laboratories in 1971. Thestructure comprised of insulating GaN (Zn doped) and an unintentionally n-type dopedGaN layer. Mg was thought to be a better p-type dopant for GaN films, but not realizedtill Amano et al.15 were able to obtain p-type GaN by low energy electron beam irra-diation of Mg-doped GaN films. The electron beam was able to dissociate the MgeHbonds, activating the Mg and resulting in p-type doping. Nakamura et al.16 working atNichia (Japan) used thermal annealing to obtain the same effect in hydride vapourphase epitaxy-grown Mg-doped GaN films. Building on the success of obtainingp-type GaN layers Nakamura et al.17 demonstrated the first GaN p-n junction lightemitting diode with emission at 430 nm. The GaN p-n junction LED was alreadymore efficient than the then-commercial blue SiC LEDs.

Thereafter, Nakamura et al.18 demonstrated InGaN-based green and yellow LEDsemitting at 525 and 590 nm, respectively. The work on III-nitride visible LEDs wascarried forward into III-nitride ultraviolet LEDs, which utilized GaN/AlGaN hetero-structures. Han et al.19 demonstrated the AlGaN/GaN MQW LED emitting at353 nm. Employing AlN, AlGaN as the buffer layers for growing high compositionAlGaN active regions, the emission wavelength could be decreased further. AlN/AlGaN superlattice structures were utilized to reduce the dislocation density and de-fects in high composition AlGaN layers grown on sapphire.20 Adivarahan et al.21 ob-tained emission at 285 nm from AlGaN/AlGaN single quantum well (SQW)structure. Taniyasu et al.22 obtained emission at 210 nm from AlN-based active re-gion. However, the external quantum efficiencies (htot) achieved for the UV LEDswere low (Plate 12 (see color plate section)). htot also decreases when the InGaNand AlInGaP composition is changed to obtain green LEDs (Plate 12 (see color platesection)). A variety of factors affect the external quantum efficiency (EQE) of III-nitride UV LEDs and green LEDs. Of these, defects and built-in polarization fieldsare the most prominent.

11.2 Polarization effects in III-nitride LEDs

III-nitrides are unique among semiconductors because of their large polarizationvalues. In the following discussion, we discuss the origin of polarization in III-nitrides and its pros and cons in optical devices such as LEDs.


11.2.1 Origin of spontaneous and piezoelectric polarization

The nitride family of semiconductors are formed from group-III metals (Al, Ga, In) andN from group of the periodic table. In GaN, the N atom has much higher electroneg-ativity than Ga. Therefore the electrons in a GaeN bond are not equally shared be-tween the two atoms, but pulled more towards the N atom. The bond is then like anelectric dipole with a dipole moment. As shown in Fig. 11.5, each Ga atom is bondedto four nitrogen atoms and vice versa forming a tetrahedron. Other IIIeV semiconduc-tors (IIIeAs/P) crystallize in the zinc-blende crystal structure, the tetrahedron formedis almost perfectly symmetric due to which the four bonds forming the tetrahedroncancel each other’s dipole moment, making the material nonpolar. On the otherhand, III-nitrides crystallize in wurtzite crystal structure (noncentrosymmetric) form-ing an imperfect tetrahedron. Therefore, there is a net polarization along the c-axis <0 0 0 1>. This equilibrium polarization is called spontaneous polarizationPsp, which exists even in the absence of strain in the material. It is similar to ferroelec-tricity, but the direction of the dipoles is frozen and cannot be altered by an externalfield.

Fig. 11.6 shows the bandgap versus lattice constants for III-nitride family of semi-conductors. It is evident that the lattice constants vary over a wide range. Duringepitaxial growth, lattice mismatch leads to strain, which further distorts the tetrahedronof bonds causing additional polarization known as piezoelectric polarization Ppz. Thetotal polarization of a III-nitride film is then the sum of spontaneous and piezoelectriccomponents (Ptot ¼ Psp þ Ppz).

The spontaneous and piezoelectric polarization values, lattice constants, and bandgaps as a function of composition for Initride semiconductors are summarized inTable 11.1.

Fig. 11.7 shows the spontaneous, piezoelectric, and total polarization of AlGaNstrained to GaN substrate, as calculated from above formulas.

N

Ga

PSP

a

c

c-pl

ane

dire

ctio

n (0

001)

Figure 11.5 GaN wurtzite crystal structure.23


11.2.2 Polarization-induced electric fields

A schematic layer structure of a III-nitride LED is shown in Fig. 11.8. A lowerbandgap InGaN QW in sandwiched between higher bandgap GaN barriers. Using asmaller bandgap active region helps in carrier confinement, improving the internalquantum efficiency (IQE) as discussed in the earlier section. Usually, the barrier andQW regions have different Ptot values, producing a polarization discontinuity at theheterointerface. From Maxwell’s boundary condition equations, we know that the po-larization discontinuity manifests as a bound sheet charge given by P1 � P2 ¼ sp.This bound interface charge of density sp as shown in Fig. 11.8 occurs as a dipoleacross the InGaN layer. The interface charge density is usually of the order of1013 cm�2 resulting in large electric fields in the QW of the order of wMV/cm asshown in Fig. 11.9 for InGaN LEDs with a GaN barrier.

The large polarization-induced electric field bends the energy bands in the QW,forcing the electrons onto one side and holes to the other side. The different effectsof this large field on LED performance are discussed below.

11.2.2.1 Red-shifted photon emission (quantum confined Starkeffect: QCSE)

As shown in Fig. 11.10, the polarization-induced field in the QW reduces the transitionenergy from first electron subband to first hole subband resulting in red-shifted emis-sion from the LED.27,28 Upon application of forward bias, excess carriers are injected

GaN

AIN

InN

3.1 3.15 3.2 3.25 3.3 3.35 3.4 3.45 3.5Lattice constant (angstrom)

6

5

4

3

2

1

0

Ban

d ga

p (e

V)

Figure 11.6 Bandgap versus lattice constants for III-nitride semiconductor family.


into the QW, which screen part of the polarization field, leading to blue shifting of thespectrum with increasing current density.27

11.2.2.2 Decrease in oscillator strength

The rate at which an electron in the conduction band state jci relaxes to an empty state(hole) in valence band jvi emitting a photon is given by Fermi golden rule for opticaltransitions as29

Rrad ¼ 2pZjVcvj2dðεc � εv � ZuÞ; [11.28]

Table 11.1 III-nitride properties as a function of composition

Spontaneous polarization(C/m2)24,25

PspAlxGa1�xN

¼ �0:090x� 0:034ð1� xÞ þ 0:019xð1� xÞ

PspInxGa1�xN

¼ �0:042x� 0:034ð1� xÞ þ 0:038xð1� xÞ

PspAlxIn1�xN

¼ �0:090x� 0:042ð1� xÞ þ 0:071xð1� xÞLattice constants (Å),24 (UsingVegards’s Law)

aAlGaNðxÞ ¼ 3:189� 0:077x

aInGaNðxÞ ¼ 3:189þ 0:356x

aAlInNðxÞ ¼ 3:545� 0:433x

cAlGaNðxÞ ¼ 5:188� 0:208x

cInGaNðxÞ ¼ 5:188þ 0:512x

cAlInNðxÞ ¼ 5:70� 0:72x

Peizoelectric polarization(C/m2)24,25

PpzGaN ¼ �0:918εþ 9:541ε2

PpzAlN ¼ �1:808εþ 5:624ε2; ε < 0

PpzAlN ¼ �1:808ε� 7:888ε2; ε > 0

PpzInN ¼ �1:373εþ 7:559ε2

PpzAlxInyGa1�x�yN

¼ xPpzAlN þ yPpz

InN þ ð1� x� yÞPpzGaN

Where basal strain, εðx; yÞ ¼ ½asub � aðx; yÞ�=aðx; yÞBandgap (eV)26 (unstrained,300K)

EAlxGa1�xNg ¼ 6:2xþ 3:4ð1� xÞ � 0:7xð1� xÞ

EInxGa1�xNg ¼ 0:7xþ 3:4ð1� xÞ � 1:4xð1� xÞ

EAlxIn1�xNg ¼ 6:2xþ 0:7ð1� xÞ � 2:5xð1� xÞ


Ptot

Psp

Ppz

0 0.2 0.4 0.6 0.8 10

1

2

3

4

5

6

7

8

9

AlGaN composition x

x 1013

Pol

ariz

atio

n (e

lect

roni

c ch

arge

/cm

2 )

Figure 11.7 Spontaneous Psp, piezoelectric Ppz, and total polarization Ptot (electronic chargeq ¼ �1.6 � 10�19 C) for pseudomorphic AlGaN on GaN.

(a) (b)

(c)

p-GaN

p-GaN

InGaN

InGaN

n-GaN

n-GaN

Ec

QW

Polarizationfield

in QW

Ev

(–)

(+)

(000

1)

Figure 11.8 (a) Basic metal-polar nitride light-emitting diode, (b) interface charges, and(c) band diagram showing polarization field.


Here, εc and εv are the energy of electron in the initial (conduction band) and final(valence band) states, respectively, Zu is the energy of the emitted photon and theDirac-d function ensures energy conservation in the transition process. From Eq.(11.28), the radiative transition rate is proportional to square of the matrix element Vcv,which quantifies the strength of interaction between electron and photon systems.Quite often in calculations, instead of using the matrix element directly, an equivalentquantity proportional to jVcvj2 called the “oscillator strength” is used. In a QW, theoscillator strength fosc of interband transition is given as,24,29

5

4.5

3.5

2.5

1.5

0.5

0.1 0.2 0.3 0.4 0.5

4

3

2

1

00

InGaN

Bui

lt-in

fiel

d (M

V/c

m)

Figure 11.9 Polarization-induced electric field in GaN/InGaN/GaN quantum well as a functionof InGaN composition. The field points in the �c direction.

Holewavefunction

Holewavefunction

Electronwavefunction

Electronwavefunction

Polarizationfieldin QW

Without field With field

QWQW EcEc

Ev

hv1 hv1hv2 hv2

Ev

<

Figure 11.10 Quantum confined stark effect in nitride light-emitting diodes.


fosc ¼ 2meZucv

��hucje:pjuvi��2��

ZfeðzÞfhðzÞdz

��2

; [11.29]

where me is mass of the electron, Z is Planck’s constant, ucv is the transition angularfrequency, and hucje:pjuvi is the momentum matrix element with uc and uv beingconduction band and valence band periodic Bloch functions, respectively.RfeðzÞfhðzÞdz is the overlap integral of electron and hole envelope wave functions. The

effect of large fields in the QW on envelope functions is shown in Fig. 11.11.The spatial separation of electrons and hole envelope functions due to the field in

the QW reduces the overlap integral.30 This reduces the radiative recombinationrate, thus allowing nonradiative processes an upper hand in the competition for carrierrecombination, leading to reduced IQE of light-emitting devices.

11.2.3 Methods for improving IQE

11.2.3.1 Increasing the overlap by using thin QWs and QD

The radiative decay time of carriers in an AlGaN/GaN QW and peak energy of emis-sion as a function of well width is shown in Fig. 11.12.30 Increasing the width of theQWs leads to a red shift in emission due to QCSE, and a corresponding decrease in theradiative decay rate due to smaller electron-hole overlap. A similar trend is seen inGaN/AlN QDs.31 By increasing the height of dots, photoluminescence decay time in-creases. This effect is due to the presence of polarization-induced electric fields in theactive region.

On the other hand, when the QW width or QD size is decreased, the electron-holesubband energies increase because of size quantization effects. Because the energyband-offset confining the carriers in the well is finite, very thin QWs or small QDs

F = 2.5 MV/cmF = 0 MV/cm

fefh

–5 50 10 15 20Thickness (nm)

Env

elop

e w

ave

func

tions

(arb

. uni

ts)

Figure 11.11 Reduction in overlap of electron and hole envelope wave functions with anincrease in polarization field.24


lead to poor confinement of carriers. This suggests that a careful design trade-off isnecessary to determine the optimal QW and QD sizes for improving the overlap of car-riers without compromising carrier confinement.

11.2.3.2 Use of semipolar/nonpolar substrates

An attractive way to improve the electronehole overlap is to reduce or completelyeliminate the polarization-induced electric field in the QW active region. This isachieved by growing III-nitrides LED heterostructures in semipolar, or nonpolar crys-tallographic orientations.32 Various crystallographic planes in III-nitrides are shown inFig. 11.13.33 Because the polarization vector is along the c-axis [0 0 0 1] of the crystal,all the crystal orientations perpendicular to it of the form (h k l 0) have no polarizationand are called nonpolar planes (m-plane: {1 �1 0 0}, a-plane: {1 1 �2 0}). A planeoriented at an angle with the c-plane other than 90� is called a semipolar plane(e.g., r-plane). LEDs grown along semipolar crystal orientations can retainpolarization-induced electric field in the QWs, but at a much smaller magnitude ascompared to the polar direction (Fig. 11.1434).

GaN/AI0.15Ga0.85N SQW’s

T = 5K

F = 350 kV/cm

GaN 0.4% strained

GaN unstrained

0.0 2.0 4.0 6.0 8.0 10.0

Well width (nm)

3.20

3.30

3.40

3.50

3.60

3.70

100

101

102

103

104

105

Dec

ay ti

me

(ns)

Pea

k en

ergy

(eV

)

Figure 11.12 Comparison of the measured energy positions (circles) and decay times (squares)of the low-energy lines in GaN/AlGaN QWs calculated based on piezoelectric fields.30


LEDs with very high efficiency have already been realized along the nonpolar andsemipolar orientations.35e37 As large area semipolar and nonpolar GaN substratesbecome available, the efficiencies of LEDs are expected to improve across the entirespectrum due to reduced polarization-induced fields.

The largest volume application of III-nitride LEDs is in solid-state lighting. Onehurdle in the development of efficient white light sources is the problem of efficiencydroop.38 Efficiency droop is the reduction in the LED electrical-to-light conversion ef-ficiency at high injection current densities. The reasons for this behavior are still a topicof vigorous debate.38 Semipolar and nonpolar orientations are expected to reduce theefficiency droop, making the realization of efficient solid-state lighting possible.39

11.2.3.3 Polarization-matched LEDs

A method to reduce fields inside the QW while retaining the polar orientation ofgrowth is to reduce the polarization discontinuity between well and barrier interface.This can be achieved by careful choice of quaternary AlInGaN barrier layers.40 The

z

xx

y

c plane (0001)polar

(1122)-

m-plane (1100) non-polar

-

a-plane (1120)non-polar

-A

B

y

Semi-polarplane

c plane (0001)z

(a) (b)

Figure 11.13 (a) Nonpolar and (b) semipolar planes in III-nitrides.33

15° 30° 45°60° 75° 90°

–0.02

–0.01

0.01

0.02

0.03

0ϑ

(4)(3)(2)(1)

ΔPz' (C

/m2 )

Figure 11.14 Computed polarization charge density in InGaN/GaN QWs as a function of tiltangle of the growth plane with respect to the c-plane.34


concept of polarization matching offers flexibility to retain a band offset of the barrierand to simultaneously tune the polarization. At the right material combination, the bar-rier and well layers become polarization matched but have different band gaps. This isshown in Fig. 11.15.40 Polarization-matching has been shown to reduce the efficiencydroop in InGaN LEDs and to decrease the blue-shift in spectrum with increasing cur-rent density.40,41 The epitaxial growth of the required material combinations iscurrently difficult but are expected to improve in the future.

11.2.4 Problem of p-type doping in nitride LEDs andpolarization-induced doping

p-type doping in III-nitrides remained elusive for a long time till the breakthroughby.15 High-performance optoelectronic devices followed soon after it was realizedthat Mg acceptors in GaN grown by chemical vapor deposition processes require acti-vation.18,42 However, even today, obtaining high hole concentrations in III-nitrides isdifficult due to large activation energy EA: 200 meV43 of Mg in GaN. The situationbecomes worse in high Al composition AlGaN alloys, because the Mg activation en-ergy in AlN is even higher (EA: 630 meV).43 Low hole concentrations combined withlow hole mobilities make the p-side of LEDs highly resistive, causing high I2R losses.In ultraviolet (UV) LEDs fabricated using AlGaN alloys, it has remained difficult tofabricate low-resistance ohmic contacts to p-AlGaN because of the low hole concen-tration. To lower the ohmic contact resistance, usually a p-GaN cap layer is used, butthis layer absorbs a portion of the UV light emitted in the active region, leading toreduced extraction efficiency of LED.

In this section, some disadvantages of polarization-induced fields in GaN LEDs weredescribed. But just like any physical feature in semiconductors, if used judiciously,

0 5

5

10

10

15

15

20

20

25

25

30

30

35

35

40

40

0

Indium composition y (%)

2.2 eV

2.4 eV

2.6 eV

2.8 eV

3.0 eV

3.2 eV3.6 eV

3.8 eV

4.0 eV

4.2 eV

4.4 eV

Alu

min

um c

ompo

sitio

n x

(%)

0.04

Cm

–2

0.02

Cm

–2

0.00

Cm

–2

0.02

Cm

–2

AlxGa1-x-yInyN grownpseudomorphically on GaN

Figure 11.15 Polarization charge and bandgap contours for quaternary AlxGa1�x�yInyN grownpseudomorphically on GaN.40


polarization also offers various advantages. It can be used to improve the p-type doping inIII-nitrides. At a sharp heterojunction, the polarization discontinuity manifests as a boundsheet charge. By a similar token, in a graded heterojunction, polarization manifests as afixed bulk charge V:P ¼ rp. If, instead of making an abrupt heterojunction, we continu-ously grade the material from one composition to other, the polarization charge is spreadover the graded layer as shown in Fig. 11.16.44 Thisfixed bulk polarization charge attractsoppositely charged mobile carriers to satisfy local charge neutrality. The creation of mo-bile charges is tantamount to doping. Because the polarization-induced doping process ispurely electrostatic, it has no thermal activation energy, and the mobile carriers do notfreeze out at low temperatures.43 p-type polarization induced doping has been demon-strated to improve light emission compared to Mg doped p-layers.43

11.3 Current status of III-nitride LEDs

High efficiency red LEDs had been developed using (AlxIn1�x)1�yGayP alloys. But thereis a cross-over fromdirect gap to indirect gap in (AlxIn1�x)1�yGayP alloys as x approaches0.53, which results in decreased LED efficiency as the wavelength is reduced from650 nm to the green part of the spectrum at 560 nm. Besides, the lack of suitable higherband gap material leads to poor carrier confinement resulting in increased electronleakage. This further reduces the EQE of AlInGaP LEDs as they inch toward emissioningreen regionof the visible spectrum.Asmentioned inSection11.1.5, once the problemsof growth and p-type doping were figured out, the field of III-nitride LEDs developedrapidly. The successful demonstration of MOCVD grown InGaN/AlGaN double hetero-structure violet LEDs and laser diodes by Nakamura et al.,45 Fig. 11.17, further broughtthe III-nitrides, and specifically the InGaN alloy system into the limelight. Besides, theInGaN alloy system spanned the entire visible spectrum. As a result, InGaN emergedas the front runner to obtain high efficiency LEDs emitting from blue to green.

11.3.1 III-nitride visible LEDs

Nakamura et al.45 utilized SQW InGaN LED structures to improve the EQE of InGaN-based blue LEDs from 2.7%45 to 7.3%.18 Prestrained growth technique in which low

Ga-faceGa-face

GaN

GaN

Graded AlGaNGraded AlGaN

GaN

GaN

Grade"down"

Grade"up"

Electrongas

Polarizationcharges

Polarizationcharges

Holegas

Figure 11.16 Three-dimensional polarization-induced n- and p-type doping.44


In content InGaN/GaN QWs are inserted before the growth of InGaN/GaN MQWactive area has also proved to be useful in boosting the IQE of InGaN LEDs.46 TheIQE, PL intensity, EL intensity at 20 mA injection current have been improved by167%, 140%, and 182%, respectively. This achievement has been assigned to weakerQCSE and weaker localization effect for carriers.

The growth of InGaN LEDs has been performed predominantly along the [0001] orGa-polar direction by MOCVD. As discussed before the large polarization field sep-arates the electron and hole wavefunctions in the QW active regions. But recently anumber of methods have been employed to improve the wavefunction overlap. Growthalong nonpolar/semipolar directions reduces the polarization-induced field and en-hances the radiative recombination efficiency.47e51 Koslow et al.51 have obtained26.5% EQE InGaN LEDs at 452 nm, whereas Zhao et al.52 have reached 52% EQEfor blue-violet InGaN LEDs on semipolar free standing GaN. Additionally, incorpora-tion of a d-AlGaN layer or d-InN in InGaN QW and staggered QWs improves thewavefunction overlap resulting in increased radiative recombination.53e63 Modifyingthe QWs into a triangular shape has also been found to enhance the radiative recom-bination efficiency.64

By increasing the indium fraction in InGaN alloy, green and yellow LEDs withEQE 2.1% and 1.2%,18 respectively, were successfully demonstrated. The decreasein EQE (Fig. 11.18) with increasing indium fraction in InGaN has been attributed todeteriorating material quality. The miscibility problem of indium and gallium leadsto phase separation in high composition InGaN and the large lattice mismatch withthe underlying GaN substrate induces more defects and dislocations compared tothe case of moderate indium composition.65,66 Insertion of low indium content InGaN

p-Electrode

n-Electrode

n-GaN

GaN buffer

Sapphire substrate

p-Al0.15Ga0.85N

n-Al0.15Ga0.85N

In0.06Ga0.94N

Figure 11.17 InGaN-AlGaN double heterostructure light-emitting diode (LED). The structurecame to form the basis for future LED work with III-nitrides.45


layer prior to the QWs can also boost the performance of green LEDs, but the physicsbehind the increase in efficiency is different than that for blue InGaN LEDs. It wasfound that the energy transfer between donoreacceptor levels of GaN towardInGaN/GaN QWs is mediated by an underlying layer of In0.063Ga0.937N.

67 The energytransfer stimulates light emission from QWs and, therefore, the LED emits 85% morelight.

The EQE of InGaN LEDs increases for low injection current densities and then de-creases gradually as the current density is increased (Fig. 11.19). The drop in efficiencywith increasing current is termed as the efficiency droop. Droop is observed univer-sally in InGaN LEDs irrespective of the InGaN composition. Even nonpolar and semi-polar plane grown LED structures show degrees of efficiency droop at alltemperatures. A number of possible physical reasons for droop have been investigated.Among them, nonradiative Auger recombination, electron leakage due to built-in po-larization, high dislocation density, and poor hole concentration and mobility havebeen argued as causes of efficiency droop.3,40,89e98. At this time, there is no clear ver-dict for the primary factor and it is likely there are various mechanisms at play.Polarization-matched active region and electron blocking layers have been shown tomitigate the droop. Zhao et al.98 have recently employed thin AlInN barriers in placeof AlGaN to decrease the carrier leakage and improve the drop in IQE with increasingcurrent density.

Besides, reduction in dislocation density (Fig. 11.20) in the structure also increasesthe efficiency and thus improves the efficiency droop (Fig. 11.21).

The radiative recombination efficiency of LEDs can also be improved by using QDsin place of QWs. QDs are 0-dimensional structures providing three-dimensionalconfinement of carriers. QD LEDs had been demonstrated in III-As semiconductors

350 400 450 500 550 600 6500

10

20

30

40

50

60

EQE

(%)

Wavelength (nm)

Figure 11.18 External quantum efficiency variation with peak emission wavelength. A changein indium fraction in InGaN results in variation of IQE.52,68e88


10.90.8

0.70.6

0.50.4

0.30.2

0.10

0 100 200 300 400 500 600 700Current density (A/cm2)


Inte

rnal

qua

ntum

effi

cien

cy ( η

IQE)

A = 1 × 107 s–1

A = 1 × 107 s–1

C = 3.5 × 10–34 cm6/s

C = 3.5 × 10–34 cm6/s

24-Å In0.28Ga0.72N / GaN

24-Å In0.28Ga0.72N24-Å In0.28Ga0.72N

24-Å In0.28Ga0.72N / 15-Å Al0.1Ga0.9N24-Å In0.28Ga0.72N / 15-Å Al0.83In0.17N

10.90.80.70.60.50.40.30.20.1

00 100 200 300 400 500 600 700

Effi

cien

cy

/ 15-Å Al0.83In0.17N

15-Å Al0.83In0.17N

GaN GaNLarge

Large

ηIQE

ηradiativeηinjection(a)

(b)

Figure 11.19 (a) Internal quantum efficiency (IQE) (hIQE) of 24-Å In0.28Ga0.72N/15-ÅAl0.83In0.17N quantum well (QW) (lw 480 nm) at 300K. IQE (hIQE), radiative efficiency (hrad),and current injection efficiency (hinj) as a function of total current density are plotted. (b) IQE(hIQE) as a function of total current density for 24-Å In0.28Ga0.72N/GaN QW, 24-ÅIn0.28Ga0.72N/15-Å Al0.1Ga0.9N QW, and 24-Å In0.28Ga0.72N/15-Å Al0.83In0.17N QW.98

(a) (b) (c) (d)

Figure 11.20 High-resolution transmission electron micrographs showing the defect distribu-tion along the light-emitting diodes (LEDs). Cross-section of full thickness of LEDs (a) onsapphire, and (c) on FS-GaN substrate; top region including the MQWs of LEDs (b) onsapphire, and (d) on FS-GaN substrate.


with the large lattice mismatch aiding the growth of QDs in Stranski Krastanow (SK)mode. Tanaka et al.99 reported an early growth of GaN QDs using antisurfactant meth-odology. The GaN QDs showed stimulated emission under optical excitation. Adel-mann et al.100 systematically studied GaN QDs nucleation on AlN and recorded theQD height dependence on the substrate temperature during growth. Due to QCSEthe peak emission wavelength from GaN QDs embedded in AlN increased as theQD height was increased. It was also found that the decay time increased with QDheight in accordance with the QCSE.31 On the other hand, Brown et al.101 studiedthe GaN QD formation under varying Ga flux. They showed that high energy peaksin the PL spectrum from GaN/AlN QD structures are obtained from the QW formedby the GaN wetting layer and corroborated it with a theoretical model.

InGaN has a large lattice mismatch with GaN and, therefore, the alloy system is asuitable candidate for forming QDs on GaN through the SK mode. It has long beensuspected that In segregation in InGaN QWs leads to the formation of QDs. It isbelieved that such dots help in radiative recombination in InGaN LEDs despite thehigh-density of dislocations.102e109 To investigate further, InGaN QDs were intention-ally grown by MOCVD using Si as antisurfactant,110 and by MBE in the SK growthmode.111 Xu et al.112 demonstrated red, green, and blue LEDs using InGaN QDs.Moustakas et al.113 have also shown that InGaN QD active region LEDs emit blueand green light. It was found that larger QDs nucleate near dislocations, whereasthe small QDs nucleate in dislocation-free areas. Therefore, the QDs mitigate the non-radiative recombination at dislocations. InGaN QD green LEDs using tunnel injectionwere found to exhibit reduced efficiency droop at high current densities compared tosimilar InGaN QW LEDs.114 Based on initial findings, III-nitride QDs are attractingincreasing interest for enhancing the EQE of InGaN LEDs and potentially overcomingthe efficiency droop problem.

0

0.2

0.4

0.6

0.8

50 100 150 200 250 300Current (mA)

Current (mA)

1.0

Nor

mal

ized

EQ

E

LEDs on FS-GaN-pulse mode

LEDs on FS-GaN-cw modeLEDs on sapphire-pulse mode

LEDs on sapphire-cw mode

LEDs on sapphireLEDs on FS-GaN

00.0

0.2

0.4

0.6

0.8

1.0

50 100 150 200 250 300

EL

inte

nsity

(a. u

.)

Figure 11.21 Normalized external quantum efficiency as a function of forward current for light-emitting diodes (LEDs) on sapphire and FS-GaN under continuous-wave (CW) and pulseoperations. The inserted plot is the output power versus current under CW mode.


11.3.2 III-nitride UV LEDs

Mukai et al.115 found that by decreasing the indium composition in the InGaN activeregion the emission wavelength could be decreased and 7.5% efficient 371 nm InGaN/AlGaN double heterostructure LEDs were fabricated. However, the InGaN alloy sys-tem cannot emit light at wavelengths less than 365 nm (GaN band gap). Moreover, thepoor carrier confinement in the QWs of low indium composition InGaN/GaN LEDsreduces the efficiency. Thus, the AlGaN alloy system has been explored in order toachieve shorter wavelength emission in the UV region, Because AlN is a direct-bandgap semiconductor with bandgap of w6.2 eV, light emitters down tow200 nm are feasible in the AlGaN material system. Although visible LEDs findapplication in displays and solid-state lighting, whereas the UV LEDs are potentialcandidate for sterilization, lithography tools, water decontamination, and novelbiosensors.

Khan et al.116 obtained the first optical emission from an AlGaN/GaN MQW struc-ture. Thereafter, the AlGaN/GaN MQW structure LED was demonstrated to emit lightat energy above the GaN band gap at 353 nm.19 The confining effects of the QW led tohigher energy emission than GaN band gap. Fig. 11.22(a) shows a typical UV LEDstructure, which consists of AlxGa1�xN/AlyGa1�yN QW/barrier active region.

Initially, the growth of III-nitride UV LED structures was performed on GaN bufferlayers grown on sapphire. To achieve higher energy emission with increased efficiencyUV transparent substrates are essential. With advances in growth techniques, AlN-on-sapphire became the substrate of choice for UV LEDs. The high composition AlGaNlayers have low lattice mismatch with the AlN buffer layer leading to reduced defectand dislocation density. Mukai et al.118 and Akita et al.74,75 showed that decrease in

1 cm

p-GaN contact layerp-AlxGa1–xN

n-AlxGa1–xN

n-AlxGa1–xN contact layer

Active region(multiple QWs)

Confinement layerelectrode

Electrode

n-contact

p-contact

Emitted light

SubstrateBuffer layer

(a)

(b)

Figure 11.22 (a) Typical ultraviolet (UV) led, (b) packaged UV led.117


dislocation density enhances the EQE of III-nitride LEDs. Besides the AlN layer iseffectively transparent to photons emitted from the AlxGa1�xN/AlyGa1�yN active re-gion. With advances in epitaxial growth methods, light emission was obtained atfurther lower wavelengths (315e210 nm).117 Yoshida et al.119 demonstrated a UVlaser diode emitting at 342 nm based on AlGaN-active region. Shatalov et al.120 ob-tained stimulated emission from AlN at 214 nm through optical excitation.

As discussed earlier, mobile hole concentrations are rather low in high compositionMg-doped AlGaN layers. This increases the resistive losses in the diode. It also posesproblems to form ohmic contacts with increasing AlGaN composition. Therefore, Mg-doped GaN layers have been used for deep UV LEDs. The p-GaN layer also reducescurrent crowding in the top layer. Because the bandgap of GaN is smaller than the en-ergy of photons emitted, this contact layer absorbs some of the emitted light and de-creases the overall efficiency of the device. To overcome the p-type deficiency inwide band gap III-nitride short-period AlGaN/AlGaN superlattices have been used.The built-in polarization of III-nitrides ionizes the Mg in GaN/Al(Ga)N superlatticeleading to efficient doping through miniband formation.121,122 AlN/AlGaN superlatti-ces have also been used to achieve thick crack-free n-type AlGaN layers on AlN bufferlayers.20 Polarization-induced doping by compositional grading of AlGaN in wideband gap semiconductors has been shown to improve the hole concentration in p-AlGaN layers.43

Sapphire is insulating in nature and AlN is also a poor conductor. Thus, lateral cur-rent crowding in UV LEDs grown on sapphire poses a major problem. Interdigitatedfinger structures and micropixel designs have been utilized successfully to facilitateuniform injection of carriers. As a result, an improvement in the UV LED power outputhas been obtained.123e125

The built-in polarization also affects the performance of UV LEDs. The use of po-larization matched active regions, electron blocking, and low-resistance contact layershas improved the efficiency droop in InGaN visible LEDs. In the case of UV LEDs,too, the use of AlInGaN alloy system and similar techniques can help in increasingtheir efficiencies.

III-nitride QD UV LEDs were demonstrated using GaN QDs grown by antisurfac-tant method.99 Self-assembled GaN QDs were grown by MBE in the SK and modified-SK growth modes.31,101,111 InAlGaN QDs grown by the antisurfactant method haveshown promising results with emission achieved at 335 nm.126 By increasing theQD density, the emission intensity was increased. By varying the height of the GaNQDs, the emission energy could be tweaked from 310 nm to 440 nm.31 Vermaet al.127 have demonstrated 261e340 nm emission from GaN QDs that are 2 ML inthickness. The QDs were grown in AlN matrix and tunneling transport of carrierswas employed in the LED structure. The first compositionally-graded polarization-induced p-doping has been achieved by Verma et al.128 for a deep-UV LED with243 nm emission. The polarization-induced doping using graded AlGaN was foundto enhance the light emission by 23 times compared to conventional nongraded p-AlGaN. 2 ML GaN/AlN QDs were used as the light emitting active region. Bayerlet al.129 have reported first-principle calculation and experimental verification ofdeep-UV emission using 1 and 2 ML GaN QWs in the AlN matrix. It was shown


that 224 nm emission can be achieved by extreme quantum confinement using 1 MLGaN QW/2 nm AlN barriers. High (w35%) IQE has been reported by Islam et al.130

for 2 ML GaN/AlN QDs emitting over 222e231 nm. The QDs were grown usingplasma-assisted MBE under modified SK modes. Recently, tunable electrolumines-cence over 232e270 nm have been reported by Islam et al.131 using 2e4 ML GaN/AlN QDs as the active region. Both the p and n regions consisted of polarization-induced graded AlGaN. 232 nm is the shortest deep-UV electroluminescence wave-length reported till date using GaN as the light-emitting material.

The AlInGaN alloy system has a large parameter space with regard to bandgap, po-larization fields, and band offset, allowing the design of the most effective LED struc-tures over a wide spectral range. It can be used to grow different quantum-confinedstructures in the form of QDs and QWs and the various techniques to improve theextraction efficiency, high efficiency UV LEDs are expected in the near future.

11.4 Modern LED designs and enhancements

In addition to the LED structures described above, novel designs are being explored toimprove the efficiency of III-nitride LEDs. We briefly review some ideas in thissection.

11.4.1 Resonant-cavity light emitting diodes (RCLEDs)

In an LED, the light is emitted by the spontaneous recombination of electrons andholes in the active region. The efficiency depends on how fast this process occurs,which is proportional to optical density of states at the frequency of emission. In anRCLED, the active region is placed inside a resonant cavity (RC) whose resonant fre-quency matches the desired emission frequency (or wavelength) of the LED. This in-creases the optical density of states near the emission frequency speeding up thespontaneous emission rate. Thus, RCLEDs result in higher optical intensities,improved LED efficiency, higher spectral purity because of the cavity, and muchmore directed far-field emission pattern.2 RCLEDs were first realized in III-Arsenide system132 but they have also been realized in III-nitrides.133e136

11.4.2 Superluminescent light-emitting diodes (SLEDs)

SLEDs are edge-emitting p-n junctions with optical gain. When the injection current inthe diode is increased to an extent that population inversion condition is achieved, thelight emitted from the active region is amplified by stimulated emission as it propa-gates in the active region, which is also a gain medium. In this way, high optical poweroutput compared to normal LEDs, can be obtained. Even though SLEDs have an activeregion with gain, they do not lase because there is no feedback mechanism available.As a consequence, SLEDs have high optical powers. Their spectrum is broadercompared to LASERs, but narrower than LEDs.2 High power blue SLEDs havebeen obtained in III-nitrides.137e139


11.4.3 Nanowire LEDs

High composition InGaNQWactive regions are susceptible to strain-related degradation.Instead, if the InGaN active regions are grown in a nanowire geometry, stress can berelaxed without generating dislocations. For this reason, InGaN nanowires hold greatpromise for realization of high-efficiency LEDs emitting in different parts of the visiblespectrum.86,140 Core/shell nanowire LEDs can also enable multicolor emitters.141

11.4.4 Polariton LEDs

A polariton is a quasi-particle formed when photons strongly couple with excitons.142

They are half-light (photon) and half-matter (excitons) and are formed when excitonsare created in a high-finesse optical cavity. Polaritons are bosons with an integral spinbut very small mass. Therefore, they can form a BoseeEinstein condensate at muchhigher temperatures, leading to highly efficient light-emitting devices with verylow-lasing thresholds.143e145 Electrically injected polariton LEDs have been realizedbased on both III-Arsenides146 and III-nitrides.147 Such devices are attracting attentionfor moving the field of semiconductor optoelectronics into the strong coupling regime,where light and matter behave as one.

References

1. Round HJ. A note on carborundum. Electr World 1907;49:308.2. Schubert EF. Light emitting diodes. Cambridge University Press; 2003.3. Shen Y-C, Mueller GO, Watanabe S, Gardner NF, Munkholm A, Krames MR. Auger

recombination in InGaN measured by photoluminescence. Appl Phys Lett 2007;91:141101e3.

4. Wang CH, Ke CC, Lee CY, Chang SP, Chang WT, Li JC, Li ZY, Yang HC, Kuo HC,Lu TC, Wang SC. Hole injection and efficiency droop improvement in InGaN/GaN light-emitting diodes by band-engineered electron blocking layer. Appl Phys Lett 2010;97:261103e5.

5. Hall RN, Carlson RO, Soltys TJ, Fenner GE, Kingsley JD. Coherent light emission fromGaAs junctions. Phys Rev Lett 1962;9:366e9.

6. Holonyak N, Bevacqua SF. Coherent (visible) light emission from Ga(As1�xPx) junctions.Appl Phys Lett 1962;1:82e3.

7. Krames MR, Shchekin OB, Mueller-Mach R, Mueller GO, Zhou L, Harbers G,Craford MG. Status and future of high-power light-emitting diodes for solid-state lighting.J Disp Tech 2007;3:160e75.

8. Kuo CP, Fletcher RM, Osentowski TD, Lardizabal MC, Craford MG, Robbins VM. Highperformance AlGaInP visible light-emitting diodes. Appl Phys Lett 1990;57:2937e9.

9. Sugawara H, Ishikawa M, Hatakoshi G. High-efficiency In-GaAlP/GaAs visible light-emitting diodes. Appl Phys Lett 1991;58:1010e2.

10. Huang KH, Yu JG, Kuo CP, Fletcher RM, Osentowski TD, Stinson LJ, Craford MG,Liao ASH. Twofold efficiency improvement in high performance AlGaInP light-emittingdiodes in the 555e620 nm spectral region using a thick GaP window layer. Appl Phys Lett1992;61:1045e7.


11. Kish FA, Steranka FM, DeFevere DC, Vanderwater DA, Park KG, Kuo CP,Osentowski TD, Peanasky MJ, Yu JG, Fletcher RM, Steigerwald DA, Craford MG,Robbins VM. Very high-efficiency semiconductor wafer-bonded transparent-substrate (AlGa) in P/GaP light-emitting diodes. Appl Phys Lett 1994;64:2839e41.

12. Krames MR, Ochiai-Holcomb M, H€ofler GE, Carter-Coman C, Chen EI, Tan IH, Grillot P,Gardner NF, Chui HC, Huang JW, Stockman SA, Kish FA, Craford MG, Tan TS,Kocot CP, Hueschen M, Posselt J, Loh B, Sasser G, Collins D. High-power truncated-inverted-pyramid (Al Ga) in P/GaP light-emitting diodes exhibiting >50% externalquantum efficiency. Appl Phys Lett 1999;75:2365e7.

13. Streubel K, Linder N, Wirth R, Jaeger A. High brightness AlGaInP light-emitting diodes.IEEE J Sel Top Quan Electron 2002;8:321e32.

14. Pankove JI, Miller EA, Richman D, Berkeyheiser JE. Electroluminescence in GaN.J Lumin 1971;4:63.

15. Amano H, Kito M, Hiramatsu K, Akasaki I. P-type conduction in Mg-doped GaN treatedwith low-energy electron beam irradiation (LEEBI). Jpn J Appl Phys 1989;28:L2112e4.

16. Nakamura S, Mukai T, Senoh M. High-power GaN P-N junction blue-light-emitting di-odes. Jpn J Appl Phys 1991a;30:L1998e2001.

17. Nakamura S, Senoh M, Mukai T. Highly P-typed Mg-doped GaN films grown with GaNbuffer layers. Jpn J Appl Phys 1991b;30:L1708e11.

18. Nakamura S, Senoh M, Iwasa N, Nagahama S. High-brightness InGaN blue, green andyellow light-emitting diodes with quantum well structures. Jpn J Appl Phys 1995;34:L797e9.

19. Han J, Crawford MH, Shul RJ, Figiel JJ, Banas M, Zhang L, Song YK, Zhou H,Nurmikko AV. AlGaN/GaN quantum well ultraviolet light emitting diodes. Appl Phys Lett1998;73:1688e90.

20. Zhang JP, Wang HM, Gaevski ME, Chen CQ, Fareed Q, Yang JW, Simin G, Khan MA.Crack-free thick AlGaN grown on sapphire using AlN/AlGaN superlattices for strainmanagement. Appl Phys Lett 2002;80:3542e4.

21. Adivarahan V, Wu S, Chitnis A, Pachipulusu R, Mandavilli V, Shatalov M, Zhang JP,Khan MA, Tamulaitis G, Sereika A, Yilmaz I, Shur MS, Gaska R. AlGaN single-quantum-well light-emitting diodes with emission at 285 nm. Appl Phys Lett 2002;81:3666e8.

22. Taniyasu Y, Kasu M, Makimoto T. An aluminium nitride light-emitting diode with awavelength of 210 nanometres. Nature 2006;441:325e8.

23. Simon J. Polarization Engineered III-V nitride heterostructure devices by molecular beamepitaxy [Ph.D. thesis]. University of Notre Dame; 2009.

24. Wood C, Jena D. Polarization effects in semiconductors: from Ab InitioTheory to deviceapplications. 1st ed. Berlin: Springer; 2007.

25. Fiorentini V, Bernardini F, Ambacher O. Evidence for nonlinear macroscopic polarizationin IIIeV nitride alloyheterostructures. Appl Phys Lett 2002;80:1204e6.

26. Piprek J. Nitride semiconductor devices: principles and simulation. 1st ed. Wiley-VCH;2007.

27. Takeuchi T, Sota S, Katsuragawa M, Komori M, Takeuchi H, Amano H, Akasaki I.Quantum-confined Stark effect due to piezoelectric fields in GaInN strained quantumwells. Jpn J Appl Phys 1997;36:L382e5.

28. Deguchi T, Sekiguchi K, Nakamura A, Sota T, Matsuo R, Chichibu S, Nakamura S.Quantum-confined Stark effect in an AlGaN/gan/AlGaN single quantum well structure.Jpn J App Phys 1999;38:L914e6.

29. Davies JH. The physics of low-dimensional semiconductors. Cambridge University Press;1998.


30. Im JS, Kollmer H, Off J, Sohmer A, Scholz F, Hangleiter A. Reduction of oscillatorstrength due to piezoelectric fields in GaN/AlxGa1-xN quantum wells. Phys Rev B 1998;57:R9435e8.

31. Renard J, Kandaswamy P, Monroy E, Gayral B. Suppression of nonradiative processes inlong-lived polar GaN/AlN quantum dots. Appl Phys Lett 2009;71:131903e5.

32. Waltereit P, Brandt O, Trampert A, Grahn HT, Menniger J, Ramsteiner M, Reiche M,Ploog KH. Nitride semiconductors free of electrostatic fields for efficient white light-emitting diodes. Nature 2000;406:865e8.

33. Konar A, Verma A, Fang T, Zhao P, Jana R, Jena D. Charge Transport in non-polar andsemi-polar III-V nitride heterostrucutres. SemicondSci Technol 2012;27:024018.

34. Romanov AE, Baker TJ, Nakamura S, Speck JS. Strain-induced polarization in wurtziteIII-nitride semipolar layers. J Appl Phys 2006;100:023522.

35. Feezell DF, DenBaars SP, Speck JS, Nakamura S. Recent performance of nonpolar andsemipolar GaN-based light emitting diodes and laser diodes. In: Compound semiconductorIntegrated Circuit Symposium 2007, IEEE; 2007.

36. Masui H, Nakamura S, DenBaars SP, Mishra UK. Nonpolar and semipolar III-nitride light-emitting diodes: achievements and challenges. IEEE TED 2010;57:88e100.

37. Nakamura S. Current status of gan-based solid-state lighting. MRS Bull 2009;34:101e7.

38. Piprek J. Efficiency droop in nitride-based light-emitting diodes. Phys Status Solidi (A)2010;207:2217e25.

39. Pan CC, Tanaka S, Wu F, Zhao Y, Speck JS, Nakamura S, DenBaars SP, Feezell D. High-power, low-efficiency-droop semipolar (20-2-1) single-quantum-well blue light-emittingdiodes. Appl Phys Expr 2012;5:062103.

40. Schubert MF, Xu J, Kim JK, Schubert EF, Kim MH, Yoon S, Lee SM, Sone C, Sakong T,Park Y. Polarization-matched GaInN/AlGaInN multi-quantum-well light-emitting diodeswith reduced efficiency droop. Appl Phys Lett 2008;93:041102.

41. Xu J, Schubert MF, Noemaun AN, Zhu D, Kim JK, Schubert EF, Kim MH, Chung HJ,Yoon S, Sone C, Park Y. Reduction in efficiency droop, forward voltage, ideality factor,and wavelength shift in polarization-matched GaInN/GaInN multi-quantum-well light-emitting diodes. Appl Phys Lett 2009;94:011113.

42. Nakamura S, Senoh M, Mukai T. Highpower InGaN/GaN double heterostructure violetlight emitting diodes. Appl Phys Lett 1993;62:2390e2.

43. Simon J, Protasenko V, Lian C, Xing H, Jena D. Polarization-induced hole doping inwideeband-gap uniaxial semiconductor heterostructures. Science 2010;327:60e4.

44. Jena D. Polarization induced electron populations in III-V nitride semiconductors:transport, growth and device applications [PhD thesis]. 2003. UCSB.

45. Nakamura S, Mukai T, Senoh M. Candela-class high-brightness InGaN/AlGaN double-heterostructure blue-light-emitting diodes. Appl Phys Lett 1994;64(1687):1994.

46. Huang C-F, Liu T-C, Lu Y-C, Shiao W-Y, Chen Y-S, Wang J-K, Lu C-F, Yang CC.Enhanced efficiency and reduced spectral shift of green light-emitting-diode epitaxialstructure with prestrained growth. J Appl Phys 2008;104:123106e7.

47. Feezell DF, Schmidt MC, DenBaars SP, Nakamura S. Development of nonpolar andsemipolar InGaN/gan visible light-emitting diodes. MRS Bull 2009;34:318e23.

48. Park SH, Ahn D, Chuang SL. Electronic and optical properties of a- and m-plane wurtziteInGaN/GaN quantum wells. IEEE J Quan Electron 2007;43:1175e82.

49. Schmidt MC, Kim K-C, Farrell RM, Feezell DF, Cohen DA, Saito M, Fujito K, Speck JS,Denbaars SP, Nakamura S. Demonstration of nonpolar m-plane InGaN/GaN laser diodes.Jpn J Appl Phys 2007;46:L190e1.


50. Farrell RM, Feezell DF, Schmidt MC, Haeger DA, Kelchner KM, Iso K, Yamada H,Saito M, Fujito K, Cohen DA, Speck JS, DenBaars SP, Nakamura S. Continuous-waveoperation of AlGaN cladding-free nonpolar m-plane InGaN/GaN laser diodes. Jpn JAppl Phys 2007;46:L761e3.

51. Koslow IL, Sonoda J, Chung RB, Pan C, Brinkley S, Ohta H, Nakamura S, DenBaars SP.High power and high efficiency blue InGaN light emitting diodes on (30-3-1) free-standingsemipolar bulk GaN substrate. Jpn J Appl Phys 2010;49:080203e5.

52. Zhao Y, Tanaka S, Pan C-C, Fujito K, Feezell D, Speck JS, DenBaars SP, Nakamura S.High-power blue-violet semipolar (2021) InGaN/gan light-emitting diodes with low effi-ciency droop at 200 a/cm2. Appl Phys Express 2011;4:082104-1e082104-3.

53. Arif RA, Ee YK, Tansu N. Polarization engineering via staggered InGaN quantum wellsfor radiative efficiency enhancement of light emitting diodes. Appl Phys Lett 2007;91:091110e2.

54. Zhao H, Arif RA, Tansu N. Design analysis of staggered InGaN quantum wells light-emitting diodes at 500-540 nm. IEEE J Sel Top Quan Electron 2009;15:1104e14.

55. Zhao H, Liu G, Li X-H, Huang GS, Poplawsky JD, Penn ST, Dierolf V, Tansu N. Growthsof staggered InGaN quantum wells light-emitting diodes emitting at 520e525 nmemploying graded growth temperature profile. Appl Phys Lett 2009;95:061104e6.

56. Zhao HP, Liu GY, Li XH, Arif RA, Huang GS, Poplawsky JD, Penn ST, Dierolf V,Tansu N. Design and characteristics of staggered InGaN quantum wells light-emittingdiodes in the green spectral regimes. IET Optoelectron 2009;3:283e95.

57. Park SH, Ahn D, Kim JW. High-efficiency Staggered 530 nm InGaN/InGaN/gan quantum-well lightemitting diodes. Appl Phys Lett 2009;94:041109e11.

58. Park SH, Ahn D, Koo BH, Kim JW. Electronic and optical properties of staggered InGaN/InGaN quantum-well light-emitting diodes. Phys Status Solidi, A Appl Mater Sci 2009b;206:2637e40.

59. Liao CT, Tsai MC, Liou BT, Yen SH, Kuo YK. Improvement in output power of a 460 nmInGaN light-emitting diode using staggered quantum well. J Appl Phys 2010;108:063107-1e063107-6.

60. Zhao H, Tansu N. Optical gain characteristics of staggered InGaN quantum well lasers.J Appl Phys 2010;107:113110. 113110-1e113110-12.

61. Park J, Kawakami Y. Photoluminescence property of InGaN Single quantum well withembedded AlGaN d layer. Appl Phys Lett 2006;88:202107e9.

62. Park SH, Park J, Yoon E. Optical gain in InGaN/GaN quantum well structures withembedded AlGaN delta layer. Appl Phys Lett 2007;90:023508e10.

63. Zhao H, Liu G, Arif RA, Tansu N. Current injection efficiency induced efficiency-droop inInGaN quantum welllight-emitting diodes. Solid State Electronics 2010;54:1119e24.

64. Yang Z, Li R, Wei Q, Yu T, Zhang Y, Chen W, Hu X. Analysis of optical gain property inthe InGaN/GaN triangular shaped quantum well under the piezoelectric field. Appl PhysLett 2009;94:061120e2.

65. Ho I-H, Stringfellow GB. Solid phase immiscibility in InGaN. Appl Phys Lett 1996;69:2701e3.

66. Teles LK, Furthm€uller J, Scolfaro LMR, Leite JR, Bechstedt F. First-principles calcula-tions of the thermodynamic and structural properties of strained InxGa1-xN and AlxGa1-xNalloys. Phys Rev B 2000;62:2475e585.

67. Xia Y, Hou W, Zhao L, Zhu M, Detchprohm T, Wetzel C. Boosting green GaInN/ganlight-emitting diode performance by a GaInN underlying layer. IEEE Trans ElectronDevices 2010;57:2639e43.


68. Nakamura S, Chichibu SF. Introduction to nitride semiconductor blue laser and lightemitting diode. London and New York: CRC Press; 2000.

69. Tadatomo K, Okagawa H, Ohuchi Y, Tsunekawa T, Jyouichi T, Imada Y, Kato M,Kudo H, Taguchi T. High output power InGaN ultraviolet light-emitting diodes fabricatedon patterned substrates using Metalorganic vapor phase epitaxy. Phys Stat Sol (A) 2001;188:121e5.

70. Yamada M, Mitani T, Narukawa Y, Shioji S, Niki I, Sonobe S, Deguchi K, Sano M,Mukai T. InGaN-based near-ultraviolet and blue-light-emitting diodes with high externalquantum efficiency using a patterned sapphire substrate and a Mesh Electrode. Jpn J ApplPhys 2002;41:L1431e3.

71. Wuu DS, Lin S-H, Han P, Horng R-H, Hsu T-C, Huo DT-C, Jou M-J, Yu Y-H, Lin A.Efficiency improvement of near-ultraviolet InGaN LEDs using patterned sapphire sub-strates. IEEE J Quan Electronics 2005a;41:1403e9.

72. Wuu DS, Wang WK, Shih WC, Horng RH, Lee CE, Lin WY, Fang JS. Enhanced outputpower of near-ultraviolet InGaN-GaN LEDs grown on patterned sapphire substrates. IEEEPhoton Tech Lett 2005b;17:288e90.

73. Kim AY, Gotz W, Steigerwald DA, Wierer JJ, Gardner NF, Sun J, Stockman SA,Martin PS, Krames MR, Kern RS, Steranka FM. Performance of high-power AlInGaNlight emitting diodes. Phys Stat Sol (A) 2001;188:15e21.

74. Akita K, Kyono T, Yoshizumi Y, Kitabayashi H, Katayama K. Improvements of externalquantum efficiency of InGaN-based blue light-emitting diodes at high current densityusing GaN substrates. J Appl Phys 2007a;101:033104-1e033104-5.

75. Akita K, Kyono T, Yoshizumi Y, Kitabayashi H, Katayama K. Characteristics of InGaNlightemitting diodes on GaN substrates with low threading dislocation densities. Phys StatSol (A) 2007b;204:246e50.

76. Leea JW, Taka Y, Kima J-Y, Honga H-G, Chaea S, Mina B, Jeonga H, Yooa J, Kimb J-R,Parka Y. Growth of high-quality InGaN/GaN LED structures on (111) Si substrates withinternal quantum efficiency exceeding 50%. J Cryst Growth 2011;315:263e6.

77. Liu Z, Wei T, Guo E, Yi X, Wang L, Wang J, Wang G, Shi Y, Ferguson I, Li J. Efficiencydroop in InGaN/GaN multiple-quantum-well blue light-emitting diodes grown on free-standing GaN substrate. Appl Phys Lett 2011;99:091104e6.

78. Zhang M, Bhattacharya P, Guo W. InGaN/GaN self-organized quantum dot green lightemitting diodes with reduced efficiency droop. Appl Phys Lett 2010;97:011103e5.

79. Sheu JK, Chi GC, Jou MJ. Enhanced output power in an InGaN-GaN multiquantum-welllight-emitting diode with an InGaN current-spreading layer. IEEE Photon Tech Lett 2001;13:1164e6.

80. Pan C-C, Lee C-M, Liu J-W, Chen G-T, Chyi J-I. Luminescence efficiency of InGaNmultiple-quantum-well ultraviolet light-emitting diodes. Appl Phys Lett 2004;84:5249e51.

81. Yamamoto S, Zhao Y, Pan C-C, Chung RB, Fujito K, Sonoda J, DenBaars SP,Nakamura S. High-efficiency single-quantum-well green and yellow-green light-emittingdiodes on semipolar (20 21) GaN substrates. Appl Phys Express 2010;3:122102e4.

82. Chang CS, Chang SJ, Su YK, Kuo CH, Lai WC, Lin YC, Hsu YP, Shei SC, Tsai JM,Lo HM, Ke JC, Sheu JK. High brightness InGaN green LEDs with an ITO on nþþ-SPSupper contact. IEEE Trans Elect Devices 2003;50:2208e12.

83. Chen CH, Chang SJ, Su YK, Chi GC, Sheu JK, Chen JF. High-efficiency InGaN-GaNMQW green light-emitting diodes with CART and DBR structures. IEEE J Sel Top QElect 2002;8:284e8.


84. Lundin WV, Nikolaev AE, Sakharov AV, Zavarin EE, Usov SO, Sizov VS, Zakgeim AL,Chernyakov AE, Tsatsul’Nikov AF. High-efficiency InGaN/GaN/AlGaN light-emittingdiodes with short-period InGaN/GaN superlattice for 530e560 nm range. Tech PhysLett 2010;36:1066e8.

85. Chang Y-L, Wang JL, Li F, Mi Z. High efficiency green, yellow, and amber emission fromInGaN/GaN dot-in-a-wire heterostructures on Si(111). Appl Phys Lett 2010;96:013106e8.

86. Guo W, Zhang M, Banerjee A, Bhattacharya P. Catalyst-free InGaN/gan nanowire lightemitting diodes grown on (001) silicon by molecular beam epitaxy. Nano Lett 2010;10:3355e9.

87. Zhang S, Cui K, Korinek A, Botton GA, Zetian M. High-efficiency InGaN/gan dot-in-a-wire red light-emitting diodes. IEEE Photon Tech Lett 2012;24:321e3.

88. Sato H, Chung RB, Hirasawa H, Fellows N, Masui H, Wu F, Saito M, Fujito K, Speck JS,DenBaars SP, Nakamura S. Optical properties of yellow light-emitting diodes grown onsemipolar (1122) bulk GaN substrates. Appl Phys Lett 2008;92:221110e2.

89. Hader J, Moloney JV, Pasenow B, Koch SW, Sabathil M, Linder N, Lutgen S. On theimportance of radiative and Auger losses in GaN based quantum wells. Appl Phys Lett2008;92:261103e5.

90. Delaney KT, Rinke P, Van de Walle CG. Auger recombination rates in nitrides from firstprinciples. Appl Phys Lett 2009;94:191109e11.

91. Pope IA, Smowton PM, Blood P, Thomson JD, Kappers MJ, Humphreys CJ. Carrierleakage in InGaN quantum well light-emitting diodes emitting at 480 nm. Appl Phys Lett2003;82:2755e7.

92. Kim M-H, Schubert MF, Dai Q, Kim JK, Schubert EF, Piprek J, Park Y. Origin of effi-ciency droop in GaN-based light-emitting diodes. Appl Phys Lett 2007;91:183507e9.

93. Hangleiter A, Hitzel F, Netzel C, Fuhrmann D, Rossow U, Ade G, Hinze P. Suppression ofnonradiative recombination by V-shaped pits in GaInN/GaN quantum wells produces alarge increase in the light emission efficiency. Phys Rev Lett 2005;95:127402.

94. Schubert MF, Chhajed S, Kim JK, Schubert EF, Koleske DD, Crawford M, Lee SR,Fischer AJ, Thaler G, Banas MA. Effect of dislocation density on efficiency droop inGaInN/GaN light-emitting diodes. Appl Phys Lett 2007;91:231114e6.

95. Monemar B, Sernelius BE. Defect related issues in the ‘current roll-off’ in InGaN basedlight emitting diodes. Appl Phys Lett 2007;91:181103e5.

96. Van der Laak NK, Oliver RA, Kappers MJ, Humphreys CJ. Role of gross well-widthfluctuations in bright, green-emitting single InGaN/GaN quantum well structures. ApplPhys Lett 2007;90:121911e3.

97. Laubsch A, Sabathil M, Bruederl G, Wagner J, Strassburg M, Baur E, Braun H,Schwarz UT, Lell A, Lutgen S, Linder N, Oberschmid R, Hahn B. Measurement of theinternal quantum efficiency of InGaN quantum wells. Proc SPIE 2007;6486. 64860J-1e64860J-10.

98. Zhao H, Liu G, Tansu N. Analysis of InGaN-delta-InN quantum wells for light-emittingdiodes. Appl Phys Lett 2010;97:131114e6.

99. Tanaka S, Iwai S, Aoyagi Y. Self-assembling GaN quantum dots on AlxGa1�xN surfacesusing a surfactant. Appl Phys Lett 1996;69:4096e8.

100. Adelmann C, Daudin B, Oliver RA, Briggs GAD, Rudd RE. Nucleation and growth ofGaN∕AlN quantum dots. Phys Rev B 2004;70:125427-1e125427-8.

101. Brown JS, Petroff PM, Wu F, Speck JS. Optical properties of gan/AlN(0001) quantumdots grown by plasma-assisted molecular beam epitaxy. Jpn J Appl Phys 2006;45:L669e72.


102. Narukawa Y, Kawakami Y, Fujita S, Makamura S. Dimensionality of excitons in laser-diode structures composed of InxGa1�xN multiple quantum wells. Phys Rev B 1999;59:10283e8.

103. O’Donnell KP, Martin RW, Middleton PG. Origin of luminescence from InGaN diodes.Phys Rev Lett 1999;82:237e40.

104. Martin RW, Middleton PG, O’Donnell KP, Van der Stricht W. Exciton localization andthe Stokes’ shift in InGaN epilayers. Appl Phys Lett 1999;74:263.

105. Chichibu S, Azuhata T, Sota T, Nakamura S. Spontaneous emission of delocalized ex-citons in InGaN single and multiquantum well structures. Appl Phys Lett 1996;69:4188e90.

106. Narukawa Y, Kawakami Y, Funato M, Fujita S, Fujita S, Nakamura S. Role of self-formedInGaN quantum dots for exciton localization in the purple laser diode emitting at 420 nm.Appl Phys Lett 1997;70:981e3.

107. Chichibu F, Azuhata T, Sugiyama M, Kitamura T, Ishida Y, Okumura H, Nakashini H,Sota T, Mukai T. Optical and structural studies in InGaN quantum well structure laserdiodes. J Vac Sci Technol B 2001;19:2177e83.

108. Lai Y-L, Liu C-P, Hsuen T-H, Lin Y-H, Chung H-C, Lin R-M, Chen Z-Q. The influence ofquasi-quantum dots on physical properties of blue InGaN/GaN multiple quantum wells.Nanotechnology 2006;17:4300e6.

109. Lai Y-L, Liu C-P, Chen Z-Q. Study of the dominant luminescence mechanism in InGaN/GaN multiple quantum wells comprised of ultrasmall InGaN quasiquantum dots. ApplPhys Lett 2005;86:121915e7.

110. Hirayama H, Tanaka S, Ramvall P, Aoyagi Y. Intense photoluminescence from self-assembling InGaN quantum dots artificially fabricated on AlGaN surfaces. Appl PhysLett 1998;72:1736e8.

111. Adelmann C, Simon J, Feuillet G, Pelekanos NT, Daudin B, Fishman G. Self-assembledInGaN quantum dots grown by molecular-beam epitaxy. Appl Phys Lett 2000;76:1570e2.

112. Xu T, Nikiforov AY, France R, Thomidis C, Williams A, Moustakas TD. Blueegreenered LEDs based on InGaN quantum dots grown by plasma-assisted molecularbeam epitaxy. Phys Stat Sol (A) 2007;204:2098e102.

113. Moustakas TD, Xu T, Thomidis C, Nikiforov AY, Zhou L, Smith DJ. Growth of III-nitridequantum dots and their applications to blue-green LEDs. Phys Stat Sol (A) 2008;205:2560e5.

114. Bhattacharya P, Zhang M, Hinckley J. Tunnel injection In0.25Ga0.75N/GaN quantum dotlight-emitting diodes. Appl Phys Lett 2010;97:251107.

115. Mukai T, Morita D, Nakamura S. High-power UV InGaN/AlGaN double-heterostructureLEDs. J Cryst Growth 1998;189e190:778e81.

116. Khan MA, Skogman RA, Van Hove GM, Krishnankutty S, Kolbas RM. Photo-luminescence characteristics of AlGaN-GaN-AlGaN quantum wells. Appl Phys Lett 1990;56:1257e9.

117. Khan A, Balakrishnan K, Katona T. Ultraviolet light-emitting diodes based on group threenitrides. Nat Photon 2008;2:77e84.

118. Mukai T, Nagahama S, Iwasa N, Senoh M, Yamada T. Nitride light-emitting diodes.J Phys Condens Matter 2001;13:7089e98.

119. Yoshida H, Yamashita Y, Kuwabara M, Kan H. A 342-nm ultraviolet AlGaN multiplequantum-well laser diode. Nat Photon 2008;2:551e4.

120. Shatalov M, Gaevski M, Adivarahan V, Khan A. Room-temperature stimulated emissionfrom AlN at 214 nm. Jpn J Appl Phys 2006;45:L1286e8.


121. Kozodoy P, Hansen M, DenBaars SP, Mishra UK. Enhanced Mg doping efficiency inAl0.2Ga0.8N/GaN superlattices. Appl Phys Lett 1999;74:3681e3.

122. Goepfert ID, Schubert EF, Osinsky A, Norris PE. Demonstration of efficient p-type dopingin AlxGa1exN/GaN superlattice structures. Electron Lett 1999;35:1109e11.

123. Khan MA, Shatalov M, Maruska HP, Wang HM, Kuokstis E. IIIenitride UV devices. JpnJ Appl Phys 2005;44:7191e206.

124. Guo X, Schubert EF. Current crowding in GaN/InGaN light emitting diodes on insulatingsubstrates. J Appl Phys 2001;90:4191e5.

125. Adivarahan V, Wu S, Sun WH, Mandavilli V, Shatalov MS, Simin G, Yang JW,Maruska HP, Khan MA. High-power deep ultraviolet light-emitting diodes based on amicro-pixel design. Appl Phys Lett 2004;85:1838e40.

126. Hirayama H, Fujikawa S. Quaternary InAlGaN quantum-dot ultraviolet light-emittingdiode emitting at 335 nm fabricated by anti-surfactant method. Phys Stat Sol (C) 2008;5:2312e5.

127. Verma J, Kandaswamy PK, Protasenko V, Verma A, Grace Xing H, et al. Tunnel-injectionGaN quantum dot ultraviolet light-emitting diodes. Appl Phys Lett 2013;102:041103.

128. Verma J, Islam SM, Protasenko V, Kumar Kandaswamy P, Xing H, et al. Tunnel-injectionquantum dot deep-ultraviolet light-emitting diodes with polarization-induced doping inIII-nitride heterostructures. Appl Phys Lett 2014;104:021105.

129. Bayerl D, Islam SM, Jones CM, Protasenko V, Jena D, et al. Deep ultraviolet emissionfrom ultra-thin GaN/AlN heterostructures. Appl Phys Lett 2016;109:241102.

130. Islam SM, Protasenko V, Rouvimov S, Xing H, Jena D. Sub-230 nm deep-UV emissionfrom GaN quantum disks in AlN grown by a modified StranskieKrastanov mode. Jpn JAppl Phys 2016;55:05FF06.

131. Islam SM, Lee K, Verma J, Protasenko V, Rouvimov S, Xing H, Jena D. 232-270 nmDeep-UV LEDs using Monolayer thin Binary GaN/AlN quantum heterostructures. ApplPhys Lett 2017;110:041108.

132. Schubert EF, Wang YH, Cho AY, Tu LW, Zydzik GJ. Resonant cavity light emittingdiode. Appl Phys Lett 1992;60:921e3.

133. Song YK, Diagne M, Zhou H, Nurmikko AV, Schneider RP, Takeuchi T. Resonant-cavityInGaN quantum-well blue light-emitting diodes. Appl Phys Lett 2000;77:1744e6.

134. Calle F, Naranjo FB, Fernandez S, Sanchez-Garcia A, Calleja E, Munoz E. NitrideRCLEDs grown by MBE for POF applications. Phys Stat Sol A 2002;192(2):277e85.

135. Dorsaz J, Carlin JF, Zellweger CM, Gradecak S, Ilegems M. InGaN/GaN resonant-cavityLED including an AlInN/GaN Bragg mirror. Phys Stat Sol (A) 2004;201:2675e8.

136. Mastro MA, Caldwell JD, Holm RT, Henry RL, Eddy CR. Design of gallium nitrideresonant cavity light-emitting diodes on Si substrates. Adv Mater 2008;20:115e8.

137. Feltin E, Castiglia A, Cosendey G, Sulmoni L, Carlin JF, Grandjean N, Rossetti M,Dorsaz J, Laino V, Duelk M, Velez C. Broadband blue superluminescent light-emittingdiodes based on GaN. Appl Phys Lett 2009;95:081107.

138. Rossetti M, Dorsaz J, Rezzonico R, Duelk M, Velez C, Feltin E, Castiglia A, Cosendey G,Carlin JF, Grandjean N. High power blue-violet superluminescent light emitting diodeswith InGaN quantum wells. Appl Phys Exp 2010;3:061002e4.

139. Kafar A, Stanczyk S, Grzanka S, Czernecki R, Leszczynski M, Suski T, Perlin P. Cavitysuppression in nitride based superluminescent diodes. J App Phys 2012;111:083106.

140. Kuykendall T, Ulrich P, Aloni S, Yang P. Complete composition tunability of InGaNnanowires using a combinatorial approach. Nat Mater 2007;6:951e6.


141. Qian F, Gradecak S, Li Y, Wen CY, Lieber CMP. Core/multishell nanowire hetero-structures as multicolor, high-efficiency light-emitting diodes. Nano Lett 2005;5:2287e91.

142. Pledran BD. Polaritronics in view. Nature 2008;453:297e8.143. Deng H, Welhs G, Snoke D, Bloch J, Yamamoto Y. Polariton lasing vs. photon lasing in a

semiconductor microcavity. PNAS 2003;100(26):15318e23.144. Christopoulos S, Hogersthal GBH, Grundy AJD, Lagoudakis PG, Kavokin AV,

Baumberg JJ, Christmann G, Butte R, Feltin E, Carlin JF, Grandjean N. Room-temperaturepolariton lasing in semiconductor microcavities. Phys Rev Lett 2007;98:126405.

145. Das A, Heo J, Jankowski M, Guo W, Zhang L, Deng H, Bhattacharya P. Room temper-ature ultralow threshold GaN nanowire polariton laser. Phys Rev Lett 2011;107:066405.

146. Tsintzos SI, Pelekanos NT, Konstantinidis G, Hatzopoulos Z, Savvidis PG. A GaAspolariton light-emitting diode operating near room temperature. Nature 2008;453:372e5.

147. Lu TC, Chen JR, Lin SC, Huang SW, Wang SC, Yamamoto Y. Room temperature currentinjection polariton light emitting diode with a hybrid microcavity. Nano Lett 2011;11:2791e5.

Further reading

1. Iso K, Yamada H, Hirasawa H, Fellows N, Saito M, Fujito K, Denbaars SP, Speck JS,Nakamura S. High brightness blue InGaN/GaN light emitting diode on nonpolar m-planebulk GaN substrate. Jpn. J. Appl. Phys. 2007;46:L960e2.

2. Kneissl M, Kolbe T, Chua C, Kueller V, Lobo N, Stellmach J, Knauer A, Rodriguez H,Einfeldt S, Yang Z, Johnson N, Weyers M. Advances in group III-nitride-based deep UVlight-emitting diode technology. Semicond.Sci. Technol. 2011;26:014036e41.

3. Maruska HP, Tietjen JJ. The preparation and properties of vapor-deposited single-crystal-lineGaN. Appl. Phys. Lett. 1969;15:327.

4. Nguyen HPT, Zhang S, Cui K, Han X, Fathololoumi S, Couillard M, Botton GA, Mi Z. p-Type modulation doped InGaN/GaN dot-in-a-wire white-light-emitting diodes mono-lithically grown on Si(111). Nano Lett 2011;11:1919e24.

5. Schwarz UT, Kneissl M. Nitride emitters go nonpolar. Phys. Stat. Sol. (RRL) 2007;1:A44e6.6. Vurgaftman I, Meyer JR. Band parameters for III-V compound semiconductors and their

alloys. J. Appl. Phys. 2001;89:5815e75.7. Yu KM, Liliental-Weber Z, Walukiewicz W, ShanW, Ager JW, Li SX, Jones RE, Haller EE,

Lu H, Schaff WJ. On the crystalline structure, stoichiometry and band gap of InN thin films.Appl. Phys. Lett. 2005;86:071910e2.



Colour tuneable LEDs andpixelated micro-LED arrays 12Yuk Fai Cheung, Zetao Ma, Hoi Wai ChoiThe University of Hong Kong, Hong Kong

12.1 Introduction: motivation for color tuning andreview of existing technologies

Over the past decade, light-emitting diodes (LEDs) based on the III-nitride materialsystem have been developed extensively with the target of replacing incandescentand gas discharge lamps as energy efficient solid-state light sources. Tremendousprogress has indeed been achieved, owing to intensive efforts carried out at researchlaboratories in institutions and in the industry alike. Major advancements in materialepitaxy, device processing, and chip packaging have contributed to the significant suc-cess, as replacements of liquid crystal display (LCD) backlighting, traffic light, streetlighting, and general lighting are gradually being replaced by LED sources in manycities around the globe, as efficiencies of these emitters move closer and closer to theirfull potentials. This is particularly encouraging in the midst of concerns overdepletionof energy resources, not to mention the long list of advantages offered by LEDs overolder technologies.

One major difference between LEDs and incandescent or fluorescent lighting liewith the emission spectrum. The monochromatic spectrum of an LED means that colorconversion of combination is always necessary for the generation of broadband light.On the other hand, the broadband emission of incandescence spares this necessity;nevertheless, the spectrum is excessively broadband with most of its spectral compo-nents in the infrared so that wasteful filtering becomes necessary. Similarly, thediscrete wavelength emission from a gas discharge calls for additional fluorescentmaterials for useful emission in the visible region.

The monochromatic nature of LED emission gives rise to the opportunity of pro-ducing energy-efficient color-tunable light emitters. Colors may be varied by tuningthe wavelength of the light source or by the addition or mixing of different colors.To date, an emissive light source that is wavelength tunable across the visible spectrumis not yet available; therefore, color mixing based on the additive principle is oftenused. For instance, colors can be mixed by projection of beams of the primary colorsonto a surface. On the other hand, devices such as liquid crystal displays (LCD) rely onthe filtering of light through a mosaic of color filters from a rear broadband lightsource.

Unlike conventional light sources based on incandescence or fluorescence, light-emitting diodes are fixed monochromatic wavelength miniature light emitters, making


https://doi.org/10.1016/B978-0-08-101942-9.00012-5

them highly suited for additive color mixing. Many applications, such as panel dis-plays, mood lighting or even biological excitation of cells, call for light sources thatare color tunable across the visible spectrum and beyond. Solutions based on thisconcept, in the form of red-green-blue (RGB) LEDs whereby chips emitting theprimary colors are bonded onto the same package adjacent to each other, are nowavailable and have been adopted on LED panel displays.1 Nevertheless, a major draw-back of this approach is spatial color variations giving rise to nonideal color mixing asemission cones from the discrete devices do not overlap with each other completely.2

Diffusers are often used to overcome this problem, although optical losses of w20%are inevitable,3 together with a loss of color sharpness and richness.

There have been pockets of demonstrations of LEDs with color-tuning capabilities,including devices based on multiple junctions, quantum dots or nanocolumns, albeitwith limited tuning ranges. To overcome such limitations, two different strategies ofproviding color-tuning capabilities to LEDs have been designed and implemented.

The first approach involves three separate RGB emitters vertically stacked on top ofeach other. LED chips of the primary colors can be physically stacked on top of eachother to produce a color-tunable LED. In such an arrangement, light rays emitted byeach LED can be combined and mixed naturally as rays from the lower LEDs passthrough the upper LEDs and mix with rays emitted by the latter.

The second approach involves arraying interconnected micrometer scale emitters ofdifferent colors. A true single-chip solution employing a group-addressable micropixe-lated emitter, in conjunction with jet-printed color-conversion pixels is realized. Beinga single-chip solution, these group-addressable micro-LED emitters are readily scal-able for integration into full-color emissive microdisplays, which certainly would bea breakthrough in emissive display technology.

12.2 Stacked LEDs

12.2.1 Initial idea

This approach is based upon optimization of the optical pathways of the LED emis-sions. Optical mixing requires that the radiation patterns from the discrete LEDs over-lap with each other. Instead of attempting to mix the beams using external optics, RGBdevices are physically placed such that their optical paths are aligned. The schematicdiagram of Fig. 12.1 illustrates this idea. In this way, the emissions from the threeLEDs are naturally mixed without the need for additional optics.

Implementation of this idea can be realized by stacking the RGB chips on top ofeach other, in a stacking topography. This is possible by virtue of the fact that GaNLED chips are grown on transparent sapphire substrates; light can transmit throughthe chip without absorption or significant attenuation. The AlInGaP red LED, with anontransparent (to visible light) GaAs substrate must be placed at the bottom of thestack (which happens to be the required sequence as explained shortly). An InGaNgreen LED is placed on top of the red LED and a blue InGaN LED is subsequentlyplaced at the top of the stack. Such a stacking strategy ensures optimal color mixing.


Adopting this stacking sequence also ensures that light emitted from lower devices(with narrower bandap) will transmit through upper devices (with wider bandgaps).The sapphire substrates of the InGaN LEDs are also transparent to visible light. Aschematic diagram of the proposed device is illustrated in Fig. 12.2. The three devicesare either connected in parallel or are individually controllable electrically. With theparallel connection method, the cathodes and anodes are interconnected, with theinsertion of appropriate resistors for adjusting the required bias voltages of each device(depending on the required proportions of red, green and blue spectral components).Such a two-terminal device acts as an all-semiconductor white light LED.

Another implementation gives user access to the individual cathodes, while inter-connecting their anodes, resulting in a four-terminal device. When all three devicesare illuminated, the optically mixed output results in polychromatic light with threespectral peaks or white light with the right proportions of red, green, and blue. Varyingthe proportions of the primary colors also offers white light of different color temper-atures. On the other hand, monochromatic light can be obtained by turning on a singledevice only. Other colors can be tuned by lighting up two or three devices simulta-neously and adjusting appropriate bias voltages. This proposed stacked design

MQWsInGaN blue LED

InGaN green LED

AllnGaP red LED

Figure 12.1 Schematic diagram depicting the transmission of light rays through the overlyinglight-emitting diode chips.

Figure 12.2 Schematic diagram illustrating the idea of light-emitting diode chip stacking.

Colour tuneable LEDs and pixelated micro-LED arrays 417

involves no color conversion in generating polychromatic light and is thus conversionloss free.

The red, green, and blue LED chips used in this study emit with center wavelengthsof 640, 510, and 470 nm, respectively, fabricated from metal organic chemical vapordeposition (MOCVD) grown AlInGaP on GaAs and InGaN on sapphire wafers. Thesapphire substrates of the nitride wafers have been thinned down to w150 mm, fol-lowed by fabrication of devices via standard microfabrication processes.

Assembly of the stack begins with attaching a red LED die to a TO-can using elec-trically conductive adhesive. A small volume of ultraviolet (UV) curing optical adhe-sive is dispensed onto the surface of the red LED chip, just enough to cover theemissive region, before the green LED chip is mounted on top using a manual diebonder. The bonding pads must not be covered by the epoxy. The blue LED chip ismounted on its top in the same manner. Once the chips are aligned in place, the stackis exposed to UV light under a deuterium lamp. The adhesive hardens and the stackwas fixed in place. Finally, the pads are wire bonded to the package. Bias voltagesare applied to the terminals to test the functionality of the assembled device. Althoughthe LEDs light up, the optical output is not as homogenous as expected, as shown inFig. 12.3, although it is much improved compared with conventional RGB devices. Asobserved, the contribution of light emission from the sidewalls of the chips has beenneglected in the design, prompting for further improvements.

Figure 12.3 An initial version of an light-emitting diode stack; sidewall emissions contribute toinhomogeneous color mixing.


12.2.2 Second generation LED stack with inclined sidewalls

A seemingly simple solution to this problem is to block sidewall emission; however,this would significantly reduce light extraction, defeating the motivation of our stackdesign of maximizing efficiency. A solution is needed to channel laterally propagatinglight into the vertical direction for emission through the top window; this can be pro-vided for by an LED of inverted pyramid geometry, with mirror coating on its inclinedsidewalls. The inclined sidewall can serve as a reflector, redirecting light rays whichwould otherwise be trapped in a cuboid LED, for extraction through the top emittingarea; examples of such rays are indicated in blue in Fig. 12.4. In the report by M.R.Krames et al., AlInGaP devices of such geometry are prepared by beveled dicing tech-nique.4 This would be difficult for GaN chips with hard sapphire substrates. Instead, asingle-step laser-micromachining approach is adopted for the formation of the angledfacets.

In traditional laser micromachining, a focused beam is directed to the wafer orthog-onally. Using a modified set-up that involves the insertion of a laser-turning mirrorbetween the focusing objective and the wafer, the beam is redirected to strike the waferat an oblique angle.5 This oblique incident beam is used for the dicing of green andblue GaN LED chips, which shapes the chips into inverted pyramids as it cuts. Thesetup for laser micromachining consists of a UV laser source, beam focusing optics,and an x-y motorized translation stage. The laser source is a third harmonic ND:YLF diode-pumped solid state laser manufactured by Spectra Physics. The laser emitsat 349 nm, although the pulse repetition rate ranges from single pulse to 5 kHz. At areference diode current of 3.2 A, the pulse energy is 120 mJ at a repetition rate of1 kHz, with a pulse width of around 4 ns. The transverse electromagnetic00 beamallows for tight focusing, offering high spatial resolution. After beam expansion andcollimation through a beam expander, the laser beam is reflected 90 degrees by adielectric laser line mirror and focused onto the horizontal machining plane to avery tiny spot several micrometers in diameter with a focusing triplet. All opticsused are made of UV fused silica and are antireflection coated. The additional featureof our set-up, as illustrated in the schematic diagram of Fig. 12.5, is the insertion of aUV mirror at an oblique angle within the optical path between the focusing optics andthe machining plane, which serves to deflect the convergent beam to strike the sampleat an oblique angle with respect to the horizontal working plane. The size of the beam

n-GaN

p-GaNMQW

Sapphire

41°

67°

23°

Figure 12.4 Typical light rayspropagating in a truncatedpyramidal geometry (TP-LED).


at the focal point is not only limited by the capability of the UV objective lens but isalso sensitive to the coaxiality of the optics. With this modified set-up, it is relativelyeasy to optimize and monitor the beam through the tube lens imaged with a charge-coupled device (CCD) camera. Once the optical setup is optimized before insertionof the tilting mirror, the mirror can be inserted without affecting the coaxiality ofthe laser beam, so that the dimension of the beam spot remains unaffected.

The beam can be effectively applied for microsectioning with nonvertical sidewallprofiles. The angle of incidence of the deflected laser beam on the wafer is 2q, where qis the angle between the plane of the mirror to the normal. This angle is readily andprecisely controlled by mounting the mirror onto a rotation stage; thus, the incidentangle can be varied over a wide range. We have used a UV objective with a focallength of 75 mm, based on two considerations. First, the focal length should be longenough to accommodate the mirror in the optical path. Second, an ideal tool for thefabrication of microstructures should have a very long penetration depth and negligiblelateral dispersion. Nevertheless, an objective lens with a longer focal length also pro-duces a larger focused beam spot. The two parameters are related via the followingequation:

d ¼ 4lM2f

pD(12.1)

where M2 quantifies the beam quality, l is the wavelength of the laser beam, f is thefocal length, and D is the diameter of the incident beam.

Chips of truncated pyramidal geometry (TP-LED) are diced and shaped byapplying four successive oblique laser cuts onto the four sides of an LED fabricated

Z stage

Tilted mirror

XYΘ stage

Sample

UV objective lens

Tube lens

Broadband visiblelight source(for observingthe sample)

CCD camera

Collimatedns laser beam

(wavelength= 349 nm)

Figure 12.5 Laser micromachining setup, incorporating a “tilted” mirror for controlling theangle of incidence of the laser beam striking the sample surface.


using standard microfabrication procedures. A scanning electron microscope (SEM)image of a TP-LED is shown in Fig. 12.6(a), whereas Fig. 12.6(b) shows the TP-LED lighted up. A layer of Ag was selectively coated onto the angled sidewall by elec-tron beam evaporation (by covering the top face with photoresist), serving as a mirrorto redirect light into the vertical direction.

A ray-tracing simulation coded in MATLAB is carried out to predict the extractionefficiencies of TP- LEDs,6 taking into account light refraction based on Snell’s law (invector form),

rr ¼ rin þ ½2ðn$� rrÞ�n (12.2)

rt ¼ ðn1=n2Þrin þn�

n1=n2Þðn$� rrÞ þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1� ðn1=n2Þ2h1� ðn$� rrÞ2

ir on

(12.3)

where rin, rr, and rt are the normalized vectors of incident ray, reflected ray, andtransmitted ray, respectively, n1 and n2 are the refractive indices, and n is thenormalized normal vector of the interface; additionally, light reflectivity and trans-missivity are evaluated based on Fresnel Equation:

Rs ¼

8><

>:

n1 cosðqiÞ � n2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1� ½ðn1=n2ÞsinðqiÞ�2

q

n1 cosðqiÞ þ n2


q

9>=

>;

2

(12.4)

Rp ¼

8><

>:

n1


q� n2 cosðqiÞ

n1


qþ n2 cosðqiÞ

9>=

>;

2

(12.5)

Figure 12.6 (a) Scanning electron microscope image showing the truncated pyramidalgeometry (TP-LED) structure. The same device is turned on as shown in (b).


where Rs and Rp are reflection coefficients of s-polarized and p-polarized light ray, andqi is the incident angle, whereby cos(qi) can be obtained by the term (n$�rr) as used in(1) and (2); and the absorption rule,

I ¼ Io expð� aLÞ (12.6)

where Io and I are the intensities at the initial and final points of the light path, a is theabsorption coefficient (aGaN ¼ 150 cm�1 was used in the simulation), and L is thedistance travelled by each light ray.

The active region of an LED is modeled by a 50 � 50 array of point sources. Thelight extraction efficiency is evaluated by summing the intensities of light raysescaping from the LED divided by total intensities of light rays emitted:

h ¼X

q;f

f�Iq;f

��X

q;f

Iq;f (12.7)

where I is the light intensity emitted at a specific angle (q, f) and f(I) is the lightintensity extracted at that angle. Light extraction efficiency as a function of inclinationangle is plotted in Fig. 12.7. As observed from the graph, the light extraction efficiencyfor a cuboid structure (90 degrees) is 18.3%, rising to 33.9% for an inclination angle of50 degrees. Further increase in inclination angle results in little further improvement,whereas the cutting efficiency is reduced. Therefore, 50 degrees is deemed to be theoptimal choice.

Another stack is assembled using these modified chips following an identical pro-cedure. Fig. 12.8(a) shows a scanning electron microscope (SEM) image of the assem-bled device, showing the trilayer topology. The contact pads of each chip are exposedfor wire bonding. Fig. 12.8(b) shows emission from the current injected device. The

40 50 60 70 80 9015

20

25

30

35

Inclination angle

Ligh

t ext

ract

ion

effic

ienc

y (%

)

Figure 12.7 Plot of light extraction efficiency as a function of the inclination angle.


chips illustrated in the device of Fig. 12.8 have not thinned down for illustrative pur-poses; in the final version, chips were thinned down tow100 mm. This time round, thecolor-mixing capability of the stacked device is improved, offering optically mixedoutput through its top output aperture. For a fair and objective comparison, both thestacked LED and the RGB LED are biased to emit with identical Commission Inter-nationale de l’Eclairage (CIE) coordinates of (0.31, 0.31) at a total current of20 mA. Due to slight dissimilarities of the component chips, the bias voltages areslightly different. Optical micrographs of both devices operated under such conditionsare shown in Fig. 12.9. It is immediately apparent that color homogeneity is furtherimproved, with single spot color-mixed emission, in stark contrast to the three spotsof spatially separated light from the RGB device. However, monochromatic emission

(a) (b)

Figure 12.8 (a) Scanning electron microscope image showing an light-emitting diode (LED)stack assembled with LED chips of inverted-pyramidal geometry. (b) The same LED stackwith the devices turned on simultaneously.

Figure 12.9 Sidewall light leakage still evident from the stacked light-emitting diode (LED)with included sidewalls as shown in (a), but significantly better optical mixing effect comparedwith a conventional red-green-blue LED as shown in (b).


is still observed at the edges of the chips, corresponding to the locations of the planarelectrodes, leaving room for further improvements.

12.2.3 Third generation tightly integrated chip-stackingapproach

It is apparent that in order to solve the color-homogeneity problem completely, chips ina stack have to be of identical dimensions and overlap with each other completely. Inview of such requirements, a new chip stacking architecture is demonstrated, over-coming the limitations described before. The RGB chips are now of identical dimen-sions, eliminating possibilities of optical leakage. This is achieved by formingchannels onto the sapphire substrates by laser micromachining,7 designed to snug-fitthe wire bonds. When assembled, the bond wires would appear to protrude from thestacked chip tower, whereas the tower maintains a planar facet. Fig. 12.10 shows aschematic diagram of the updated design.

As before, red, green, and blue LED chips are used. The sapphire substrates of thenitride wafers have been thinned down to w150 mm, followed by fabrication ofdevices via standard microfabrication processes, subsequently diced into 1 mm2 chipsby laser micromachining using a nanosecond diode-pumped solid-state ultraviolet(349 nm) laser source. Channels are formed at the locations of wire bonds of thechip beneath; they are micromachined with the same laser as used for dicing. The chipsto be machined are placed on an x-y motorized platform with the sapphire surface fac-ing up. The laser beam, expanded and collimated by a beam expander, trepans acrossthe surface to form a two-dimensional channel of desired dimensions. Fig. 12.11shows an SEM image of one such channel formed on the backside sapphire face ofan LED chip.

The stack assembly begins with adhering the bottom n-contact of a red AlInGaPvertical LED chip to a TO-can using electrically conductive epoxy; the topp-electrode is wire-bonded to a lead on the TO-can. The green InGaN LED chipwith laser micromachined bottom channel is aligned to cover the red LED in its

Figure 12.10 Schematic diagram of the tightly integrated stacked light-emitting diode withembedded wire bonds.


entirety and that the wire bonds of the red LED fit snugly into the trench; in fact, thissnap-in action automatically aligns the chips. Between chips optical epoxy is appliedto secure them in position. The p-electrodes on the green LED are then wire-bonded tothe package. Similarly, the blue InGaN LED chip is piled on top of the green LED; thestacking procedure is illustrated in the schematic diagram of Fig. 12.12(a). The resul-tant trilayer tower structure is illustrated in the optical microphotograph of

Figure 12.11 Scanning electron microscope image showing a laser micromachined channel onthe backside sapphire face of a 1 mm2 InGaN light-emitting diode chip.

Figure 12.12 (a) Schematic diagram showing the layer sequence and chip structure. An opticalmicrophotograph showing the assembled stacked light-emitting diode is shown in (b).


Fig. 12.12(b). The optical measurements are performed by mounting the packagedLEDs onto the input port of a 2-in. integrating sphere, fiber coupled to a radiametri-cally calibrated optical spectrometer.

Color homogeneity of the stacked device is evaluated being the primary goal of thisdesign. The red, green, and blue chips in a stack are biased at currents of 79, 109, and38 mA in order to emit white light with CIE coordinates of (0.3, 0.3) when measured inthe normal direction. Measurement of angular emission profiles is the one of the meth-odologies for assessing color homogeneity. For comparison, the same set of measure-ments is performed on a commercial RGB LED (Avago ASMT-QTC0-0AA02). Anoptical fiber, coupled to a spectrometer, is rotated about the central axis of the devicebeing tested. The blue, green, and red chips in the stack tower are turned on sequen-tially, the optical intensity at each angle between 0 and 180 degrees in step of 1 degreerecorded (90 degrees being the normal direction). The data collected from the stackedLED are plotted onto left hemisphere of the polar graph, Fig. 12.13(a), whereas theright hemisphere shows data from the conventional RGB LED. The shapes of theangular plots are self-explanatory: the emission graphs of the stacked tower overlapwith each other, as if they are emitted from the same chip. On the other hand, the emis-sion from individual chips in the conventional RGB LED exhibits distinct direction-ality, giving rise to an overall nonhomogeneous appearance.

Although the quantitative measurements presented should be convincing enough,the visual appearances of emission from the devices paint an even clearer picture.Fig. 12.14(a) and (b) shows optical photographs of the stacked LED and the conven-tional RGB LED, respectively, captured in the normal direction using a color CCDcamera. Both devices are biased to emit a range of different polychromatic colorsby mixing appropriate proportions of red, green, and blue light. Emission from thestacked tower always appears as a single color, visual proof of satisfactory internal co-lor mixing. On the other hand, red, green, and blue spots of light remain clearly visiblefrom the RGB LED.

To understand the consequences of stacking to optical performances, L-I character-istics of the chips on different layers of the stack are measured. For the LED chips inthe stack, only one of the three chips is turned on for each set of measurements. For afair comparison, identical RGB chips are mounted side-by-side onto an identical pack-age, equivalent to the conventional planar RGB LED configuration; the correspondingchip is turned on and measured. The measured L-I data for the blue, green and red

0 180180

30 150150

60

0

30

6090 90

120120(a) (b)

Figure 12.13 Polar emission plots for the red (gray in print versions), green (light gray in printversions), and blue (dark gray in print versions) spectral components of the (a) commercial red-green-blue (RGB) light-emitting diode (LED) and (b) stacked RGB LED.


devices are plotted in Fig. 12.15(a)e(c); the curves formed by square symbols repre-sent data points for the planar RGB devices, whereas those with circular symbolsrepresent the stacked devices.

At all measured currents, the emitted light intensity by LED chips mounted in aplanar configuration is higher than chips integrated into the stacked structure, althoughto varying extents, due to a combination of thermal, absorption, and reflection effects.For the red LED at the lowest layer of the stack structure, the emitted power drops byw24% at 300 mA (all subsequent comparisons are based on a bias current of 300 mA);this is attributed to a combination of thermal loading and interface reflections. Theextents of optical reductions for chips in the stack due to thermal effects can be esti-mated from the deviation of the L-I curves to the linearly interpolated dotted curvesin Fig. 12.5(a)e(c). As expected, thermal effects on the red LED is minimal becausethe chip is attached directly to the package, allowing efficient conductive heat sinking.The remaining optical drops are due to optical losses along the red light optical path,mainly in the form of interface reflection losses. Red light passes through the green andblue QWs, together with the sapphire substrates, with minimal absorption. For thegreen LED sandwiched between the red and blue LED chips, reduction in opticalpower is most severe of the three at w34%; thermal effects are more severe due asthe generated heat has to be channeled away through the red LED chip. Light emitteddownward from the green chip is almost completely absorbed by the red QWs,whereas upward emitting light encounters interface reflection losses. This is exacer-bated by partial absorption of the green light by the blue QWs from the chip abovedue to spectral overlap between the blue and green QWs. Fortunately, such lossescan easily be avoided by picking blue and green LED chips with a larger separationof central wavelengths. The blue LED, being at the top, suffers an optical drop of

(a) (b)

Figure 12.14 Optical microphotographs showing the (a) stacked light-emitting diode (LED)and (b) planar red-green-blue LED emitting different colors, highlighting the effectiveness ofcolor mixing via chip stacking.


w27%; this originates from a combination of thermal effects, and absorption of down-ward emitted blue light by the green QWs. Being exposed at the top, convective heatdissipation is possible, reducing the heat burden. The measured data provide insight onthe major optical loss mechanisms, so that suitable remedies can be applied. In partic-ular, absorption of downward propagated light from the blue and green chips may beeliminated by the coating of a wavelength selective distributed Bragg reflector on thebottoms of the chips, enabling selective reflection and transmission of light.

Phosphor-free white light emission is a challenging yet important topic in LEDtechnology, promising to take solid-state lighting to the next level. In the literature,there have been several reports of achieving broadband emission using variousmethods summarized below:

• Dual wavelength InGaN/GaN multiquantum well (MQW) LEDs have been reported by Y.D.Qi et al.8 The dual QWs consisted of different well widths and barrier widths, with designedemission in the blue and green regions;

Current (mA)

0 50 100 150 200 250 300

EL

inte

nsity

(a.u

.)

(a)

(b)

(c)

Figure 12.15 L-I characteristics of (a) blue (b) green, and (c) red light-emitting diodes (LEDs)in a stack (circle symbols) and in a planar configuration (square symbols). The plot in the insetof (b) shows the optical EL spectrum of the green LEDs in a stack (dotted line) and in theplanar configuration (continuous line).


• In the paper by H. Sekiguchi et al.,9 nanocolumn arrays with diameters ranging from 137 to270 nm were grown by rf-molecular beam epitaxy on top of a patterned Ti mask. In thereport, each device emits a single wavelength, although theoretically the idea can beextended by attempting growth of nanocolumns of different diameters within the same wafer,which may pose to be a challenging growth problem;

• The paper by H.W. Lin et al.10 reports a similar approach, although the nanorods were self-aligned during plasma-assisted molecular beam epitaxy growth, so that the diameters werebeyond dimensional control, and thus emission spectrum unpredictable;

• Quantum dot is yet another reported approach towards modifying emission wavelength.MQWs incorporates indium-rich InGaN connected-dot nanostructures with a height ofw1.0 nm were reported by C.B. Soh et al.,11 enabling cool-white phosphor-free emissionfrom a single chip.

The stacked tower also functions as a conversion-free white-light LED and in fact, acorrelated color temperature (CCT)-variable white-light LED. Fig. 12.16 showsemission spectra of the stacked LED operated as cool white (CCT of 7332K, drivenat currents of 79, 120, and 45 mA in the order of RGB), neutral white (5999K at 79,110, and 38 mA), and warm white (2362K at 150, 121, and 29 mA) light sources,respectively, The luminous efficacies of the device operated at the three stated CCTsare 19.23, 20.19, and 20.70 lumens per watt, respectively, being respectable figuresfor a prototypic device. For comparison, the planar RGB LED assembled using identicalchips perform as follows: 32.02, 32.78, and 37.22 lumens per watt at CCTs of 7141,6104, and 2401K, respectively. In other words, the efficacy of the stacked LED isapproximately 38% lower than its planar counterpart; however, the additional function-alities far overweigh such losses. It is also worth noting that the luminous efficacy isnearly independent of CCT, allowing efficient operation at low CCTs, a desirable colorfor indoor lighting; such characteristics is obviously superior to phosphor-convertedwhite LEDs at low CCT due to the low efficiencies of longer wavelength phosphors.

Inte

nsity

400 500 600 700Wavelength

Cool whiteNeutral whiteWarm white

Figure 12.16 Optical emission spectra of the stacked light-emitting diode (LED) functioning asa phosphor-free white LED emitting at cool, neutral, and warm white.


12.3 Group-addressable pixelated micro-LED arrays

12.3.1 Concept of natural color mixing using miniature RGBpixels

An alternative way of color-mixing RGB is to have an array of miniaturized RGBemitters, each of which is so small that cannot be resolved by the human eye. The ma-jor advantage of this approach is the ability to integrate multiple color-tunable emittersonto a single wafer, thus the possibility of building single-chip full-color displays. Inthis section, the ongoing development work toward realization of such a display isdescribed. The implementation of such an idea can be achieved through themicrolight-emitting diode concept, whereby the active region of the LED is sectionedinto multiple micrometer-scaled regions, also called pixels.12 Color conversion ele-ments are introduced onto each of these pixels, as depicted in Fig. 12.17. As a result,each pixel will emit at a different wavelength according to the type of fluorescent ma-terial coated.

Because each pixel is of micrometer scale, the overall output appears opticallymixed as the dimensions of the pixels are beyond the resolution limit of the unaidedhuman eye, as illustrated in Fig. 12.18. Pixels with the same color of emission areinterconnected via metal lines. By varying the intensity of the blue, green, and redemitting regions by varying the bias voltage to the three cathodes of the device, theoutput wavelength (color) can be continuously tuned. For example, if we apply abias voltage to the red cathode only, the red pixels will light up and the overall devicewill emit red color, as shown in Fig. 12.19(a). However, if both the red and blue cath-odes are connected to a supply voltage as shown in Fig. 12.19(b), both the blue and redelements will be illuminated and the overall device will appear purple due to colormixing at the microscale level.

Figure 12.17 Schematic diagram illustrating the micropixelated red-green-blue light-emittingdiode (LED).


12.3.2 Implementation of the group-addressable micro-LEDarray

A group-addressable blue LED forms the basis of this color-tunable LED. The activeregion of the device is microsectioned into multiple individual emitters, each of whichshould be of dimensions in the vicinity of tens of microns. Pixels of like colors areinterconnected via metallization, so that they can be addressed as a group. There arethree groups of pixels of the primary colors. Because the starting wafer emits bluelight, the pixels designated for blue emission do not require additional coating of fluo-rescent materials. Green and red fluorescent materials are coated onto the other groupsof pixels, pumped by the high-energy blue light emitter beneath to generate the respec-tive colors.

The group-addressable micro-LED will take the architecture similar to that of a reg-ular parallel-addressed13 or matrix-addressed14 micro-LED, whereby individual pixelsare patterned by photolithography and etched by plasma etching. Fig. 12.20 shows across-sectional view of a typical micro-LED structure. To maximize packing density,the emitters are shaped as hexagons in a closed-packed fashion. The dimensions of in-dividual pixels are in the range of 50e100 microns. Also, pixels emitting the samecolors should not be directly adjacent to each other. Metal interconnects are depositedfor interconnection between pixels of like colors, terminating at the edge of the chip as

Figure 12.18 Illustration of color mixing using micron-scale pixels. As the dimensions of pixelsshrinks, they become unresolvable to the unaided eye and appears as white color.

Figure 12.19 (a) The device with (a) red (gray in print versions) pixels and (b) both red (gray inprint versions) and blue (dark gray in print versions) pixels turned on.


a bond pad. Due to the complexity of the interconnect routing, multilayer metallizationis required. Fig. 12.21(a) shows a schematic diagram of the device layout, where re-gions shaded in red are the individual micropixels, green being the metallic intercon-nects and bond pads, whereas an isolation oxide layer is deposited in the grey region toallow dual-layer interconnections. The fabricated device is as shown in Fig. 12.21(b).The group-addressable behavior of the fabricated LEDs is illustrated in Fig. 12.22.When an electrical bias voltage is applied between the common cathode and one setof anode, one-third of the micro-LED pixels light up.

12.3.3 Forming the RGB pixels by jet-printing

The remaining, yet most challenging step, relates to selective deposition of fluorescentmaterials onto individual pixels. Having explored photolithographic methods withoutsuccess, it became apparent that direct deposition of fluorescent materials is the poten-tial solution. Inkjet printing was thus adopted based on its ability to dispense liquidswith volumes as low as picoliters.15 To be compatible with inkjet printing, the fluores-cent particles must be soluble in a solvent, have dimensions much smaller than thenozzle, and be of spherical shape to prevent clogging up of the printhead. In our

Sapphire

AluminumTitaniumSilicon dioxide

n-GaN

Figure 12.20 Cross-sectional schematic view of two adjacent microlight-emitting pixels.

Figure 12.21 (a) Schematic diagram showing the layout of a group-addressable microlight-emitting diode (micro-LED) and (b) optical microphotograph of the fabricated device.


experiments, fluorescent microspheres with diameters of about 0.39 mm were used.Two types of fluorescent microspheres are used to produce green and orangeeredemissions.16,17 The internally dyed green nanospheres are maximally excited at468 nm and produces broadband green emission with two spectral peaks at 525 and560 nm. The internally dyed orangeered nanospheres absorb maximally at 558 nmand fluoresce at the longer wavelength region with spectral peaks at 570 and610 nm. Fig. 12.23(a) illustrates fluorescence from a closed-packed array containinggreen and orangeered microspheres.

The microspheres suspended in deionized (DI) water are printed onto the group-addressable micro-LED using a Jetlab 4 printer from Microfab. During the printing

Figure 12.22 The group-addressable microlight-emitting diode (micro-LED) in action: one ofthree groups of pixels lights up at a time.

Sapphire

AluminumTitanium

Silicon dioxiden-GaN

p-GaNMQW

ITO

n-contact

p-contact

Fluorescentmicrosphere

Figure 12.23 Cross-sectional view of a single pixel with a fluorescent microsphere coating; anoptical micrograph of a close-packed array of green and orangeered fluorescent microspheresis shown in the inset.


process, the inkjet nozzle was driven by an AC-voltage waveform to eject stable andreproducible microdrops in the drop-on-demand mode. After the microdrops weredeposited onto assigned pixels according to computer aided design pattern correspond-ing to the micropixel layout, the DI water evaporates, leaving behind fluorescent mi-crospheres sitting on the top of the pixels, as illustrated in Fig. 12.23(b). The formationof microdrops from the nozzle could be observed via an integrated stroboscopic cam-era system, an image capture of which is shown in Fig. 12.24. Typically, the radii ofdrops are approximately equal to the radii of the nozzle that ejected them. The oper-ating parameters for ejecting an optimal microdrop are experimentally optimized,depending upon the nature of particles/fluid (viscosity, surface tension, and concentra-tion), size of particles/nozzle aperture, and the working environment (pressure, temper-ature, humidity).

12.3.4 Current status of development

Fig. 12.25 shows a completed monolithic color-tunable LED with the jet-printed redand green pixels. To deliver sufficient levels of red or green light emission, the dropswere repeatedly deposited over the same pixel to form a microsphere stack of morethan 100 mm thick; this requires hundreds of drops to be deposited at nearly on thesame spot. Inevitably, the deposited microspheres may be displaced from theirintended locations affecting adjacent ones. Such occurrences may be minimized byheating up the substrate to accelerate evaporation. However, because the nozzle isin close proximity to the substrate with a separation of about 200 mm to ensure preciselanding, substrate heating may cause the solvent inside the tip of the nozzle to dry outquickly, so that fluorescent microspheres aggregate near the tip of the nozzle, impedingthe burst of microdrops.

The group-addressable pixels can be controlled independently by adjusting suitablebias voltages/currents to generate different colors or color temperatures, as demon-strated in Fig. 12.26(a)e(d), which illustrates the (a) blue, (b) green, (c) red, and

Figure 12.24 Close-up view of jet-printing of fluorescent microspheres on top of pixels.


Figure 12.25 Microphotograph of a completed micropixelated microlight-emitting diode(micro-LED) color-tunable device with deposited red (gray in print versions)/green (white inprint versions) fluorescent microsphere stacks.

Figure 12.26 The (a) blue, (b) green, and (c) red set of pixels illuminated. In (d), all pixels areturned on simultaneously.


(d) a combination of blue/green/red pixels being turned on. When all three groups ofRGB pixels are turned on and viewed without magnification, the individual pixels willno longer be resolvable and appear as a single point source, producing a natural colormixing effect.

From these initial devices, two major issues have been identified, the first of whichis a dominant blue background emission clearly visible from Fig. 12.26(a)e(d). Frominvestigations by confocal microscopy, the cause of the blue background has beenidentified as light channeling along the planar GaN and sapphire layers, subsequentlyextracted from the surfaces due to surface roughnesses.18 Having identified the sourceof background emission, the design of the device architecture can be modified accord-ingly to overcome the problem. As optical channeling occurs in the GaN and sapphirelayer,19 and that the sapphire substrate is significantly thicker than the GaN epilayer, itshould be adequate to remove the sapphire substrate from the devices. Incidentally,sapphire removal by LLO is an established process for fabricating thin-film GaNLEDs, which has been developed primarily for improving heat-sinking in high-powers LEDs due to the low thermal conductivity of sapphire.20 The same techniquecan now be employed for the fabrication of thin-film micropixel LEDs. A cross-sectionschematic diagram of the regular and thin-film versions of the same micro-LED deviceis shown in Fig. 12.27(a) and (b), respectively, whereas Fig. 12.27(c) and (d) shows amicrophotograph of the illuminated devices. Compared with the regular micro-LED, amarked reduction of the background blue emission is observed. Cross-section emis-sion profiles of the devices, obtained from confocal z-stacks,18 are plotted inFig. 12.27(e) and (f). Compared to the regular micro-LED, the massive amount of lighttrapped beneath the pixel is no longer observed because the light-trapping sapphirelayer has been removed and the light rays are instead reflected by the metallic mirrorat the interface.

On the other hand, color-conversion efficiency of the jet-printed fluorescent micro-spheres leaves much to be desired. However, conventional color-conversion materialssuch as phosphors cannot be coated by jet printing due to their dimensions, nonunifor-mities, and geometries. Recently, high quantum efficiency quantum dots have beensuccessfully jet-printed onto selective pixels of the micro-LED devices, as shown inFig. 12.28. The red and green light emissions seen in the figure represent fluorescencefrom CdSe/ZnS quantum dots printed on the treated pixels. In terms of brightness, theyare a significant step up from the fluorescent microspheres shown in Fig. 12.25. Opti-mization of the quantum dot printing process is still in progress, but this represents apromising solution toward realizing the group-addressable pixelated micro-LEDarrays.

12.4 Conclusions

The prospects of having color-tunable emitters are exciting, expanding the functional-ities of LEDs beyond conventional lighting applications. To fully implement this idea,new forms of LEDs must be developed. Two designs of color-tunable LEDs have been


proposed and described in this chapter. The vertically stacked LED, based on aligningthe optical paths of three RGB chips on top of each other, is a solution geared towardsindividual high-power color-tunable emitters. Such devices can be used for assemblinglarge area LED display panels. The micropixelated design, on the other hand, enablesmonolithic integration of multiple devices and is thus an ideal platform for single-chipfull-color emissive microdisplays, promising to challenge similar offerings fromOLED technologies.

n-GaN

p-GaNn-GaN

p-GaN

CopperSapphire

AI mirror

ITO MQWITO MQW

Metal pad

SiO2

(a)(b)

(c) (d)

(e) (f)

Figure 12.27 (a) Cross-sectional schematic diagrams depicting architectures of (a) regularmicrolight-emitting diode (micro-LED) and (b) thin-film micro-LED. Microphotographsshowing emission from a group of pixels from the respective devices in (c) and (d). Cross-sectional emission intensity maps (142 � 1000 mm2 along x-z plane) of the respective devicesplotted on linear scale in (e) and (f).


Acknowledgments

The authors are grateful to Dr Xianghua Wang of Hefei University of Technology for contribu-tions to the work on jet-printing of quantum dots.

References

1. Shlayan N, Venkat R, Ginobbi P, Singh AK. Energy efficient RGBW pixel configuration forlight-emitting displays. J Display Tech 2009;5:418.

2. Wu H, Narendran N, Gu Y, Bierman A. Improving the performance of mixed-color whiteLED systems by using scattered photon extraction technique. Proc SPIE 2007;6669:666905.

3. Sun C-C, Chien W-T, Moreno I, Hsieh CT, Lin M-C, Hsiao S-L, Lee X-H. Calculatingmodel light transmission efficiency diffusers attached a lighting cavity. OSA Opt Express2010;18:6137.

4. Krames MR, Ochiai-Holcomb M, Hofler GE, Carter-Coman C, Chen EI, Tan IH, Grillot P,Gardner NF, Chui HC, Huang JW, Stockman SA, Kish FA, Craford MG, Tan TS,Kocot CP, Hueschen M, Posselt J, Loh B, Sasser G, Collins D. High-power truncated-inverted-pyramid (AlxGa1�x)0.5In0.5P/GaP light-emitting diodes exhibiting >50% externalquantum efficiency. Appl Phys Lett October 1999;vol. 75:2365e7.

5. Wang XH, Lai PT, Choi HW. Laser-micromachining Opt Microstructures Inclined SidewallProfile. J Vacuum Sci Technol B May 2009;72:1048e52.

6. Fu WY, Hui KN, Wang XH, Wong KKY, Lai PT, Choi HW. Geometrical shaping InGaNlight-emitting diodes by laser micromachining. IEEE Photon Technol Lett 2009;21:1078.

7. Mak GY, Lam EY, Choi HW. Liquid-immersion laser micromachining of GaN grown onsapphire. Appl Phys A 2011;102:441.

Figure 12.28 A microlight-emitting diode (micro-LED) device with jet-printed red and greenquantum dots on selective pixels.


8. Qi YD, Liang H, Tang W, Lu ZD, Lau KM. Dual wavelength InGaN/GaN multi-quantumwell LEDs grown by metalorganic vapor phase epitaxy. J Cry Growth 2004;272:333.

9. Sekigucki H, Kishino K, Kikuchi A. Emission color control from blue to red with nano-column diameter of InGaN/GaN nanocolumn arrays grown on same substrate. Appl PhysLett 2010;96:231104.

10. Lin H-W, Lu Y-J, Chen H-Y, Lee H-M, Gwo S. InGaN/GaN nanorod array white light-emitting diode. Appl Phys Lett 2010;97:073101.

11. Soh CB, Liu W, Teng JH, Chow SY, Ang SS, Chua SJ. Cool white III-nitride light emittingdiodes based on phosphor-free indium-rich InGaN nanostructures. Appl Phys Lett 2008;92:261909.

12. Jeon CW, Choi HW, Gu E, Dawson MD. High-density, matrix-addressable AlInGaN-based368 nm micro-array light-emitting diodes. IEEE Photon Technol Lett 2004;16:2421.

13. Choi HW, Jeon CW, Martin MD, Edwards PR, Martin RW. Fabrication perform parallel-addressed InGaN micro-led arrays. IEEE Photon Technol Lett 2003;15:510.

14. Jeon CW, Choi HW, Dawson MD. Fabrication matrix-addressable InGaN-based micro-displays high array density. IEEE Photon Technol Lett 2003;15:1516.

15. Amemiya I, Nomura Y, Mori K, Yoda M, Takasu I, Uchikoga S. LED packaging by ink-jetmicrodeposition high-viscosity resin phosphor dispersion. J Soc Inf Display 2008;16:475.

16. Hui KN, Lai PT, Choi HW. Spectral converse fluorescent microspheres light emitting di-odes. OSA Opt Express 2008;16:13.

17. Hui KN, Fu YF, Ng WN, Leung CH, Lai PT, Wong KKY, Choi HW. Polychromatic Light-emitting Diodes with a Fluorescent Nanosphere Opal Coating. Nanotechnology 2008;19:355203.

18. Li KH, Feng C, Choi HW. Analysis of micro-lens integrated flip-chip InGaN light-emittingdiodes by confocal microscopy. Appl Phys Lett 2014;104:051107.

19. David A, Fujii T, Sharma R, McGroddy K, Nakamura S, DenBaars SP, Hu EL,Weisbuch C, Benisty H. GaN light-emitting diodes with Archimedean lattice photoniccrystals. Appl Phys Lett 2006;88:061124.

20. Li KH, Cheung YF, Zhang Q, Choi HW. Optical and thermal analyses of thin-film hex-agonal micro-mesh light-emitting diodes. IEEE Photon Technol Lett 2013;25:374.



Reliability of nitride LEDs 13Tzung-Te Chen 1, Chun-Fan Dai 1, Chien-Ping Wang 2,Han-Kuei Fu 1, Pei-Ting Chou 1, Wen-Yung Yeh 1

1Industrial Technology Research Institute (ITRI), Hsinchu, Taiwan; 2Chung Yuan ChristianUniversity, Chung-Li, TaiwanRevised by Shyh-Chiang Shen

13.1 Introduction

With the widespread applications of nitride-based white LEDs for the lighting and thedisplay, the reliability analysis of LEDs has an increased importance for creating high-quality products. Common reliability studies of such devices include the methodolo-gies of lifetime and reliability tests, characterizations of failure phenomena, and theanalysis of failure mechanisms related to photonic devices. The objective is to enhancethe reliability of LEDs by reducing possible causes of device failure. While typical reli-ability study in the semiconductor industry focuses mainly on electrical characteristics,the reliability tests for LEDs involve characterizations of various properties throughelectrical (current vs. voltage), optoelectronic (lumen vs. current/voltage), and optical(chromaticity, etc.) aspects. In particular, monitoring the time-dependent evolution ofthe light output properties is the most rigorous basis for assessing the reliability andlifetime of LEDs. In addition, unlike other semiconductor devices, the luminescencefrom an LED can also yield insights to the effective failure analysis. In this chapter,we will explore these properties and offer recommendations and solutions to theseissues.

13.2 Reliability testing of nitride LEDs

The reliability tests for nitride LEDs generally include the power and the temperaturecycling test (JESD22-A105), the salt atmosphere test (JESD22-A107), the mechanicalshock test (JESD22-B110, JESD22-B104), the thermal shock test (JESD22-A106), thevibration test (JESD22-B103), the temperature cycling test (JESD22-A104), thehigh and low temperature storage test (JESD22-A119, JESD22-A103), the hightemperature and humidity operating lifetime test (JESD22-A110), the moistureresistance test (JESD22-A118), the electrostatic discharge (ESD) robustness test(JESD22-A114), and the LED lumen maintenance test (IES LM-80-08). The lifetimetest also encompasses the estimation of the device lifetime. These are commonly usedtests by LED manufacturers to evaluate the reliability of LEDs. Since these tests aredestructive by nature, device sampling is commonly adopted for such tests.


https://doi.org/10.1016/B978-0-08-101942-9.00013-7

In this chapter, we will focus on the ESD robustness and LED lumen maintenancetests, which are the main reliability tests for nitride LEDs. In 2008, the IlluminatingEngineering Society (IES) published a standard entitled “IES LM-80-08: MeasuringLumen Maintenance of LED Light Sources.” This standard specifies requiredmeasurements of the lumen maintenance testing for LED light sources that includeLED packages, arrays, and modules. Another standard entitled “IES TM-21-11:Projecting Long Term Lumen Maintenance of LED Light Sources” was subsequentlypublished in 2011. Most manufacturers today employ these two standards as bench-mark testing methods for LEDs. When applying such standards, there are two precau-tions to bear in mind:

1. The standards mentioned so far are only for the output luminous flux through a lumenmaintenance test. The lumen maintenance, however, is not entirely representative of thelifetime of an LED. Nevertheless, lumen maintenance is commonly used to determine thelifetime of an LED device.

2. There are no separate lifetime testing standards for LED dies or packaging materials. Thelifetime test of an LED die can only be conducted within its simplest packaged form.

The ESD robustness tests for LEDs include the human body model (HBM), themachine model (MM), and the charged device model (CDM). The HBM and theMM simulate the discharge of the static electricity from a human or a machine to anLED device. On the other hand, the LED itself carries static charges and dischargesto the surrounding environment in the CDM tests. For latest LED products, includingthe die itself and the packaged components, HBM damage is the most commonlyobserved, and therefore, the HBM is the primary test employed to evaluate ESDrobustness at the level of the die and the finished products for LEDs.

13.2.1 Methods for life testing and lifetime estimation

The life testing methods for LEDs adhere primarily to the LM-80-08 standard. TheLM-80-08 standard stipulates that the driving current employed in LED life testsmust be representative of commonly used applications, with three case temperatures:55, 85�C and a manufacturer-specified temperature. In order to acquire sufficient datato estimate the lifetime of an LED accurately, the life test must be performed for at least6000 h. Furthermore, at intervals of at least once every 1000 h during the LED life test,variations in the light output, chromaticity coordinates, the forward biased voltage, etc.must be measured. The LM-80-08 standard does not specify the percentage to whichthe lumen has decayed before an LED is deemed to have failed, and define the lumenmaintenance life to be the time for the LED lumen to decay to 70% or 50% of its initialvalue.

The LM-80-08 standard also does not provide any guidelines on how to estimate thelifetime based on the 6000-h test data. According to the TM-21-11 standard, thelifetime estimation is based on a projection of the LED lumen maintenance ratefrom the data obtained through the LM-80-08 tests. For an LED, the lumen decaytypically follows a simple exponential model, because the LED component materialsundergo diffusion over time and the rate of diffusion increases with operating


temperature.1 Based on this model, the lifetime estimate for LED lumen decay givenby the TM-21-11 standard is expressed as:

FðtÞ ¼ B$e�a$t (13.1)

where F is the lumen output at time t, B is an initial constant and a is the lumen decayrate constant. From this exponential decay equation, one can estimate the LED lifetimefor a fixed temperature. If one wishes to estimate the LED lifetime at differenttemperatures, it is necessary to introduce the Arrhenius equation,1 which describeshow the component materials of an LED and chemical reactions at different temper-atures affect the lumen decay rate:

a ¼ A$e�EakT (13.2)

Ea is the activation energy, k the Boltzmann constant, T the temperature and A aconstant related to the properties of the material. In particular, one can obtain Ea/k byplugging in the empirical lifetimes at two different temperatures:

Ea

k

lnða1Þ � lnða2Þ1T2

� 1T1

(13.3)

where a1 and a2 are the lumen decay rate constants at temperatures T1 and T2,respectively. Substituting the Ea/k obtained from Eq. (13.3) back into Eq. (13.2), theArrhenius equation, one can obtain the lumen decay rate constants at differenttemperatures. From Eq. (13.1), one can obtain the relation between the lumen decayand the time at intermediate temperature. In other words, the lifetime can be estimatedat these intermediate temperatures. The lifetime estimation defined by LM-80-08 andTM-21-11 standards has enabled LED manufacturers to adopt a universal set of criteriafor lifetime testing and evaluation.

13.2.2 Methods for electrostatic discharge testing

The ESD takes place between two objects at different electrostatic potentials with atransfer of energy. Fig. 13.1 depicts the equivalent circuit used to simulate the ESDprocess. In addition to the device under test (DUT), the ESD configuration consistsof a power supply and a circuit network that consists of capacitors and resistors.The power supply provides the test voltage for the ESD. Typical test voltage rangesfrom 0 to 4000 V for LEDs. When the capacitor is fully charged and is electrostaticallybalanced, instantaneous switching of the circuit via a switch will lead to a discharge ofthe electric current from a capacitor through a resistor to the DUT to simulate an actualESD scenario. Of the three most commonly used discharge modes, namely HBM,MM, and CDM, HBM damage is the most commonly observed in LED devices.Therefore, LED manufacturers will generally begin with the HBM ESD robustness

Reliability of nitride LEDs 443

level test as a basis for product classifications according to their reliability. Two mainsets of specifications are adopted in HBM tests. In MIL-STD-883G and JEDECJESD22-A114 series, the equivalent capacitance of a human body is set at 100 pFand the corresponding equivalent discharge resistance is set at 1.5 kU. In the IEC-61000-4-2 specifications, the equivalent capacitance of a human body is set at150 pF and the corresponding equivalent discharge resistance is set at 330 U. Othercommonly adopted specifications include JESD22-A115 series for the MM modeand JESD22-C101 series for the CDM mode.

13.3 Evaluation of LED degradation

Due to factors such as intrinsic defects in an LED die and the uneven distribution ofcurrent and temperature, the degradation of an LED die tends to occur on specific re-gions of a device. Since some of these degraded regions are only a few tens or hun-dreds of nanometers in physical dimension, sophisticated analytical techniques willbe required to augment a microscope inspection for valid failure analysis. These toolsinclude the emission microscopy (EMMI), the electroluminescence (EL) mapping incombination with an optical microscopy (OM), the energy dispersive spectroscopy(EDS), and the electron energy loss spectroscopy (EELS), in combination with a scan-ning electron microscopy (SEM) and a transmission electron microscopy (TEM), etc.These characterization techniques are employed to identify the regions of interest forthe failure mechanism analysis in an LED die. We will not intend to explain in detailthe principles of various failure characterization techniques, and rather will merelyintroduce the application of several commonly used characterization techniques inLED failure analysis. An exemplary LED failure analysis flow chart is presented inSection 13.3.2.

13.3.1 Failure evaluation techniques

Since the failure criteria for LEDs are not specified in the LM-80-08 standards, LEDmanufacturers typically define the failure of an LED as the point when the light output

Switch

Capacitance

DUT

Resistance

Powersupply

Figure 13.1 Equivalent circuit for ESD.


of an LED at its operating voltage drops under 70% of the initial value, or when theleakage current exceeds 1 mA at�5 V of the reverse-biased voltage. The characterizedregions and detectable signals for various characterization techniques applied duringthe LED failure evaluation are listed in Table 13.1. The failure analysis can be broadlydivided into three categories. The first category is the morphology characterization(e.g., OM images, X-ray images, SEM images, TEM images, etc.), where externalstructural damage is observed with the incident visible light, X-rays, or electrons.The second category uses electrical and optical imaging methods such as EL mappingand EMMI to obtain luminescence images of an LED under the forward- and reverse-biased conditions, respectively. These methods help characterize specific devicefailure in regions where no external structural damages exist (e.g., the active layer,

Table 13.1 Detectable signals and characterization regions for variouscharacterization techniques

Characterizationtechnique Characterization region Detectable signals

OM SurfaceDie/wire bonding

Strength of reflectedvisible light

X-ray image SurfaceDie/wire bonding

Strength of transmittedX-ray

EL mapping SurfaceActive regionElectrode

Forward-biased emission

EMMI SurfaceLeakage pathSidewall

Reverse-biased emission

SEM Die structureElectrodeTransparent conductive layer

Secondary electrons

EDS Die structureElectrodeTransparent conductive layer

Characteristic X-rayspectrum

TEM Die structureQuantum well (QW) structureInterface quality

Transmission electrons

EELS Die structureQuantum well (QW) structureInterface quality

Transmission electrons

Secondary ionmass spectroscopy(SIMS)

Die structureQuantum well (QW) structureInterface quality

Secondary ions


the surface leakage pathway, etc.). The third category uses a composition mapping andchemical bond imaging methods (e.g., EDS and EELS coupled with SEM or TEM) toverify the failure of an LED. In some instances, the cross-section of an LED can beprepared using a focused ion beam (FIB) method for a subsequent TEM analysis.

13.3.2 Failure evaluation flow chart

In Fig. 13.2, we reorganize the context of Table 13.1 into a failure evaluation flowchart for typical LED die analysis. The flow chart progresses from the top to thebottom, and in principle, nondestructive characterization procedures are performedbefore destructive characterization procedures are carried out. Likewise, macroscopiccharacterization procedures are performed prior to microscopic characterization

Yes

NoSatisfy failurecriteria

Macroscopic characterization of diesurface (OM, X-ray images)

Macroscopic characterization of diesurface (EL mapping)

Macroscopic characterization of diesurface (EMMI)

Macroscopic characterization of diesurface (SEM, EDS)

Characterization of die crosssection (SEM, EDS)

Characterization of die crosssection (TEM, EDS, EELS)

Characterization of die crosssection (SIMS)

Main location Failure type•Surface

•Surface

•Surface

•Die surface

•Die structure

•Die structure•QW structure

•Die structure•QW structure

•Electrode

•Electrode

•TCL

•TCL

•Active region

•Depletion region•Sidewall

•Electrode

•Die/wire bonding•Die crack•Bonding degradation•Solder degradation

•Electrode degradation

•Electrode degradation•TCL degradation

•Electrode degradation•Metal diffustion/alloy reaction

•Metal diffustion/alloy reaction

•Metal diffustion/alloy reaction•Impurity diffusion•Doping migration

•TCL degradation

•Surface degradation

•Defect generation



•Threading dislocation

•Leakage current region

Remove of the cap (if necessary)

Non

dest

ruct

ive

Des

truct

ive

Reliability testing(life testing, ESD testing)

Electrical and optical measurement(I-V curve, illumination flux)

FIB

TEM specimen

Figure 13.2 LED failure evaluation flow chart.2


procedures. Nondestructive characterizations mainly entail the use of techniquessuch as the EL mapping and the EMMI to inspect the electrodes and the transparentconductive layer (TCL) for possible damages. Detailed characterizations in regionswith weaker luminescence or current leakage are also performed for further study.When one uses the EL mapping or the EMMI for macroscopic characterization onthe surface of an LED die, it may be necessary to remove the encapsulating cap ifthe silicon or phosphor on the surface of the die is found to be interfering with thecharacterization. However, this procedure is likely to damage the die or the bond wires,thereby rendering EL mapping and EMMI next to impossible. When performing SEMand EDS analyses of the LED die surface, there should be no device encapsulation onthe surface of interest. For destructive characterizations, FIB can be used, based on thesurface images obtained with the EMMI or an SEM, to make longitudinal cuts inregions with obvious current leakage or sites with apparent surface defects. TheEDS and SEM can be used to observe the failure conditions in the cross-section ofa sample. Finally, the FIB can be used to produce a TEM sample of a specific regionfor further characterizations. When the TEM method is used to observe defect-relatedcauses of failure such as the threading dislocation and additional defect generation, themeasurements should be made on the same sample to verify the variations before andafter the appearance of the defects that resulted in the failure of an LED. A secondaryion mass spectroscopy (SIMS) can be used to obtain the composition distribution ofthe LED sample at different depths after the device encapsulation is removed.

13.4 Degradation mechanisms

The degradation mechanisms for most nitride LEDs are intimately related to intrinsicdefects, the operating current density, and the junction temperature. During extendedLED lifetime tests, the defect density may increase with the test time. It helps revealapparent LED performance degradation. When an LED suffers from an ESD stressfracture, instantaneous failure and different forms of degradation may result. Ingeneral, a high-power LED chip operates at an electrical power consumption of around1e3 W. It is equivalent to an operating current of 350e1000 mA for a 1 mm2 LED dieover the course of operation. The junction temperature can be estimated as:

Junction temperature ¼ Ambient temperatureþ Thermal resistance

� Electrical power

The thermal resistance is dependent on the packaging. During the actual LEDoperation, due to factors such as the layer structures and the device’s physical design,the current may be unevenly distributed across an LED die. The current density and thejunction temperature in certain regions of the device often exceed the average values ofan LED die due to the current crowding effects. As a result, the device degradationtends to concentrate in specific regions. Building on prior discussions of the reliabilitytests in Section 14.2 and the reliability analysis techniques discussed in Section 14.3,we may further explore the device degradation mechanisms in the following section.


13.4.1 Degradation mechanisms found in LED life testing

Various failure mechanisms resulted from extended hours of usage can affect the LEDlifetime and the luminescence efficiency. As described in Section 14.2.1, most LEDlife testing methods that are currently in use adhere to the specifications of theLM-80-08 standards published by the IES. These tests are conducted under specificconditions to gauge the reliability of LEDs through extended aging tests to analyzethe physics behind various types of failures. However, the composition of an LEDis extremely complex. Such devices includes components including conductors, semi-conductors, and nonconductive materials in a highly integrated and packaged form. Inaddition to the electrical and thermal stress failures that are commonly observed in adiode, the failure mechanisms of an LED may further include photon-induced failures.For example, the deterioration of the plastic encapsulation material may result in a dropof the optical transmittance; the physics and the failure modes related to the conversionefficiency degradation of the phosphor appear to be more complex too. The manifes-tation of these LED failures is a drop in the photon output power, usually quantified bythe luminous efficacy in the units of lm/W. Most of the failure phenomena also varywith a change in the leakage current and the forward bias, as shown in Figs. 13.3and 13.4.

The current versus voltage of a diode is given by:

Ijunction ¼ I � V � IRs

Rp¼ Is exp

�eðV � IRsÞnideal KT

�; (13.4)

Zone I Zone II Zone III

–5 –4 –3 –2 –1 0 1 2 3Voltage (V)

1E–14

1E–12

1E–10

1E–18

1E–16

1E–14

0.01

Cur

rent

(A)

1A,85°C1A,55°C0.7A,85°C0.7A,55°C0.7A,25°C0.35A,25°CInitial

Figure 13.3 Semi-log currentevoltage curve measured after long-term aging tests.3

Copyright 2011, Japan Society of Applied Physics.


where Ijunction is the current across a p-n junction (PNJ), I is the total input current, Rp isthe parallel resistance, Rs is the series resistance, and nideal is the ideality factor.Referring to Fig. 13.3, Zone I is where the carrier recombination effect occurs; Zone IIis where the ideality factor influences the ideal diode equation; and Zone III contributesto the diode’s series resistance including the ohmic contacts at the anode and thecathode.4

The variations of the current versus voltage and the related parameters extractedfrom the aging test can be used as a mean of the failure mechanism observation.The factors that cause a decay of LED’s luminous efficacy include a reduction ofthe internal quantum efficiency that occurs partly due to an increase in the nonradiativerecombination defect density, an increase in the ohmic contact resistance due to theaging of materials, the formation of point defects by hydrogen and magnesium ions,etc.5

The concentration of electrical stress or heat stress owing to uneven currentspreading in an LED may also exacerbate the aforementioned causes of failure. Forexample, the reverse-biased current of an LED increases by several orders of magni-tude with higher voltage and current stressing, also with longer stressing time, asshown in Fig. 13.4. Possible causes of failures include an increase in the dislocationdensity across the PNJ6 or increased nonradiative surface leakage paths due to themesa etching damage in LED fabrication.

In general, the physics of LED failure modes can be classified as (1) rapid degra-dation in the initial stage, (2) gradual degradation, and (3) catastrophic degradation.

0 2000 4000 6000Time (h)

0.01

1E–3

1E–4

1E–5

1E–6R

ever

se c

urre

nt (A

)

1A,85°C1A,55°C0.7A,85°C0.7A,55°C0.7A,25°C0.35A,25°C

Figure 13.4 Reverse current as a function of aging time under various conditions.3

Copyright 2011, Japan Society of Applied Physics.


13.4.1.1 Rapid degradation in the initial stage

The underlying causes of such rapid device degradation are line defects of the die,which are usually caused by the lattice mismatch in epitaxial layers. Most of thecommon line defects arise from the surface defects of the die. They grow along thesurface defects after the application of external electrical or thermal stress.

13.4.1.2 Gradual degradation

As described in Section 13.2.1, most LED efficiency degradation models exhibitgradual degradation, with a rate of decay dependent on the activation energy andthe temperature of a specific device design.

13.4.1.3 Catastrophic degradation

Catastrophic degradation is usually a result of the electrical damage, which can takeplace during the operation or the packaging of the device. The majority of such failurescan be identified from the dark spots or dark regions in the PNJ as observed by the ELmapping. Another possible cause is the short-circuiting of the PNJ, which can beobserved from the current versus voltage relation in the low biasing conditions.

13.4.2 Degradation mechanisms found in LED ESD testing

During the electrostatic discharge, most of the damage that leads to the failure of anLED die results from a transient peak discharge and a transient high temperature. Arelatively large peak current is accompanied by a large energy release that bringspermanent damages to the components. In view of the destructive effects of theESD on LED components, a better understanding of the sources of the device failure,as well as the preventive and remedial steps to be taken in LED designs will be criticalto the success of LED manufacture.

The main causes of ESD failures are defect density, electrical conductivity, andthermal dissipation. Under different ESD stresses, the causes of ESD-induced LEDcomponent failures can be classified as follows:7

• Hard failure: When an LED die is subject to large transient current or voltage pulses, it tendsto rupture internally in the epitaxial layers, leading to the creation of undesired conductivepaths: the generation of defects enhances the leakage current. The temperature and thecurrent crowding effect also cause the material to melt at the interface of the metal andthe semiconductor, indirectly contributing to the diffusion of the dopants and the metalatoms. In some instances, this effect may cause a short circuit of the component, leadingto permanent device failure.7

• Soft failure: When an LED die is subject to a small transient current or voltage pulse, theannihilation of some defective pathways7 or the conductivity fluctuation of the leakagepathways8 may lead to an instability of the leakage current. During this time, the LEDwill still function normally, but its performance will gradually deteriorate.

When the ESD stressing with different electrode polarities are applied to an LEDdevice, it is found that the reverse-biased voltage has a greater impact on the leakage


current than those test conditions with the forward bias voltage stressing. It can beunderstood by two distinct current flow mechanisms in LED devices under differentbiasing conditions. In the forward biased condition, the electron current flows in theconduction band. When an external forward ESD bias is applied to an LED device,the ESD transient discharge current is carried through the conduction band of thedevice and no associated defects are created. Under a reverse-biased condition, mostof the current is concentrated in the leakage pathways, and the electrons in the valenceband of the materials will be driven by large electric potential that creates highprobability of these electrons to be promoted to the conduction band via the defectlevels or through the impact ionization. This process consequently creates more defectlevels in the PNJ and leads to larger leakage currents as the time evolves.

In addition to the polarity effect of the ESD stress test, the ESD robustness of anLED is also related to its physical device design. The current conduction in mostLED devices assumes a horizontal structure, as most of the LEDs are built on noncon-ducting sapphire substrates with poor thermal conductivity. As shown in Fig. 13.5, theanode (the p-type contact) and the cathode (the n-type contact) are located on the leftand right in the graph, respectively. These electrodes are placed on the same side (topside) of a sapphire substrate. When an LED is subject to an ESD stress, the currentcrowding effect will lead to a mix of thermal, potential difference, and light emissionphenomena. This will render the die prone to local fixed-point failures,9,10 and is themost common cause of failure during a transient electrostatic discharge.

It is important to perform an optimization of the electrode placement in the design ofthe physical device as it plays a role in reducing the current crowding.7 For example,when the separation between the two electrodes is large, the differences in theresistance of various current conduction pathways will cause some of the device re-gions prone to the current crowding effect. The energy will be dissipated in the formof heat, making it more likely for an LED to experience regional failure under anESD stress.9 As shown in a schematic drawing of the current conduction pathwaysin Fig. 13.5, the active region, the n-GaN area, and the contact area between the contactlayer and the p-GaN area are the three major areas in an LED where the heat accumu-lation may be of major concern due to the current crowding effect. The poor thermal

Substrate

Figure 13.5 Horizontal structure of an LED device.10


conductivity (35 W/mK) of the sapphire substrate will result in the accumulation ofheat within the device, leading to a diffusion of the dopants. At the same time, themelting of the metallic contact may also occur, creating permanent failure of LEDs.

13.5 Conclusion

We described two characterization methods that are commonly used by LED manufac-turers to evaluate the reliability of LEDs, and outlined the standards for comparingdifferent products. Among the tests employed are the lifetime test and the ESD robust-ness test after extended operation at different levels of the temperature and the electriccurrent. During the failure analysis phase, one may follow a failure analysis flow chartoutlined in Section 13.3.2, and perform nondestructive failure analysis prior to thedestructive failure analysis. One may further confirm the failure modes using lumines-cence images under forward and reverse bias, high-resolution electron microscope im-ages, EDS composition images, EELS binding energy images, etc. to ascertain thefailure regions and root causes of these failure mechanisms. LED-specific failuremechanisms are often related to intrinsic defects, current density, junction temperature,etc. Factors such as current crowding and regional hot spots tend to shorten the oper-ating lifetime of an LED, with failure usually concentrated in specific regions.

The main indicators for LED reliability have traditionally been based on criteriasuch as light output, operating voltage, the magnitude of the reverse leakage current,etc. In the future, monitoring of the junction temperature should be added to the list ofcriteria for reliability evaluation. Similar to what has been established in the measure-ment of other criteria (e.g., light output or luminous flux, operating voltage, etc.), amethod of quick evaluation of the junction temperature should be developed to providecomprehensive inspections and grading of future LED products. For the lifetimetesting, one may consider using high current density stress or high temperatures toaccelerate the test schedule for timely evaluation of the LED’s reliability. Regardingthe ESD robustness test, the development of comprehensive nondestructive ESD char-acterization technologies to replace destructive testing methods should become thefocus of the reliability test development in the near future. In addition, due to the diver-sification and rapid development of LED products, the establishment of testing andevaluation criteria that extend current standards to new LED products such as ACLEDs, high voltage LEDs, micro-LED arrays, etc. are yet to be developed. These de-velopments will be critically needed as these new LED technologies have quickly tran-sitioned from research interests to mass production these days.

References

1. Lanza C, et al. Aging effects in GaAs electroluminescent diodes. Solid-St Electron 1967;10:21e31.

2. Osamu Ueda. Reliability and degradation of III-V optical devices. Boston: Artech House;1996.


3. Yang SC, et al. Accelerated degradation of high power light-emitting diode resulted frominhomogeneous current distribution. Jpn J Appl Phys 2011;50. 034301-1e034301-6.

4. Meneghesso G, et al. Reliability of visible GaN LEDs in plastic package. MicroelectronReliab 2003;43:1737e42.

5. Kozodoy P, et al. Depletion region effects in Mg-doped GaN. J Appl Phys 2000;87:770e5.6. Yang SC, et al. Failure and degradation mechanisms of high-power white light emitting

diodes. Microelectron Reliab 2010;50:959e64.7. Matteo M, et al. Soft and hard failures of InGaN-based LEDs submitted to electrostatic

discharge testing. IEEE Elec Dev Lett 2010;31:579e81.8. Chen NC, et al. Damage of light-emitting diodes induced by high reverse-bias stress. J Cryst

Growth 2009;311:994e7.9. Shim J-I. Design and characterization issues in GaN-based light emitting diodes. Proc SPIE

2008;7135:71350C. 71351C-1e71350C-9.10. Meneghesso G, et al. Electrostatic discharge and electrical overstress on GaN/InGaN light

emitting diodes. In: Electrical overstress/electrostatic discharge symposium; 2001.p. 247e52.



Physical mechanisms limiting theperformance and the reliability ofGaN-based LEDs

14Carlo De Santi 1, Matteo Meneghini 1, Alberto Tibaldi 2, Marco Vallone 3,Michele Goano 2,3, Francesco Bertazzi 2,3, Giovanni Verzellesi 4,Gaudenzio Meneghesso 1, Enrico Zanoni 11University of Padova, Padova, Italy; 2Istituto di Elettronica e di Ingegneriadell’Informazione e delle Telecomunicazioni, Consiglio Nazionale delle Ricerche, Torino,Italy; 3Politecnico di Torino, Torino, Italy; 4Universit�a di Modena e Reggio Emilia, ReggioEmilia, Italy

Introduction

GaN-based optoelectronic devices are the market standard for light emission in theblue-green visible range, and the emission wavelengths are rapidly approachingthe UV part of the electromagnetic spectrum. Given the complexity of theirstructure, composed of multiple quantum wells (MQWs), carrier blocking layers,nucleation layers, and strain relief structures for the reduction of the strain causedby the heteroepitaxial growth, the analysis of the mechanisms influencing theirperformance is not straightforward, and often involves the use of computer-assistedsimulations.

The dynamics of the loss mechanisms may significantly change during the opera-tion of the devices, resulting in the worsening of the overall optical and electrical prop-erties. The generation of defects inside the active region, enhanced by the temperatureand the bias level, affects the trap-assisted tunneling current components and theamount of non-radiative recombination. The diffusion of impurities/point defects,often originating from the p-side of the device, may lead to an increase in the non-radiative recombination components.

The first part of this chapter describes the main processes that influence the opticaland electrical behavior of the devices below and above the optical turn-on, along withtheir theoretical framework and a brief review of the literature on the topic. In thesecond part, we present a comprehensive analysis of diffusion-related degradationprocesses, based on previous literature reports. Finally, we present a discussion onthe possible physical models able to explain experimental data on degradation ofGaN-based optoelectronic devices.


https://doi.org/10.1016/B978-0-08-101942-9.00014-9

14.1 Modeling the performance-limiting effects inGaN-based LEDs

Modeling of GaN-based active devices is among the most challenging topics in theoptoelectronics community, as demonstrated by the number of research institutionscurrently involved in the development of computer-aided design (CAD) tools basedon semiconductor physics. Semiconductor device simulators consist of a transportmodel providing a description of the motion of charges, and of electromagnetic modelsdescribing the induced and external fields. The relevant quantities for the characteriza-tion of the device performance, such as current or carrier distributions, are obtained bydiscretizing and solving these models, which must be coupled self-consistently.1 ForGaN LED analysis, the electromagnetic model can be reduced to the quasi-static Pois-son equation. The description of carrier transport, at least in the MQW active region,should be addressed within a genuine quantum framework,2e4 which is essential to un-derstand such open problems as the nature of droop, that is, the non-thermal decline ofthe internal quantum efficiency (IQE) in III-N based LEDs at high current densities.However, since the application of full-quantum approaches to an entire device is un-feasible with present-day computational resources, we still need to rely on semi-classical pictures. Presently the most popular choice by far is the drift-diffusion model,a simplification of the “master semiclassical model”dthe Boltzmann transport equa-tion (BTE).5,6 Several strategies aimed at preserving the computational advantages ofthe drift-diffusion model while including descriptions of different quantum phenom-ena are currently investigated. In the following, we present two complementary exam-ples of quantum corrections, specifically relevant below and above the optical turn-on.The former regards the inclusion of forward-bias trap-assisted tunneling effects inGaN-based LED structures described realistically at material, technology, and geomet-rical levels.7e9 The latter consists of the modeling of quantum effects in the opticalactive region, where radiative recombination occurs, in view of striving for a reliableestimate of the LED IQE.

14.1.1 Excess subthreshold forward current induced bynon-local trap-assisted tunneling mechanisms

Although not directly related to bias conditions adopted for light emission, GaN LEDsbehavior in the subthreshold forward-bias regime contains useful information thatmakes it worth investigating. In fact, the excess current in this operating region is asensitive indicator of device growth quality,10 and its increase is correlated to devicedegradation.11,12 The deviations of the current-voltage (IV) relationship from idealityhave been attributed to trap-assisted tunneling (TAT) mechanisms, which some au-thors consider to be involved in the efficiency droop phenomenon at high drivingcurrents.13e16 Several studies (most of them dealing with devices grown on sapphiresubstrates) have pointed out the possible role of trap-assisted tunneling in causing theexcess subthreshold forward current typically observed in GaN-based LEDs.10,11,17e32

Most of these works are based on the analysis of the large ideality factor valuesextracted from experimental IV curves.


Under forward bias, tunneling contributions to LED current can arise as a result ofthe following processes. Modulation of the junction potential barrier can make the con-duction band on the n-type side of the LED structure energetically aligned to trap statesspatially located on the other side of the quantum well in the p region, as well asvalence band on the p-type side energetically aligns to trap states in the n region.33

As a result, electrons can tunnel from the n-region into p-region traps, whereas holescan tunnel from the p-region into n-region traps. Non-radiative recombinationprocesses can then occur at those trap states involving the trapped carrier and a freemajority carrier. Traps, besides behaving as centers of standard Shockley-Read-Hall(SRH) recombination, can therefore support the carrier flow across the LED activeregion through the described nonlocal, trap-assisted tunneling (TAT) process. Theresulting current contributions are characterized by large ideality factors, thus domi-nating the IV curve only at low-to-moderate forward bias, before radiative recombina-tion settles on. We were able to account for the above effects within the framework of adrift-diffusion transport simulator, by using the “tunneling-into-traps” model availablein the commercial device simulator Sentaurus Device (Synopsys Inc.34). Specifically,the additional non-radiative recombination channel associated with the above,nonlocal TAT mechanism is included by properly modeling the electron and holecapture cross-sections of traps, namely by expressing them as functions of the carriertunneling probability by means of a Wentzel-Kramers-Brillouin formalism. Bothelastic and inelastic contributions to tunneling can be accounted for. A comprehensivediscussion of the adopted models can be found in Refs. 7, 8, along with a sensitivityanalysis of model predictions with respect to different material and trap parameters.Figs. 14.1 and 14.2 show selected results obtained on two different blue InGaN/GaN

1.E+31.E+21.E+11.E+01.E-11.E-21.E-31.E-4

1.E-61.E-7

1.E-5

1.E+21.E+11.E+01.E-11.E-21.E-31.E-41.E-51.E-61.E-71.E-8

1 2 3 4 1.5 2.5 3.5

Forward bias (V) Forward bias (V)

Cur

rent

den

sity

(A/c

m2 )

Cur

rent

den

sity

(A/c

m2 )

LED-Si

LED-SiC

exp

exp

sim - w/o TAT

sim - w/o TAT

sim - w/ TAT

sim - w/ TAT

Figure 14.1 Room temperature experimental and simulated IV characteristics of an (a) LED ona SiC substrate (LED-SiC) and (b) a Si substrate (LED-Si). Simulations are shown in the twocases with and without non-local trap-assisted tunneling effects. Assumed trap parameters(energy, density) in the different LED regions are as follows: (EEBL ¼ EC � 1.2 eV,NEBL ¼ 9.9 � 1015 cm�3), (Espacer ¼ EC � 1.4 eV, Nspacer ¼ 2 � 1017 cm�3),(En,QB ¼ EC � 1.7 eV, Nn,QB ¼ 3 � 1017 cm�3), (Ep,QB ¼ EV þ 1.5 eV,Np,QB ¼ 5 � 1016 cm�3).

Physical mechanisms limiting the performance and the reliability of GaN-based LEDs 457

single quantum well (SQW) LED test devices grown on highly conductive SiC(LED-SiC) and Si (LED-Si) substrates, respectively. SQW LEDs were specificallyadopted to put into evidence the nonlocal TAT mechanism in its simplest configurationdescribed above, whereas MQW LED structures would require TAT processesinvolving multiple traps to be accounted for. Radiative, Auger, and conventionalSRH recombination mechanisms were also included in the simulations; thermionicemission, as well as spontaneous and piezoelectric polarization charges, were activatedat heterointerfaces. Simulated, room-temperature forward-bias IV curves are comparedto experimental data in Fig. 14.1(a) and (b) for the LED-SiC and LED-Si devices,respectively. Simulations are reported for the two cases with and without nonlocalTAT activated. As can be noted, simulations without the nonlocal TAT deviate signif-icantly from experiments in the subthreshold regime for both LED-SiC and LED-Sidevices. On the other hand, when non-local TAT is taken into account in the simula-tions, a good agreement between experiments and simulations is achieved over a >5order-of-magnitude current interval, from on conditions down to the 10�4 A/cm2 currentregime. Simulation results were specifically obtained by assuming electron TAT fromthe n-GaN region into traps located in the EBL and spacer regions in the case ofLED-SiC, and both electron and hole nonlocal TAT processes into traps distributedwithin the quantum barrier in the case of LED-Si.

Additional insight can be gained by fitting the IV curves measured at different tem-peratures to the simplified Shockley diode equation I(V) ¼ IS exp (Ve/E0) (e being theelementary charge) and extracting the energy parameter E0. E0 values greater (smaller)than 90 meV are assumed to be the signature of electron (hole) TAT.10,18,23 The resultsof this analysis are plotted in Fig. 14.2, reporting the fitting values of E0 as a functionof temperature for both LED devices. The bias range adopted for the extraction of E0 is1.9e2.45 V for LED-SiC, while, for LED-Si, the fitting procedure has been appliedseparately to two voltage intervals in the low and intermediate bias regimes, namely1.5e2.3 and 2.3e2.8 V. As can be noted, E0 values in the 130e140 meV range areobtained for LED-SiC, denoting that electron TAT is the dominant tunneling contribu-tion for this device. On the other hand, for LED-Si, E0 values in the 90e130 and70e80 meV range are extracted in the low and intermediate bias range, respectively.

150 250 350 4500

50

100

150

200

Temperature (K)

LED-SiC, 1.9–2.45 VLED-Si, 1.5–2.3 VLED-Si, 2.3–2.8 V

E0

(meV

)

Figure 14.2 Energy parameter E0 (see text for definition) as a function of temperature extractedfrom experimental IV curves of the LED-SiC and LED-Si devices.


This suggests the contribution of hole TAT in the intermediate bias range and providesjustification for the need of accounting for both electron and hole nonlocal tunnelingeffects in the simulations of the LED-Si device in order to reproduce experimental IVcurves.

14.1.2 Modeling GaN LEDs above the optical turn-on: towardsan improved quantum description of the active region

The reconciliation of experiments and simulations in GaN-based LEDs becomes evenmore challenging when approaching the optical turn-on, due to the complex interplaybetween transport, material, and optical properties, and to the large uncertainties (per-taining, for example, to composition fluctuations, Auger transitions, polarizationfields, and other heterointerface properties, threading dislocations, etc.4) that affectIII-nitride materials to a much larger extent than conventional IIIeV light-emittingdevices.

As an example of limitation and possible improvement of the present modelingframework, we focus here on the description of carrier transport across the QW activeregion. The standard treatment of heterojunctions, based on thermionic emission,35

does not provide a complete picture of the physical mechanisms occurring in presenceof quantum wells, where the allowed energy levels of the confined electrons and holesare quantized. A remarkable progress was presented by Grupen et al. in the early 1990sin the context of the Minilase simulator applied to the modeling of quantum well la-sers.36 This seminal work and its successive developments summarized in Ref. 37inspired other researchers, such as Witzigmann38 and Streiff,39 from the optoelec-tronics group of IIS-ETH, who developed comprehensive simulation codes foredge-emitting and vertical-cavity surface-emitting lasers.a

According to this framework, a realistic model of the active region can be obtainedby applying a quantum description just for the portion of the device where quantumeffects are most relevant. To this aim, carriers are separated into two subsets: bulkcarriers (n3D, p3D), whose transport is described by the standard carrier continuityequations (or by higher-order semi-classical models, e.g., energy-balance or hydrody-namic), and bound carriers (n2D, p2D), which are distributed along the confinementdirection according to an envelope function. In order to keep current spreading intoaccount, 2D carriers can move freely in the lateral directions r, parallel to the inter-faces; hence their motion can be correctly described by a semi-classical model.

The envelope function of bound states is obtained by solving the Schr€odinger andPoisson equations in a “quantum box” that includes the active region. Appropriateboundary conditions must be defined at the edges of the quantum box, in order to avoidabrupt discontinuities in the resulting charge profiles.

Focusing the discussion on electrons (since computations for holes are completelyanalogous), the 3D and 2D carrier density distributions can be written as

a The commercial simulator TCAD Sentaurus Device from Synopsys has inherited these features, butdocumentation and support are currently limited.34


n3D r; z� �

¼Z N

Et

g3D Eð ÞfFD E;E3DF;n

�r; z��

dE

n2D r; z� �

¼X

m

jzm zð Þj2Z Et

Em

g2D Eð Þ fFD E;E2DF;n

�r��

dE

|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}N2D

m

; (14.1)

where g3DðEÞ and g2DðEÞ are the density of states of the bulk and bound populations,respectively40,41

g3D Eð Þ ¼ 4ph3

2m�e

� �3=2 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiE � EC

pg2D Eð Þ ¼ 4pm�

e

h2X

k

Q E � Ekð Þ (14.2)

and h, m�e and EC are the Planck constant, the electron effective mass and the bottom

edge of the conduction band. The Heaviside step function is indicated by QðEÞ andzmðzÞ is the envelope function [Ref. 6, Eq. (2.27)]. The function

fFDðE;EFÞ ¼�1þ exp

E � EF

kBT

��1

(14.3)

is the Fermi-Dirac distribution with energy E and Fermi level EF. In Eq. (14.1), theupper integral bound Et identifies the top of the quantum well, i.e., the EC level whichwould occur in absence of the heterostructure. Such a formulation, which is graphi-cally represented in Fig. 14.3, has been tailored to prevent carriers double counting. By

Ec

Cn

Et

Ew

cap

Figure 14.3 Description of the quantum capture model. The solid blue (gray in print versions)and dashed purple (dark gray in print versions) curves are examples of the conduction band edgerelative to the bulk and bound carriers, respectively. The circular markers indicate the drift-diffusion simulator mesh points. The bound-bulk carriers interaction described by the net capturerate Ccap

n occurs in the central node of the quantum well, indicated in red (light gray in printversions).


substituting Eq. (14.2) and (14.3) in Eq. (14.1), the following explicit expression forN2Dm is obtained:

N2Dm ¼ 4pkBTm�

e

h2

2

4ln

1þ exp

E2DF;n � Em

kBT

!!

� ln

1þ exp

E2DF;n � Et

kBT

!!3

5

(14.4)

The additional carrier populations are new unknowns of the drift-diffusion model,which lead to the requirement of additional continuity equations, for bound carrier, inorder to close the system. As in all BTE-derived models, the equations associated todifferent populations are coupled through collision terms, which reduce, in the drift-diffusion picture, to generation/recombination rates.6 This concept lies also behindthe carrier exchanges between bulk and bound continuity equations, which aredescribed by a net capture rate Ccap

n ¼ Rn � Gn. The recombination term can be writ-ten in terms of joint probability as follows,36,40,42

Rn ¼Z N

Et

Z Et

Eb

g3DðEcÞfFD�Ec;E

3DF;n

�g2DðEwÞ

�1� fFD

�Ew;E

2DF;n

��

� SðEc;EwÞdEwdEcxhSðEt;EbÞiZ N

Et

g3DðEcÞfFD�Ec;E

3DF;n

�dEc

|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}n3D

�

2

6664

Z Et

Eb

g2DðEwÞdEw

|fflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflffl}N2

�Z Et

Eb

g2DðEwÞfFD�Ew;E

2DF;n

�dEw

|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}N2D

3

7775

(14.5)

where hSðEt;EbÞi is an expected value of the bulk-bound scattering rate [Ref. 40,Chapter 7], whereas N2 (N2D) is the sum of all available (populated) bound states. Byvery similar considerations the following expression for the generation rate could beobtained

GnxhSðEb;EtÞi�n3DN2 � n3DN2D (14.6)

By exploiting

fFD�Ew;E

2DF;n

��1� fFD

�Ec;E

3DF;n

��exp

Ew � Ec

kBT

¼�1� fFD

�Ew;E

2DF;n

��fFD�Ec;E

3DF;n

�exp

E2DF;n � E3D

F;n

kBT

!


which relates the expected values in Eqs. (14.5) and (14.6) [Ref. 42 p. 32], thefollowing expression of the net capture rate can be obtained:

Ccapn ¼

1� exp

E2DF;n � E3D

F;n

kBT

!!1� N2D

N2

n3D

scape(14.7)

The net capture rate is proportional to the bulk carrier density n3D; at thermody-namic equilibrium, no splitting of the quasi-Fermi levels occurs, so the net capturerate is equal to zero; capture is zero also when N2D approaches N2, i.e., when all boundstates tends to be filled. The coefficient scape ¼ 1

�ðN2hSðEb;EtÞiÞ has dimensions oftime, so it is usually referred to as capture time.

The presence of N2D ¼PmN2Dm in Eq. (14.7) suggests that it should be the unknown

of the additional continuity equations. However, it is to be remarked that the contribu-tions N2D

m can be evaluated by inverting Eq. (14.4), knowing the bound state energylevels fEmg and E2D

F;n, which is a unique energy level representing all bound carriers.By using an appropriate sign convention, the term acts as a drain for 3D charges and asa source for the 2D carriers,6 modeling the capture as a generation/recombination forthe bound/bulk populations. The net capture term (Eq. 14.7) exhibits no dependenceon the confinement direction z. Hence, a single equation is added for each node trans-verse to z in the discretized model, as also indicated in Fig. 14.3, where a single meshpoint is reported in the well. As a convention, this term is usually introduced in thecentral z coordinate of the quantum region in the bulk continuity equations.37,43 There-fore, in the case of a 1D bulk drift-diffusion analysis, the quantum correction consistsof a single additional equation requiring the balance of capture with the other recom-bination rates (e.g., SRH, radiative, Auger).6 Then, from the bound states, 2D carrierscan either scatter back into the continuum or recombine.

For what concerns the electrostatic Poisson equation, it must be solved self-consistently with the quantum-corrected transport model by including the 2D chargecontribution in the right-hand side term:

� V$½εsVf� ¼ e�p3D þ p2D � �n3D þ n2D

�þ NþD � N�

A

; (14.8)

where εs is the position-dependent static dielectric constant, NþD and N�

A are the ionizeddoping concentrations and the 2D charge contributions n2D, p2D are defined as inEq. (14.1) and following.

14.1.2.1 A many-body formulation for capture time in InGaN/GaN QWs

The capture time scape introduced in Eq. (14.7) can be adopted as a fitting/phenomeno-logical parameter of the active region model. However, in the literature it is possible tofind several formulations aimed at providing an educated guess starting from thematerial and structural parameters of the semiconductor system.37 As an example, a


formulation based on many-body physics has been recently proposed,44 briefly sum-marized and commented below.

Electron capture in QWs may take place through electron-electron (ee) scattering,relaxation on defects, longitudinal optical (LO) phonon emission (e-ph), multi-phononemission, tunneling, etc. (see e.g., Refs. 45, 46 and references therein). Among theseprocesses, capture times via LO-phonon emission se�ph and via ee scattering seeconcur to the determination of the total QW capture time to a considerable extent.46e48

A QW capture process via ee scattering consists of an electron belonging to a barrierstate Jk1 with energy E and wavevector k1 that interacts with a second electron withwavevector k2 ending in the nth QW state fn;k1�q, with the exchange of a virtualphoton with wavevector q and frequency um. Correspondingly, the capture processvia e-ph can be described as an electronic transition from an initial barrier state Jk1to a final QW state fn;k1�q through the emission of a phonon of wavevector q and fre-quency um: as an example, the capture mechanism e-ph is sketched in Fig. 14.4.

The e-ph and the ee elementary interactions are described respectively by the un-screened Fr€olich and Coulomb potentials Vph and VN

ee , given by49

Vph ¼ �M2jqjD0ðumÞ; VN

ee ¼ Vee

εN¼ 4pe2

ε0εN��q��2; (14.9)

3D barrier state

2D QW state

2D QW state

⏐k1, E⟩

⏐k1, E⟩

⏐k1', E'⟩

⏐k1', E'⟩

LOphononemission

E-axis

⏐q, ћ m⟩ω

ћ mω

k1

Latticeion

q

ΔE

Figure 14.4 Electron capture process for the case of phonon emission (the ee scattering issimilar): an electronic state jk1;Ei in the barrier’s conduction band (CB) makes a transition tothe QW CB state

��k01;E0� emitting a quantum of energy Zum and momentum q.


where M2jqj ¼ ð1=2ÞKεuLOVee is the square of the unscreened electron-phonon matrix

elementKε ¼ ε�1N � ε

�1s , εN and εs are the dynamic and static dielectric con-

stants,D0ðumÞ ¼ 2uLO��

u2m � u2

LO

�is the unscreened phonon propagator,49,50 uLO

is the polar LO-phonon frequency, q is the virtual phonon or photon wavevector, um

its (bosonic) frequency and ε0 is the vacuum dielectric permittivity.At the lowest perturbative order, two electrons can interact exchanging a virtual

LO-phonon or photon, as shown in Fig. 14.5(a). At higher perturbative orders, thetwo considered scattering processes become deeply connected, as shown, for example,by the second order Feynman diagrams in Fig. 14.5(b).49,51 Thus, considering dia-grams with a single polarization bubble, at the nth perturbative order there are 2nþ1

possible arrangements, and the random phase approximation (RPA) effective eeplus e-ph interaction Veff ¼ Vee;s þ Vph;s is found summing up to infinite perturbativeorder all possible n-bubble diagrams (Fig. 14.5(c)), obtaining

Vee;s ¼ VNee

ε

; Vph;s ¼Vph

ε2�1�M2

jqjD0P=ε� ; (14.10)

+

+

+ + + + + +

+ +

...

LE: Check the 1st order (2nd order)

=

Arrow:

Waved line:

G0 = free particlepropagator

The effective interaction,dynamically screenedthrough the RPA dielectricfunction

ΣRPA

ΣRPA

= —— Σ

QW capture time:

Veff (q, m,n3D)

∞ ∞

∞–=

=

ωm 0ω

ω

∫ dqG0 ( m,q,k1,E )

τecap ⎛

⎝

⎛⎝2

ћIm

–1

(a)

(b)

(c)

(d)

Figure 14.5 First (a) and second (b) order Feynman diagrams for e-ph or ee scattering, withphonon (dashed line) or photon (wavy line) exchange. The dressed RPA interaction (doublewavy line) (c), and selfenergy

P

RPA(d), with a scheme of its calculation and how scape is related to

the imaginary part ofP

RPA.


where P ¼ ð1� εÞ�VNee is the polarization of the electron gas and ε is the RPA single

plasmon pole (SPP) dielectric function, given by ε�1 ¼ 1þ U2

pl

.�u2m � u2

q

�, con-

taining the n3D- and jqj-dependent effective plasmon and plasma frequencies uq andUpl.

49,52

The capture time scape is given by scape ¼ Z=2ð ÞIm P�1

RPA, where

P

RPA¼P

eeþ Pe�ph

is the total, complex RPA selfenergy due to ee and e-ph interactions

in presence of screening carrier plasma, whose density in barrier and QW are respec-tively n3D and N2D from Eq. (14.1) and following. Its expression49,53 is proportional tothe matrix element of Veff times G0, summed over um, and integrated in q, as sketchedin Fig. 14.5(d). In particular, its ee and e-ph contributions are given respectively by

Xee

k1;k2

Eð Þ ¼ �1bZ

X

q;k02;um

Jk02fn;k1�q

��Vee;s��Jk1Jk2

D EG0 um; q; k1;Eð Þ (14.11)

Xe�ph

k1

Eð Þ ¼ �1bZ

X

q;um

�fn;k1�q

��Vph;s��Jk1

�G0 um; q; k1;Eð Þ (14.12)

Here b is the inverse temperature in energy units,

G0ðum; q; k1;EÞ ¼ 1

ium þ iu� Z

��k1 � qj22m�

e� E2D

F;n

Z

(14.13)

is the Matsubara’s single-particle propagator49,50,53 where i is the imaginary unit,

u ¼ E=Z, and Z ¼ h=2p is the reduced Planck’s constant. The selfenergyPee

k1;k2

and similarlyPe�ph

k1

!

was evaluated by the following procedure based on complex

integration: (1) the k02 summation was performed by exploiting a Dirac-d factorstemming from the calculation of the potential matrix element, representing themomentum conservation at each vertex; (2) the summation over the frequency um wasdone following the Matsubara formalism (the Fermi nF and Bose nB occupation factorsappear during the summation thanks to the bosonic character of um)

49; (3) theq-summation was converted into an integral by exploiting the QW-plane translationalinvariance: considering for q and k1 their orthogonal and in-plane components�qz; qjj!

�and

�k1;z; k1k

�!�, the integration could be carried out analytically using

residue theorems, after having extended the integration to a complex domain, first in


qz, then in qjj!, without any truncation unlike in some numerical approaches.54e56 In theend, for ee scattering we obtained

ImXee

k1;k2

Eð Þ ¼ pam�eZc

εN

Z 2p

0Iee qð Þn;k1;k2

U2pl

uq

1þ nB � nFð ÞQ Eresð Þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2m�

eEres � Z2k21sin2 qð Þ

q dq;

(14.14)

where a is the fine structure constant, c is the speed of light, k1;2 ¼��k1;2k��!��, q ¼

��qjj!��, q

is the angle between qjj! and k1k�!

, Eres ¼ E � E2DF;nQ

�E2DF;n

�� Zuq, and the form factor

IeeðqÞ comes from the eigenfunctions overlap integral. The corresponding result for thee-ph process is

ImXe�ph;�

k1

Eð Þ ¼ 2pKεu2LOam

�eZc

Z 2p

0Ie�ph qð Þn;k1F�

1þ nB � nFð ÞQ E�res

� �

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2m�

eE�res � Z2k21sin

2 qð Þq dq;

(14.15)

where E�res ¼ E � E2D

F;nQ�E2DF;n

�� Zu�, Ie�phðqÞ is a form factor ensuing from the

wavefunctions overlap integral, F� ¼�u2� þ U2

pl � u2q

�2.h2u�

�u2� � u2

q

�

�u2� � u2

H

�i, and the upper or lower signs in

Pe�ph;�

k1refer to the emission of a

phonon-plasmon mode of frequency u�, where u2� ¼ u2

q � u2LO

�

2�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�u2q � u2

LO

�2 þ 4Kεu2LOU

2pl

r �2.

Since the two e-ph self-energiesPe�ph;�

k1correspond to the emission of two possible

and distinct LO-phonon-plasmons, two distinct quantum capture times 1�s�e�ph are

possible.In order to test the model, we considered two sets of experimental data. The first

(referred to as set A in the following) was obtained by W. H. Fan et al.47 employingtime-resolved differential transmission spectroscopy to evaluate the dependence of

the overall capture time scape ¼ 1��

1�sþe�ph þ 1

�s�e�ph þ 1

�see�

from n3D, for a

nominally intrinsic 2.5 nm/7.5 nm In0.08Ga0.92N/GaN well-barrier system. The second(set B in the following) was presented by A. David et al.57: with a small-signal analysisof a 4 nm/30 nm In0.09Ga0.91N/GaN single QW-barrier heterostructure, the authorsobtained indications of a scape with two components: a fast one, around 1 ps and attrib-uted to the e-ph process, and a second one, proportional to 1

�n3D, much slower in the


experimentally explored interval of n3D (1012 < n3D < 1014 cm�3), with a probablesignature of ee scattering. Adopting material parameter values reported in Refs. 50,58, we tested the present combined ee plus e-ph model against set A and set B data,separating the individual capture rate contributions and considering only the captureto the QW ground state.

Fig. 14.6 shows the calculated capture times for set A and set B for the three distinctprocesses: e-ph via modes u� and uþ, and ee scattering. In Fig. 14.6 two intervals ofn3D can be identified: in the low density regime, scape is mainly determined by LO-phonon emission as a u� phonon-plasmon mode, with characteristic time s�e�ph.

When the carrier density is increased, see and sþe�ph progressively reduce, competing

with s�e�ph when n3D is abovez1017 cm�3. In the interval of n3D corresponding to set

A experimental points (Fig. 14.6(a)) and also typical of LED operation, the capturetime is mainly given by contributions coming from see and sþe�ph at similar extent,

therefore investigations about IQE droop in LEDs should exclude neither of thesetwo mechanisms, and the customary approximation of the overall capture time scape

with a constant value (see e.g., Ref. 59) cannot be considered realistic.

14.2 Degradation of LEDs under electrical and thermalstress

In the past, several research groups investigated the different causes for the degrada-tion of LEDs and laser diodes submitted to accelerated lifetime tests. Typical

10–13

1012 1014 1016 1018 1012 1014 1016 1018

10–12

10–11

10–10

10–9

10–8

10–7

10–6

10–5

10–13

10–12

10–11

10–10

10–9

10–8

10–7

10–6

10–5

Cap

ture

tim

e (s

)

Cap

ture

tim

e (s

)

τ–e-ph

τ+e-ph

τcape

τeeset A

τ–e-ph

τ+e-ph

τee

τeeset B

(a) (b)

Figure 14.6 Electron capture times s�e�ph and see, calculated as functions of n3D, for set A(panel (a)) and set B (panel (b)). Experimental points are shown as symbols.Reproduced from Vallone M, Goano, M, Bertazzi, F, Ghione, G. Carrier capture in InGaN/GaNquantum wells: role of electron-electron scattering. J Appl Phys 2017;121(12):123107. https://dx.doi.org/10.1063/1.4979010, with the permission of AIP Publishing.


https://dx.doi.org/10.1063/1.4979010

https://dx.doi.org/10.1063/1.4979010

catastrophic failure modes include cracking of the semiconductor due to the thermalmismatch between different layers60e63 and damage caused by electrostatic discharges(ESD)64,65 or by electrical overstress (EOS).66,67 Besides catastrophic processes, thedegradation can also be ascribed to a gradual mechanisms. In this case, the degradationof the ohmic contacts68e70 affects mostly the electrical behavior of the devices,whereas the generation of defects inside the active region causes a significant decreasein the optical performance due to an enhanced non-radiative SRH recombination. Thedefects responsible for SRH recombination may be generated locally due to the com-bined effect of temperature and current flow or to the electric field,71,72 can be inducedby electromigration of metal from the contacts,73e76 or may propagate through theactive region via a diffusion process. The latter case will be reviewed in detail inthe following, focusing the analysis not only on LEDs but also on laser diodes, wherethe higher current density significantly enhances current- and temperature-drivendegradation processes.

The possible role of diffusion in the degradation of GaN-based devices was alreadyproposed in early works by Nam et al.77 and by Takeya et al.78 on laser diodes in 2003.The first report providing a possible evidence of a diffusion process in GaN optoelec-tronic devices was presented the following year by Nam et al.79 They investigated theperformance and reliability of violet laser diodes, and detected the diffusion of the Mgdopant from the p-side to the active region in the stressed samples. The analysis wascarried out on larger-area LED samples fabricated by using the same epitaxial wafersof the laser diodes, due to the large spot size of the secondary ion mass spectrometer(SIMS). They suggested that dislocations are a likely path for the diffusion process, asmentioned by Marona et al. in a subsequent study on laser diodes.80

This hypothesis was supported by the data collected by Orita et al. who listed valuesof the diffusion coefficients of several impurities in bulk or through dislocations fromsome studies in the literature.81 In the same paper, the authors provided a discussion onthe expected dependence of the threshold current variation on stress time when a diffu-sion process is responsible for the degradation, based on the solution of Fick’s secondlaw in one dimension.

14.2.1 Diffusion over time: solution of Fick’s second law in onedimension

In this model, a constant N0 density of impurities far away from the active region isdiffusing toward the quantum wells. After some amount of time t, the number ofimpurities Ndiff that can be found at position x can be expressed as

Ndiffðx; tÞ ¼ N0erfc

x

2ffiffiffiffiffiDt

p; (14.16)

where erfc is the error function and D the diffusion coefficient.81 The local increase inimpurity density, acting as non-radiative recombination deep centers, causes adecrease in the non-radiative lifetime snr . The variation over time can be described as


D

1snr

¼ KN0

w

Z w

0erfc

x

2ffiffiffiffiffiDt

pdx; (14.17)

where K is a coefficient that is related to carrier capture rate and w the width of thespace charge region.

Extrapolation of the non-radiative lifetime from experimental measurements in anLED is not trivial, given the competing role of the radiative lifetime in the final e�-hþ

pair lifetime and, therefore, in the optical power. The variation of snr during the agingof the device is easier to track in a laser diode, thanks to the direct effect on thresholdcurrent according to

Ith ¼ qV

hinj

Nth

sn(14.18)

1=sn ¼ BNth þ 1=snr; (14.19)

where q is the electron elementary charge, V the volume of the active region, Nth thethreshold carrier density, and B the bimolecular recombination coefficient. For thisreasondunder the assumption that the threshold carrier density does not vary overtimedwhen a diffusion process affects the performance and reliability of a laser diode,the variation in the threshold current should follow a square-root dependence onoperating time.

14.2.2 Signatures of diffusion-like processes in laser diodedegradation

In the aforementioned experimental reports, the presence of a diffusion process wassuggested or detected, but never verified by a specific design of experiment. The firsttests were carried out by De Santi et al. on unpackaged blue InGaN-based laserdiodes.82 Stress tests at constant current were found to cause a significant increasein the threshold current of the devices and a correlated decrease in subthreshold slope(see Fig. 14.7(a)). The authors attributed the degradation to the diffusion of impuritiestoward the active region, based on the square-root dependence of the stress kinetics.An additional remarkable finding is the spatial distribution of the damage, whichwas found to affect an area wider than the ridge by means of top-view cathodolumi-nescence imaging reported in Fig. 14.7(b). This effect was reported also by Marioliet al. the following year.83 Possible reasons for this behavior are (1) a non-optimal cur-rent confinement under the ridge, causing the spreading current to degrade a widerarea, (2) the temperature profile below the ridge, or (3) the damage caused by diffusionof impurities, which follows the concentration gradient and therefore is not strictly tiedto the p-side to n-side vertical direction.

Interestingly, the authors indicated that purely thermal stress (T ¼ 120�C, no cur-rent applied) results in different degradation kinetics with respect to current-drivenstress. Two main differences may be noticed: Fig. 14.8(a) shows that purely thermal


storage causes the strongest effect on the current-voltage curves, inducing changes inthe forward current. These changes may be related to variations of the series resistance,which is largely influenced by the resistivity of the p-type layer, due to the suboptimalMg activation and ionization. These variations are stronger under purely thermalstress, sincedduring constant-current stressdself-heating is stronger in the activeregion than at the p-GaN contact.

Conversely, the variation in the threshold current, shown in Fig. 14.8(b), is strongerin the constant current stress. Its value reflects the quality of the active region and of thefacets, which are submitted to more intense stimuli when the device is biased, withrespect to the case of purely thermal stress.

Opt

ical

pow

er (a

.u.)

(a)

(b)

0 20

1E-4

1E-3

0.01

0.1

1 55 mA, 25°C Untreated 1 min 2 min 4 min 8 min 16 min 32 min 64 min

0.0

0.5

1.0

1.5

2.0

Opt

ical

pow

er (a

.u.)

40

128 min 256 min 500 min 750 min 1350 min 2000 min 2500 min 3000 min

Current (mA)

Current (mA)

60 80

0 20 40 60 80 100

100

Stress condition:

Figure 14.7 (a) Increase in threshold current and decrease in subthreshold slope and (b) top-view cathodoluminescence signal at 420 nm of a device stressed at 55 mA, 25�C.Reprinted from De Santi C, Meneghini M, Marioli M, Buffolo M, Trivellin N, Weig T, Holc K,K€ohler K, Wagner J, Schwarz UTT, Meneghesso G, Zanoni E. Thermally-activated degradationof InGaN-based laser diodes: effect on threshold current and forward voltage. MicroelectronReliab 2014;54(9e10):2147e50. https://doi.org/10.1016/j.microrel.2014.07.073, Copyright(2014), with permission from Elsevier.


https://doi.org/10.1016/j.microrel.2014.07.073

In summary, this set of experiments points out an important role of temperature inthe degradation, even when no current flows through the device. The degradationkinetics are similar to the biased-stress tests, even though the magnitude is different.The broad degraded region (Fig. 14.7) suggests that the degradation may be relatedto a diffusion process, to the temperature profile below the ridge, or to an excess lateralcurrent spreading.

A more detailed analysis of the area affected by the degradation was provided by DeSanti et al. in Ref. 84. They tested several commercially available green laser diodesunder various bias conditions but at the same junction temperature, evaluated bymeans of the forward voltage method.85 The degradation kinetics followed a

ffiffit

pdependence, suggesting the presence of a diffusion process. By means of hyperspectral

0 1000 2000 3000 4000 50000.4

0.5

0.6

0.7

0.8

0.9

1.0

Stress time (min)

Pur

ely

ther

mal

sto

rage

0.7

0.8

0.9

1.0

Forward current (V = 6 V) variation, normalized

Con

stan

t cur

rent

stre

ss

0 1000 2000 3000 4000 5000

0.99

1.00

1.01

1.02

Threshold current variation, normalized

Stress time (min)

Pur

ely

ther

mal

sto

rage

1.0

1.1

1.2

1.3

1.4

Con

stan

t cur

rent

stre

ss

(a)

(b)

Figure 14.8 Variation in (a) forward current (at V ¼ 6V) and (b) threshold current duringstresses at constant current or under no bias at 120�C.Reprinted from De Santi C, Meneghini M, Marioli M, Buffolo M, Trivellin N, Weig T, Holc K,K€ohler K, Wagner J, Schwarz UTT, Meneghesso G, Zanoni E. Thermally-activated degradationof InGaN-based laser diodes: effect on threshold current and forward voltage. MicroelectronReliab 2014;54(9e10):2147e50. https://doi.org/10.1016/j.microrel.2014.07.073, Copyright(2014), with permission from Elsevier.



cathodoluminescence, they analyzed the shape and wavelength dependence of thedegraded region at the front facet and at a focused ion beam (FIB) cut. The degradationis visible as a semicircular dark area, symmetric along the p-to-n axis and originatingfrom the p-side. Even the cladding layer, waveguide, and bulk GaN are affected by it.All these effects are compatible with the diffusion of point defects toward the activeregion: extrinsic defects, such as magnesium or hydrogen, have been taken intoaccount as possibly responsible for degradation.

An extensive set of tests was reported by De Santi et al. on green commercialInGaN-based laser diodes at various junction temperatures and values of the biascurrent.86

The dots in Fig. 14.9 summarize the variation in the threshold current for sixdevices stressed at the same bias current (160 mA) and different junction temperature,along with a stress at the highest temperature and no applied bias. The dashed lines arethe fits according to the diffusion Eq. (14.17); the good accuracy confirms that thedegradation may be explained by a diffusion process. From the fits it is possible toextrapolate the diffusion coefficient at every junction temperature, and by plottingthese values in an Arrhenius plot a 1.98 eV activation energy for the diffusion coeffi-cient was computed (see Fig. 14.10).

This value is similar to the 1.93 eV reported by Seager et al. for H diffusion in pndiodes, and even closer to the 2.03 eV they computed from first principles for the sumof the activation energies for diffusion and binding to magnesium acceptors of ionizedhydrogen (Hþ).87 Since temperature alone is not able to induce any degradation in thesamples under test (see the purple squares in Fig. 14.9), the effect of the bias current isnot negligible, confirming the results in Fig. 14.8(b). For this reason, the authors

1 10 100

0%

10%

20%

30%

40%

50%

60% 150 °C, no bias

Stressed at 160 mA,junction temperature:

80°C 90°C 106°C 116°C 125°C 147°C

Thre

shol

d cu

rren

t inc

reas

e

Stress time (h)

Figure 14.9 Dots: variation of the threshold current in laser diodes stressed at 160 mA andvarious junction temperatures. Dashed lines: fitting according to the diffusion law (Eq. 14.17).Reprinted from De Santi C, Meneghini M, Meneghesso G, Zanoni E. Degradation of InGaNlaser diodes caused by temperature- and current-driven diffusion processes. MicroelectronReliab September 2016;64:623e26. https://doi.org/10.1016/j.microrel.2016.07.118, Copyright(2016), with permission from Elsevier.



carried out an additional set of aging tests at constant junction temperature andincreasing bias current.

The results are reported in Fig. 14.11, by using the same conventions as inFig. 14.9. Even in this case the fitting quality according to the diffusion Eq. (14.17)

27 28 29 30 31 32 33e–50

e–48

e–46

e–44

e–42

e–40

e–38

Diff

usio

n co

effic

ient

(cm

2 /s)

q/kT (eV–1)

Ea ~ 1.98 eV

Figure 14.10 Extrapolated diffusion coefficient versus thermal energy at the junction and itsactivation energy.Reprinted from De Santi C, Meneghini M, Meneghesso G, Zanoni E. Degradation of InGaNlaser diodes caused by temperature- and current-driven diffusion processes. MicroelectronReliab September 2016;64:623e26. https://doi.org/10.1016/j.microrel.2016.07.118, Copyright(2016), with permission from Elsevier.

1 10 100 1000

0%

10%

20%

30%

40%

50%

60%

70%

80%

Stressed at:Junction temperature 80°CBias current:

45 mA 100 mA 130 mA 145 mA 160 mA 180 mA

Thre

shol

d cu

rren

t inc

reas

e

Stress time (h)

Figure 14.11 Dots: variation of the threshold current in laser diodes stressed at 80�C junctiontemperature and various bias current. Dashed lines: fitting according to the diffusion law(Eq. 14.17).Reprinted from De Santi C, Meneghini M, Meneghesso G, Zanoni E. Degradation of InGaNlaser diodes caused by temperature- and current-driven diffusion processes. MicroelectronReliab September 2016;64:623e26. https://doi.org/10.1016/j.microrel.2016.07.118, Copyright(2016), with permission from Elsevier.




is very good, supporting the hypothesis on the role of a diffusion process in the degra-dation of the device. An important finding originates from the analysis of the extrap-olated diffusion coefficients, as shown in Fig. 14.12: when the bias current is higher,the diffusion coefficient is larger and the degradation is stronger. The effect of currentflow and electric field on the diffusion coefficient is usually overlooked in the litera-ture, and only one experimental report is available for the gallium nitride materialsystem by Seager et al.87 They show that Hþ impurities may be affected by drift underelectric field, supporting the extrapolation of the diffusion coefficient reported inFig. 14.12.

To confirm the assumptions, the paper by De Santi et al.86 reports a comprehensivesummary of the theoretical and experimental reports on the diffusion coefficient of Mgand H in the gallium nitride literature. The experimental data are in good agreementwith data for hydrogen (see Fig. 14.14), whereas magnesium is not a likely candidatein this case (see Fig. 14.13). This is not surprising: owing to its lower mass; hydrogenshould diffuse faster than magnesium, playing a stronger role in the degradation of thedevice.

The most striking information contained in Figs. 14.13 and 14.14 is the differencebetween experimental reports (in the green regions) and theoretical computations fromfirst principles (in the red regions), which significantly overestimate the speed of thediffusion. This suggests that the physical processes used in the literature for themodeling of the diffusion may not be completely accurate, lacking, for example, anadequate contribution from bias current and electric field. This conclusion assumesthat the degradation is related only to diffusion of impurities, and does not considerthe possible role of native defects. A more detailed discussion on native defects isgiven in Section 14.2.4.4.

50 100 150 200

6x10–22

7x10–22

8x10–22

9x10–22

Diff

usio

n co

effic

ient

(cm

2 /s)

Current (mA)

Figure 14.12 Extrapolated diffusion coefficient versus bias current.Reprinted from De Santi C, Meneghini M, Meneghesso G, Zanoni E. Degradation of InGaNlaser diodes caused by temperature- and current-driven diffusion processes. MicroelectronReliab September 2016;64:623e26. https://doi.org/10.1016/j.microrel.2016.07.118, Copyright(2016), with permission from Elsevier.



8 16 24 32 401E-43

1E-39

1E-35

1E-31

1E-27

1E-23

1E-19

1E-15

1E-11

1E-7

Experimental: [9] [10] [11] [12]

Calculated: [13] [14] [15] [5]M

g di

ffusi

on c

oeffi

cien

t D0

(cm

2 /s)

q/kT (eV–1)

This work

Figure 14.13 Experimental (green region (gray in print versions)) and calculated (red region(light gray in print versions)) diffusion coefficients for magnesium extrapolated from theliterature.Reprinted from De Santi C, Meneghini M, Meneghesso G, Zanoni E. Degradation of InGaNlaser diodes caused by temperature- and current-driven diffusion processes. MicroelectronReliab September 2016;64:623e26. https://doi.org/10.1016/j.microrel.2016.07.118, Copyright(2016), with permission from Elsevier, reference numbers according to the list in that paper.

24 32 401E-431E-391E-351E-311E-271E-231E-191E-151E-11

1E-7

Experimental: [7] [8]

Calculated: [16] [17] [5] [18] [19]

H d

iffus

ion

coef

ficie

nt D

0 (c

m2 /

s)

q/kT (eV–1)

This work

Figure 14.14 Experimental (green region (gray in print versions)) and calculated (red region(light gray in print versions)) diffusion coefficients for hydrogen extrapolated from the literature.Reprinted from De Santi C, Meneghini M, Meneghesso G, Zanoni E. Degradation of InGaNlaser diodes caused by temperature- and current-driven diffusion processes. MicroelectronReliab September 2016;64:623e26. https://doi.org/10.1016/j.microrel.2016.07.118, Copyright(2016), with permission from Elsevier, reference numbers according to the list in that paper.




14.2.3 Diffusion in LEDs

In the previous sections, all the experimental reports focused on the degradation oflaser diodes. In the literature, few reports on diffusion processes in LEDs are present,and they are mostly related to tests for laser diodes79 or to diffusion during growth.88,89

One paper describing the role of the diffusion in the degradation of light emittingdiodes was presented by La Grassa et al.90 The authors investigated the changesinduced by constant current stresses at high temperature of InGaN-based LEDs, bymonitoring the electrical and optical characteristics, the variation in the capacitance-voltage curve, and the capacitance deep level transient spectroscopy (C-DLTS)profiles. They extrapolated the variation in the “A” SRH non-radiative recombinationcoefficient by means of the method developed by van Opdorp and t’Hooft,91 detectinga clear increase due to the generation of deep levels inside the active region. Thekinetic, reported in Fig. 14.15, closely follows the

ffiffit

pdependence, suggesting the pres-

ence of a diffusion process.The same time dependence can be noticed in the capacitance integrated between 1.5

and 2 V (Fig. 14.16). This value corresponds to the total free charge scanned by theborder of the space charge region when its position is varied by the different voltages,and therefore shows an increase in free charge in a position close to the quantum wells,which may be caused by a higher density of defects acting as generation centers.

The apparent charge profile inside the device can be extrapolated from the CeVmeasurements under the assumption of a highly doped p-side, leading to a pþ-njunction. Its shape and time variation is reported in Fig. 14.17. The free charge densityclearly increases during stress, with a stronger variation closer to the quantum wells.

1.35

1.30

1.25

1.20

1.15

1.10

1.05

1.00

A co

effic

ient

(nor

mal

ized

)

0 10 20 30 40 50Stress time (h)

Figure 14.15 Increase in the “A” non-radiative radiation coefficient over stress time and itssquare-root dependence.Reprinted from La Grassa M, Meneghini M, De Santi C, Zanoni E, Meneghesso G. Degradationof InGaN-based LEDs related to charge diffusion and build-up.Microelectron Reliab September2016;64:614e6. https://doi.org/10.1016/j.microrel.2016.07.131, Copyright (2016), withpermission from Elsevier.



Both this effect and the square root of time dependence are consistent with a diffusionprocess of impurities originating from the p-side and moving toward the n-side.

The increase in the concentration of deep levels inside the active region isconfirmed by the C-DLTS analysis (Fig. 14.18). The stress caused the annealing of

8.4E-10

8.3E-10

8.2E-10

8.1E-10

8.0E-10

7.9E-10

C in

tegr

ated

0 2 4 6 8Square-root of stress time (h0.5)

Figure 14.16 Capacitance integrated in the 1.5e2 V range, showing a square-root of timedependence.Reprinted from La Grassa M, Meneghini M, De Santi C, Zanoni E, Meneghesso G. Degradationof InGaN-based LEDs related to charge diffusion and build-up.Microelectron Reliab September2016;64:614e6. https://doi.org/10.1016/j.microrel.2016.07.131, Copyright (2016), withpermission from Elsevier.

1.4×1018

1.2×1018

1.0×1018

8.0×1017

6.0×1017

Cha

rge

dens

ity (c

m–3

)

40 60 80 100 120Depletion width (nm)

Figure 14.17 Apparent free charge profile extrapolated from CeV curves, showing an increasein a well-defined region.Reprinted from La Grassa M, Meneghini M, De Santi C, Zanoni E, Meneghesso G. Degradationof InGaN-based LEDs related to charge diffusion and build-up.Microelectron Reliab September2016;64:614e6. https://doi.org/10.1016/j.microrel.2016.07.131, Copyright (2016), withpermission from Elsevier.




a preexisting deep level with activation energy E1 ¼ 0.4 eV, which should lead to animprovement in the electrical and optical performance of the device. Unfortunately, asecond deep level with activation energy E2 ¼ 0.9 eV was created in the active region.Its higher activation energy places it deeper in the energy gap, where it acts as a moreefficient non-radiative recombination center compared to the more shallow E1. As aconsequence, even though one deep level was no more detectable, the rise of a secondand more efficient one caused the degradation in the optical power.

14.2.4 Diffusion mechanisms

A discussion on the possible diffusion models able to explain the experimental data inSection 14.2.2 is given in the same paper by De Santi et al.86

Part of the analysis relies on some background on the presence and role of magne-sium and hydrogen in gallium nitride. Magnesium is the most common p-type dopant,but only atoms in a substitutional position in the crystal lattice are able to contribute tothe current flow in the device as acceptors. A relevant portion of Mg atoms may bondwith hydrogen, whose presence in the growth chamber cannot be completely elimi-nated, and occupy an interstitial position in the final lattice. The MgeH bond canbe broken by the combined effect of temperature and energetic carriers, a processreferred to as “activation,” increasing the effective doping level and leaving some mag-nesium and hydrogen atoms able to move inside the lattice.92,93 Therefore, the mostcommon impurities found in an interstitial position in gallium nitride are MgeH com-plexes and hydrogen atoms in the Hþ or H0 species.87,94e96

The diffusion of substitutional atoms is typically slower than the diffusion of inter-stitials, because substitutionals need to be moved in interstitial position before

0.0

–2.0×10–5

–4.0×10–5

–6.0×10–5

–8.0×10–5

–1.0×10–4

100 200 300 400 500Temperature (K)

Before stressAfter stress

e2

e1

e1

DLT

S s

igna

l (ΔC

/C)

Figure 14.18 Variation of the capacitance DLTS signal as a consequence of stress.Reprinted from La Grassa M, Meneghini M, De Santi C, Zanoni E, Meneghesso G. Degradationof InGaN-based LEDs related to charge diffusion and build-up.Microelectron Reliab September2016;64:614e6. https://doi.org/10.1016/j.microrel.2016.07.131, Copyright (2016), withpermission from Elsevier.



diffusing, thus slowing the overall process. This reasoning is confirmed by some theo-retical predictions.97,98 For these reasons, the analysis reported in Ref. 86 focuses onlyon diffusion of interstitials or native defects.

14.2.4.1 Dopant activation

A first model to explain the degradation kinetics considers the local effects of thedopant activation process. Once the magnesium atom moves to substitutional position,its coupled hydrogen atom(s) (red circles in Fig. 14.19) is left in an interstitial position.As a consequence, the local concentration of mobile impurities after the activationincreases (from left to right in the sketch), as does the concentration gradient, leadingto a stronger diffusion. The amount of broken MgeH bonds is directly proportional tothe amount of electrons travelling in the p-side, that is, to the current flow, confirmingthe results and the linear dependence in Fig. 14.12. Unfortunately, in the case of thepaper under analysis this process is not the root cause for the degradation, because acti-vation of residual dopants is usually completed after few hours.99,100

14.2.4.2 Kinetic energy transfer

The second model is based on the overall effects of the current flow through the device.The moving electrons gain kinetic energy, which could be transferred to the diffusingimpurities through collisions. The electron energy might reach values in the order offew electronvolts,101,102 even higher in the high-energy tail of the electron distribu-tion.103,104 This value is significantly higher than the average thermal phonon energy(z25 meV at room temperature) or LO phonon energy (z90 meV, Refs. 105e110)driving the diffusion.

Again, this process cannot explain the experimental data. Electrons (blue circles inFig. 14.20) flow from the n-side to the p-side; therefore, they would act as a diffusion-reducing factor, limiting impurities movement toward the active region from thep-side, but this is not consistent with Fig. 14.12.

It is worth noticing that the sole increased temperature due to self-heating at higherbias levels cannot strengthen in a sufficient way the diffusion, because no change isdetected in Fig. 14.9 after purely thermal stress.

14.2.4.3 Drift of the impurities

In the first two models the role of the electric field in the space charge region is nevertaken into account. The higher mass of magnesium, with respect to hydrogen, suggests

p-side QW p-side QW

Figure 14.19 Sketch of a diffusion process driven by the dopant activation.


that only the drift of the latter should cause an effect on the optical properties, sinceboth of them are usually present in similar concentration inside the device.87,111 Asmentioned before, atomic hydrogen is usually present in two states: H0, unaffectedby drift due to its null electrical charge, and Hþ, which diffuses from the p-side tothe n-side (red arrow in Fig. 14.21) but drifts from the n-side to the p-side due tothe built-in electric field in the space charge region (blue arrow in the sketch). There-fore, in the space charge region the drift opposes the diffusion process, consistent withprevious reports.87

At higher bias current levels (from left to right in Fig. 14.21), in the simpler case of ap-n junction the electric field in the space charge region is reduced, leading to a weakerdrift and to a stronger diffusion.

14.2.4.4 Native defects

Magnesium and hydrogen (or other impurities in the growth chamber) are not the onlyatoms that could be found in an interstitial position. During growth, gallium and nitro-gen atoms, which should occupy a well-defined position in the lattice, have a nonzeroprobability of reaching a different site, becoming gallium (Gai) and nitrogen (Ni)interstitials. These interstitials could diffuse, as could do gallium and nitrogen atomsin substitutional position (GaN and NGa). Additionally, even though no actual atomis moving and therefore it is not a diffusion process per se, the propagation of Galliumvacancies (VGa), nitrogen vacancies (VN), and dislocations is usually referred to asdiffusion in the literature.112,113

All these native defects are not detectable through SIMS, because they cannot bedistinguished from the atoms in the lattice. For this reason, it is not possible to evaluate

p-side QW

Figure 14.20 Sketch of a diffusion process driven by the transfer of kinetic energy from the biascurrent.

p-side QW p-side QW

Figure 14.21 Sketch of a diffusion process influenced by the drift of the diffusing impurities.


their diffusion coefficient or its activation energy, and compare them with the exper-imental data. Gallium vacancies are known non-radiative recombination centers114das are Mg88 and H115dand therefore their role in the degradation of the devices undertest could not be confirmed or excluded. Foreign impurities (as oxygen or hydrogen)may also make chemical bonds with vacancies, thus resulting in the creation of non-radiative recombination centers.

14.3 Conclusions

In summary, most of the structural issues affecting performance and reliability of GaN-based optoelectronic devices were solved by iterations of the technological process anddesign improvements. At the present time, the main limiting factors are the efficiencydroop and the presence of defects.

The study of these topics is theoretically complex and the design of suitable exper-imental tests is not trivial, leading to the need for accurate models and improvedcomputer-assisted tools for the simulation of the current conduction mechanismdespecially in the subthreshold forward current regimedand for the quantum descrip-tion of the active region.

If numerical simulations can enhance our understanding of the variation in perfor-mance caused by the defects, experimental tests are required to understand their origin.Diffusion of impurities or native defects is a common process in LEDs and laserdiodes, and can significantly affect the electrical and optical behavior of the devices,leading to a decrease in their lifetime. The role of the bias current and/or the electricfield cannot be disregarded in this analysis, because it was found to have a strongimpact on the diffusion coefficient.

References

1. Ghione G, Benvenuti A. Discretization schemes for high-frequency semiconductor devicemodels. IEEE Trans Antennas Propag 1997;AP-45(3):443e56.

2. Aeberhard U. A microscopic theory of quantum well photovoltaics. Eidgen€ossischeTechnische Hochschule Z€urich; 2008.

3. Shedbalkar A, Andreev Z, Witzigmann B. Simulation of an indium gallium nitridequantum well light-emitting diode with the non-equilibrium Green’s function method.Phys Status Solidi 2016;253(1):158e63. https://doi.org/10.1002/pssb.201552276.

4. Goano M, Bertazzi F, Zhou X, Mandurrino M, Dominici S, Vallone M, Ghione G,Tibaldi A, Calciati M, Debernardi P, Dolcini F, Rossi F, Verzellesi G, Meneghini M,Trivellin N, De Santi C, Zanoni E, Bellotti E. Challenges towards the simulation ofGaN-based LEDs beyond the semiclassical framework. In: Proceedings of SPIE ethe International Society for Optical Engineering; 2016. https://doi.org/10.1117/12.2216489.

5. Fischetti M, VandenbergheWG. Advanced physics of electron transport in semiconductorsand nanostructures. Berlin: Springer-Verlag; 2016.


https://doi.org/10.1002/pssb.201552276

https://doi.org/10.1117/12.2216489

https://doi.org/10.1117/12.2216489

6. Bertazzi F, Goano M, Ghione G, Tibaldi A, Debernardi P, Bellotti E. Electron transport. In:Piprek J, editor. Handbook of optoelectronic device modeling and simulation. Boca Raton,FL: CRC Press; 2017. p. 35e80 [chapter 2].

7. Mandurrino M, Verzellesi G, Goano M, Vallone M, Bertazzi F, Ghione G, Meneghini M,Meneghesso G, Zanoni E. Physics-based modeling and experimental implications of trap-assisted tunneling in InGaN/GaN light-emitting diodes. Phys Status Solidi May 2015;212(5):947e53. https://doi.org/10.1002/pssa.201431743.

8. Mandurrino M, Goano M, Vallone M, Bertazzi F, Ghione G, Verzellesi G, Meneghini M,Meneghesso G, Zanoni E. Semiclassical simulation of trap-assisted tunneling in GaN-based light-emitting diodes. J Comp Electron 2015;14(2):444e55. https://doi.org/10.1007/s10825-015-0675-3.

9. Mandurrino M, Goano M, Dominici S, Vallone M, Bertazzi F, Ghione G, Bernabei M,Rovati L, Verzellesi G, Meneghini M, Meneghesso G, Zanoni E. Trap-assisted tunnelingcontributions to subthreshold forward current in InGaN/GaN light-emitting diodes. In:SPIE OpticsþPhotonics, 14th International Conference on solid state lighting and LED-based illumination systems, vol. 9571; 2015, https://doi.org/10.1117/12.2187443.95710Ue1e95710Ue6.

10. Kim J, Tak Y, Kim J, Chae S, Kim J-K, Park Y. Analysis of forward tunneling current inInGaN/GaN multiple quantum well light-emitting diodes grown on Si (111) substrate.J Appl Phys 2013;114(1):13101. https://doi.org/10.1063/1.4812231.

11. Liu L, Yang J, Teng D, Qi S, Wang G. An explanation for invalidity of working currents’derating on improving light-emitting diode devices’ reliability. J Appl Phys 2013;114(2):23102. https://doi.org/10.1063/1.4813092.

12. Yang T-K, Kim S-B. Quantitative analysis of initial short-term aging behavior and itsimplication for the efficiency droop in blue light-emitting diodes. IEEE Trans ElectronDevices 2012;59(10):2656e61. https://doi.org/10.1109/TED.2012.2209181.

13. Bochkareva NI, Zhirnov EA, Efremov AA, Rebane YT, Gorbunov RI, Shreter YG.Tunnel-recombination currents and electroluminescence efficiency in InGaN/GaN LEDs.Semiconductor 2005;39(5):627e32. https://doi.org/10.1134/1.1923571.

14. Bochkareva NI, V Voronenkov V, Gorbunov RI, Zubrilov AS, Lelikov YS, Latyshev PE,Rebane YT, Tsyuk AI, Shreter YG. Defect-related tunneling mechanism of efficiencydroop in III-nitride light-emitting diodes. Appl Phys Lett 2010;96(13):133502. https://doi.org/10.1063/1.3367897.

15. Verzellesi G, Saguatti D, Meneghini M, Bertazzi F, Goano M, Meneghesso G, Zanoni E.Efficiency droop in InGaN/GaN blue light-emitting diodes: physical mechanisms andremedies. J Appl Phys 2013;114(7):71101. https://doi.org/10.1063/1.4816434.

16. Calciati M, Goano M, Bertazzi F, Vallone M, Zhou X, Ghione G, Meneghini M,Meneghesso G, Zanoni E, Bellotti E, Verzellesi G, Zhu D, Humphreys C. Correlatingelectroluminescence characterization and physics-based models of InGaN/GaN LEDs:pitfalls and open issues. AIP Adv 2014;4:67118. https://doi.org/10.1063/1.4882176.

17. Casey HC, Muth J, Krishnankutty S, Zavada JM. Dominance of tunneling current and bandfilling in InGaN/AlGaN double heterostructure blue light-emitting diodes. Appl Phys Lett1996;68(20):2867e9. https://doi.org/10.1063/1.116351.

18. Perlin P, Osinski M, Eliseev PG, Smagley VA, Mu J, Banas M, Sartori P. Low-temperaturestudy of current and electroluminescence in InGaN/AlGaN/GaN double-heterostructureblue light-emitting diodes. Appl Phys Lett 1996;69(12):1680. https://doi.org/10.1063/1.117026.


https://doi.org/10.1002/pssa.201431743

https://doi.org/10.1007/s10825-015-0675-3

https://doi.org/10.1007/s10825-015-0675-3

https://doi.org/10.1117/12.2187443

https://doi.org/10.1063/1.4812231

https://doi.org/10.1063/1.4813092

https://doi.org/10.1109/TED.2012.2209181

https://doi.org/10.1134/1.1923571

https://doi.org/10.1063/1.3367897

https://doi.org/10.1063/1.3367897

https://doi.org/10.1063/1.4816434

https://doi.org/10.1063/1.4882176

https://doi.org/10.1063/1.116351

https://doi.org/10.1063/1.117026

https://doi.org/10.1063/1.117026

19. Chitnis A, Kumar A, Shatalov M, Adivarahan V, Lunev A, Yang JW, Simin G, Khan MA.High-quality p-n junctions with quaternary AlInGaN/InGaN quantum wells. Appl PhysLett 2000;77(23):3800e2. https://doi.org/10.1063/1.1331084.

20. Cao XA, Stokes EB, Sandvik PM, LeBoeuf SF, Kretchmer J, Walker D. Diffusion andtunneling currents in GaN/InGaN multiple quantum well light-emitting diodes. IEEEElectron Device Lett 2002;23(9):535e7. https://doi.org/10.1109/LED.2002.802601.

21. Hirsch L, Barriere a S. Electrical characterization of InGaN/GaN light emitting diodesgrown by molecular beam epitaxy. J Appl Phys 2003;94(8):5014e20. https://doi.org/10.1063/1.1605252.

22. Cao XA, Teetsov JM, D’Evelyn MP, Merfeld DW, Yan CH. Electrical characteristics ofInGaN/GaN light-emitting diodes grown on GaN and sapphire substrates. Appl Phys Lett2004;85(1):7e9. https://doi.org/10.1063/1.1767280.

23. Reynolds CL, Patel A. Tunneling entity in different injection regimes of InGaN lightemitting diodes. J Appl Phys 2008;103(8):86102. https://doi.org/10.1063/1.2906326.

24. Averkiev NS, Chernyakov AE, Levinshtein ME, V Petrov P, Yakimov EB, Shmidt NM,Shabunina EI. Two channels of non-radiative recombination in InGaN/GaN LEDs. PhysRev B 2009;404:4896e8. https://doi.org/10.1016/j.physb.2009.08.252.

25. Yan D, Lu H, Chen D, Zhang R, Zheng Y. Forward tunneling current in GaN-based bluelight-emitting diodes. Appl Phys Lett 2010;96(8):83504. https://doi.org/10.1063/1.3327332.

26. Nana R, Gnanachchelvi P, Awaah MA, Gowda MH, Kamto AM,Wang Y, Park M, Das K.Effect of deep-level states on current-voltage characteristics and electroluminescenceof blue and UV light-emitting diodes. Phys Status Solidi 2010;207(6):1489e96. https://doi.org/10.1002/pssa.200925596.

27. Lee KB, Parbrook PJ, Wang T, Ba J, Ranalli F, Airey RJ, Hill G. The origin of the highideality factor in AlGaN-based quantum well ultraviolet light emitting diodes. Phys StatusSolidi 2010;247(7):1761e3. https://doi.org/10.1002/pssb.200983617.

28. Huang S, Xian Y, Fan B, Zheng Z, Chen Z, Jia W, Jiang H, Wang G. Contrary lumi-nescence behaviors of InGaN/GaN light emitting diodes caused by carrier tunnelingleakage. J Appl Phys 2011;110(6):64511. https://doi.org/10.1063/1.3642955.

29. Lee J-M, Kim S-B. Analysis of current components and estimation of internal quantumefficiency in light-emitting diodes. IEEE Trans Electron Devices 2011;58(9):3053e7.https://doi.org/10.1109/TED.2011.2158826.

30. Kang B, Kim S-B. Temperature dependence of the component currents and internalquantum efficiency in blue light-emitting diodes. IEEE Trans Electron Devices 2013;60(3):1060e7. https://doi.org/10.1109/TED.2013.2242470.

31. Meyaard DS, Cho J, Schubert EF, Han S-H, Kim M-H, Sone C. Analysis of the temper-ature dependence of the forward voltage characteristics of GaInN light-emitting diodes.Appl Phys Lett 2013;103(12):121103. https://doi.org/10.1063/1.4821538.

32. Auf der Maur M, Galler B, Pietzonka I, Strassburg M, Lugauer H, Di Carlo A. Trap-assisted tunneling in InGaN/GaN single-quantum-well light-emitting diodes. Appl PhysLett 2014;105(13):133504. https://doi.org/10.1063/1.4896970.

33. del Alamo JA, Swanson RM. Forward-bias tunneling: a limitation to bipolar devicescaling. IEEE Electron Device Lett 1986;7(11):629e31. https://doi.org/10.1109/EDL.1986.26499.

34. Version M-2016.12 Sentaurus Device User Guide. Mountain View, CA: Synopsys Inc.;December 2016.


https://doi.org/10.1063/1.1331084

https://doi.org/10.1109/LED.2002.802601

https://doi.org/10.1063/1.1605252

https://doi.org/10.1063/1.1605252

https://doi.org/10.1063/1.1767280

https://doi.org/10.1063/1.2906326

https://doi.org/10.1016/j.physb.2009.08.252

https://doi.org/10.1063/1.3327332

https://doi.org/10.1063/1.3327332




https://doi.org/10.1063/1.3642955

https://doi.org/10.1109/TED.2011.2158826

https://doi.org/10.1109/TED.2013.2242470

https://doi.org/10.1063/1.4821538

https://doi.org/10.1063/1.4896970

https://doi.org/10.1109/EDL.1986.26499

https://doi.org/10.1109/EDL.1986.26499

35. Schroeder D. Modelling of interface carrier transport for device simulation. Wien:Springer-Verlag; 1994.

36. Grupen M, Kosinovsky G, Hess K. The effect of carrier capture on the modulationbandwidth of quantum well lasers. In: Proc. IEDM; 1993. 23.6.1e23.6.4.

37. Grupen M, Hess K. Simulation of carrier transport and nonlinearities in quantum-well laserdiodes. J Quan Electron 1998;34(1):120e40. https://doi.org/10.1109/3.655016.

38. Witzigmann B, Witzig A, Fichtner W. A multidimensional laser simulator for edge-emitters including quantum carrier capture. IEEE Trans Electron Devices 2000;47(10):1926e34. https://doi.org/10.1109/16.870574.

39. Streiff M, Witzig A, Pfeiffer M, Royo P, Fichtner W. A comprehensive VCSEL devicesimulator. IEEE J Sel Top Quan Electron 2003;9(3):879e91.

40. Hess K. Advanced theory of semiconductor devices. Englewood Cliffs, NJ: Prentice-Hall;1988.

41. Sze SM, Ng KK. Physics of semiconductor devices. 3rd ed. Hoboken, NJ: John Wiley &Sons; 2007.

42. Kosinovsky GA. Threshold current and modulation response of semiconductor lasers.University of Illinois at Urbana-Champaign; 1995.

43. Steiger S. Modelling nano-LEDs. Eidgen€ossische Technische Hochschule Z€urich; 2009.44. Vallone M, Goano M, Bertazzi F, Ghione G. Carrier capture in InGaN/GaN quantum

wells: role of electron-electron scattering. J Appl Phys 2017;121(12):123107. https://doi.org/10.1063/1.4979010.

45. K�alna K, Mo�sko M. Electron capture in quantum wells via scattering by electrons, holes,and optical phonons. Phys Rev B 1996;54(24):17730e7. https://doi.org/10.1103/PhysRevB.54.17730.

46. Sizov DS, Bhat R, Zakharian A, Song K, Allen DE, Coleman S, Zah C. Carrier transport inInGaN MQWs of aquamarine and green-lased diodes. IEEE J Sel Top Quan Electron2011;17(5):1390e401. https://doi.org/10.1109/JSTQE.2011.2116770.

47. Fan WH, Olaizola SM, Wells J-PR, Fox AM, Wang T, Parbrook PJ, Mowbray DJ,Skolnick MS. Femtosecond studies of electron capture times in InGaN/GaN multiplequantum wells. Appl Phys Lett April 2004;84(16):3052e4. https://doi.org/10.1063/1.1707226.

48. Fan WH, Olaizola SM, Wells J-PR, Fox AM, Wang T, Parbrook PJ, Mowbray DJ,Skolnick MS. Erratum: “Femtosecond studies of electron capture times in InGaN/GaNmultiple quantum wells”. Appl Phys Lett 2007;91:99901. https://doi.org/10.1063/1.2778361.

49. Mahan GD. Many-particle physics. 2nd ed. New York: Plenum Press; 1990.50. Vallone M, Bertazzi F, Goano M, Ghione G. Model for carrier capture time through

phonon emission in InGaN/GaN quantum wells. Phys Status Solidi 2015;252(5):971e6.https://doi.org/10.1002/pssb.201451580.

51. Fetter AL, Walecka JD. Quantum theory of many-particle physics. 3rd ed. New York:Dover Publications; 2003.

52. Haug H, Koch SW. Quantum theory of the optical and electronic properties of semi-conductors. Singapore: World Scientific Publishing; 2004.

53. Haug H, Koch SW. Semiconductor laser theory with many-body effects. Phys Rev A 1989;39(4):1887e98.

54. Sotirelis P, von Allmen P, Hess K. Electron intersubband relaxation in doped quantumwells. Phys Rev B 1993;47(19):12744e53.

55. Sotirelis P, Hess K. Electron capture in GaAs quantum wells. Phys Rev B 1994;49(11):7543e7.


https://doi.org/10.1109/3.655016

https://doi.org/10.1109/16.870574

https://doi.org/10.1063/1.4979010

https://doi.org/10.1063/1.4979010

https://doi.org/10.1103/PhysRevB.54.17730

https://doi.org/10.1103/PhysRevB.54.17730

https://doi.org/10.1109/JSTQE.2011.2116770

https://doi.org/10.1063/1.1707226

https://doi.org/10.1063/1.1707226

https://doi.org/10.1063/1.2778361

https://doi.org/10.1063/1.2778361


56. Levetas SA, Godfrey MJ. Calculation of capture of carriers by quantum wells. Phys Rev BApril 1999;59(15):10202e7.

57. David A, Hurni CA, Young NG, Craven MD. Carrier dynamics and Coulomb-enhancedcapture in III-nitride quantum heterostructures. Appl Phys Lett 2016;109(3):33504.https://doi.org/10.1063/1.4959143.

58. Chiaria S, Furno E, Goano M, Bellotti E. Design criteria for near-ultraviolet GaN-basedlight-emitting diodes. IEEE Trans Electron Devices 2010;57(1):60e70. https://doi.org/10.1109/TED.2009.2034792.

59. Saguatti D, Bidinelli L, Verzellesi G, Meneghini M, Meneghesso G, Zanoni E,Butendeich R, Hahn B. Investigation of efficiency-droop mechanisms in multi-quantum-well InGaN/GaN blue light-emitting diodes. IEEE Trans Electron Devices 2012;59(5):1402e9. https://doi.org/10.1109/TED.2012.2186579.

60. Krost A, Dadgar A. GaN-based optoelectronics on silicon substrates.Mater Sci Eng BMay2002;93(1e3):77e84. https://doi.org/10.1016/S0921-5107(02)00043-0.

61. Sarua A, Rajasingam S, Kuball M, Younes C, Yavich B, Wang WN. High temperatureannealing of AlGaN: stress and composition changes. Phys Status Solidi 2003;(1):568e71.https://doi.org/10.1002/pssc.200390115.

62. Barton DL, Osinski M, Perlin P, Helms CJ, Berg NH. Life tests and failure mechanisms ofGaN/AlGaN/InGaN light emitting diodes. In: 1997 IEEE international reliability physicssymposium proceedings. 35th annual; 1997. p. 276e81. https://doi.org/10.1109/RELPHY.1997.584273.

63. Blake AH, Caselli D, Durot C, Mueller J, Parra E, Gilgen J, Boley A, Smith DJ, Tsong IST,Roberts JC, Piner E, Linthicum K, Cook JW, Koleske DD, Crawford MH, Fischer AJ.InGaN/GaN multiple-quantum-well light-emitting diodes grown on Si(111) substrateswith ZrB2(0001) buffer layers. J Appl Phys February 2012;111(3):33107. https://doi.org/10.1063/1.3684557.

64. Tsai CM, Sheu JK, Wang PT, Lai WC, Shei SC, Chang SJ, Kuo CH, Kuo CW, Su YK.High efficiency and improved ESD characteristics of GaN-based LEDs with naturallytextured surface grown by MOCVD. IEEE Photon Technol Lett June 2006;18(11):1213e5. https://doi.org/10.1109/LPT.2006.875063.

65. Dal Lago M, Meneghini M, De Santi C, Barbato M, Trivellin N, Meneghesso G, Zanoni E.ESD on GaN-based LEDs: an analysis based on dynamic electroluminescence measure-ments and current waveforms.Microelectron Reliab September 2014;54(9e10):2138e41.https://doi.org/10.1016/j.microrel.2014.07.122.

66. Dal Lago M, Meneghini M, Trivellin N, Mura G, Vanzi M, Meneghesso G, Zanoni E.‘Hot-plugging’ of LED modules: electrical characterization and device degradation.Microelectron Reliab September 2013;53(9e11):1524e8. https://doi.org/10.1016/j.microrel.2013.07.054.

67. Meneghini M, De Santi C, Buffolo M, Munaretto A, Meneghesso G, Zanoni E. Towardshigh reliability GaN LEDs: understanding the physical origin of gradual and catastrophicfailure. In: 2015 12th China international forum on solid state lighting (SSLCHINA); 2015.p. 63e6. https://doi.org/10.1109/SSLCHINA.2015.7360690.

68. Meneghini M, Trevisanello L, Zehnder U, Zahner T, Strauss U, Meneghesso G, Zanoni E.High-temperature degradation of GaN LEDs related to passivation. IEEE Trans ElectronDevices December 2006;53(12):2981e7. https://doi.org/10.1109/TED.2006.885544.

69. Meneghini M, Trevisanello L-R, Zehnder U, Meneghesso G, Zanoni E. Reversibledegradation of ohmic contacts on p-GaN for application in high-Brightness LEDs. IEEETrans Electron Devices 2007;54(12):3245e51. https://doi.org/10.1109/TED.2007.908900.


https://doi.org/10.1063/1.4959143

https://doi.org/10.1109/TED.2009.2034792

https://doi.org/10.1109/TED.2009.2034792

https://doi.org/10.1109/TED.2012.2186579

https://doi.org/10.1016/S0921-5107(02)00043-0


https://doi.org/10.1109/RELPHY.1997.584273


https://doi.org/10.1063/1.3684557

https://doi.org/10.1063/1.3684557

https://doi.org/10.1109/LPT.2006.875063




https://doi.org/10.1109/SSLCHINA.2015.7360690

https://doi.org/10.1109/TED.2006.885544

https://doi.org/10.1109/TED.2007.908900

70. Stafford L, Voss LF, Pearton SJ, Wang HT, Ren F. Improved long-term thermal stability ofInGaN∕GaN multiple quantum well light-emitting diodes using TiB2- and Ir-based p-Ohmic contacts. Appl Phys Lett 2007;90(24):242103. https://doi.org/10.1063/1.2748306.

71. Meneghini M, Zehnder U, Hahn B, Meneghesso G, Zanoni E. Degradation of high-Brightness green LEDs submitted to reverse electrical stress. IEEE Electron Device LettOctober 2009;30(10):1051e3. https://doi.org/10.1109/LED.2009.2029129.

72. De Santi C, Meneghini M, Buffolo M, Meneghesso G, Zanoni E. Experimentaldemonstration of time-dependent breakdown in GaN-based light emitting diodes.IEEE Electron Device Lett May 2016;37(5):611e4. https://doi.org/10.1109/LED.2016.2543805.

73. Barton DL, Zeller J, Phillips BS, Chiu P-C, Askar S, Lee D-S, Osinski M, Malloy KJ.Degradation of blue AlGaN/InGaN/GaN LEDs subjected to high current pulses. In: 33rdIEEE international reliability physics symposium; 1995. p. 191e9. https://doi.org/10.1109/RELPHY.1995.513674.

74. Osin

ski M, Zeller J, Chiu P-C, Scott Phillips B, Barton DL. AlGaN/InGaN/GaN blue lightemitting diode degradation under pulsed current stress. Appl Phys Lett August 1996;69(7):898. https://doi.org/10.1063/1.116936.

75. Kim H, Yang H, Huh C, Kim S-W, Park S-J, Hwang H. Electromigration-induced failureof GaN multi-quantum well light emitting diode. Electron Lett 2000;36(10):908. https://doi.org/10.1049/el:20000657.

76. Hsu C-Y, Lan W-H, Wu YS. Effect of thermal annealing of Ni/Au ohmic contact on theleakage current of GaN based light emitting diodes. Appl Phys Lett September 2003;83(12):2447. https://doi.org/10.1063/1.1601306.

77. Nam OH, Ha KH, Kwak JS, Lee SN, Choi KK, Chang TH, Chae SH, Lee WS, Sung YJ,Paek HS, Chae JH, Sakong T, Park Y. Recent progress of high power GaN-based violetlaser diodes. Phys Status Solidi C December 2003;0(7):2278e82. https://doi.org/10.1002/pssc.200303334.

78. Takeya M, Mizuno T, Sasaki T, Ikeda S, Fujimoto T, Ohfuji Y, Oikawa K, Yabuki Y,Uchida S, Ikeda M. Degradation in AlGaInN lasers. Phys Status Solidi C December 2003;0(7):2292e5. https://doi.org/10.1002/pssc.200303324.

79. Nam OH, Ha KH, Kwak JS, Lee SN, Choi KK, Chang TH, Chae SH, Lee WS, Sung YJ,Paek HS, Chae JH, Sakong T, Son JK, Ryu HY, Kim YH, Park Y. Characteristics of GaN-based laser diodes for post-DVD applications. Phys Status Solidi A 2004;201(12):2717e20. https://doi.org/10.1002/pssa.200405114.

80. Marona L, Riemann T, Christen J, �Swietlik T, Franssen G, Wi�sniewski P, Leszczy�nski M,Prystawko P, Grzegory I, Suski T, Porowski S, Czernecki R, Perlin P. Towards identifi-cation of degradation mechanisms in InGaN laser diodes grown on bulk GaN crystals.Phys Status Solidi A 2006;203(7):1778e82. https://doi.org/10.1002/pssa.200565316.

81. Orita K, Meneghini M, Ohno H, Trivellin N, Ikedo N, Takigawa S, Yuri M, Tanaka T,Zanoni E, Meneghesso G. Analysis of diffusion-related gradual degradation of InGaN-based laser diodes. IEEE J Quan Electron September 2012;48(9):1169e76. https://doi.org/10.1109/JQE.2012.2203795.

82. De Santi C, Meneghini M, Marioli M, Buffolo M, Trivellin N, Weig T, Holc K, K€ohler K,Wagner J, Schwarz UTT, Meneghesso G, Zanoni E. Thermally-activated degradation ofInGaN-based laser diodes: effect on threshold current and forward voltage. MicroelectronReliab 2014;54(9e10):2147e50. https://doi.org/10.1016/j.microrel.2014.07.073.

83. Marioli M, Meneghini M, Rossi F, Salviati G, De Santi C, Mura G, Meneghesso G,Zanoni E. Degradation mechanisms and lifetime of state-of-the-art green laser diodes. PhysStatus Solidi A 2015;212(5):974e9. https://doi.org/10.1002/pssa.201431714.


https://doi.org/10.1063/1.2748306

https://doi.org/10.1109/LED.2009.2029129

https://doi.org/10.1109/LED.2016.2543805

https://doi.org/10.1109/LED.2016.2543805



https://doi.org/10.1063/1.116936

https://doi.org/10.1049/el:20000657

https://doi.org/10.1049/el:20000657

https://doi.org/10.1063/1.1601306






https://doi.org/10.1109/JQE.2012.2203795

https://doi.org/10.1109/JQE.2012.2203795



84. De Santi C, Meneghini M, Gachet D, Mura G, Vanzi M, Meneghesso G, Zanoni E.Nanoscale investigation of degradation and wavelength fluctuations in InGaN-based greenlaser diodes. IEEE Trans Nanotechnol March 2016;15(2):274e80. https://doi.org/10.1109/TNANO.2016.2520833.

85. Xi Y, Schubert EF. Junction-temperature measurement in GaN ultraviolet light-emittingdiodes using diode forward voltage method. Appl Phys Lett 2004;85(12):2163e5.https://doi.org/10.1063/1.1795351.

86. De Santi C, Meneghini M, Meneghesso G, Zanoni E. Degradation of InGaN laser diodescaused by temperature- and current-driven diffusion processes. Microelectron ReliabSeptember 2016;64:623e6. https://doi.org/10.1016/j.microrel.2016.07.118.

87. Seager CH, Myers SM, Wright AF, Koleske DD, Allerman AA. Drift, diffusion, andtrapping of hydrogen in p-type GaN. J Appl Phys 2002;92(12):7246e52. https://doi.org/10.1063/1.1520719.

88. Kӧhler K, Stephan T, Perona A, Wiegert J, Maier M, Kunzer M, Wagner J. Control of theMg doping profile in III-N light-emitting diodes and its effect on the electroluminescenceefficiency. J Appl Phys 2005;97(10):104914. https://doi.org/10.1063/1.1901836.

89. Kwon MK, Park IK, Kim JY, Kim JO, Kim B, Park SJ. Gradient doping of Mg in p-typeGaN for high efficiency InGaN-GaN ultraviolet light-emitting diode. IEEE PhotonTechnol Lett 2007;19(23):1880e2. https://doi.org/10.1109/LPT.2007.908659.

90. La Grassa M, Meneghini M, De Santi C, Zanoni E, Meneghesso G. Degradation of InGaN-based LEDs related to charge diffusion and build-up. Microelectron Reliab September2016;64:614e6. https://doi.org/10.1016/j.microrel.2016.07.131.

91. van Opdorp C, t’Hooft GW. Method for determining effective nonradiative lifetime andleakage losses in double-heterostructure lasers. J Appl Phys June 1981;52(6):3827e39.https://doi.org/10.1063/1.329845.

92. Amano H, Kito M, Hiramatsu K, Akasaki I. P-type conduction in Mg-Doped GaN treatedwith low-energy electron beam irradiation (LEEBI). Jpn J Appl Phys December 1989;28(Part 2, No. 12):L2112e4. https://doi.org/10.1143/JJAP.28.L2112.

93. Li X, Coleman JJ. Time-dependent study of low energy electron beam irradiation of Mg-doped GaN grown by metalorganic chemical vapor deposition. Appl Phys Lett 1996;69(11):1605. https://doi.org/10.1063/1.117045.

94. Myers SM, Wright AF. Theoretical description of H behavior in GaN p-n junctions. J ApplPhys 2001;90(11):5612e22. https://doi.org/10.1063/1.1413950.

95. Neugebauer J, Van De Walle CG. Hydrogen in GaN: novel aspects of a common impurity.Phys Rev Lett 1995;75(24):4452e5. https://doi.org/10.1103/PhysRevLett.75.4452.

96. Miyachi M, Tanaka T, Kimura Y, Ota H. The activation of Mg in GaN by annealing withminority-carrier injection. Appl Phys Lett 1998;72(9):1101e3. https://doi.org/10.1063/1.120936.

97. Harafuji K, Kawamura K. Magnesium diffusion at dislocation in wurtzite-typeGaN crystal. Jpn J Appl Phys 2005;44(9A):6495e504. https://doi.org/10.1143/JJAP.44.6495.

98. Harafuji K, Tsuchiya T, Kawamura K. Molecular dynamics of magnesium diffusion inwurtzite-type GaN crystal. Jpn J Appl Phys 2004;43(2):522e31. https://doi.org/10.1143/JJAP.43.522.

99. Liu L, Ling M, Yang J, Xiong W, Jia W, Wang G. Efficiency degradation behaviors ofcurrent/thermal co-stressed GaN-based blue light emitting diodes with vertical-structure.J Appl Phys 2012;111:93110. https://doi.org/10.1063/1.4712030.

100. De Santi C, Meneghini M, Carraro S, Vaccari S, Trivellin N, Marconi S, Marioli M,Meneghesso G, Zanoni E. Variations in junction capacitance and doping activation


https://doi.org/10.1109/TNANO.2016.2520833

https://doi.org/10.1109/TNANO.2016.2520833

https://doi.org/10.1063/1.1795351


https://doi.org/10.1063/1.1520719

https://doi.org/10.1063/1.1520719

https://doi.org/10.1063/1.1901836

https://doi.org/10.1109/LPT.2007.908659


https://doi.org/10.1063/1.329845

https://doi.org/10.1143/JJAP.28.L2112

https://doi.org/10.1063/1.117045

https://doi.org/10.1063/1.1413950

https://doi.org/10.1103/PhysRevLett.75.4452

https://doi.org/10.1063/1.120936

https://doi.org/10.1063/1.120936

https://doi.org/10.1143/JJAP.44.6495




https://doi.org/10.1063/1.4712030

associated with electrical stress of InGaN/GaN laser diodes. Microelectron ReliabSeptember 2013;53(9e11):1534e7. https://doi.org/10.1016/j.microrel.2013.07.053.

101. Kolník J, O�guzman _IH, Brennan KF, Wang R, Ruden PP, Wang Y. Electronic transportstudies of bulk zincblende and wurtzite phases of GaN based on an ensemble Monte Carlocalculation including a full zone band structure. J Appl Phys July 1995;78(2):1033e8.https://doi.org/10.1063/1.360405.

102. Herbert DC, Uren MJ, Hughes BT, Hayes DG, Birbeck JCH, Balmer R, Martin T,Crow GC, Abram RA, Walmsley M, Davies RA, Wallis RH, Phillips WA, Jones S. MonteCarlo simulations of AlGaN/GaN heterojunction field-effect transistors (HFETs).J Phys Condens Matter April 2002;14(13):3479e97. https://doi.org/10.1088/0953-8984/14/13/307.

103. Iveland J, Martinelli L, Peretti J, Speck JS, Weisbuch C. Direct measurement of augerelectrons emitted from a semiconductor light-emitting diode under electrical injection:identification of the dominant mechanism for efficiency droop. Phys Rev Lett 2013;110:177406. https://doi.org/10.1103/PhysRevLett.110.177406.

104. Iveland J, Piccardo M, Martinelli L, Peretti J, Choi JW, Young N, Nakamura S, Speck JS,Weisbuch C. Origin of electrons emitted into vacuum from InGaN light emitting diodes.Appl Phys Lett August 2014;105(5):52103. https://doi.org/10.1063/1.4892473.

105. De Santi C, Meneghini M, La Grassa M, Galler B, Zeisel R, Goano M, Dominici S,Mandurrino M, Bertazzi F, Robidas D, Meneghesso G, Zanoni E. Role of defects in thethermal droop of InGaN-based light emitting diodes. J Appl Phys March 2016;119(9):94501. https://doi.org/10.1063/1.4942438.

106. K€ummler V, Br€uderl G, Bader S, Miller S, Weimar A, Lell A, H€arle V, Schwarz UT,Gmeinwieser N, Wegscheider W. Degradation analysis of InGaN laser diodes. Phys StatusSolidi A 2002;194(2):419e22. https://doi.org/10.1002/1521-396X(200212)194:2<419::AID-PSSA419>3.0.CO;2-B.

107. Paskov PP, Holtz PO, Monemar B, Kamiyama S, Iwaya M, Amano H, Akasaki I. Phonon-assisted photoluminescence in InGaN/GaN multiple quantum wells. Phys Status Solidi BDecember 2002;234(3):755e8. https://doi.org/10.1002/1521-3951(200212)234:3<755::AID-PSSB755>3.0.CO;2-0.

108. Estrin Y, Rich DH, Keller S, DenBaars SP. Observations of exciton-surface plasmonpolariton coupling and exciton-phonon coupling in InGaN/GaN quantum wells coveredwith Au, Ag, and Al films. J Phys Condens Matter July 2015;27(26):265802. https://doi.org/10.1088/0953-8984/27/26/265802.

109. Kalliakos S, Lefebvre P, Zhang XB, Taliercio T, Gil B, Grandjean N, Damilano B,Massies J. The effects of localization and of electric fields on LO-phonon-exciton couplingin InGaN/GaN quantum wells and quantum boxes. Phys Status Solidi A March 2002;190(1):149e54. https://doi.org/10.1002/1521-396X(200203)190:1<149::AID-PSSA149>3.0.CO;2-I.

110. Graham DM, Soltani-Vala A, Dawson P, Godfrey MJ, Smeeton TM, Barnard JS,Kappers MJ, Humphreys CJ, Thrush EJ. Optical and microstructural studies ofInGaN∕GaN single-quantum-well structures. J Appl Phys April 2005;97(10):103508.https://doi.org/10.1063/1.1897070.

111. Castiglia A, Carlin JF, Grandjean N. Role of stable and metastable Mg-H complexes in p-type GaN for cw blue laser diodes. Appl Phys Lett 2011;98(21):2011e4. https://doi.org/10.1063/1.3593964.

112. Meneghini M, De Santi C, Trivellin N, Orita K, Takigawa S, Tanaka T, Ueda D,Meneghesso G, Zanoni E. Investigation of the deep level involved in InGaN laser



https://doi.org/10.1063/1.360405

https://doi.org/10.1088/0953-8984/14/13/307

https://doi.org/10.1088/0953-8984/14/13/307

https://doi.org/10.1103/PhysRevLett.110.177406

https://doi.org/10.1063/1.4892473

https://doi.org/10.1063/1.4942438

https://doi.org/10.1002/1521-396X(200212)194:2<419::AID-PSSA419>3.0.CO;2-B




https://doi.org/10.1002/1521-3951(200212)234:3<755::AID-PSSB755>3.0.CO;2-0




https://doi.org/10.1088/0953-8984/27/26/265802

https://doi.org/10.1088/0953-8984/27/26/265802

https://doi.org/10.1002/1521-396X(200203)190:1<149::AID-PSSA149>3.0.CO;2-I




https://doi.org/10.1063/1.1897070

https://doi.org/10.1063/1.3593964

https://doi.org/10.1063/1.3593964

degradation by deep level transient spectroscopy. Appl Phys Lett 2011;99(9):2011e4.https://doi.org/10.1063/1.3626280.

113. Rossetti M, Smeeton TM, Tan WS, Kauer M, Hooper SE, Heffernan J, Xiu H,Humphreys CJ. Degradation of InGaNGaN laser diodes analyzed by micro-photoluminescence and microelectroluminescence mappings. Appl Phys Lett 2008;92(15):1e3. https://doi.org/10.1063/1.2908919.

114. Nykӓnen H, Suihkonen S, Kilanski L, SopanenM, Tuomisto F. Low energy electron beaminduced vacancy activation in GaN. Appl Phys Lett 2012;100(12):122105. https://doi.org/10.1063/1.3696047.

115. Reshchikov MA, Iqbal MZ, Huang D, He L, Morkoç H. Surface-related photo-luminescence effects in GaN.MRS Proc January 2002;743:L11.2. https://doi.org/10.1557/PROC-743-L11.2.


https://doi.org/10.1063/1.3626280

https://doi.org/10.1063/1.2908919

https://doi.org/10.1063/1.3696047

https://doi.org/10.1063/1.3696047

https://doi.org/10.1557/PROC-743-L11.2

https://doi.org/10.1557/PROC-743-L11.2


Chip packaging: encapsulationof nitride LEDs 15Xiaobing Luo, Run HuHuazhong University of Science and Technology, Hubei, China

15.1 Functions of LED chip packaging

Considered to be the next generation of light sources, light-emitting diodes (LEDs) arewidely used because of their extraordinary features compared to traditional light sour-ces, such as high luminous efficiency, high reliability, long lifetime, environmentallyfriendly, low-power consumption, etc. A bare LED chip may light up but it cannot beused due to its poor long-term reliability. LED chip packaging not only ensures goodperformance of LED devices by enhancing reliability and optical characteristics, but itis used to control and adjust the final performance. As a prerequisite of LED use, LEDchip packaging plays an important role in determining the final optical and thermalperformance of LED devices. Like electronic packaging, LED chip packaging hassix main functions, as shown in Fig. 15.1.

15.1.1 Encapsulation and protection

One of the purposes of LED chip packaging is to provide a physical housing to protectthe LED chip and bonding wires from the hostile environment. A nitride LED chip isvulnerable to its environment, especially the real working conditions include moistureand dust in the air, vibration, shock (mechanical shock and thermal shock), thermalcycling, etc. Moisture and dust in the air are the two main direct causes of LED device

Encapsulation and protection

Electrical interconnection

Light extraction

Heat dissipation

System testing

Design for X

Functions ofLED chip

packaging

Figure 15.1 Functions of LED chip packaging.


https://doi.org/10.1016/B978-0-08-101942-9.00015-0

failure.1e3 Moisture can invade the interfaces of the packaging and induce delamina-tion, which will increase contact thermal resistance and degrade reliability signifi-cantly. Dust can block light output and also increase the thermal resistance of anLED package. Other hostile mechanical or thermal conditions may cause the failureof solder joints, delamination, degradation of packaging materials, etc. With LEDchip packaging, most of the interconnections and the chip are protected, and failuretends to be more usually attributed to the packaging of the devices rather than to thedevices themselves. The encapsulant material needs to have chemical inertness, rigid-ity, high-temperature stability, high optical transparency and a high refractive index.

15.1.2 Electrical interconnection

With LED chip packaging, an LED chip is usually bonded onto a copper heat sink orlead frame structure using solder or a conductive die adhesive. Various LED chipstructures are shown in Fig. 15.2. Chip bonding is dependent on the position of theelectrodes. With wire bonding between the chip and lead frame, as shown inFig. 15.3, electrical power is input to the chip through the lead frame, which acts assecondary electrodes, providing electrical paths for power and signal distribution.

15.1.3 Light extraction

Light extraction is very important in high-brightness LED packaging. A high luminousefficiency requires high light extraction from an LED package. When blue light

P-GaN

P-GaN

N-GaN

N-GaN

N-GaN

Sapphire

Sapphire

Sapphire

MQW

MQW

P-GaN

MQW

P-electrode

P-electrode N-electrode

N-electrode(a) (b)

(c)

Figure 15.2 Three types of high-power LED chip: (a) a conventional chip, (b) a verticalinjection chip and (c) a flip chip. MQW, multi-quantum well.


emanates from the multi-quantum wells of an LED chip and reaches the interface be-tween the chip and the air, the total internal reflection (TIR) phenomenon wouldhappen. The consequence of TIR is that some of the blue light cannot escape fromthe chip structure, but instead it oscillates inside the chip structure until it is absorbedby the chip materials.

According to Snell’s law, at the interface:

nchip sin qchip ¼ nair sin qair (15.1)

where nchip and nair are the refractive indices of the LED chip and air, respectively.qchip and qair are shown in Fig. 15.4. The refractive index of GaN material nchip is

Figure 15.3 Wire bonding for electrical interconnection.

Air

Chip Chip

Silicone

Airθ

θ

air

θ

θair

silicone

chip θchip

(a) (b)

Figure 15.4 Refraction at the (a) chip and air interface and (b) chip, silicone and air interfaces.

Chip packaging: encapsulation of nitride LEDs 493

normally around 2.4 and nair usually equals 1.0. With total internal reflection, whenqair equals 90� qchip is the half angle of the escape cone and it can be calculated as:

qchip ¼ arcsin

nairnchip

sin qair

!

¼ arcsin

�12:4

�z 24:6� (15.2)

However, when an LED chip is packaged, there is a thin layer of silicone betweenthe LED chip and the air. The refractive index of the silicone layer is 1.5. Thereforewith total internal reflection, when qsilicone equals 90�, qchip is calculated as

qchip ¼ arcsin

nsiliconenchip

sin qsilicone

!

¼ arcsin

�1.52.4

�z 38:7� (15.3)

Compared with Eq. (15.2), it can be seen that qchip has increased, which means morelight can escape from the chip structure and light extraction is enhanced. In short, LEDchip packaging can enhance light extraction.

15.1.4 Heat dissipation

Although LEDs are considered as cold light sources compared with conventional lightsources, this does not mean that heat dissipation for LEDs is unnecessary. Due to thelow photoelectric conversion efficiency of current LED devices, most of the inputelectrical power is converted into heat, thus good solutions for heat dissipation inLED applications are of great importance.

Since the dimensions of a 1 W LED chip are usually very small compared to thesubstrate, the heat flux is almost 70 W/cm2, which is much higher than a conventionalmicroprocessor chip.4 The high-power density can generate significant heat inside thesmall chip, which may lead to a high junction temperature, thus increasing non-radiative recombination of holes and electrons, decreasing luminous efficiency,degrading packaging materials, etc.5e7 Therefore, the junction temperature of anLED usually should not exceed 120�C. Thermal management is a critical consider-ation for overall LED device performance, reliability and lifetime.

The thermal resistance of the package from the junction to the lead frame is the mostcommon parameter for evaluating the heat dissipation of LED packages. The essenceof heat management is to make the heat dissipation path as short as possible with asurface as large as possible. In other words, the purpose of thermal management isto reduce the thermal resistance of an LED package. Currently, a typical efficientdesign may have a thermal resistance of 8�C/W, which means that the LED junctionwill be 8�C hotter than the lead frame. To reduce thermal resistance, some LED chippackaging designs have a copper heat sink.

15.1.5 System testing

The fast development of the LED industry has resulted in higher demands on thermaldesigns, optical illumination designs, manufacturing testing, etc. A feedback mechanism


that can provide accurate and effective information based on the overall performance ofLED products is necessary in the development of holistic designs and manufacturingprocesses for LEDs. Once LED chips are packaged, the system as a whole needs tobe rigorously tested to check performance and quality. System testing LEDs cancomprehensively provide all of the performance parameters. The system test resultscan be used to determine the locations that are inconsistent with overall requirementsand to find solutions for improving the packaging processes, device design, materialselection, etc. This can further improve the overall characteristics of LEDs and decreasethe cost of manufacturing and mass production.

15.1.6 Design for X

One of the main functions of LED chip packaging is that it can help to realize a rangeof performance characteristics and functions. Design for X, where X is a variable withmany values, is a widely used method in industry, and it mainly includes:8

• design for performance (high brightness, uniform illumination, specific light intensitydistribution, etc.)

• design for reliability• design for cost• design for manufacturability• design for assembly• design for testing• design for the environment.

Consider the light intensity distribution designs shown in Fig. 15.5 as an example.The light intensity distribution of a typical LED chip is usually of the Lambertian type,but we can realize batwing and side-emitting types using different packages (morespecifically, using different lens designs) according to the different requirements ofthe LED applications.9,10 A Lambertian lens is the configuration mostly widely adop-ted in LED packaging. Lambertian radiation can be used in applications such as roadlights and MR16 lamps. Batwing lenses and side-emitting lenses are suitable for appli-cations such as backlighting and cell phones.

15.2 Basic structure of LED packaging modules

To realize the different requirements for heat dissipation and luminescence efficiencyof LEDs in different applications, a large variety of types of LED packaging structureshave been proposed. LED packages can be divided into four types: lamp LEDs,surface-mounted devices (SMDs), power LEDs and high-power LEDs.11 Fig. 15.6shows the development of packaging structures for LED modules. Although thereare very many different LED packages and technologies, basically an LED modulehas either a low-power or a high-power packaging structure.


Lambertian Batwing Side-emitting

100

75

50

25

0–90 –60 –30 0 30 60 90

Radiation angle (°)

Rel

ativ

e in

tens

ity (%

)

100

75

50

25

0–90 –60 –30 0 30 60 90


Rel

ativ

e in

tens

ity (%

)

100

75

50

25

0–120 –60 –40 0 40 80 120


Rel

ativ

e in

tens

ity (%

)

(a) (b) (c)

Figure 15.5 Light intensity distribution of three different packages: (a) Lambertian, (b) batwing and (c) side-emitting.

496Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

15.2.1 Low-power LED packaging

Fig. 15.6 is a schematic of a low-power LED package. Currently, a low-power LED isan LED packaged with epoxy resin. The chip is bonded onto the lead frame with solderor silver paste, and the top of the chip is bonded on the other part of the lead frame withbonding wires. The whole lead frame is immersed into a mold filled with epoxy resin.The size of the metal lead frame for a low-power LED is small and the thermal con-ductivity of epoxy resin is also small, thus the heat dissipation ability is limited. Asa result, the maximum driving current is limited to w20 mA and the typical forwardvoltage is w3.2 V. These characteristics limit the LED power to 0.1 W and the lightoutput rarely exceeds 2e3 lm. Consequently, low-power LEDs are mostly used asindicators.

15.2.2 High-power LED packaging

Since low-power LEDs cannot satisfy all application requirements, high-power LEDswere produced with the development of semiconductor material and packaging tech-nologies. As well as the electrical and optical paths, high-power LED packages shouldalso have a thermal dissipation path. There are many kinds of high-power LED pack-aging structures, including lead frame LEDs, silicon packaging LEDs, printed circuitboard (PCB) packaging LEDs and ceramic packaging LEDs. These structures arealmost the same and the only difference lies in the substrate materials. Therefore,we will use the lead frame LED as an example for describing the basic structure ofLED packaging modules. Luxeon, which was proposed by Lumileds Co in 1998

Chip

Solder/paste/conductive epoxy

Cathode lead

Epoxy

Bond wire

Anode lead

Figure 15.6 Low-power LED package.


and is widely used, is a typical lead frame LED (Fig. 15.7). In the Luxeon structure, thechip is bonded onto the heat sink with solder or silver paste. The material for the heatsink is usually copper, which has high thermal conductivity. The electrodes of the chipare connected to the lead frame by the bonding wires. Silicone gel phosphor of a spe-cific concentration is dispersed onto the chip surface to form a phosphor layer. The lensis embedded into the molding compound using a mechanical structure. Silicone gelfills the interspace between the lens and the molding compound to protect the chipand bonding wires. High-power LED packaging has better heat dissipation, thus thepower can be more than 10 times that of low-power LED packaging. Normally, ahigh-power LED is driven at a much higher current, typically 350, 700 or 1000 mA.The power can be increased to 1e5 W and commercially available 1 W LED packagescan produce up to 231 lm of light.12

15.3 Processes used in LED packaging

Production of an LED starts from an LED chip and progresses to an LED package, anLED module and then to a system. It is obvious that LED packaging is necessarybefore the applications of LEDs. The LED packaging processes affect the final opticalperformance and reliability of LED devices, thus they are critical. The typical pack-aging processes for a high-power LED are shown in Fig. 15.8 and are described brieflybelow.

15.3.1 Chip bonding

The chip is bonded onto the substrate with solder or silver paste. The solder or silverpaste is pre-coated onto the substrate and then the chip is mounted onto the solder orpasted with slight pressure. The solder is melted and cooled by reflow soldering, or thesilver paste is cured at high temperature.

LED chip

Phosphor layer

Bonding wire

Lead frame

Lens

Silicone

Copper slug

Molding compound

Figure 15.7 Typical high-power LED package.


15.3.2 Wire bonding

The electrodes of the chip are connected to the lead frame by the wire bonding. Thewire bonding process should be very careful, otherwise bonding failure may occurdue to bond pad cratering, peeling or cracking below the bond pad.13 The wire bondingprocesses depend on how the electrodes are situated on the chips. The materials of thebonding wires for LEDs are usually gold or copper.

15.3.2.1 Gold wire

Gold wire bonding is widely used for LEDs since gold wire can be bonded easily byheat, pressure and ultrasonic energy, which is referred to as thermosonic bonding. Inaddition, the junction size, bond strength and electrical and thermal conductivity ofgold wire are also suitable for LED applications.

15.3.2.2 Copper wire

Compared with gold wire, copper wire can achieve greater mechanical stability. Stan-dard bond strength tests, such as the wire pull test and the ball shear test, have demon-strated that copper wire bonds are 25%e30% stronger than comparable gold wirebonds. However, copper wires have significant disadvantages compared with goldwires, and the copper wire bonding process is not yet well understood and it is imma-ture. Copper can oxidize at a relatively low temperature and the bonding parametersare harsher (a higher bond force and more ultrasonic energy are required). Therefore,copper wire bonding is still being studied and optimized.14

Substrate

LED chip

Solder

Reflector

Gold wire

Lens

Silicone

Phosphor

(a)

(c)

(e)

(b)

(d)

Figure 15.8 LED packaging processes: (a) chip bonding, (b) wire bonding, (c) phosphorcoating, (d) lens laying, (e) silicone injection and curing.


15.3.3 Phosphor coating

A phosphor layer is dispersed onto the chip surface manually or with a dispenser. Thephosphor layer will absorb some of the blue light emanating from a blue LED chip andre-emit yellow light, so that white light is produced as a mix of the blue light andyellow light. The white-light quality is sensitive to the thickness, concentrationand location of the phosphor layer. As shown in Fig. 15.9, there are three phosphorcoating methods.

15.3.3.1 Freely dispersed coating

The freely dispersed coating method is the simplest one that is massively applied in thetraditional LED packaging industry. A phosphor embedded silicone matrix isdispensed onto a chip without a mold to restrict the flow until a surface force balanceis achieved. The phosphor layer is normally convex and the thickness of the centralzone on a chip is larger than that of the edge zone. This variation in thickness maycause spatially non-uniform distribution of the transmitted blue and yellow light,resulting in the final white light of poor angular color uniformity (ACU). Colored ringsmay appear in the edge of the final white light pattern. In addition, due to the variationof the mixture volume in the dispensing process, the average correlated color temper-ature (CCT) may vary from package to package. This results in a low yield and anincreased cost.

Everlight Seoulsemicon

XLampLuxeon

LedEngin GELcore

(a)

(b)

(c)

Figure 15.9 Phosphor coating technologies: (a) freely dispersed coating and representativeproducts (Everlight and Seoul Semicon), (b) conformal coating and representative products(Luxeon from Lumileds and XLamp from Cree), (c) remote coating and representativeproducts (LedEngin and GELcore).


15.3.3.2 Conformal coating

A conformal coating is produced by a progressive packaging process, which canrealize an extraordinarily thin phosphor layer. To control the final color output, theconcentration of the phosphor layer is usually high due to its thinness. This methodovercomes the problem with the non-uniform thickness of the phosphor layerproduced by the freely dispersed method, and the angular color uniformity of the whitelight can be greatly enhanced. The conformal coating method was first developed byLumileds, which used an electrophoretic method15,16 to deposit charged phosphorparticles onto the chip surface. The thickness of the phosphor film can be adjustedby controlling the voltage and deposition time. Therefore, a conformal coating caneasily have micron precision. Other approaches, such as slurry, settling,17 evaporatingsolvent,18 wafer-level coating,19 capillary-assisted coating20 and direct white light,21

can also be used to conformally coat a phosphor. As shown in Fig. 15.9, representativeproducts with a conformal coating of phosphor are Luxeon from Lumileds and XLampfrom Cree. However, it has been confirmed experimentally that approximately50%e60% of the light is back-scattered by the phosphor layer.22,23 The back-scattered light rays are re-absorbed by the chip and some of the energy is lost dueto absorption by the packaging materials. In addition, localized heating caused by ahigh-power chip can induce thermal quenching and reduce the quantum efficiencyof the phosphor.24

15.3.3.3 Remote coating

Remote phosphor coating, in which the phosphor layer is some distance above theLED chip, was proposed for reducing the amount of light trapped in LED chips andimproving the luminous efficiency of white LEDs.25e27 This LED structure can lowerthe temperature of the phosphor layer and enhance the stability of the color, becauseless heat is transferred from the LED chip to the phosphor layer.28 As shown inFig. 15.9, representative products with a remote coating of phosphor include LedEnginand GELcore. The main disadvantage of remote coating is that the shape of the thinphosphor layer is not easy to control and realize. Due to surface tension of the liquid,the pre-cured encapsulant materials and phosphor layer normally have concave sur-faces. The curvature of these surfaces is dependent on the dimensions and surfaceroughness of the reflector, the viscosity of the silicone phosphor, operating tempera-ture, etc. Since the remote phosphor layer is away from the LED chip, its area is usu-ally much larger than the area of the LED chip. The large phosphor layer will disorderthe propagating light and disrupt the pattern of light re-emission significantly since theoptics is usually designed for chips or small light sources.

15.3.4 Lens laying

The lens is embedded into the substrate using a mechanical structure. The lenses, as thedominant optics, are the key components for realizing the different optical require-ments of applications. A conventional lens is a hemisphere, but its optical performance


is poor. Freeform lenses have better optical performance and they can be designed tosuit various illumination requirements. A detailed introduction to the optical design offreeform lenses will be given in Section 15.6.

15.3.5 Silicone injection and curing

Silicone gel is injected into the interspace between the lens and molding compound toprotect the chip and bonding wires. The whole module is put into an oven to cure thesilicone at a relatively high temperature for a short time. After curing, the silicone willadhere to the lens and the molding compound and keep moisture and dust away.

15.4 Optical effects of gold wire bonding

As an optoelectronic device, an LED can be optically designed for many differentapplications, such as backlighting, displays, projectors, automotive lighting, roadlighting, landscape lighting and indoor lighting. The quality of the light from anLED is important for applications, and includes the light pattern, light uniformityand light color. Since the packaging processes will affect the light output of anLED, their optical design must be optimized.

Due to the wetting effect, the phosphor gel spreads along the gold wire surfaces. Asshown in Fig. 15.10, the gold wires influence the geometry of the phosphor layer.Fig. 15.11 shows how the profiles of phosphor layers are influenced by the gold wires

Left view

Right view

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5

Sample 5Sample 4Sample 3Sample 2Sample 1

Figure 15.10 Left and front views of phosphor layers in LED module samples.

Profile of phosphor layer without gold wiresProfile of phosphor layer with gold wires

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5

Figure 15.11 Shape and height of the gold wires of the five samples and their phosphor profiles.


based on experimental results. Compared with the phosphor layer without gold wires,it is clear that the shapes of the phosphor layers for Sample 2, Sample 3 and Sample 5were significantly influenced by the gold wires, in particular, near the gold wires. InSample 3, the gold wires are lower than in Sample 1, and the external parts of the wiresare closer to the surface of the phosphor, hence more phosphor gel spreads along thegold wire surfaces until the surface energy for the air, phosphor gel and gold wirebecomes balanced. Obviously, this reduces the height of the phosphor layer andchange its curvature. For Sample 1 and Sample 4, the effect of the gold wires onthe phosphor layer geometry is tiny, and the left and front views of the phosphor layersare nearly the same.

As shown in Fig. 15.12, to quantify the optical effect of the gold wire, the angularCCT of the five samples at points (r, q, f) on plane Cq � (q þ 180�) of the top hemi-sphere is recorded. Points are defined by (r, q, f) in the coordinate system, where theorigin is fixed at the center of the top surface of the copper heat sink. r is the radialdistance. Inclination angle 4 is measured from the fixed zenith direction Z. Azimuthangle q is measured from the azimuth axis A. Plane Cq � (q þ 180�) is defined asthe plane passing through the origin and zenith direction Z, and where the azimuthangle of points on that plane is q.

Fig. 15.13 shows the results of measuring the spatial angular CCT distributions ofdifferent samples, using a CCT recorder. For all samples, the angular CCT fluctuationfor plane C0e180� is smaller than for plane C90e270� in the package. The reasonmay be that the geometries of the phosphor layers for plane C90e270� in all samplesare affected by the gold wires more than for plane C0e180�. Taking Sample 5 forinstance, the angular CCT decreases dramatically from 7512K (at f ¼ �90�) to5050K (at f ¼ �40�) for plane C90e270�, while it decreases from 5719K (atf ¼ �90�) to 5125K (at f ¼ �40�) for plane C0e180�. For Sample 1 and Sample 4,the angular CCT distributions for plane C0e180� and plane C90e270� are similar,whereas, for Sample 2, Sample 3 and Sample 5, the angular CCT deviates sharplyas the inclination angle f changes from 0 to �90� for the two planes. This is becausethe shape of the phosphor gel for the two planes varies less for Sample 1 and Sample 4,as we mentioned for Fig. 15.11.

CCT recorder

Z

A

Plane C 0° – 180°

Plane Cθ – (θ

θ

ϕ

+ 180°)

(r,θ,ϕ)

Figure 15.12 Angular CCT measurement.


10 000

9000

8000

7000

6000

5000

Spatial plane C0°–180°


Sample 1

Sample 2

Sample 3

Ang

ular

CC

T (K

)

–100 –80 –60 –40 –20 0 20 40 60 80 100

Inclination angle ϕ (degree)

10 000

9000

8000

7000

6000

5000



Sample 4

Sample 5

Ang

ular

CC

T (K

)

–100 –80 –60 –40 –20 0 20 40 60 80 100

Inclination angle ϕ (degree)

(a)

(b)

Figure 15.13 Angular CCT distributions for plane C0e180� and plane C90e270�: (a) Sample1, Sample 2 and Sample 3; (b) Sample 4 and Sample 5.


These experimental results for the angular CCT distributions correspondwell with thegeometries of the phosphor layers. CCTdistributions of LEDmodules, encapsulatedwithsilicone gel and with a lens, also show a similar trend. The results show that the gold wirebonding process has an important effect on the angular CCT distribution of an LED pack-age. Thus, when manufacturing LED products, the shape of the gold wire should be welldesigned and optimized to improve the optical performance of the products.29

15.5 Optical effects of phosphor coating

As a bulk scattering material, the phosphor absorbs short-wavelength emissions fromthe primary LED chip and down-converts them to long-wavelength emissions. Forinstance, the most widely used white LEDs have a blue GaN LED pumping aYAG:Ce3þ yellow phosphor. Mixing the yellow phosphorescence and the escapedblue emission produces white light. The phosphor layer plays an important role indetermining the final optical performance of LED devices. Many studies have inves-tigated the effect of phosphor properties on packaging performance.25,30e37 The loca-tion, thickness, concentration and geometry of the phosphor layer are the mostimportant factors affecting an LED’s optical performance.

15.5.1 Phosphor location

The location of the phosphor is the primary consideration for LED packaging. Asmentioned above, there are three main ways to realize the phosphor layer, i.e., a freelydispersed coating, conformal coating and remote coating. It is obvious that the phos-phor locations realized by these three methods are different. By changing the phosphorlayer from being close to the chip to away from the chip, the propagation path and lightenergy will be affected by scattering and absorption by the phosphor, reflection fromthe substrates, absorption in the chip, refraction by the lens, etc. Absorption in thephosphor and chip will influence the output optical power. Scattering by the phosphorwill disorder the propagating light. The rays could converge to central zones throughreflection from the reflector and refraction by the lens or diverge to edge zones. Thesewill induce variations in light extraction and CCT.35

Kim et al.25 compared the light extraction efficiency of four different placementsand arrangements for the phosphor as shown in Fig. 15.14(a). They used a new pack-aging method with a diffuse reflector cup and the remote phosphor configuration toenhance light extraction, as shown in Fig. 15.14(a4). The light extraction efficiencyof the four configurations was compared. The surface of the reflector is specular inFig. 15.14(a3), and diffuse in Fig. 15.14(a4). They conducted ray-tracing simulationsto confirm the benefits of the proposed methods and the results are shown inFig. 15.14(b). The light extraction efficiency was improved by using the remote phos-phor arrangement by 36% for a specular reflector cup, and 75% for a diffuse reflectorcup, compared with the phosphor-in-cup arrangement for a specular reflector cup. Forthe phosphor-in-cup arrangement, the use of a diffuse reflector cup results in a 47%improvement over the specular reflector cup.


(1) Reflectorcup

Encapsulant

Encapsulant

Phosphor

Phosphor

(2)LED chip

1 2 3

Specularreflector cup

Diffusereflector cup

Diffusereflectionpattern

(3)

(4)

1 2 3

80

70

60

50 Specular cup +phosphor in cup

Diffuse cup + remote phosphor

Diffuse cup + phosphor in cup

Specular cup + remote phosphor

0 0.5 1 1.5 2 2.5Height of reflector cup h (mm)

Ext

ract

ion

effic

ienc

y (%

)

LED

ab b

hθ

(a)

(b)

Figure 15.14 (a) Arrangement of phosphor in white LEDs: (1) conformal distribution directlyon LED chip; (2) uniform distribution in reflector cup (phosphor-in-cup); (3) uniform distri-bution in thin layer above LED chip (remote phosphor); (4) remote phosphor distribution indiffuse reflector cup. (b) Calculated light extraction efficiency as a function of the height of thereflector cup, obtained from ray-tracing simulations.Reproduced with permission from Kim JK, Luo H, Schubert EF, Cho J, Sone C, et al. Stronglyenhanced phosphor efficiency in GaInN white light-emitting diodes using remote phosphorconfiguration and diffuse reflector cup. Jpn J Appl Phys 2005;44:L649e51. Copyright 2005,Japan Society of Applied Physics.


15.5.2 Phosphor thickness and concentration

The thickness and concentration of the phosphor are secondary considerations forwhite LED packaging. This is because the luminous flux and the color of an LEDare adjusted mainly through changing the phosphor thickness and concentration afterthe phosphor converters have been chosen. The phosphor thickness and concentrationcan be varied during manufacturing and they affect the optical consistency of whiteLEDs.38e44 The optical consistency depends on the ability to control fluctuations inoptical performance so that the luminous efficiency, CCT and color rendering index(CRI) remain within the desired range. The optical consistency is an important param-eter for the quality of the light from an LED.

Sommer et al.41 found that changing the phosphor thickness and concentration canaffect the spatial color distribution. Tran et al.42 experimentally studied the effects ofphosphor thickness and concentration on LED luminous flux and correlated color tem-perature. Their results showed that a package with a lower phosphor concentration anda higher phosphor thickness had a lower trapping efficiency and caused less backscat-tering of the light, and thus had higher luminous efficacy. When the CCT value wasaround 4000K, the experimental results showed that the lumen output for a 1.8 mmthick phosphor package was 23% higher than for a 0.8 mm thick phosphor package.Hence, they also found that the brightness or luminous efficiency of white LEDs high-ly depends on the phosphor thickness and concentration.

Our group systematically analyzed the effects of YAG:Ce phosphor thickness andconcentration on the optical performance of phosphor-converted white LEDs, consid-ering light extraction, luminous efficiency and CCT.43e45 Five LED packagingmethods with different phosphor locations are presented for comparison, as shownin Fig. 15.15. In Methods I, II and III we conformally coated the phosphor to replicatethe shape of the chip. The difference is that there was a small gap between the phos-phor and the chip in Methods II and III whereas the phosphor was directly dispensedonto the chip surface in Method I. In Method IV, the phosphor had a planar shape butthe location was farther than that for Methods II and III. In all methods, the surfaces ofthe board and reflector were coated with silver to provide high reflection. For eachmethod, the phosphor thickness and concentration were varied and the five casesare listed in Table 15.1.

The results of the Monte Carlo ray-tracing simulations are shown in Figs. 15.16 and15.17. To explain the results, we have to define the luminous efficiency (h) and colormixing fraction (f) as:

h ¼ 683 lm=W� Ppc�BVpc�B þ Ppc�YVpc�Y

Pelec(15.4)

f ¼ Ppc�Y=ðPpc�B þ Ppc�YÞ (15.5)

where Pelec is the consumed electrical power, and Vpc�B and Vpc�Y are eye sensitivitycoefficients (for blue light spectra and converted-yellow light spectra, respectively.Ppc�B and Ppc�Y are the extracted optical power of the blue light and converted-yellowlight, respectively.


From Figs. 15.16 and 15.17, we can see that the brightness and the light color of thefive packaging methods change as the phosphor thickness and concentration varies.Generally, for each case, that is for fixed phosphor thickness and concentration, theluminous efficiency and the color mixing fraction of Method IV increase more slowlythan for the other methods. The luminous efficiencies of Method I and Method Vincrease more rapidly than the others, and the luminous efficiency of Method V ishigher than that of Method I. The color mixing fractions of Method I and V are alwaysclose, indicating similar variations of their light color. The luminous efficiencies andcolor mixing fractions of Methods II and III vary similarly and their variations aremoderate among the five packaging methods.

15.5.3 Phosphor geometry

The above discussions cover the phosphor location, thickness and concentration, butfor specific cases, which are not sufficient to give an understanding of the influenceof phosphor geometry on the luminous flux. Yu et al.46 investigated the effect ofchanging phosphor geometry using Monte Carlo ray-tracing simulations. Fig. 15.18shows the five phosphor geometries: dispersed-coating geometry (the conventional

Chip

Phosphor

Silicone

Board

R RL

L L

L

R R

RH

D

I II

III IV V

Figure 15.15 Five packaging methods used in the analysis. L gives the location of the phosphor.In Methods II, III and IV, L is the gap between the phosphor and the chip. In Method V, L is theradius of the phosphor. The radius (R) of the lens was 4 mm. The baseline diameter (D) of thereflector was 3 mm and the height (H) was 2 mm.

Table 15.1 Phosphor thickness and concentration for the five cases

Case 1 2 3 4 5

Phosphor thickness (mm) 60 100 140 40e200

Phosphor concentration (g/cm3) 0.4e1.4 0.6 1


100

90

80

70

60

50

40

9080

70

60

50

40

30

20

100

Lum

inou

s ef

ficie

ncy

(Im/W

)C

olor

mix

ing

fract

ion

(%)

0.4 0.6 0.8 1.0 1.2 1.4 0.4 0.6 0.8 1.0 1.2 1.4 0.4 0.6 0.8 1.0 1.2 1.4Phosphor concentration (g/cm2) Phosphor concentration (g/cm2) Phosphor concentration (g/cm2)

Method I Method II Method III Method IV Method V

(a)

(d)

(b)

(e)

(c)

(f)

Figure 15.16 Dependencies of luminous efficiency and color mixing fraction of the five packaging methods on phosphor concentration for (a and d)Case 1, (b and e) Case 2, and (c and f) Case 3.

Chip

packaging:encapsulation

ofnitride

LEDs

509

100

90

80

70

60

50

90

75

60

45

30

Lum

inou

s ef

ficie

ncy

(Im/W

)

Method IMethod IV

Method II Method IIIMethod V

60 90 120 150 180 60 90 120 150 180Phosphor thickness (μm) Phosphor thickness (μm)

Col

or m

ixin

g fra

ctio

n (%

)(a) (b)

(c) (d)

Figure 15.17 Dependencies of luminous efficiency and color mixing fractions of the fivepackaging methods on the phosphor thickness for (a and c) Case 4, and (b and d) Case 5.

Encapsulant

Phosphor

Chip

Board

(a) (b) (c)

(d) (e)

Figure 15.18 Cross sections of phosphor-converted LEDs with (a) dispersed-coating phosphorgeometry, (b) two-flat phosphor geometry, (c) top-convex phosphor geometry, (d) bottom-convex phosphor geometry and (e) two-convex phosphor geometry.


coating method), two-flat geometry (planar layer), top-convex geometry (top interfacehas a convex shape), bottom-convex geometry (bottom interface has a convex shape)and two-convex geometry (both top and bottom interfaces have a convex shape). In thesimulations, they used the mean free path (MFP), which is the average distance thatphotons travel between collisions with phosphor particles, for the analysis instead ofthe phosphor concentration because the MFP is inversely proportional to the phosphorconcentration.47 They normalized the luminous fluxes of the phosphor-convertedLEDs with different phosphor geometries with respect to the dispersed-coating case.They confirmed that the remote phosphor configuration has a higher luminous fluxthan conventional dispersed-coating LEDs. The remote-phosphor LED having a phos-phor layer with a hemispherical top surface improved by more than 12% comparedwith the conventional dispersed-coating case, and a 5% improvement comparedwith the two-flat cases. Won et al.48 fabricated high-power white LEDs by combiningblue LEDs and green (Ba, Sr)2SiO4:Eu

2þ and red CaAlSiN3:Eu2þ phosphors with

various phosphor geometries. The results showed that the luminous efficiencyimproved because of a decrease in the reabsorption of green light by the red phosphorowing to a difference of refractive indices. The white LED had a very high luminousefficiency of 51 lm/W and a high color rendering index of 95 under 350 mA.

15.6 Optical effects of freeform lenses

The direct output of an LED chip is usually a circular spot with non-uniform illumi-nation, which struggles to meet the illumination requirements. Appropriate opticsincluding primary lenses or secondary lenses are essential for obtaining high-qualityLED illumination.49e53 Many solid-state lighting (SSL) applications use a secondaryoptic to couple light from the LED into a desired beam shape. There has been a trendfor using freeform lenses in LED optical design because of advantages such as highdesign freedom, small size and precise light irradiation control.50 Many differentmethods have been proposed for freeform lens design, such as the tailored freeformlens surface design method, the simultaneous multiple-surface lens design method,the discontinuous freeform lens design method and the continuous freeform lensdesign method.

15.6.1 Tailored freeform surface design method

In 2002, Harald Ries and Julius Muschaweck54 proposed the tailored freeform surfacedesign method, where a freeform surface is constructed through numerical solution ofpartial differential equations. The method uses wavefronts, which are surfaces of con-stant phase in the electromagnetic field. The local normal to the wavefront is the prop-agation direction of the field, which is the connection between wave optics and ray orgeometrical optics.55 For a given wavefront, a set of rays can be defined using the localsurface normals. In a similar way, a set of rays can be used to construct the equivalentwavefronts. This method can be used to obtain an optical surface by solving a series of


non-linear partial differential equations. It can ensure the local smoothness of thesurface by adopting the continuous Gaussian curvature of surface, and it can obtainan ideal distribution of illumination in an area with small angles.

However, it still has limitations:

• The calculations are complex and difficult for light sources with arbitrary light intensitydistribution.

• It is only suitable for light sources with a small volume and does not consider rays with largeemerging angles.

• It has low utilization efficiency.

Therefore, the theory and design need further development. Fig. 15.19 shows a free-form lens designed by the tailored method and its performance. A detailed descriptionof this method can be found in Ries and Muschaweck.54

15.6.2 Simultaneous multiple surfaces (SMS) method

This method can be used for the simultaneous design of multiple optical surfaces. Theoriginal idea came from Minano and it was further developed by Benitez, therefore theSMS method is also called the MinanoeBenitez design method.56e58 The SMSmethod is a procedure for designing two optical surfaces such that two given normalcongruencies Wi1 and Wi2 are transformed (by a combination of refractions and/orreflections at these surfaces) into another two given normal congruencies Wo1 andWo2, as shown in Fig. 15.20. The SMS method generates an optical system withtwo freeform surfaces that deflect the rays of the input bundles into the rays of the

10 5 0Horizontal position (relative units)

Verti

cal p

ositi

on (r

elat

ive

units

)

–5 –10

–10

–5

0

5

10(a) (b)

Figure 15.19 (a) Freeform lens designed based on the tailored method and (b) its numericalillumination performance.Reproduced with permission from Ries H, Muschaweck J. Tailored freeform optical surfaces.J Opt Soc America A 2002;19(3):590e95. Copyright 2002, Optical Society of America.


corresponding output bundles and vice versa. At present, only the SMS method candeal with an extended light source effectively. The SMS method can be used toproduce LED lighting using the effective design method of freeform lenses. Plate13 (see color plate section) show the multichannel optics for an LED designed usingthe SMS method.56

The advantages of the SMS method are:

• It can be used to design more than one optical surface simultaneously.• It is suitable for designing an optical system with an extended source.• It is applicable for rays with large emerging angles.• It makes greater utilization of the light source.

However, in this method the given illumination distributions need to be convertedinto wavefronts. It also needs to solve second-order non-linear MongeeAmp�ere equa-tions through a complicated and lengthy calculation.

15.6.3 Discontinuous freeform lens method

Both the SMS method and the tailored method mentioned above are based on thecoupling of input wavefronts and output wavefronts. A practical design method isto establish light energy mapping relations between the light source and the target.The strategy for designing a freeform lens based on energy mapping is as follows.Firstly, it is assumed that all of the light emitted from the light source radiates ontothe target plane, which means the energy of the light source is equal to that of the targetplane. Both the light source and target plane are divided into many cells, where thelight energy is constant over a cell. According to the edge ray principle, the

Two optical surfaces

Optical path length I2

Optical path length I1

Wo1Wo2

Wi1Wi2

Figure 15.20 The simplest version of the SMS 3Dmethod generates two surfaces that transformtwo input congruencies onto two output ones.Reproduced with permission from Minano JC, Benitez P, Liu JY, Infante J, Chaves J, et al.Applications of the SMS method to design of compact optics. Proc SPIE 2010;7717:771701.Copyright 2010, SPIE.


corresponding relation between each energy cell of the light source and the target planeis established by solving the energy conservation-based differential/integral equationor through direct correspondence. Finally, according to Snell’s law and methods forconstructing curved surfaces, points on the freeform surface can be calculated andan integrated optical surface can be constructed. This surface has to be validatedand modified through numerical simulation.

The introduction of discontinuities into a lens surface can effectively reduce thenormal deviation, but there is no guarantee that all light rays will be transmitted to cor-responding positions on the irradiance plane. The cliffs between the discontinuous sur-faces will distort a small portion of the light.59 The total energy transmission ratio fromthe point light source to the target plane is as high as 95%. Fig. 15.21 shows anexample of a discontinuous freeform lens and the simulation results of its irradianceon the target plane.60

A discontinuous freeform lens was designed, fabricated and tested by us.61 Wefound that during mass production of the lens (i.e., injection molding), manymanufacturing factors, such as the surface morphology of the mold, the injection mold-ing temperature and pressure, and the viscosity of the liquid, affect the surfacemorphology of the discontinuous freeform lens and thereby affect its optical perfor-mance. As shown in Fig. 15.22, the surface roughness and smooth transitions betweenthe discrete sub-surfaces are two of the most common manufacturing defects found indiscontinuous freeform lenses. The light pattern for a lens made from BK7 opticalglass at 70 cm away from the LED is shown in Fig. 15.23. It is obvious that the illu-mination is poor: the distribution of light energy on the target plane is non-uniform andthe shape of the light pattern is not rectangular. There are obvious dark and brightstripes on the light pattern, especially in the center and along the diagonals. Therefore,when designing a discontinuous freeform lens, we have to pay attention to themanufacturing defects and add feedback to the initial designs.

15.6.4 Continuous freeform lens method

It is difficult to fabricate discontinuous freeform lenses. A continuous freeform lenshas high illumination quality and is easy to fabricate. A continuous freeform lens isdesigned by a similar method as a discontinuous freeform lens. The only differencelies in the error control in the normal directions of the refracted rays through thelens surface. The details of the design method are presented in Refs. 62e64. Usingthe method, we designed freeform lenses for road lighting, automotive headlamps,MR16 lamps, etc. Fig. 15.24 shows a continuous freeform lens for LED road lightingand its illumination performance. From Fig. 15.24, we can see that surface of the lensis quite smooth and the light pattern is better than that shown in Fig. 15.23. The com-parison demonstrates that the continuous freeform lens method is an effective way todesign a freeform lens for LED lighting. Plate 14 (see color plate section) shows a free-form lens for an MR16 lamp and its lighting performance. The illumination uniformityis good but there is obvious light deterioration at the edge of the light pattern, whichneeds further optimization.


15.7 Thermal design and processing of LED packaging

As mentioned above, temperature plays a crucial role in the reliability of LED pack-ages; therefore it is necessary to consider thermal aspects during design. The primarygoal of thermal design is to maintain the junction temperature of the LED device belowthe critical temperature. Exceeding the critical temperature can lead to accelerated lightoutput degradation and even to catastrophic failure. The critical temperature is

1

2

–250 –200 –150 –100 –50 0 50 100 150 200 250

–250 –200 –150 –100 –50 0 50 100 150 200 250

250

200

150

100

50

0

–50

–100

–150

–200

–250

250

200

150

100

50

0

–50

–100

–150

–200

–250

3 4

(a)

(b)

1 2

Figure 15.21 (a) Discontinuous freeform lens used to form an “E” light pattern and (b) asimulation of its illumination.Reproduced with permission from Wang L, Qian KY, Luo Y. Discontinuous free-form lensdesign for prescribed irradiance. Appl Opt 2007;46(18):3716e23. Copyright 2007, OpticalSociety of America.


essential for basic semiconducting properties, including non-radiative recombinationvia deep levels, surface combination and carrier loss over heterostructure barriers.Light output decreases with an increase of junction temperature.65 High temperaturesreduce the internal quantum efficiency of an LED chip and the quantum efficiency ofthe phosphor. The lifetime of LEDs decreases with an increase of junction

(a) (b)

(c) (d)

Figure 15.22 Micrographs of different parts of a PMMA discontinuous freeform lens.

Dark stripes

Figure 15.23 Light pattern for a BK7 optical glass lens at 70 cm away from LED.


temperature.66 The wavelength of the emitted light shifts with an increase of junctiontemperature,67 which leads to variations in the correlated color temperature and colorrendering index. Another goal of thermal design is to minimize the temperaturethroughout the assembly during operation. Because of the mismatch of the coefficientsof thermal expansion (CTE), an elevated temperature induces thermal stresses in thepackaging components, leading to cracks, delamination and other failures. The reli-ability of LED applications decreases with an increase of junction temperature. An in-crease of phosphor temperature may lead to degradation of phosphor efficiency, whichis known as phosphor thermal quenching.68 Phosphor thermal quenching will decreaselight output through an increase of the non-radiative transition probability. Therefore,it is of great importance to conduct thermal design for LED packaging.

15.7.1 Thermal design of packaging

Because of the small area and relatively low temperature of LEDs, only a small amountof heat is dissipated by radiation. Conduction and convection are the dominating heattransfer modes in LED packaging products. Since LED chips are encapsulated, thereare two paths for heat dissipation in LED packages. The first is the upper path throughthe encapsulant to the ambient air, and the other is the lower path through the substrateto the ambient air. Since encapsulants are usually polymers with low thermal conduc-tivities, most of the heat is conducted through the substrate and then dissipated to theambient air. In this case, the lower path is the main heat dissipation path. Fig. 15.25(a)shows the thermal resistance network of the lower path for a complete Luxeon LEDmodule. From this thermal resistance network, it can be seen that there are many in-terfaces in LED packages and the total thermal resistance is the sum of the series ther-mal resistances, which is called the system thermal resistance. A thermal resistancenetwork is a good way to evaluate heat dissipation. A low system thermal resistanceimplies that heat can be conducted to the environment efficiently and therefore thejunction temperature will be low at the ambient temperature.

Instead of using separate heat sink/lead frame assembly packaging as shown inFig. 15.25(a), another approach for high-power LED solutions is the chip-on-board(CoB) technology, in which the chip is directly mounted onto the board with an

Figure 15.24 (a) Continuous freeform lens for LED road lighting and (b) its illuminationperformance in the laboratory.


appropriately designed circuit. Fig. 15.25(b) shows the thermal resistance network of aCoB package. From Fig. 15.25(b), it can be seen that the one of the thermal interfacesbetween the chip and the heat sink has been eliminated in CoB packaging and the sizeis more compact, therefore the packaging density when using CoB can be significantlyhigher. The direct contact between the LED and the board allows for optimal thermalmanagement with a high packaging density and results in long-lasting and high-performance LEDs. CoB technology can be used to produce LED arrays for differentapplications. The advantages of CoB LEDs include:

• compactness• high intensity, particularly at close distances• high uniformity even at a close working distance• better thermal performance (long lifetime, high stability, etc.)• clustering of LEDs on a circuit board• reduction of manufacturing cost for the same power

Following the analysis of the heat dissipation paths of LED packages, we willdiscuss thermal problems inside LED packages from top to bottom as shown in

Rchip

Rchip

Rinterface1

Rinterface1

Rinterface2

Rinterface2

Rinterface3

Rheatsink

Rheatsink

Rsubstrate

Rsubstrate

Rslug

(a)

(b)

Figure 15.25 Thermal resistance network of: (a) a Luxeon LED; (b) a chip-on-board (CoB)package.


Fig. 15.25. The LED chip is not the only heat source e there is also self-heating of thephosphor layer. Hence, we will discuss the self-heating of phosphor layer first, andthen contact thermal resistance, spreading thermal resistance and cooling solutionsin turn.

15.7.2 Self-heating of the phosphor coating

Since not all of the absorbed blue light is converted to yellow light by the phosphorlayer, some of the absorbed light must turn into heat. Therefore, there is local heatingof the phosphor particles, which also contributes to heat generation inside an LEDpackage. Arik et al.69 studied the effects of localized heat generation in particlesand layers using the finite element technique. They found that 3 mW of heat generatedin a 20 mm diameter spherical phosphor particle might lead to excessive temperatures,which could be a major source of light output degradation and reliability problems forhigh-power LEDs. Yan et al.70 found that the junction temperature, which character-izes the thermal behavior of monotonic color LED emitters, cannot be used on its ownto characterize the thermal behavior of white LEDs. They also found that the phosphortemperature is critical in determining the lumen performance and reliability of whiteLEDs.

Studying the heat generated in a phosphor layer is a complex coupled probleminvolving both photonics and thermal aspects. A combination of optical simulationand finite element simulation was used to investigate this problem.71 In the MonteCarlo optical simulation, the heat accumulated through optical absorption by the pack-aging materials was calculated. The heat was loaded into the finite element model andthe temperature field of the LED package was simulated. The location of hotspots in aphosphor-converted white LED package for two different kinds of phosphor coating (adirect coating and a remote coating) were compared, while the phosphor concentrationwas changed. Plates 15 and 16 (see color plate section) show the results. From Plate15, it can be seen that on increasing the phosphor concentration from 0.05 to0.35 g/cm3, the temperature of the LED packages increased and the hotspot wasmore obvious. The hotspot was close to the phosphor layer in Plate 15 regardless ofthe phosphor concentration. As shown in Plate 16, for the remote coating, the locationof the hotspot had shifted. When the phosphor concentration was low (0.05 g/cm3), thehotspot was located at the chip; when the phosphor concentration increased, the hot-spot shifted to the phosphor layer. It was concluded that the location of hotspots inremote phosphor coating packages depends on the phosphor concentration, while therewas no dependence for direct phosphor coating packages. In summary, the location ofhotspots depends on phosphor concentration as well as the packaging method.

15.7.3 Contact thermal resistance and the thermal interfacematerial

LEDs are composed of many components, which are mounted on top of each other aslayers. There are many interfaces between the layers, such as the chip-to-copper slug


interface, the copper slug-to-substrate interface and the substrate-to-heat sink interface.Basically, there are two kinds of interface:

• a permanent interface, such as a solder or die adhesive• a non-permanent interface, e.g., where a component is mounted onto a heat sink or between

an assembled module and a chassis.

No real surface is perfectly smooth, and surface roughness is believed to play a crit-ical role in determining the contact thermal resistance. In an interface, only a small areamechanically makes contact between the two surfaces, thus the contact thermal resis-tance is an important part of the overall thermal resistance. Fig. 15.26 is a basic sche-matic of heat transfer between two materials. There are two principal contributions tothe heat transfer at the joint:

• solid-to-solid conduction at the points of contact• conduction through entrapped air in the void spaces created by the contact.

The second contribution is believed to be the major cause of resistance to heatflow, since the thermal conductivity of air is quite small in comparison to that ofsolids.72 The total contact thermal resistance is therefore the two contributions takenin parallel, as shown in Fig. 15.26, where Rc is the thermal resistance of the solid-to-solid conduction and Rg is the thermal resistance of the entrapped air or an infillingthermal interface material (TIM). The TIM is used to expel the air and fill in thevoid at the interface to reduce the contact thermal resistance.73,74 Compared withthe other packaging materials along the heat flow path, the TIM has the lowest thermalconductivity. Consequently, the TIM acts as a bottleneck in the heat flow path andthus affects the heat flow rate. Categories and general properties of TIMs are listedin Table 15.2.

d Solid 1

Solid 2

T1 T1

T2

Rc Rg

T2Contact

point

Air gapor TIM

Figure 15.26 Contact face of an interface. TIM, thermal interface material.


15.7.4 Spreading thermal resistance and packaging substrate

Generally, LED devices are mounted on a broad substrate. Heat generated by an LEDdie conducts through the package and transfers to the substrate. As LED dies and pack-ages are usually much smaller than the substrate, the heat dissipation processes can betreated as a heat flux from a surface conducting into a larger plate. When an LEDcomes into contact with the base of a large heat sink, the hotspot phenomenon appears.A large spreading thermal resistance contributes to this phenomenon. Spreading ther-mal resistance is the main contributor to the overall thermal resistance when heat isconducted from a small area into a large plate, especially for LEDs. Therefore it isvery important to reduce the spreading thermal resistance for LED packaging.

For many LED applications, to meet the demands for illumination and reduce costs,several chips are packaged into one module. For these multi-chip packages, the

Table 15.2 Categories and general properties of TIMs

Category General properties

Thermal grease • Usually consists of two primary components, i.e., a polymer baseand a ceramic or metallic filler. Silicone is usually used as thepolymer base material

• Typical fillers: alumina, AlN, ZnO, SiO2, BN, silver, aluminumpowders, etc.

Phase change • Low-temperature thermoplastic adhesive

Materials (PCMs) • All-metal phase change materials based on low melting alloys andshape memory alloys

Thermallyconductive

• Typical PCMs: thermal pads, low melting alloys, shape memoryalloys, exfoliated clay, fusible/non-fusible fillers

Elastomers (gels) • Generally consist of a silicone elastomer filled with thermallyconductive ceramic particles

Adhesives • Thermal resistance depends on thickness, clamping pressure andbulk thermal conductivity

• Typically silver particles in a cured epoxy matrix

Solders • Low thermal conductivity

• High thermal conductivity

• High processing temperature

Carbon-based TIMs • Challenge: voids underneath the chips

• Unique thermal and rheological properties

• Typical materials: carbon fibers/nano-fibers, graphite flakes,carbon nanotubes


spreading thermal resistance is large. The heat from different chips interacts and sincethe locations of the chips affects the spreading thermal resistance, it is essential to opti-mize the design. By changing the arrangement of the chips on the substrate, thespreading resistance can be reduced and the temperature distribution can be madeuniform. Plate 17 (see color plate section) shows the temperature before and afterthe optimization of the position of LED chips.75

Another method for reducing spreading thermal resistance is to use a heat pipe typesubstrate such as a flat-plate vapor chamber. Fig. 15.27 shows how a flat-plate vaporchamber works. A flat-plate vapor chamber is a vacuum vessel with a wick structure onthe inside walls. The chamber is partly filled with a working liquid. As heat is applied,the liquid in the evaporation region absorbs the heat from the heat source and evapo-rates, and the resulting vapor flows to fill the empty part of the cavity. Whenever thevapor comes into contact with a cooler wall surface, it will condense and release thelatent heat of vaporization to the heat sink. Finally the condensed liquid returns tothe evaporation region via the capillary force of the wick. The area of the heat sinkis usually several-fold or several tenfold that of the heat source. The net effect of aflat-plate vapor chamber is to transport heat from the evaporation region to thecondenser region so the heat generated by the heat source is diffused to the heatsink where the heat flux density decreases considerably. Because of the flowing vaporinside the chamber, the temperature of the top surface of the chamber is uniform intheory, thus the spreading thermal resistance can be reduced significantly. Moreover,the dissipation efficiency of a heat sink coupled with a flat-plate vapor chamber can beincreased. Huang et al.76 experimentally studied the thermal performance of a vaporchamber module applied to high-power LEDs and found that the spreading thermalresistance of a flat-plate vapor chamber at 30 W was lower than that of a copper plateby 34%. We fabricated a flat-plate vapor chamber coupled to a fin heat sink and used itto cool a 20 W LED light source. We found that the flat-plate vapor chamber was agood way to alleviate the hotspots and decrease the spreading thermal resistance.77

Heat sink

WicksHeat source

Condenser region

Evaporation region

Evaporation

Figure 15.27 Working mechanism of flat-plate vapor chambers.


15.7.5 Cooling solutions for LED applications

Cooling solutions for LED applications either use passive or active cooling. Passivecooling occurs when heat is transferred without any artificially imposed force and extraenergy consumption, such as free air convection. Active cooling needs an imposedforce or input power, such as forced air cooling, liquid cooling (microchannel coolingor micro-jet cooling), semiconductor refrigeration, ultrasonic heat dissipation andsuperconducting cooling.

Generally speaking, the heat dissipation ability of passive cooling is limited sincethe convection heat-transfer coefficient h for air is usually less than 10 W/(m2K).Because of the small h, according to Newton’s law of cooling, a large heat exchangearea A is needed for sufficient heat transfer. As a result, engineers usually improve theheat dissipation performance of a plate fin heat sink by increasing the number of fins toincrease the heat exchange area A or designing the space between the fins andincreasing the height of the fins to increase the heat transfer coefficient h. So far, platefin heat sinks have worked quite well with general LED lighting products.78

Compared with passive cooling, active cooling can enhance the heat transfer abilitysignificantly, but its application to LED products is limited because of reliability andcost requirements. When the total power of an LED is very high, passive cooling isinadequate and active cooling must be used. There are many kinds of active coolingsolutions; here we only consider forced convection and micro-jet cooling as examplesto demonstrate active cooling solutions. Forced convection is driven by an artificialforce or power, such as a fan or pump, which accelerates the air flow and substantiallyincreases the heat transfer coefficient. Forced convection can improve the heat ex-change rate significantly compared with free convection. Another kind of active liquidcooling technology is the micro-jet.79e81 Fig. 15.28 shows a closed-loop LED micro-jet array cooling system. It has three parts: a micro-jet array device, a micro-pump anda mini fluid container with a heat sink. In the closed-loop system, water or another fluidis driven into the micro-jet array device through an inlet by a micro-pump. There aremany micro-jets inside the jet device, which directly impinge onto the bottom plate ofthe LED array. Since an impinging jet has a very high heat transfer coefficient, the heatcreated by the LEDs is easily removed by the fluid recycling in the system. The fluid

Bottom plate of LED array

Bottom cavity of jet device

Outlet

Fan

Heat sinkMini water container

Micro-pump

Micro-jet array

Jet device inlet

Figure 15.28 Closed-loop micro-jet array cooling system.


heats up and its temperature increases after it flows out of the jet device, and the heatedfluid enters the mini fluid container. The heat sink, which has a fan, cools the fluid andthe heat dissipates into the external environment. The cooled fluid is delivered back tothe jet device to cool the LED array, again driven by the force of the micro-pump in thesystem.

15.8 Conclusion

In this chapter, issues with the chip packaging for nitride LEDs were reviewed,including the functions, structures, processes, and the optical and thermal design ofLEDs. Packaging processes have an important role in determining the final opticaland thermal performance of LEDs, thus they should be well designed optically andthermally. The effects of the gold wire bonding process, the phosphor coating processand lenses on optical performance were discussed. Heat dissipation for high-powerLEDs is a big challenge and the LED packages should be thermally well designed.A thermal design should take into account the packaging structure and phosphorself-heating. The effects of contact thermal resistance, spreading thermal resistanceand external cooling solutions were also discussed.

References

1. Hu JZ, Yang LQ, Shin MW. Thermal effects of moisture-inducing delamination in light-emitting diode packages. Electronic components technology conference. 2006. p. 1957e62.

2. Luo XB, Wu BL, Liu S. Effects of moist environments on LED module reliability. IEEETrans Device Mater Reliability June 2010;10(2):182e6.

3. Hu JZ, Yang LQ, Shin MW.Mechanism and thermal effect of delamination in light-emittingdiode packages. Microelectronics J 2007;38(2):157e63.

4. Arik M, Weaver S. Chip scale thermal management of high brightness LED packages. ProcSPIE 2004;5530:214e23.

5. Chen Q, Luo XB, Zhou S, Liu S. Dynamic junction temperature measurement for highpower light emitting diodes. Rev Scientific Instr 2011;82:084904.

6. Gu YM, Narendran N. Anon-contact method for determining junction temperature ofphosphor converted white LEDs. Proc SPIE 2004;5187:107e14.

7. Xi Y, Xi JQ, Gessmann T, Shah JM, Kim JK, et al. Junction and carrier temperaturemeasurements in deep-ultraviolet light-emitting diodes using three different methods. ApplPhys Lett 2005;86:031907.

8. Liu S, Luo XB. LED packaging for lighting applications: design, manufacturing andtesting. USA: John Wiley & Sons; 2011.

9. Jiang JB, To S, Lee WB, Cheung B. Optical design of a freeform TIR lens for LEDstreetlight. Optik 2010;121:1761e5.

10. Chang JG, Liao LD, Hwang CC. Enhancement of the optical performances for the LEDbacklight systems with a novel lens-cap. Proc SPIE 2006;6289:62890X.

11. Steigerwald DA, Bhat JC, Collins D. Illumination with solid state lighting technology. IEEEJ Selected Topic Quan Electronics 2002;18(2):310e20.


12. http://www.cree.com/press/press_detail.asp?i¼1304945651119.13. Harman G. Wire bonding in microelectronics. 3rd ed. McGraw Hill; 2010.14. Toyozawa K, Fujita K, Minamide S, Maeda T. Development of copper wire bonding

application technology. IEEE Trans Components, Hybrids, Manufacturing Technol 1990;13(4):667e72.

15. Collins WD, Krames MR, Verhoeckx GJ, Leth NJM. Using electrophoresis to produce aconformal coated phosphor-converted light emitting semiconductor. US Patent, lumiledslighting, US 6576488. 2001.

16. Collins WD, Krames MR, Verhoeckx GJ, Leth NV.Method for conformally coating a lightemitting semiconductor structure with a phosphor by electrophoresis. European Patentapplication, lumileds lighting, EP 1267424A2. 2002.

17. Yum JH, Seo SY, Lee S, Sung YE. Comparison of Y3Al4O12:Ce0.05 phosphor coatingmethods for white light-emitting diode on gallium nitride. In: Solid State Lighting andDisplays, Proceedings of SPIE, San Diego, CA, USA; 2001. p. 60e9.

18. Loh BP, Medendorp Jr NW, Andrews P, Fu Y, Laughner M, et al. Method of uniformphosphor chip coating and LED package fabricated using method. US Patent, Cree Inc.,20080079017 A1. 2008.

19. Braune B, Peterson K, Strauss J, Kromotis P, KaempfM. A newwafer level coating techniqueto reduce the color distribution of LEDs. In: Light-emitting diode: research, manufacturing,and applications XI, proceedings of SPIE, San Jose, CA, USA; 2007. 64860X.

20. Zheng H, Luo XB, Hu R, Cao B, Fu X, et al. Conformal phosphor coating using capillarymicrochannel for controlling color deviation of phosphor-converted white light-emittingdiodes. Opt Express 2012;20(5):5092e8.

21. Liu S. Method to fabricate white light-emitting diode. Chinese Patent application,200610029858.6. 2006.

22. Yamada K, Imai Y, Ishii K. Optical simulation of light source devices composed of blueLEDs and YAG phosphor. J Light Vis Environ 2003;27(2):70e4.

23. Zhu Y, Narendran N, Gu Y. Investigation of the optical properties of YAG: Ce phosphor.In: Sixth International Conference on solid state lighting, San Diego, CA, USA, Pro-ceedings of SPIE; 2007. 63370S.

24. Arik M, Seltur A, Weaver S, Haitko D, Petroski J. Chip to system thermal needs andalternative thermal technologies for high brightness LEDs. J Electron Packaging 2007;129(3):328e38.

25. Kim JK, Luo H, Schubert EF, Cho J, Sone C, et al. Strongly enhanced phosphor efficiencyin GaInN white light-emitting diodes using remote phosphor configuration and diffusereflector cup. Jpn J Appl Phys 2005;44:L649e51.

26. Luo H, Kim JK, Schubert EF, Cho J, Sone C, et al. Analysis of high-power packages forphosphor-based white-light-emitting diodes. Appl Phys Lett 2005;86(24):243505.

27. Luo H, Kim JK, Xi Y, Schubert EF, Cho J, et al. Analysis of high-power packages for white-light-emitting diode lamps with remote phosphor. Mater Res Soc 2006;892. FF09-FF07.

28. Fan BF, Wu H, Zhao Y, Xian YL, Wang G. Study of phosphor thermal-isolated packagingtechnologies for high-power white light-emitting diodes. IEEE Photon Technol Lett 2007;19(15):1121e3.

29. Wu BL, Luo XB, Zheng H, Liu S. Effect of gold wire bonding process on angular correlatedcolor temperature uniformity of white light-emitting diode. Opt Express 2011;19(24):24115e21.

30. Liu ZY, Li C, Yu BH, Wang YH, Liu S, et al. Effects of YAG: Ce phosphor particle size onoptical performance of white LEDs. In: 12th International Conference on electronicpackaging technology and high density packaging; 2011. p. 1066e71.


http://www.cree.com/press/press_detail.asp?i=1304945651119

http://www.cree.com/press/press_detail.asp?i=1304945651119

31. Hu R, Luo XB, Liu S. Effect of the amount of phosphor silicone gel on optical property ofwhite light-emitting diodes packaging. In: 12th International Conference on electronicpackaging technology and high density packaging; 2011. p. 1081e4.

32. Hu R, Luo XB, Feng H, Liu S. Effect of phosphor settling on the optical performance ofphosphor-converted white light-emitting diodes. J Lumin 2012;132:1252e6.

33. Sommer C, Reil F, Krenn JR, Hartmann P, Pachler P, et al. The impact of light scattering onthe radiant flux of phosphor-converted high power white light-emitting diodes. J LightwaveTechnol 2011;29(15):2285e91.

34. Sommer C, Wenzl FP, Reil F, Krenn JR, Pachler P, et al. A comprehensive study on theparameters effecting color conversion in phosphor converted white light-emitting diodes.In: Tenth International Conference on solid state lighting, proceedings of SPIE, 7784;2010. 77840D.

35. Sommer C, Hartmann P, Pachler P, Schweighart M, Tasch S, et al. A detailed study on therequirements for angular homogeneity of phosphor converted high power white LED lightsources,. Opt Mater 2009;31:837e48.

36. Uchida Y, Taguchi T. Lighting theory and luminous characteristics of white light-emittingdiodes. Opt Eng 2005;44(12):124003.

37. Keppesn A, Denijs S, Wouters S, Rychaert WR, Deconinck G, et al. Modelling the spatialcolour distribution of phosphor-white high power light-emitting diodes. Proc SPIE 2010;7717:77170J.

38. Liu ZY, Liu S, Wang K, Luo XB. Effects of phosphor’s location on LED packagingperformance. In: International Conference on electronic packaging technology and highdensity packaging, Shanghai, China; 2008. 4606982-1-7.

39. Liu ZY, Liu S, Wang K, Luo XB. Status and prospects for phosphor-based white LEDpackaging. Front Optoelectronics China 2009;2(2):119e40.

40. Liu ZY, Liu S, Wang K, Luo XB. Studies on optical consistency of white LEDs affected byphosphor thickness and concentration using optical simulation. IEEE Trans ComponentsPackaging Technology 2010;33(4):680e7.

41. Sommer C, Wenzl FP, Hartmann P, Pachler P, Schweighart M, et al. Tailoring of the colorconversion elements in phosphor-converted high-power LEDs by optical simulations. IEEEPhoton Technol Lett 2008;20(5):739e41.

42. Tran NT, Shi FG. Studies of phosphor concentration and thickness for phosphor-basedwhite light-emitting-diodes. J Lightwave Technology 2008;26(21):3556e9.

43. Liu ZY, Liu S, Wang K, Luo XB. Optical analysis of color distribution in white LEDs withvarious packaging methods. IEEE Photon Technol Lett 2008;20(24):2027e9.

44. Liu ZY, Liu S, Wang K, Luo XB. Measurement and numerical studies of optical propertiesof YAG: Ce phosphor for white light-emitting diode packaging. Appl Opt 2010;49(2):247e57.

45. Hu R, Luo XB, Liu S. Study on the optical properties of conformal coating light-emittingdiode by Monte Carlo simulation. IEEE Photon Technol Lett 2011;23(20):1673e5.

46. Yu RY, Jin SZ, Cen SY, Liang P. Effect of the phosphor geometry on the luminous flux ofphosphor-converted light-emitting diodes. IEEE Photon Technology Lett 2010;22(23):1765e7.

47. Zhu Y, Narendran N. Optimizing the performance of remote phosphor LEDs. Jpn J LightVis Environ 2008;32(2):115e9.

48. Won YH, Jang HS, Cho KW, Song YS, Jeon DY, et al. Effect of phosphor geometry on theluminous efficiency of high-power white light-emitting diodes with excellent colorrendering property. Opt Lett 2009;34(1):1e3.

49. Feng ZX, Luo Y, Han YJ. Design of LED freeform optical system for road lighting withhigh luminance/illuminance ratio. Opt Express 2010;18(21):22020e31.


50. Ding Y, Liu X, Zheng ZR, Gu PF. Freeform LED lens for uniform illumination. OptExpress 2008;16(17):12958e66.

51. Parkyn WA. The design of illumination lenses via extrinsic differential geometry,. ProcSPIE 1998;3428:154e62.

52. Wang K, Wu D, Chen F, Liu ZY, Luo XB, et al. Angular color uniformity enhancement ofwhite light-emitting diodes integrated with freeform lenses. Opt Lett 2010;35(11):1860e2.

53. Wang K, Wu D, Qin Z, Chen F, Luo XB, et al. New reversing design method for LEDuniform illumination. Opt Express 2011;19(S4):A830e40.

54. Ries H, Muschaweck J. Tailored freeform optical surfaces. J Opt Soc America A 2002;19(3):590e5.

55. Chaves J. Introduction to nonimaging optics. CRC press, Taylor & Francis Group; 2008.56. Minano JC, Benitez P, Liu JY, Infante J, Chaves J, et al. Applications of the SMS method to

design of compact optics. Proc SPIE 2010;7717:771701.57. Benitez P, Minano JC, Blen J, Mohedano R, Chaves J, et al. Simultaneous multiple surface

optical design method in three dimensions. Opt Eng 2004;43(7):1489e502.58. Dross O, Cvetkovic A, Chaves J, Benitez P, Minano JC. LED headlight architecture that

creates a high quality beam pattern independent of LED shortcomings. Proceeding of SPIE2005;5942:59420D.

59. Wang K, Liu S, Chen F, Qin Z, Liu ZY, et al. Freeform LED lens for rectangularly pre-scribed illumination. J Opt A: Pure Appl Opt 2009;11:105501.

60. Wang L, Qian KY, Luo Y. Discontinuous free-form lens design for prescribed irradiance.Appl Opt 2007;46(18):3716e23.

61. Wang K, Liu S, Chen F, Liu ZY, Luo XB. Effect of manufacturing defects on opticalperformance of discontinuous freeform lenses. Opt Express 2009;17(7):5457e65.

62. Wang K, Chen F, Liu ZY, Luo XB, Liu S. Design of compact freeform lens for applicationspecific light-emitting diode packaging. Opt Express 2010;18(2):413e25.

63. Wang S, Wang K, Chen F, Liu S. Design of primary optics for LED chip array in roadlighting applications. Opt Express 2011;19(S4):A716e24.

64. Chen F, Liu S, Wang K, Liu ZY, Luo XB. Free-form lenses for high illumination qualitylight-emitting diode MR16 lamps. Opt Eng 2009;48(12):123002.

65. Tang YS, Hu SF, Lin CC, Bagkar NC, Liu RS. Thermal stable luminescence of KSrPO4:Euphosphor for white light UV light-emitting diodes. Appl Phys Lett 2007;90(15):151108.

66. Arik MP, Weaver J. Thermal challenges in the future generation solid state lightingapplications: light emitting diodes. In: International Society Conference on thermalphenomena, san Diego, USA; 2002. p. 113e20.

67. Tsao JY. Solid state lighting-lamps-chips and materials for tomorrow. IEEE CircuitsDevices Mag 2004;20(3):28e37.

68. DeLuca JA. An introduction to luminescence in inorganic solids. J Chem Education 1980;57(8):541e8.

69. Arik M, Weaver S, Becker C, Hsing M, Srivastava A. Effects of localized heat generationdue to the color conversion in phosphor particles and layers of high brightness light emittingdiodes. In: Proceedings of International Electronic Packaging Technical Conference andExhibition, Maui, Hawaii, USA; July 6e11, 2003. p. 611e9.

70. Yan B, Tran NT, You JP, Shi FG. Can junction temperature alone characterize thermalperformance of white LED emitters? IEEE Photon Technol LettMay 1, 2011;23(9):555e7.

71. Hu R, Luo XB. Will the hotspot location in the high power phosphor converted white light-emitting diode package shift?. In: International Conference on Advanced Electromaterials,Jeju, Korea; November 7e10, 2011.


72. Holman JP. Heat transfer. Beijing, China: McGraw-Hill Inc. and China Machine Press;2005.

73. Prasher R. Thermal interface materials: historical perspective, status, and future directions.Proc IEEE 2006;94(8):1571e86.

74. Sarvar F, Whalley DC, Conway PP. Thermal interface materials e a review of the state ofthe art. In: Electronics System Integration Technology Conference, Dresden, Germany;2006. p. 1292e302.

75. Cheng T, Luo XB, Huang SY, Liu S. Thermal analysis and optimization of multiple LEDpackaging based on a general analytical solution. Int J Therm Sci 2010;49(1):96e201.

76. Huang HS, Chiang YC, Huang CK, L Chen S. Experimental investigation of vapor chambermodule applied to high-power light-emitting diodes. Exp Heat Transfer 2009;22:26e38.

77. Luo XB, Hu R, Guo TH, Zhu XL, Chen W, et al. Low thermal resistance LED light sourcewith vapor chamber coupled fin heat sink. In: Electronic Components and TechnologyConference; 2010. p. 1347e52.

78. Luo XB, Xiong W, Cheng T, Liu S. Design and optimization of horizontally-located platefin heat sink for high power LED street lamps. In: Electronic Components and TechnologyConference, San Diego, CA, USA; 2009. p. 854e9.

79. Luo XB, Liu S. A microjet array cooling system for thermal management of high-brightnessLEDs,. IEEE Trans Adv Packaging August 2007;30(3):475e84.

80. Luo XB, Chen W, Sun RX, Liu S. Experimental and numerical investigation of a microjet-based cooling system for high power LEDs. Heat Transfer Eng 2008;29(9):774e81.

81. Luo XB, Liu S, Jiang XP, Cheng T. Experimental and numerical study on a micro jetcooling solution for high power LEDs. Sci China Ser E: Technol Sci 2007;50(4):478e89.


Part Three

Applications of nitride LEDs


White LEDs for lightingapplications 16Richard KotschenreutherOSRAM GmbH, Munich, Germany

16.1 White LEDsddefinition of area

Dependent on the application, different areas of “White” are used. Table 16.1shows boundaries and intersections of white areas in Fig. 16.1 for commonapplications.

16.2 Why “white LEDs”?

Professional applications prefer white LED light. Industry, offices, laboratories, pub-lic areas, streets, and other applications need illumination with white light in order toconduct tasks like reading, writing, detection, recognition, and handling. Nearly allIEC standards are written for white light sources. LEDs comprise the color fieldsdesignated as “warm white” (<3.300K), “neutral white” (3.300e5.300K), and“cool light”/“daylight” (>5.300K). The German main occupational organizationssay: “For illumination, supplementing daylight with artificial light, lamps are recom-mended which have a CCT (correlated color temperature) of 4.000K or more. Lampswith a light color neutral white and daylight white affect positively the vigilance(consciousness) due to their spectral composition, thus increasing the workingefficiency of the employees. Therefore, lamps with CCT 4.000K or more arerecommended.”4

Although this chapter focuses on white LEDs, some adjacent sectors are important,for example, blue for human centric lighting (HCL). Another interesting examplewhere colored LEDs are used is the agro-industrial application (see Section 16.4)with crop-dependent colors like “deep-blue,” “hyper red,” and “far red.”

16.3 The three-side-approach for lighting applications

16.3.1 Market needs

Development of blue LEDs opened general lighting business for white LED andcreated market needs for illumination by white LED. These needs are described


https://doi.org/10.1016/B978-0-08-101942-9.00016-2

Table 16.1 x-y coordinates determining white areas in different applications2

CIE white Aa CIE white Ba UN R48b EU reg. 1194c

Signal lights Signal lights Automotive HCL Special purpose

Boundaries

Purple y ¼ 0.047 þ 0.762 x y ¼ 0.047 þ 0.762 x y ¼ 0.050 þ 0.750 x Similar to CIE WhiteA, but for specialapplications alsoadjacent Blue

x < 0.270 or x > 0.530;y < �2.3172x2

þ 2.3653x � 0.2199 ory > 2.3172x2 þ 2.3653x� 0.1595

Blue x ¼ 0.300 x ¼ 0.300 x ¼ 0.310

Green y ¼ 0.150 þ 0.640 x y ¼ 0.150 þ 0.640 x y ¼ 0.150 þ 0.640 x

Green y ¼ 0.340 þ 0.140 x

Yellow/Green

x ¼ 0.440 y ¼ 0.440

Yellow x ¼ 0.500 x ¼ 0.500

Red/Purple

e y ¼ 0.382 x ¼ 0.382

Intersections

x ¼ 0.300, y ¼ 0.342 x ¼ 0.300, y ¼ 0.342 x ¼ 0.310, y ¼ 0.348

x ¼ 0.440, y ¼ 0.432 x ¼ 0.440, y ¼ 0.432 x ¼ 0.453, y ¼ 0.440

x ¼ 0.440, y ¼ 0.382 x ¼ 0.500, y ¼ 0.440 x ¼ 0.500, y ¼ 0.440

x ¼ 0.300, y ¼ 0.276 x ¼ 0.500, y ¼ 0.382 x ¼ 0.500, y ¼ 0.382

x ¼ 0.440, y ¼ 0.382 x ¼ 0.443, y ¼ 0.382

x ¼ 0.300, y ¼ 0.276 x ¼ 0.310, y ¼ 0.283

aDIN 6163:2015.1bUN R48.2cEU Regulation 1194.3

532Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

in terms of safety, performance, life time, efficiency, serving the customer, well-being, exchangeability, and nowadays of being “smart” and “connected.” In thefollowing, these prominent terms are considered.

0,8

510

0,7

0,6

0,5

0,4

0,3

0,2

0,1

0 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8x

y

520530

540

550

500

560

570

2 360

P 2 000

1 900

2 856

4 000

580

590

600610

620630

640650660680

690 bis 780

Rot

6 000

10 000

Gelb

Weiß BWeiß A

Grϋn

490

Blau440

470

480

380 bis 410460450

430420

X◦∞

Figure 16.1 x-y diagram of color locations.1

1 Wiedergegeben mit Erlaubnis von DIN Deutsches Institut f€ur Normung e.V. Mabgebend f€ur dasAnwenden der DIN-Norm ist deren Fassung mit dem neuesten Ausgabedatum, die bei der Beuth VerlagGmbH, Am DIN Platz, Burggrafenstrabe 6, 10787 Berlin, erh€altlich ist.

White LEDs for lighting applications 533

16.3.1.1 Safety

Safety takes the highest priority and the creation of safe products deserves ourutmost attention. This comprises requirements and tests for electrical, thermal,and photometrical parameters. Although not specific for white LEDs, the examplebelow is typical for safety checks of light sources and is extracted from the IECstandard of double-capped LED lamps for retrofit purposes.

IEC 62776,5 extract from contents page.

As a consequence of safety as highest priority, safety standards of an LED productare developed before deploying performance standards.

16.3.1.2 Performance

Performance is specified with requirements and tests for light output (luminous flux,luminous intensity, peak/beam intensity) and colorimetric quantities (chromaticitycoordinates, CCT, CRI). The example below shows typical standardized tests whichwere introduced first time as a complete set for testing LED modules. Angular coloruniformity (sometimes called color spatial uniformity) of luminance is in discussionin IEC (International Electrotechnical Commission).


IEC 62717,6 extract from contents page.

16.3.1.3 Life time

Critical to be determined is the life time, nevertheless being one of the most importantfeatures. Even if the life time of LED packages itself cannot be predicted precisely, theknowledge of life time of semiconductor products, together with the experience ofLED, for general lighting permits prediction of maintenance, when measurements oflight output over a limited period of time is available. From these empirically recordeddata, factors for extrapolation are derived. An example is given in Fig. 16.2.

In Fig. 16.2, curves were measured at different temperatures and the achievedvalues being inserted in a double-logarithm-diagram with the x-axis ¼ time and they-axis ¼ luminous flux maintenance. The higher temperature is causing a shorterLED life time, that is, the LED arrives quicker at low values of luminous flux main-tenance, compared with the low temperature measurement result. A time factor iscalculated between both curves. This factor is averaged and applied to the high tem-perature curve, yielding an accelerated (extrapolated) value of luminous fluxmaintenance for the low temperature curve.

Another method is used in IES. Based on real testing time of at least 6.000 h anddata (IES LM-80-088), an exponential function is fitted to the data (IES TM-21-119). The projection of life time from the last measurement point is limited to afactor of 6.


16.3.1.4 Efficiency

Efficiency, calculated by dividing the output luminous flux by the input power to aluminaire, is a metric to assess the money and energy to be spent for a certain amountof light. If the luminaire performs more than emitting light, for example, needingenergy for providing included sensors, bluetooth speakers, outlets etc. the efficiencycalculation becomes odd when comparing such a luminaire with another one withoutthese features. Therefore, for comparison purposes, the additional devices have to bestripped off when measuring the consumed energy.10

The energy discussion, predominant when evaluating lighting products entering themarket, might be only one argument for future market needs when regarding thenonvisual effects that lighting can exert on humans.

16.3.1.5 Serving the customer’s well-being

Apart from the energy aspect, living in an aging society, the intensity and color of lightplay an important role since the eyes of elder people require another quality of lightthan the eyes of younger ones, keyword: “ambient assisted living.”11

100

95

90

85100 1000 10000 100000

Time (hrs)

Iv (%

)

Calculated acceleration

Iv (%) Accelerationfactor

9594,794,393,693,493,192,992,992,6

9,111,511,99,1

10,010,0

11,112,510,5

Mean value = 10,6Std. dev. = 1,2

Extrapolated lifetime basedon mean acceleration factor

Figure 16.2 Acceleration of aging by means of temperature.VDE-AR 2715-1,7 translated2.

2 Bild B1 aus VDE-AR-E:2012e11, f€ur die angemeldete limitierte Auflage wiedergegeben mit Genehmi-gung 362.016 des VDE Verband der Elektrotechnik Elektronik Informationstechnik e.V. F€ur weitereWiedergaben oder Auflagen ist eine gesonderte Genehmigung erforderlich. Mabgebend f€ur die Anwen-dung der Normen sind deren Fassungen mit dem neuesten Ausgabedatum, die bei der VDE VERLAGGMBH, Bismarckstr. 33, 10,625 Berlin, www.vde-verlag.de erh€altlich sind.


http://www.vde-verlag.de/

Some years ago, human needs with regard to lighting did not mean more thanhaving a certain amount of lux for a certain kind of work at a working place. Afterdiscovery of the nonvisual receptor in the eye, this simple view has changed. Thecircadian body clock is triggered by these receptors, as well as melatonin production.One of the HCL findings is that depending on the light color the attention varies. Interms of application, HCL-based illumination enters schools and many workingplaces. It should therefore be noted that illumination of working places and HCLhave different aims. Both illuminations at one place may lead to having an effect oneach other.

Different to the comfort provided by HCL is the comfort to operate the light sourcewireless (Figs. 16.3 and 16.4). Typically, components like wireless socket outlet,wireless key, WLAN Gateway are used. Example:

Lightify

Figure 16.4 Innovative wireless lighting solutions with cloud.OSRAM.

(a) (b) (c) (d)

Figure 16.3 Components of a wireless system. (a) Gateway WLAN þ lamp (b) luminaire(c) plug (d) switch.OSRAM.


16.3.1.6 Exchangeability

Exchangeability is an ambivalent feature: The market may wish that the LED lamp isexchangeable against an incandescent lamp with the same cap. This may work, if thereis no further electric/electronic component between lamp and mains. It gets to bedifficult, if a dimmer is involved. Cooperation between the IEC committee for lightsources and the IEC committee for dimmers is going to fix mutual functioning ofboth products in their standards in parallel. Extract from the scope of the IECproject:

This Technical Report is meant to help the control gear / integrated lamp designers tomake

products suitable to operate with phase-cut dimmers. It describes the possible voltagesignals

and the expected behavior of the control gear / integrated lamps.IEC/TR 63037.12

Exchangeability can also take the form of replacing a fluorescent lamp by an LEDtube, and guiding the safety of this kind of application is covered by anotherstandard, IEC 62776.5 The installation base for fluorescent lamps is quite big, andin order to serve this market, in addition to the lamp exchange and keeping theexisting caps and lampholders (G5, G13), new types of caps and lampholders(e.g., GX16t-5,13) are brought to the market, which, of course requires newluminaires (Fig 16.5).

(a) (b)

Figure 16.5 Tubular LED lamps with different caps. (a) LED tube with cap G13 (b) LED tubewith cap GX16t-5.(a) OSRAM (b) BJB.


16.3.1.7 Smart and connected

“Smart” in the context with lighting seems to be synonymous for efficient lightingindoor and outdoor, in homes, streets, and cities. The expectation of saving energyis combined with a “smart,” “clever,” or “intelligent,” way how to achieve this, forexample, with “smartmeters.”

Controllability of LED is a necessary feature in a connected world. What shall beconnected? Several luminaires could be combined in lighting scenes, luminaires couldbe combined with sensors (presence, daylight/blinds, traffic, etc.). Interconnection ispossible with other controllable electrical devices (timer, heating/cooling, etc.)(Fig 16.6).

16.3.2 Manufacturers’ abilities

In an open market, any variation of LED power, luminous flux, CCT, voltage binningetc. is possible. Standardization was and will be helpful in providing reference tospecific values and tests. This allows manufacturers to establish certain productcategoriesdin contrast to infinitesimal variationsdand serve the volume market ata reasonable price and with products of typical parameter values. An example ofdedicated values (here: CCT and chromaticity coordinates) is provided in the LEDlamp standard, IEC 62612,14 see Table 16.2 below.

Here is the schematic publication and project time scale of selected LED standardsin IEC (International Electrotechnical Commission) (Fig. 16.7).

These and other LED standards are the landmarks of good lighting; for moreinformation see Annex 1. Apart from IEC, consortia like Zhaga are working on theperformance of LED products.

Trivially, market needs are not always compatible with manufacturers’ abilities. Forexample, the pair of parameters, CRI and efficiency, are not easy to be optimized at the

Figure 16.6 Schematic picture of luminaires connected in a lighting scene.R. Kotschenreuther (author).


same time. The market wants both at best level, the manufacturer can provide productsat highest CRI, highest efficiency, or at a balanced value of both. The reason forcompromise is the composition of the CRI consists of, for example, red values.However, red is not the wavelength at which the eye is most efficient.

Also, the use of LED in high temperature applications is in contrast to theirlife time. At high temperatures, the life time may be shortened whereas at lowtemperatures, the life time may be longer. Hence, the same product may be used indifferent climatic ambients, well understood: with different life time expectations.IEC 627176 provides the following Table 16.3:

In all these cases where the advantage of one feature is put against another one, themanufacturer has to make a choice, and standardized values are helping.

16.3.3 Regulators’ views

The existence of white LED applications is not only a matter of free play of marketforces and availability of standards but a “fence” is drawn around products brought

2006 2016

Controlgear

Lamp,module

Luminaire

SystemFigure 16.7 Selected IEC LEDpublications and expected workalong time axis.R. Kotschenreuther (author).

Table 16.2 Preferred values of LED lamp CCT and chromaticitycoordinates

Color marking CCT (Tc)

Chromaticity coordinates

x y

F 6500 6400 0.313 0.337

F 5000 5000 0.346 0.359

F 4000 4040 0.380 0.380

F 3500 3450 0.409 0.394

F 3000 2940 0.440 0.403

F 2700 2720 0.463 0.420

P 2700 2700 0.458 0.410

The letters in the color marking designation stand for: F ¼ Values from IEC 60081, Annex D, P ¼ Value close to thePlanckian curve.IEC 62612.14


to the market for political reasons: directives and regulations aim, for example, atreducing the use and import of energy. Main viewpoints of regulators are safety(e.g., in Europe the so-called “Low Voltage Directive” (2014/35/EU15)) and efficiency(e.g., in Europe the Energy Related Products Directive (2009/125/EC16)). In addition,European regulators require fulfilling the performance characteristics of harmonizedstandards, as well as ecological features. In particular, life time being one of theperformance characteristics is not easy to be verified, because testing time 6.000 his given in standards which are accepted in the European Union. Some authoritiescannot even find a laboratory, which agrees conducting long-term tests.

Annex IV,3 Table 9: Verification procedure for market surveillance purposes.

In the United States, the Department of Energy (DoE) is the national body topublish rules on energy efficiency for example, incandescent lamp replacements,and the Energy protection agency (EPA) is drafting the second version of EnergyStar under which many products are registered.

Energy efficiency is a key policy in many countries. Some examples:

Japan (Ministry of Economy, Trade and Industry) http://www.meti.go.jp/english/policy/energy_environment/energy_efficiency/China (Ministry of Finance and the National Development and Reform Commission,published by the China National Institute of Standardisation CNIS http://www.energylabel.gov.cn/en/KeyProgramsandProjects/DiscountProgram/detail/594.htmlSouth Korea (Ministry of Trade, Industry and Energy) http://english.motie.go.kr/?p¼5452Australia (Australian Government) http://www.australia.gov.au/information-and-services/environment/energy-efficiency

Table 16.3 LED module life time information

tp temperature (�C) measured at the tp-point XXa XXa XXa

Rated life time (h) XX XXXa XX XXXa XX XXXa

aValues to be declared by the LED module manufacturer.IEC 62717.6


http://www.meti.go.jp/english/policy/energy_environment/energy_efficiency/

http://www.meti.go.jp/english/policy/energy_environment/energy_efficiency/

http://www.energylabel.gov.cn/en/KeyProgramsandProjects/DiscountProgram/detail/594.html

http://www.energylabel.gov.cn/en/KeyProgramsandProjects/DiscountProgram/detail/594.html

http://english.motie.go.kr/?p=5452

http://english.motie.go.kr/?p=5452

http://www.australia.gov.au/information-and-services/environment/energy-efficiency

http://www.australia.gov.au/information-and-services/environment/energy-efficiency

Summarizing, it can be said that not every kind of application may be installed, butmostly those which conserve energy and those which do not harm the environment. Inmany regions, the regulators are keeping close contact to the manufacturing industry inorder to not overstress the energy argument, to paying attention for the existence ofenergy saving products and to having in mind that the new products can be obtainedat a reasonable price.

The above mentioned HCL could be a future pillar, aside the energy aspect, whichmay achieve regulators’ attention and consideration.

16.4 Fields of application

After more than 10 years of market penetration and ever higher values of luminousflux, improved color rendering, minimized tolerances in manufacture, more andmore reliable statements of life time etc., LEDs, incorporated in modules, lamps,luminaires, are installed in nearly each kind of application. Some examples illustratethe wide range of tasks LEDs are accomplishing (Figs. 16.8e16.15).

16.4.1 Street lighting

Figure 16.8 Example of a street lighting LED product.OSRAM.


16.4.2 Automotive lighting

16.4.3 Office lighting

(a) (b)

Figure 16.10 Examples of office lighting with LED products. (a) LED luminaire, DALI(b) officedreception.OSRAM.

(a) (b)

Figure 16.9 Examples of automotive lighting LED products: Fog and daytime running light andLEDriving. (a) LED fog and daytime running light (b) LEDriving.OSRAM.


16.4.4 Shop lighting/accent lighting

16.4.5 Industry/factory/public lighting

Figure 16.11 Example of shop and accent lighting LED product.OSRAM.

Figure 16.12 Example of an industry/factory/public lighting LED product and public place ofapplication.OSRAM.


16.4.6 Sports lighting

16.4.7 Outdoor lighting

Figure 16.13 Example of afloodlight LED product.OSRAM.

Figure 16.14 Example of a wall washerlighting LED product.OSRAM.


16.4.8 Residential lighting

16.5 LED light sources in the connected world

Application of white LED in a single point, as stand-alone solution or even organizedin scenes in a room may lose its meaning.

What is the aim of connecting LED products to a point external to them? And howis this done?

In short words: Lighting equipment is an attractive element amongst other electricdevices. Combining all these devices may minimize the electricity used (e.g., officelight is switched on only when sensing someone is present) or increase the comfort(e.g., the alarm clock triggers soft morning light, low-volume music, and first run ofthe coffee machine). More than only light dimming can be done by installing a pairof dimmer and light source IEC 6275617 enabling (see introduction and scope) “simplecontrol of brightness, colour, colour temperature, and other parameters” “(via) digitalsignals over the load side mains wiring.”

Apart from the efficient and comfortable use of energy with LEDs for lighting,LEDs’ radiation can be the carrier of information. For example, sensing the car head-lamp light can be of use in directing street traffic in order to minimize traffic densityand bypassing accidents.

Catching the stream of energy and recording/transmitting the signals is a task ofsensors. Processing this information and finally analyzing it requires usually gatewaysand software. What kind of information or data do we expect in lighting applicationswith white LED?

• Availability of light: the level of artificial light is set, dependent on the detected daylight(“daylight harvest”), in combination with blinds and under the condition of people beingpresentdinformation for building management

• Access from everywhere with mobile commandsdcustomer service• Configuration: mode of operation, for example, PWMdinformation for development• Interface: connected partners, access to internetdinformation for system engineer

Figure 16.15 Examples of residential lighting LED products.OSRAM.


• Application place: where and how (e.g., height of a street lamp)dinformation fordevelopment and marketing

• Geography: longitude, latitude, climatedinformation for development and marketing• Diagnosis: failures, life timedlow light levels be adjusted by increasing the current or

changing the PWM pauses• Statistics: number of light pointsdinformation for marketing• Security status: information of trial of breaches

16.6 Outlook

16.6.1 Platforms, alliances, and consortia

How would it be, if everybody spoke the same language? No translation would benecessary. We know that this would be an ideal situation. Conveying the idea to thetechnical area, where ideally “things” in the internet understood each other, we findmeanwhile a lot of different languages, like in form of proprietary platforms oralliancesdparallel worlds, which should communicate across their borders. Thetrouble lies with the customer whose devices belong to the one or the other alliance,based on different, noncompatible interfaces. Examples of alliances/consortia

Name Aim Output

Zhagahttp://www.zhagastandard.org/

“to simplify LEDluminaire design andmanufacturing, and toaccelerate the adoptionof LED lightingsolutions”

Consortia standards(“books”) for “LEDlightengines and associatedcomponents,” i.e., LEDmodules with andwithout integratedcontrol gear

TALQhttp://www.talq-consortium.org/

“aims to set a globallyaccepted standard formanagement softwareinterfaces to control andmonitor heterogeneousoutdoor lightingnetworks” and“generates answers tothe main challenges ofoutdoor lighting asreducing energyconsumption and CO2

emissions worldwide,increasing costefficiency andaccelerating theintroduction of LEDluminaires in road andurban lighting”

“The specification definesthe application layerprotocol between acentral managementsystem and outdoorlighting networks.”

Continued


http://www.zhagastandard.org/

http://www.talq-consortium.org/

http://www.talq-consortium.org/

Name Aim Output

TCLAdThe ConnectedLighting Alliance

http://www.theconnectedlightingalliance.org/

“advocate of wirelessconnectivity in lightingapplications” and“promote the globaladoption and growth ofwireless lightingsolutions by supportingopen standards”

“endorses ZigBee 3.0 asthe preferred commonopen standard forresidential connectedlighting applications,simplifying choicesfor both lightingcompanies andconsumers” and “isevaluating therequirements of theprofessional indoorlighting market, andwill make proposals foran open wirelessconnectivity standard”

Threadhttps://www.threadgroup.org/

“to create the best way toconnect and controlproducts in the home”

“The thread stack is anopen standard forreliable, cost-effective,low-power, wirelessD2D (device-to-device)communication. It isdesigned specificallyfor connected homeapplications where IP-based networking isdesired and a variety ofapplication layers canbe used on the stack.”

Fairhair Alliancehttps://www.fairhair-alliance.org/

“enabling lighting andbuilding automation viathe internet of things”

“will collect therequirements of thelighting and buildingautomation industriesand use this informationto develop a set oftechnical specificationsfor a common IP-basedinfrastructure, based onopen IEEE and IETFstandards. The alliancewill liaise with therelevant eco-systems topromote and supportadoption of the Fairhairspecifications.”





https://www.threadgroup.org/

https://www.fairhair-alliance.org/

https://www.fairhair-alliance.org/

It is questionable whether at least the biggest alliances will have a mutual understand-ing. The main aspects that should be covered under this understanding should be: safety,energy efficiency, security, and saving of natural resources. The role of standardizationin this giant task is to fix a minimum set of requirements which are acceptedworldwide. The Zhaga consortium, for example, started to transfer their books to IEC.

The traditional tasks of standardizationdshaping the basis for testing andcomparing products, providing tools for market surveillance and test reports andhelping the customer to understand for which application the LED product is suited,are certainly required in the “IoT” future as well.

16.6.2 Certification

Obstacles to proliferation and application of LED products should be removed as far aspossible. What increasingly can be seen is a special kind of “industry”: the certifica-tion. Certification by a third party is a valuable tool; so far, it is voluntary and the pro-cedure/result is accepted by the customer. However, several countries have discoveredcertification as a source of income. This stops delivery in time and is an extra burdenfor manufacturers having achieved and paid for certificates of conformity in their homecountry. Reliable certification in the manufacturers’ country should be enough.

16.6.3 Technical/mathematical challenges

From a technical point of view, one of the challenges is the precise prediction of the lifetime of LED package, LED module, LED lamp, and LED luminaire. Due to differentmaterials used, procedures applied, and components assembled, it is not easy to pro-vide a single projection method for all. At a more consolidated status, such life timeview might be more promising.

Abbreviations and Acronyms

CIE Commission Internationale de l’�EclairageEU European Union (mostly used in the context of Directives and Regulations of the

European Parliament and European Council)HCL Human Centric LightingIEC International Electrotechnical CommissionIES Illuminating Engineering SocietyUN United Nations (used in the context of automotive products; former UNECE United

Nations Economic Commission for Europe)VDE Verband der Elektrotechnik Elektronik Informationstechnik e.V.

References

1. DIN 6163:2015-01, Farben und Farbgrenzen f€ur Signallichter bei der Eisenbahn und im€offentlichen Nahverkehr.

2. UN R48, UN Vehicle Regulationsd1958 Agreement, Installation of lighting and light-signalling devices (http://www.unece.org/trans/main/wp29/wp29regs41-60.html).


http://www.unece.org/trans/main/wp29/wp29regs41-60.html

3. Commission Regulation (EU) No 1194/2012 of 12 December 2012 implementing Directive2009/125/EC of the European Parliament and of the Council with regard to ecodesignrequirements for directional lamps, light emitting diode lamps, and related equipment.

4. BGR 131e132:2006, clauses 4.1.4 and 5.1.5 (translated), Nat€urliche und k€unstlicheBeleuchtung von Arbeitsst€atten, published by HVBG Hauptverband der gewerblichenBerufsgenossenschaften.

5. IEC 62776:2014, Double-capped LED lamps designed to retrofit linear fluorescentlampsdsafety specifications.

6. IEC 62717:2014, LED modules for general lightingdperformance requirements.7. VDE-AR-E 2715e1:2012-11, Measurement and prediction of reduction in luminous flux of

LEDs.8. IES LM-80-08, IES Approved method for measuring Lumen maintenance of LED light

sources.9. IES LM-21-11, Projecting long-term lumen maintenance of LED light sources.10. IEC 62442e62443:2014, Energy performance of lamp controlgeardPart 3: controlgear for

halogen lamps and LED modulesdmethod of measurement to determine the efficiency ofthe controlgear.

11. ISO 24502:2010, Guidelines for all people including elderly persons and persons withdisabilitiesdvisual signs and displaysdspecification of age-related relative luminance andits use in assessment of light.

12. IEC/TR 63037, Electrical interface specification for self-ballasted lamps and controlgear inphase cut dimmed lighting systems.

13. IEC 62931 (in preparation), GX16te5 capped tubular LED lampdsafety specification.14. IEC 62612:2013, self-ballasted LED lamps for general lighting services by voltage >50

Vdperformance requirements.15. Directive 2014/35/EU of the European Parliament and of the European Council of 26

February 2014 on the harmonization of the laws of the member states relating to the makingavailable on the market of electrical equipment designed for use within certain voltagelimits.

16. Directive 2009/125/EC of the European Parliament and of the European Council of 21October 2009 establishing a framework for the setting of ecodesign requirements forenergy-related products.

17. IEC 62756e1:2015, digital load side transmission lighting control (DLT)dPart 1: basicrequirements.

Further reading

1. OSRAM light-engines-and-modules.2. IEC 63013 (in preparation), LED packagesdLong term luminous flux maintenance

projection.


Annex 1

List of IEC LED related publications

Terms 1. IEC 62504:2014, General lightingdLight emitting diodes (LED)products and related equipmentdTerms and definitions

Designations 2. IEC 61231:2010, International lamp coding system (ILCOS)

Controlgear

Safety 3. IEC 61347-2-13 (plus basic part IEC 61347-1):2006, LampcontrolgeardPart 2e13: Particular requirements for DC. or ACsupplied electronic controlgear for LED modules

Performance 4. IEC 62384:2006, DC or AC supplied electronic controlgear forLED modulesdperformance requirements

DALI 5. IEC 62386e207:2009, Digital addressable lighting interfacedPart207: particular requirements for controlgeardLED modules (devicetype 6)

Efficiency 6. IEC 62442e3:2014, Energy performance of lamp controlgeardPart 3: controlgear for halogen lamps and LED modulesdmethodof measurement to determine the efficiency of the controlgear

Components

Reliability 7. IEC TS 62861 (in preparation), guide to principal componentreliability testing for LED light sources and LED luminaires

Packageperformance

8. (in preparation), LED packagesdPerformance informationrequirements

Maintenanceprojection

9. IEC 63013 (in preparation), LED packagesdlong-term luminousflux maintenance projection

Binning 10. IEC 62707-1:2013, LEDdPart 1: general requirements and whitecolor grid

Intensitymeasurement

11. IEC/TR 61341:2010, method of measurement of center beam in-tensity and beam angle(s) of reflector lamps

Lamp

Safety >50 V 12. IEC 62560:2011, Self-ballasted LED lamps for general lightingservices by voltage >50 Vdsafety specifications

Safety <¼ 50 V 13. IEC 62838:2015, LEDsi lamps for general lighting services withsupply voltages not exceeding 50 V a.c. r.m.s. or 120 V ripple freeDCdsafety specifications

Continued


List of IEC LED related publicationsdcont’d

Performance>50 V

14. IEC 62612:2013, self-ballasted LED lamps for general lightingservices by voltage >50 Vdperformance requirements

Performance<¼ 50 V

15. IEC 63063 (in preparation), semi-integrated LED lamps for generallighting services with supply voltages not exceeding 50 V A C.R.M.S. or 120 V ripple free DCdperformance requirements

Tubular safetynew cap

16. IEC 62931 (in preparation), GX16t-5 capped tubular LEDlampdsafety specification

Retrofit safety 17. IEC 62776:2014, double-capped LED lamps designed to retrofitlinear fluorescent lampsdsafety specifications

Retrofitperformance

18. (in preparation) Double-capped LED lamps designed to retrofitlinear fluorescent lampsdperformance requirement

Module

Safety 19. IEC 62031:2008, LED modules for general lightingdsafetyspecifications

Performance 20. IEC 62717:2014, LED modules for general lightingdperformance requirements

Luminaire

Safety 21. IEC 60598e1:2014, luminairesdPart 1: general requirements andtests

Performance 22. IEC 62722-2-1:2014 (plus basic part IEC 62722-1), luminaireperformancedPart 2-1: particular requirements for LEDluminaires

System

Dimming 23. IEC TR 63037 (in preparation), electrical interface specificationfor self-ballasted lamps and control gear in phase cut dimmedlighting systems

Informationtransfer

24. IEC 62756-1:2015, digital load side transmission lighting control(DLT)dPart 1: basic requirements


Ultraviolet LEDs 17Hideki HirayamaRiken, Saitama, Japan

17.1 Research background of deep ultravioletlight-emitting diodes

The development of semiconductor light sources operating in the deep ultraviolet(DUV) region, such as DUV light-emitting diodes (LEDs) and laser diodes (LDs),is quite an important subject because they are required for a wide variety of applica-tions. Fig. 17.1 summarizes the potential applications of high-efficiency DUV-LEDsand LDs. DUV-LEDs and DUV-LDs with emission wavelengths in the range of230e350 nm are expected to be used in applications such as sterilization, waterpurification, medicine, and biochemistry, and as light sources for high-density opticalrecording, white-light illumination, fluorescence analytical systems, and related

Sterilization High-speed dissociation of

pollutant materials

Laser for high-densityoptical recording

Medical use

Semiconductor illumination

(260–280 nm)

•Skin therapy (315 nm)Determination of cancerposition

Deep-UV light source(260–350 nm)

(Long-lifetime, high color-rendering light)

Wavelength:260–340 nm

High-brightnesswhite light

White lightphosphor

Deep UV-LD(250–300 nm)DVD recorder

using UV-laser

Long lifetime:several tens of

years

Titaniumoxide

Water pollutants:

dioxin, PCB,pesticides UV LED array

260–340 nm

(Clean water)

.Purification of rivers,lakes

.Industrial wasteprocessing

Laserspot

Short wavelength High densityUV DVD diskUV LED array

Other application fields:••••

Sterilization, household air cleanersHigh speed purification of automobile exhaust gassesOptical sensing (luminescence analysis, surface analysis, UV sensing)Chemical and biochemical industry

Figure 17.1 Potential applications of deep ultraviolet (DUV) light-emitting diodes (LEDs) andlight diodes (LDs).


https://doi.org/10.1016/B978-0-08-101942-9.00017-4

information sensing fields. They are also very important for air purification equipmentand for zero emission automobiles.1,2

The wavelength range between 260 and 280 nm is most suitable for sterilization orwater purification with direct UV light. For UV purification using a titanium oxide(TiO2) catalyst, the wavelength range between 320 and 380 nm is also useful. Forwhite LED illumination using UV-LEDs with a mixture of red-green-blue phosphors,wavelengths around 340 nm are considered to be most suitable, taking into accountboth the efficient absorption by the phosphors (<350 nm) and the high-efficiencywavelength range of AlGaN UV-LEDs. For sterilization and several kinds of opticalsensing, DUV-LEDs can be useful even though the process efficiency is not veryhigh. On the other hand, for LED lamps, the efficiency and output power are requiredto be quite high.

Because of their wide direct transition energy range in the UV, covering the regionbetween 6.2 (AlN) and 3.4 eV, AlGaN and quaternary InAlGaN are attractingconsiderable attention as candidate materials for the realization of DUV-LEDs andDUV-LDs.2 Fig. 17.2 shows the relation between the direct transition bandgap energy

IR

Red

Blue

UV

GaN

InN

AIN

Region ofUV-LEDdevelopment

0

1.0

2.0

Ban

dgap

ene

rgy

(eV

)

3.0

4.0

5.0

6.0

1.5 μm1 μm

700 nm

500 nm

400 nm

300 nm

200 nm

Gas lasers

Excimerlasers

308 nm

248 nm

193 nm

KrF

ArF

325 nm XeCIHeCd

257 nmArSHG

Wavelength

3.0 4.0

Lattice constant (Å)

Figure 17.2 Relation between the direct transition bandgap energy and the lattice constant forthe wurtzite (WZ) InAlGaN material system and the lasing wavelengths of various gas lasers.


and the lattice constant for the wurtzite (WZ) InAlGaN material system and the lasingwavelengths of various gas lasers. The main advantages of using AlGaN or InAlGaNfor DUV light sources are: (1) the possibility of obtaining high-efficiency optical emis-sion from quantum wells (QWs), (2) the possibility of producing both p- and n-typesemiconductors in the wide bandgap spectral region, (3) their physical properties,i.e., nitrides are mechanically hard and the devices have long lifetimes and (4) thefact that the materials are free from harmful arsenic, mercury, and lead.2

Fig. 17.3 shows the current status of the external quantum efficiency (EQE) of nitrideUV-LEDs measured at room temperature (RT). Research into AlGaN-based UV-LEDsfor wavelengths shorter than 360 nm, i.e., wavelengths of 330e355 nm,3e5 was initi-ated by several research groups between 1996 and 1999. In the United States, workon deep UV light sources is coordinated by DARPA’s Semiconductor Ultraviolet Op-tical Sources program. The group at the University of South Carolina produced the first250e280-nm AlGaN-based DUV-LEDs between 2002 and 2006.6e8 In 2006, thegroup at NTT Basic Research Laboratories produced the shortest wavelength(210 nm) LED to date using an AlN emitting layer; however, the EQE was quite low(of the order of 2 � 10�6%) because they did not obtain sufficient electron hole confine-ment or efficient carrier injection due to using the highest bandgap AlN emitter.9

We commenced research into AlGaN-based deep UV-LEDs in 1997. We producedthe first efficient DUV (230 nm) photoluminescence (PL) from AlGaN QWs,10 and a333 nm AlGaN-QW UV-LED on SiC in 1999.4 We have also developed high-efficiency UV-LEDs using quaternary InAlGaN-based QWs.2,11,12 We have achieved

0

10

20

30

40

50

Ext

ende

d qu

antu

m e

ffici

ency

(%)

200 250 300 350 400

Wavelength (nm)

NTT210 nm

Nichia

Target

RIKEN:222–352 nm3.8% @ 2701.8% @ 247 DOWA

EQE: 6%

UVcraftoryEQE:5%

SET(2%–4%)

Nitek(2%–4%)Crystal IS(2%–4%)

Shortest LD336 nm

(Hamamatsuphotonics)

Meijo Univ.EQE = 6.7%

@ 345nm

NichiaEQE = 26%@ 365nm

AIGaN InGaN

Figure 17.3 Current status of the external quantum efficiency (EQE) of nitride ultraviolet light-emitting diodes (UV-LEDs) measured at room temperature.

Ultraviolet LEDs 555

a high internal quantum efficiency (IQE) of 47% for 338 nm quaternary InAlGaNQWs and demonstrated several milliwatts of continuous wave (CW) operation with340e350 nm InAlGaN-based UV-LEDs on GaN single-crystal substrates13 andsapphire substrates.14,15

The development of AlGaN-based LEDs to obtain higher efficiency has recentlybecome extremely competitive. We developed a new method for obtaining low thread-ing dislocation density (TDD) AlN templates on sapphire substrates,16 and achievedhigh IQE values (>60%) for AlGaN-QW DUV emission.17 We also achieved highelectron injection efficiencies (EIEs) using a multiquantum barrier (MQB) design18

and produced AlGaN- and InAlGaN-based DUV-LEDs with a wide emission range(222e351 nm) and high EQE (a maximum of 3.8%).19e22 Sensor electronic technol-ogy (SET) have developed commercial DUV-LEDs with wavelengths of240e360 nm.23 Crystal IS has developed DUV-LEDs on AlN single-crystal substratesfabricated by a sublimation method.24 The US companies, SET, Crystal IS, and Nitek,developed high EQE (2%e4%) DUV-LEDs with wavelengths of 260e280 nm duringthe period between 2009 and 2011.23e26 Also, the Japanese companies, UV Craftoryand Nichia, demonstrated EQEs of about 5% and 3% for DUV-LEDs in 2010.27,28 Theshortest wavelength achieved for a DUV-LD is 336 nm,29 which was achieved byHamamatsu Photonics.

The next targets in deep UV device research are to develop EQEs of several tens ofpercent for 220e350 nm DUV-LEDs and to achieve 250e330 nm DUV-LDs. How-ever, the realization of high-EQE DUV-LEDs with wavelengths below 360 nm is stillchallenging, as shown in Fig. 17.3 because of some major problems. The sudden dropin efficiency of DUV-LEDs below 360 nm is mainly due to the following three factors:

• The IQE of AlGaN is quite sensitive to the TDD.• The hole concentration of p-AlGaN is quite low and the EIE is low.• The light extraction efficiency (LEE) is low because of the absorption of UV light in p-GaN

contact layers and in p-electrodes.

The development of low TDD AlN templates on sapphire substrates is very impor-tant because the IQE of AlGaN QWs is as low as 1% if the QWs are fabricated onconventional AlGaN/AlN templates with a high TDD (>2 � 1010 cm�2). We foundthat the IQE of an AlGaN QW can be increased to several tens of percent by reducingthe TDD below 5 � 108 cm�2. In order to fabricate such a low TDD AlN template,innovative crystal growth techniques must be used. We achieved low TDD AlN tem-plates by using an ammonia (NH3) pulse-flow multilayer (ML) growth technique.16

Using low TDD AlN templates, we obtained high-IQE (>60%) emission from AlGaNQWs.

To realize high-efficiency UV-LEDs and UV-LDs, quaternary InAlGaN with a fewpercent of indium is considered to be quite effective because efficient DUV emissioncan be obtained due to the indium-segregation effect.11,12 The emission from a local-ized electronehole pair in the indium segregation region of quaternary InAlGaN isquite effective for obtaining a high IQE. Due to this effect, quaternary InAlGaN isvery promising as the active layer of 200e350 nm band LDs and LEDs. The advantage


of InAlGaN is that the emission efficiency of quaternary InAlGaN is less sensitive tothe TDD due to the effect of localized carriers in the indium segregation area.

The device properties of AlGaN DUV-LEDs strongly depend on the properties ofthe p-AlGaN. The hole concentration of p-AlGaN with high aluminum content(aluminum > 60%) is quite low (as low as 1014 cm�3) due to its very deep acceptorlevel, which is reported to be 240 meV (GaN) to 670 meV (AlN). As a result, theEIE of a DUV-LED is reduced due to the leakage of electrons to the p-side layers.The high series resistance of p-type layers is also a problem for the device properties.We demonstrated that the leakage of electrons can be dramatically suppressed byreflecting the electrons using an MQB design.

Due to the lack of high-hole density p-type AlGaN, we must use p-GaN contactlayers. A p-GaN contact layer causes a significant reduction in LEE due to the strongabsorption of UV light by the p-GaN layer. The value of LEE for a DUV-LED is typi-cally below 10% due to absorption through the p-GaN contact layer, as well as the lackof a transparent p-type electrode or a highly reflective electrode that can be used in theDUV spectral region.

The current value of EQE for 270 nm DUV-LEDs from our group is approximately4%, which is determined by 60% IQE, 80% EIE, and 8% LEE. Further improvementsin EQE are expected as we move toward the production of commercial DUV-LEDs.Techniques for increasing each of these efficiencies are described in the followingsections.

17.2 Growth of low TDD AlN layers on sapphire

In order to realize high-efficiency DUV-LEDs, it is necessary to develop a low TDDAlGaN/AlN template. The TDD of a conventional AlN buffer layer on a sapphire sub-strate, fabricated using a low-temperature-AlN buffer, was greater than2 � 1010 cm�2. On the other hand, a TDD of 108e109 cm�2 is required in order toobtain high IQE values of several tens of percent from AlGaN QWs. Several fabrica-tion methods have been used to obtain high-quality AlN buffers; for example, the useof AlN/AlGaN superlattices (SLs) grown with alternating gas feeds,6 AlGaN bufferlayers deposited by epitaxial lateral overgrowth (ELO)30 and a combination of GaN/AlN SLs and AlGaN produced by alternate source-feeding epitaxy on SiC.31

We grew AlN layers directly onto sapphire substrates at a high growth temperatureafter initial nitridation with NH3. The growth temperature was around 1300�C, and theV/III ratio was relatively low. Fig. 17.4 shows the relation between the full width athalf maximum (FWHM) of X-ray diffraction (10�12) and (0002) u-scan rockingcurves (XRCs) and the nitridation time of AlN layers grown on sapphire substrates.As the nitridation time increased from 5 to 10 min, the FWHM of (10�12) XRCreduced to 560 arcsec. The value of the FWHM of (10�12) XRC corresponds tothe edge-type TDD. We found that larger AlN nuclei formed in the initial stages ofthe growth process with longer nitridation times, and that edge dislocations werereduced by embedding them in a thick AlN layer. However, heavy nitridation on sap-phire caused an inversion from aluminum to nitrogen polarity, which led to the


generation of abnormally large nuclei on the AlN surface. We also found that a longnitridation time led to cracks on the AlN surface.

It is necessary to satisfy several conditions to achieve high-quality AlGaN/AlNtemplates that can be used for DUV emitters, i.e., a low TDD, crack-free, atomicallyflat surfaces and stable aluminum (þc) polarity. We used an ammonia (NH3) pulsed-flow ML growth method to fabricate AlN layers on sapphire. Fig. 17.5 shows the gasflow sequence and a schematic view of the growth control method with pulsed- andcontinuous-flow gas feeding.

The samples were grown on sapphire (0001) substrates by low-pressure metal-organic chemical vapor deposition (LP-MOCVD). First, an AlN nucleation layerand a “burying” AlN layer were deposited, both by NH3 pulsed-flow growth. Thetrimethylaluminum (TMAl) flow was continuous during the NH3 pulsed-flowsequence, as shown in Fig. 17.5. Low-TDD AlN can be achieved by promotingcoalescence in the AlN nucleation layer. After the growth of the first AlN layer, thesurface is still rough because of the low growth rate during the pulsed-flow mode.We added a high-growth-rate continuous-flow mode to reduce the surface roughness.By repeating the pulsed- and continuous-flow modes, we obtained a crack-free, thickAlN layer with an atomically flat surface. NH3 pulsed-flow growth is effective forobtaining high-quality AlN because of the enhancement of precursor migration.Furthermore, it is effective for obtaining stable aluminum (þc) polarity, which isnecessary for suppressing polarity inversion from aluminum to nitrogen, by maintain-ing aluminum-rich growth conditions.

400

600

800

1000

0 5 10 15 20

Nitridation time (min)

560 arcsec

40

60

80

100

120

140

160

(000

2) (a

rcse

c)

(101

2) (a

rcse

c)–

XRC FWHM of AIN

Sapphire

AIN (3.3 μm)

Figure 17.4 Full width at half maximum (FWHM) of X-ray diffraction (10e12)u-scan rockingcurves (XRC) for AlN layers grown on sapphire substrates as a function of initial nitridationtime.


Figure 17.5 Gas flow sequence and the growth control method used for an NH3 pulsed-flow multilayer (ML)-AlN growth technique.

Ultraviolet

LEDs

559

As described above, we used three growth modes. The initial deposition, whichfabricated an AlN nucleation layer, was through NH3 pulsed-flow growth with agrowth pressure of 200 Torr and a temperature of 1300�C, with an average V/III ratioof approximately 60. Following this, migration enhancement epitaxy, which was usedfor the coalescence process, was performed using NH3 pulsed-flow growth at 76 Torrand 1200�C with an average V/III ratio of approximately 750. An AlN layer was thengrown at a high rate using a conventional continuous flow process at 76 Torr and1200�C with a V/III ratio of approximately 25. The growth rates in the pulsed- andcontinuous-flow modes were approximately 0.6 and 6 mm/h, respectively.

The advantage of using ML-AlN for a DUV-LED is that a low TDD AlN can beobtained without the need for AlGaN layers, yielding a device structure with minimalDUV absorption. An AlGaN-free buffer is believed to be important for realizing sub-250 nm band high-efficiency LEDs.

Fig. 17.6 shows the FWHM of the X-ray diffraction (10�12)u-scan rocking curves(XRC) for various stages of ML-AlN growth. The FWHM of XRC (10�12) for AlNreduced from 2160 to 550 arcsec with two repetitions of the NH3 pulsed–flowML-AlNgrowth. Fig. 17.7 shows atomic force microscope (AFM) images of the surface ofML-AlN on sapphire after various stages of ML-AlN growth. We can observe thatthe surface was improved by growing MLs of AlN and we can confirm that therewas an atomically flat surface. The root mean square of the surface roughness of theML-AlN layer obtained from the AFM image was 0.16 nm.

Fig. 17.8 shows (a) the schematic structure and (b) the cross-sectional transmissionelectron microscope (TEM) image of an AlGaN/AlN template with a five-step ML-AlN buffer layer grown on a sapphire substrate. The total thickness of the ML-AlNbuffer was typically 4 mm. The minimum values of the FWHM of the (0002) and(10�12) XRCs of the ML-AlN were approximately 150 and 250 arcsec, respectively.

Continuous-flow AIN

Productionof nucleation

AIN layer

Productionof nucleation

AIN + 1

1 + 2 1 + 2 + 3 + 4

0

1000

2000

3000

4000

5000

6000

XR

D (1

0–12

) ω-s

can

FWH

M (a

rcse

c)

MultilayerAIN

1

2

3

4 Continuous-flow AIN 1 μm

Continuous-flow AIN 1 μm

NH3 pulse-flow AIN 0.3 μm

NH3 pulse-flow AIN 0.3 μm

Nucleation AIN layerSapphire substrate

Figure 17.6 Reduction in the full width at half maximum (FWHM) of the X-ray diffraction(10e12) u-scan rocking curve (XRC) for various stages of multilayer (ML)-AlN growth.


The minimum values of the edge- and screw-type dislocation densities of the ML-AlNwere approximately 3 � 108 and 4 � 107 cm�2, respectively, as measured from thecross-sectional TEM image.

17.3 Marked increases in IQE

We observed a remarkable enhancement of the DUV emission of AlGaN QWs byfabricating them on low TDD AlN templates.17 Fig. 17.9 shows a cross-sectionalTEM image of the QW region of an AlGaN multiquantum well (MQW) DUV-LEDwith an emission wavelength of 227 nm fabricated on an ML-AlN buffer. We useda thin QW to obtain a high IQE because it suppressed the effects of the polarizationfield in the well. This is believed to be particularly important for obtaining theatomically smooth heterointerfaces that are necessary for a high IQE from such athin QW. The atomically flat heterointerfaces of the 1.3 nm thick three-layer QWsare shown in the cross-sectional TEM image.

Fig. 17.10 shows PL spectra of AlGaN QWs fabricated on ML-AlN templates withvarious values of XRC (10�12) FWHM, as measured at RT. The peak emission wave-lengths of the QWs were around 254 nm. The QWs were excited with a 244 nm argonion second harmonics generation (SHG) laser. The excitation power density was fixedat 200 W/cm2. The PL emission intensity of the AlGaN QWs was significantlyincreased by improving the XRC (10�12) FWHM, as shown in Fig. 17.10. We cansee from Fig. 17.10 that the emission efficiency of AlGaN depends strongly on theedge-type TDD.

Fig. 17.11 shows the PL peak intensity as measured at RT for 254-nm-emission AlGaNQWs as a function of XRC (10�12) FWHM. The PL intensity increased by approximately100 times by reducing the XRC (10�12) FWHM from 1400 to 500 arcsec. The PL inten-sity increased rapidly when the FWHM of the XRC was reduced to 500e800 arcsec. Therapid increase in the PL intensity can be explained by a reduction of the nonradiativerecombination rate as the distance between threading dislocations becomes greatercompared with the carrier diffusion length in the QW.We obtained a similar enhancementof the emission from AlGaN QWs with QWs of various wavelengths.

RMS 21.4 nm 8.2 nm 1.63 Å

1 + 2 + 3 + 41 + 2Production of

nucleation AIN

Figure 17.7 Atomic focal microscopy (AFM) images of the surface of multilayer (ML)-AlNwith an area of 5 � 5 mm2 square after various stages of growth.


MultilayerAIN buffer(5-step)3.8 μm

5-stepmultilayerAIN buffer

3.8 μm

NH3 pulsed-flow AIN0.18 μm

AI0.88Ga0.12N0.2 μm

AI0.76Ga0.24N; Si0.2 μm

AI0.88Ga0.12N; Si

AI0.76Ga0.24N 2.45 μm

Sapphire substrate

Nucleation AIN layer(NH3 pulsed flow)

Continuous-flow AIN 0.56 μm





LED layers

Sapphire1 μm

(a)

(b)

Figure 17.8 (a) Structure and (b) cross-sectional transmission electron microscope (TEM)image of an AlGaN/AlN template with a five-step multilayer (ML)-AlN buffer layer grown ona sapphire substrate.


FWHM of XRC(10–12) ω-scan

1410arcsec

899arcsec

501 arcsec

571 arcsec

AlGaN QW

λ = 255 nm

240 260 280 300 320Wavelength (nm)

PL

inte

nsity

(a.u

.)

103

104

105

106

107

Figure 17.10 Photoluminescence (PL) spectra of AlGaN quantum wells (QWs) on multilayer(ML) AlN templates with various values of u-scan rocking curve (XRC) (10e12) full width athalf maximum (FWHM).

10 nm

p-AI0.98Ga0.02Ne-blocking layer(15 nm)

AI0.87Ga0.13Nbarrier (21 nm)

AI0.79Ga0.21Nwell (1.3 nm)/AI0.87Ga0.13Nbarrier (7 nm)3-layer MQW

n-AI0.87Ga0.13N

Figure 17.9 Cross-sectional transmission electron microscope (TEM) image of the quantumwell region of an AlGaN multiquantum well (MQW) deep ultraviolet (DUV)-light-emittingdiode (LED).


Fig. 17.12 shows the temperature dependence of the integrated PL intensitymeasured for an AlGaN MQW with an emission wavelength at 288 nm fabricatedon an ML-AlN template. The IQE can be roughly estimated from the temperaturedependence of the integrated PL intensity if we assume that the nonradiative

0 100 200 300

Temperature (K)

0

0.2

0.4

0.6

0.8

1

Nor

mal

ized

PL

inte

grat

ed in

tens

ity

250 350300 400Wavelength (nm)

PL

inte

nsity

(a.u

.)

λ = 288 nm

AlGaN/AlGaN QWon AIN/sapphire

Excited withAr-SHG laser(λ = 257 nm)200 W/cm2)

Figure 17.12 Temperature dependence of integrated photoluminescence (PL) intensity for anAlGaN three-layer multiquantum well (MQW) grown on an multilayer (ML)-AlN template.

0 500 1000 1500

FWHM of XRC (10–12) ω-scan

PL

inte

nsity

(a.u

.)

Measured at RT

λ = 255 nmAlGaN QWon ML-AIN

Figure 17.11 Photoluminescence (PL) intensity of AlGaN quantumwells (QWs) as a function ofu-scan rocking curve (XRC) (10e12) full width at half maximum (FWHM) of AlGaN buffers.


recombination rate is quite low at low temperature. The estimated IQE for the 288-nm-emission AlGaN QW was approximately 30% at RT.

The quaternary alloy InAlGaN is attracting considerable attention as a candidatematerial for realizing DUV-LEDs because efficient UV emission as well as higherhole concentrations can be realized2 due to indium incorporation effects. The incorpora-tion of a few percent of indium into AlGaN is considered to be quite effective for obtain-ing a high IQE, because efficient DUV emission can be obtained due to the indiumsegregation effect,11,12 which has already been investigated for the ternary InGaN alloy.

Fig. 17.13 show a schematic image of a band diagram of a quaternary InAlGaNalloy with radiative recombination by localized carriers into the indium segregationregion, and a surface cathodoluminescence (CL) image obtained from a quaternaryInAlGaN layer.2 Emission fluctuations in the submicron region are clearly visible inthe surface CL image. The dark spots in the bright area of the surface CL imagewere confirmed to correspond to the position of threading dislocations (TDs). Thefluctuations in the emission are considered to be due to carrier localization in theindium segregation area. Surface CL images obtained for quaternary InAlGaN werevery similar to those obtained for InGaN films. Electronehole pairs localized in the

TD

In-rich area

1 μm

Figure 17.13 Band diagram of a quaternary InAlGaN alloy and radiative recombination bylocalized carriers into the indium-segregation region, and a surface cathodoluminescence (CL)image obtained from a quaternary InAlGaN layer.


low-potential valley emit before they are trapped in nonradiative centers induced bydislocations. Therefore, the advantage of indium incorporation is that the emissionefficiency is less sensitive to the TDD.

Fig. 17.14 shows the temperature dependence of the integrated PL intensity measuredfor an InAlGaN/InAlGaN MQW with an emission wavelength of 338 nm fabricated ona high-temperature (HT)-AlN buffer on sapphire. The TDD of the HT-AlN was approx-imately 2 � 1010 cm�2. The estimated IQE from Fig. 17.14 was approximately 47% atRT. We found that a high IQE can be obtained for InAlGaN QWs in the wavelengthrange 310e380 nm, even when using a high TDD template.2,11,12


18K35K50K73K95K110K130K150K170K190K210K230K250K270K290KP

L in

tens

ity (a

.u.)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Nor

mal

ized

PL

inte

nsity

InAlGaN/InAlGaN3-layer MQW

AlGaN

HT-AIN

Sapphire

47% at RT

Excited withAr-SHG laser(λ = 257 nm)500 W/cm2

Temperature (K)0 50 100 150 200 250 300

Figure 17.14 Temperature dependence of integrated photoluminescence intensity measured foran InAlGaN/InAlGaN multiquantum well (MQW) with an emission wavelength of 338 nmfabricated on a high-temperature (HT)-AlN buffer on sapphire.


Then we took up the challenge of developing crystal growth to produce high-quality InAlGaN alloys emitting at the “sterilization” wavelength (280 nm).21 Thecrystal growth of high-aluminum-composition quaternary InAlGaN is relativelydifficult, because indium incorporation becomes more difficult with increasing growthtemperature, which is required to maintain the crystal quality of high-aluminum-content AlGaN. We achieved high-quality quaternary InAlGaN layers with high-aluminum-content (>45%) using epitaxy with a relatively low growth rate, i.e.,0.03 mm/h. The emission intensity of a 280 nm band quaternary InAlGaN QW atRT increased by five times on reducing the growth rate from 0.05 to 0.03 mm/h.

Fig. 17.15 shows the PL spectra of a quaternary InAlGaN QWmeasured at 77K andat RT. We obtained extremely high intensity PL emission at RT. The ratio of the in-tegrated intensity of the RT PL against the 77K PL was 86%. Thus, high IQE was ob-tained from the quaternary InAlGaN-QW at RT. Fig. 17.16 summarizes thewavelength dependence of the ratio of the integrated PL intensity (PL measured atRT against PL measured at low temperature, usually 10K), which is directly relatedto the IQE. The IQE of 340 nm band InAlGaN QWs is 30%e50%, even with ahigh TDD template (2 � 1010 cm�2). However, the IQE was below 2% for short wave-length (280 nm) QWs. On the other hand, we achieved high values of IQE by fabri-cating QWs on low TDD ML-AlN templates. The ratios of the integrated PLintensity for 280 nm band QWs were approximately 30% and 86% for an AlGaNQW and an InAlGaN QW, respectively.

PL

RT(λ = 291 nm)

77K(λ = 284 nm)

4 4.5Energy (eV)

PL

inte

nsity

(a.u

.)

InAlGaN/InAlGaN-2QW

Excited byAR-SHG laser

(244 nm)200 W/cm2

Figure 17.15 Photoluminescence (PL) spectra of a quaternary InAlGaN quantum well (QW)measured at 77K and at room temperature (RT).


17.4 Aluminum gallium nitride-based DUV-LEDsfabricated on high-quality aluminum nitride

Aluminum gallium nitride (AlGaN) and quaternary InAlGaN DUV-LEDs were fabri-cated on low TDD ML-aluminum nitride (AlN) templates.16e22 Fig. 17.17 shows aschematic of the structure and emission of an AlGaN-based DUV-LED fabricatedon a sapphire substrate. Table 17.1 shows typical design values for the aluminum

250 300 350 400 450 500Wavelength (nm)

0

10

20

30

40

50

60

70

80

90

100

InAIGaN QW(using ML-AIN)

InAIGaN QWs(usual AIN buffer)

InGaN QW(usual GaN buffer)

AIGaN QW(using ML-AIN)

ML-AIN:Usual AIN:

TDD(edge) ∼7 × 108 cm–2

TDD(edge) ∼2 × 1010 cm–2R

atio

of i

nteg

rate

d P

L in

tens

ityP

L(R

T)/P

L(LT

) (%

)

Figure 17.16 Wavelength dependence of the ratio of the integrated photoluminescence (PL)intensity (PL measured at room temperature [RT] against PL measured at low temperature).

Ni/Au

ML-AIN buffer(NH3 pulsed-flow method)

Sapphire substrate

UV output

n-AI0.75Ga0.25N;Sibuffer (2 μm)

AI0.75Ga0.25N/AI0.60Ga0.40N3-layer MQW

AI0.75Ga0.25N;Mg

GaN;Mg

AI0.95Ga0.05N;Mge-blocking layer

Ni/Au electrode

Figure 17.17 AlGaN-based deep ultraviolet (DUV) light-emitting diode (LED) fabricated on asapphire substrate and UV emission.


composition (x) in the AlxGa1�xN wells, the buffer and barrier layers, and the electronblocking layers (EBLs) that were used for the 222e273 nm AlGaN-MQW LEDs.High-aluminum-composition AlGaN layers were used to obtain short-wavelengthDUV emissions. A typical LED structure consisted of an approximately 4-mm-thickundoped ML-AlN buffer layer grown on sapphire, a 2-mm-thick silicon-dopedAlGaN buffer layer, followed by a three-layer undoped MQW region consisting of1.3-nm-thick AlGaN wells and 7-nm-thick AlGaN barriers, a 20-nm-thick undopedAlGaN barrier, a 15-nm-thick magnesium-doped AlGaN EBL, a 10-nm-thickmagnesium-doped AlGaN p-layer and an approximately 20-nm-thick magnesium-doped GaN contact layer. The QW thickness varied within the range 1.3e2 nm.Thin QWs are preferable for AlGaN QWs to suppress the effect of the large piezoelec-tric fields in the well. Ni/Au electrodes were used for both the n-type and p-type elec-trodes. The typical size of the p-type electrode was 300 � 300 mm2. The output powerthat radiated into the back of the LED was measured using a silicon photodetectorlocated behind the LED sample, which was calibrated to measure the luminous fluxfrom LED sources using an integrated-spheres system. The LEDs were measured un-der bare wafer or flip-chip conditions. The forward voltages of the bare wafer and theflip-chip samples were 20e30 V and 7e10 V, respectively.

Fig. 17.18 shows the electroluminescence (EL) spectra of the fabricated AlGaN andInAlGaN MQW LEDs with emission wavelengths of 222e351, all measured at RTwith an injection current of around 50 mA. As can be seen, single-peak operationwas obtained for each sample. The deep level emission was negligible for every LED.

Fig. 17.19 shows the EL spectra of a 227 nm AlGaN LED on a log scale. We ob-tained single-peaked EL spectra, even for sub-230 nm wavelength LEDs. The deeplevel emissions with wavelengths at around 255 and 330e450 nm were more thantwo orders of magnitude smaller than the main peak. These peaks may correspondto deep level emissions associated with magnesium acceptors or other impurities.The output power of the 227 nm LED was 0.15 mW at an injection current of30 mA, and the maximum EQE was 0.2% under RT pulsed operation. The pulse widthand the repetition frequency were 3 ms and 10 kHz, respectively.

Table 17.1 Typical design values for the aluminum composition (x) inAlxGa1LxN wells, and buffer, barrier, and electron blocking layers(EBLs) for 222e273 nm AlGaN MQW LEDs

Wavelength (nm) Well (x) Barrier and buffer layers (x) EBL (x)

222 0.83 0.89 0.98

227 0.79 0.87 0.98

234 0.74 0.84 0.97

248 0.64 0.78 0.96

255 0.60 0.75 0.95

261 0.55 0.72 0.94

273 0.47 0.67 0.93


Fig. 17.20 shows the EL spectra for various injection currents, and the currentversus output power (IeL) and EQE (hext) characteristics for a 222 nm AlGaNMQW LED measured under RT pulsed operation. Single-peaked operation was real-ized: this is the shortest reported wavelength for an AlGaN LED on a sapphire

200 250 300 350 400 450

Nor

mal

ized

inte

nsity

AlGaN QWDUV-LEDs

DUV-LEDs

222 nm Pulsed227 nm Pulsed234 nm CW240 nm CW248 nm CW255 nm CW261 nm CW

282 nm CW342 nm CW351 nm CW

InAlGaN QW

Measured at RT

Wavelength (nm)

Figure 17.18 Electroluminescence (EL) spectra of fabricated AlGaN and InAlGaNmultiquantum well (MQW) light-emitting diodes (LEDs) with emission wavelengths of222e351 nm, all measured at room temperature (RT) with injection currents of around 50 mA.

RT pulsed

45 mA

20 mA

4 mA

AlGaN QW UV-LEDon AIN/sapphire

λ = 227 nm

200 300 400

Wavelength (nm)

EL

inte

nsity

(a.u

.)

Figure 17.19 Electroluminescence (EL) spectra on a log scale of a 227 nm AlGaN light-emitting diode (LED).


substrate. The output power of the 222 nm LED was 14 mW at an injection current of80 mA, and the maximum EQE was 0.003% under RT pulsed operation.

Fig. 17.21 shows (a) current versus output power (IeL) and (b) current versus EQE(hext) for 250 nm band AlGaN MQW LEDs under RT CW operation. We fabricatedtwo types of samples with different aluminum compositions in the AlGaN EBLs,one at 90% and the other at 95%. The corresponding barrier heights of the EBLs inthe conduction band were 280 and 420 meV, respectively. As seen in Fig. 17.21,the EQE of the LEDs was significantly increased with a higher electron blockingheight. This indicates that electron overflow is significantly reduced due to electronreflection by the EBL, and therefore the EIE into the QW is increased. The maximum

80 mA60 mA40 mA20 mA

λ = 222 nmRT pulsed

RT pulsed


EL

inte

nsity

(a.u

.)

0.01

0.008

0.006

0.004

0.002

0806040200

0

5

10

15AlGaN MQW DUV-LED

on AIN/sapphire

Current (mA)

Out

put p

ower

(μW

)

EQ

E (%

)

Figure 17.20 Electroluminescence (EL) spectra for various injection currents, and currentversus output power (IeL) and external quantum efficiency (EQE) (hext) characteristics for a222 nm AlGaN multiquantum well (MQW) light-emitting diode (LED) measured under roomtemperature (RT) pulsed operation.


output power and EQE were 2.2 mW and 0.43%, respectively, for an LED with anemission wavelength of 250 nm under RT CW operation.

Fig. 17.22 shows the wavelength dependence of the output power of 245e260 nmAlGaN MQW LEDs, for various edge-type TDDs of the AlN templates and electronbarrier heights of the EBLs. A marked increase in EQE was observed on reducing theTDD and increasing the EBL height. The EQE of the 250 nm band LED increasedfrom 0.02% to 0.4% and the output power increased by more than 30 times onreducing the TDD from 3 � 109 cm�2 to 7 � 108 cm�2. We also found that a higherelectron blocking height is effective for obtaining high output power.

E-blocking layer1 Al0.9Ga0.1N2 Al0.95Ga0.05N

ML-AIN

Δ Ec = 280 meVΔ Ec = 420 meV

n-AlGaN

MQW p-GaN

0.5

0.4

0.3

0.1

0.2

00

00 100100 200200 300300 400400

Current (mA)Current (mA)

RT CW

RT CW

EQ

E g

ext (

%)

ΔΕEBL = 420 meV

ΔΕEBL = 420 meV

ΔΕEBL : blocking height of EBL

ΔΕEBL = 280 meV

ΔΕEBL = 280 meV

λ = 250 nm

λ = 250 nm

λ = 254 nm

λ = 254 nm

1

2

Out

put p

ower

(mW

)

AlGaN-3MQWDUV-LED

on ML-AIN

(a) (b)Figure 17.21 (a) Current versus output power (IeL) and (b) current versus external quantumefficiency (EQE) (hext) for 250 nm-band AlGaN-multiquantum well (MQW) light-emittingdiodes (LEDs) under room temperature (RT) continuous wave (CW) operation.


Fig. 17.23 shows the EL spectra of 225 nm band AlGaN QW DUV-LEDs withvarious QW thicknesses, as measured under RT pulsed operation. The well thicknesseswere in the range 1.6e4 nm. Intense emission was obtained for the thin QWs. Fromthis experiment, we confirmed that thin QWs are suitable for AlGaN QWs becausethey suppress the effect of the large piezoelectric fields.

It has been reported that emission in the normal c-axis direction (vertical emission)is difficult to obtain from an AlN (0001) or a high-aluminum-content AlGaN surface,because the optical transition between the conduction band and the top of the valenceband is mainly only allowed for light that has its electric field parallel to the c-axis di-rection of AlN (E/c).7 The suppression of the vertical emission is a significant problemfor AlGaN-based DUV-LEDs, because it results in a significant reduction in the LEE.Several groups have reported that vertical c-axis emission is suppressed for high-aluminum content AlGaN QWs.32,33 Banal et al. showed that the critical aluminumcomposition for polarization switching could be expanded to approximately 0.82 byusing a very thin (1.3 nm) QW, when AlGaN-QW was fabricated on an AlN/sapphiretemplate.32

Fig. 17.24 shows the radiation angle dependence of the emission spectra of a222 nm AlGaN QWLED on AlN/sapphire. We demonstrated that normal c-axis-direc-tion emission (vertical emission) can be obtained for short wavelength (222 nm) LEDs,even when the aluminum composition range of the AlGaN QW was as high as 83%.20

We fabricated quaternary InAlGaN-based DUV-LEDs to increase the IQE and EIEof DUV-LEDs. Fig. 17.25 shows the schematic structure and a cross-sectional TEMimage of an InAlGaN QW DUV-LED. We confirmed that the surface roughness of

TDD = 3 × 109 cm–2

TDD = 7 × 108 cm–2

TDD = 7 × 108 cm–2

ΔΕEBL = 420 meV

ΔΕEBL = 280 meV

ΔΕEBL = 420 meV

RT CW0.01

0.1

1

245 255 260 265250Wavelength (nm)

EQ

E

ext(

%)

Figure 17.22 Wavelength dependence of external quantum efficiency (EQE) (hext) of245e260 nm AlGaN multiquantum well (MQW) light-emitting diodes (LEDs) for variousedge-type threading dislocation densities (TDDs) of the AlN template and electron barrierheights of the electron blocking layer (EBL).


the InAlGaN layer was significantly improved by introducing a silicon-doped InAl-GaN buffer layer. The InAlGaN-based DUV-LED is considered to be attractive forachieving high EQE due to the higher IQE and higher hole concentration obtainedby indium segregation effects. Fig. 17.26 shows the EL spectrum and the currentversus output power (IeL) and EQE of an InAlGaN-based QW DUV-LED with anemission wavelength of 282 nm. The maximum output power and EQE were10.6 mW and 1.2%, respectively, under RT CW operation. From these results, wefound that quaternary InAlGaN QWs and p-type InAlGaN are quite useful forachieving high-efficiency DUV-LEDs.

hv hv

AlGaN MQW DUV-LEDs on AIN/sapphire

Well thickness = 1.6 nm (l = 40 mA)

2.6 nm (l = 40 mA)

4 nm (l = 80 mA)

Measured at RTpulsed

EL

inte

nsity

(a.u

.)

210 220 230 240 250 260 270

Wavelength (nm)

Figure 17.23 Electroluminescence (EL) spectra of 225 nm-band AlGaN-quantum well (QW)deep ultraviolet (DUV) light-emitting diodes (LEDs) with various quantum well thicknesses,measured under room temperature (RT) pulsed operation.


AlGaN QW LED222 nm

220 240 260 280 300200

Wavelength (nm)

Inte

nsity

(a.u

.) 30°

0°

50°

60°

70°

θ

θ

=

Figure 17.24 Radiation angle dependence of the emission spectra for a 222 nm AlGaNquantum well (QW) light-emitting diode (LED).

p-InAlGaNp-InAlGaN

p-InGaN contact

n-AlGaN

InAlGaN well

i-InAlGaN interlayer

InAlGaN:Si (20 nm)

InAlGaN cap (10 nm)

e-block layer (7 nm)

(1.7 nm)

(7 nm) 2QW/InAlGaN:Si barrier

(3 nm)

Sapphire (0001)

AIN (NH3 pulsed-flowmultilayer growth)

10 nm

Figure 17.25 Structure and cross-sectional transmission electron microscopy (TEM) image ofan InAlGaN quantum well (QW) deep ultraviolet (DUV) light-emitting diode (LED).


17.5 Increase in EIE and LEE

High IQEs of 30%e80% have been realized for AlGaN and InAlGaN QWs with emis-sion wavelengths of 220e350 nm.17,21 The low EQE figures for AlGaN DUV-LEDscompared with those for InGaN blue LEDs are a result of low EIE into the QWs due toelectron leakage caused by low hole concentrations in the p-type AlGaN layers, as wellas inferior light extraction efficiencies (lower than 8%) due to strong UV absorption inthe p-GaN contact layer and the p-side electrode. The EIEs for 250e280 nm bandAlGaN-based DUV-LEDs were roughly estimated to be 10%e30% by numerical cal-culations.2 We added a multiquantum barrier (MQB) as an EBL in an AlGaN QWLED and consequently achieved a marked increase in EIE.18

Fig. 17.27 shows a schematic image of the enhancement of EIE caused by using anMQB in AlGaN DUV-LEDs. In order to obtain a high EIE, an EBL is effective forsuppressing the overflow of electrons above the QW into the p-type AlGaN layers.A high barrier is required for an EBL in order to obtain a sufficiently high EIE. Wetried using AlN or high-aluminum-composition (aluminum > 0.95) AlGaN layersfor the EBL18e21; however, the barrier height of these EBLs was still not sufficientto obtain the desired high EIE. Indeed, the EIE was particularly low for short-wavelength AlGaN LEDs (<250 nm), because the electron-barrier heights of theEBLs in these devices was low in relative terms. This material limitation can be over-come by enhancing the effective barrier height through the introduction of an MQB,which causes multireflection effects in the wave functions. The MQB was predictedtheoretically by Iga et al. in 1986,34 and the effects were demonstrated experimentallyin GaInP/AlInP red LDs in 1991.35 It has been reported that the effective electronbarrier of an MQB in comparison with a bulk potential barrier is as much as 30%higher for GaAs/AlAs and 50% higher for GaInAs/InP MQBs. It is believed that


Inte

nsity

(a.u

.)

InAlGaN-2QWUV-LED

RT CW

l = 50 mAPeak = 282 nm 10

5

00

0

0.5

1.5

100 200 300

1

Current (mA)

RT CW

InAlGaN-2QW-LEDon AIN/AIGaN

/sapphire substrate

λ = 282 nm

Out

put p

ower

(mW

)

Ext

erna

l qua

ntum

effi

cien

cy (%

)

Figure 17.26 Electroluminescence (EL) spectrum and current versus output power (IeL) andexternal quantum efficiency (EQE) of an InAlGaN-based quantum well (QW) deep ultraviolet(DUV) light-emitting diode (LED) with emission wavelength at 282 nm.


AlN/AlGaN or AlGaN/AlGaN MQBs would be quite effective for increasing theeffective barrier height of an EBL and, as a result, would contribute to the realizationof high EQE AlGaN DUV-LEDs.

Fig. 17.28 shows a conduction band diagram (right) and the electron transmittance(left) for an AlGaN/AlGaN MQB and a conventional single-barrier EBL calculated fora 250 nm band AlGaN QW LED. The multireflection effects in the heterostructureswere analyzed using a transfer matrix method. It was shown that the effective electronbarrier of an MQB in comparison with a conventional single-barrier EBL was as muchas twice as high for an AlGaN/AlGaN MQB using barriers with thickness modulation.

Fig. 17.29 shows the schematic structure and a cross-sectional TEM image of afabricated 250 nm AlGaN QW DUV-LED with an MQB. We investigated an

Electron

Electron

MQB1

1

–1

–1

–20

–20

0

0

0

0

20

20

40

40

Ene

rgy

E (e

V)

Ene

rgy

E (e

V)

Si-doped

Si-doped

Mg-doped

Mg-doped

Undoped

Undoped

Distance (nm)

Distance (nm)

Leakage

Conventionalsingle barrier

Figure 17.27 Enhancement of electron injection efficiency by using an MQB in AlGaN deepultraviolet (DUV) light-emitting diodes (LEDs).


1

1

–100

0

100.5 20 30 40 50

Distance (nm)

Modulated MQB(Barrier/well)

Al0.95Ga0.05N/Al0.70Ga0.30N

3.8/0.5/2.5/0.5/2.5/0.5/2.5/0.5/2.5/0.5/1.8/0.5/1.8/0.5/1.8/0.5/1.8/0.5/1.8/0.5/1.3/0.5/1.3/0.5/1.3/0.5/0.8/0.5/0.8/0.5/0.8/0.5/ (nm)

Transmittance T

2

Ene

rgy

E (e

V)

Figure 17.28 Conduction band diagram (right) and electron transmittance (left) of an AlGaN/AlGaN multiquantum barrier (MQB) and a conventional single-barrier electron blockinglayers (EBL) calculated for a 250 nm-band AlGaN quantum well (QW) light-emitting diode(LED).

p-GaN;Mgcontact layerp-Al0.77Ga0.23N;Mg

n-Al0.77Ga0.23N;Si

Al0.77Ga0.23N;Mg

Al0.77Ga0.23N;Mg

Al0.62Ga0.38N(1.5 nm)/Al0.77Ga0.23N(6 nm)

Al0.95Ga0.0.5N;Mg/

5-layer MQB

3-layer QW

buffer layer

UV output50 nm

Sapphire (0001)

Multilayer AIN (NH3pulse flow growth)

Ni/Aup-electrodeNi/Au

n-electrode

Figure 17.29 Structure and cross-sectional transmission electron microscope (TEM) image of250 nm AlGaN deep ultraviolet (DUV) light-emitting diode (LED) with an multiquantumbarrier (MQB).


appropriate MQB structure experimentally for use with 250 nm based DUV-LEDs.We found that the insertion of an initial thick barrier is important for reflecting lower-energy electrons. We also found that thin barriers contribute to the reflection of higherenergy electrons. The optimized MQB structure for a 250 nm AlGaN QW LED was afive–layer Al0.95Ga0.05N/Al0.77Ga0.23N MQB with thicknesses of 7/4/5.5/4/4/2.5/4/2.5/4 nm, in which the bold letters are for the barriers and the normal letters are forvalleys. The total thickness of the MQB should be less than 40 nm, because a coherentlength exists for the multireflection effect of an MQB.

Fig. 17.30 shows the current versus output power (IeL) and current versus EQE(hext) for a 250 nm AlGaN MQW LED with an MQB or a single EBL, both measuredunder RT CW operation. Significant increases in output power and EQE wereobserved when the single EBL was replaced by the MQB. The maximum outputpowers of the 250 nm LED with the MQB and the single EBL were 15 and2.2 mW, respectively. The EQE of a 250 nm LED increased by approximately fourtimes with the introduction of the MQB. From Fig. 17.30, we estimated that theEIE of the 250 nm LED improved from approximately 25% to more than 80% by us-ing the MQB.

Fig. 17.31 shows the current versus output power (IeL) for a 237 nm AlGaN-MQW LED with an MQB or a single EBL, both measured under RT CW operation.The enhancement of the EIE with the MQW was found to be extremely high for short-wavelength DUV-LEDs. The output power of the LED increased by approximately 12times when the single EBL was replaced by the MQB.

Fig. 17.32 summarizes the wavelength dependence of the EQE of AlGaN DUV-LEDs with MQBs and single EBLs. The enhancement factors of the EQE by

Single barrier

Modula

ted M

QB

00

200100 300 400Current / (mA) Current / (mA)

15 mW

Ext

erna

l qua

ntum

effi

cien

cy g

ext (

%)

5

10

15

AlGaN MQW LEDson AIN/sapphire

λ = 250 nm

0 50 1000

Single barrier

1

2

EQE: 1.5%Modulated MQB

Measured at RTCW

Out

put p

ower

(mW

)

Figure 17.30 Current versus output power (IeL) and current versus external quantumefficiency (EQE) (hext) for 250 nm AlGaN multiquantum well (MQW) light-emitting diodes(LEDs) with an MQB and with a single electron blocking layers (EBLs).


With single-barrier EBL

RT CW

AIGaN DUV-LEDs

With MQB

220 240 260 280

Wavelength (nm)

0.01

0.1

0.05

0.5

1

5

10

Ext

erna

l qua

ntum

effi

cien

cy

ext(

%)

Figure 17.32 Wavelength dependence of external quantum efficiency (EQE) of AlGaN light-emitting diode deep ultraviolet (DUV) light-emitting diodes (LEDs) with MQBs and singleelectron blocking layers (EBLs).

6

5

4

3

2

1

00 100 300200

Current / (mA)

AlGaN DUV-LEDsRT CW

λ = 237 nmwith MQB

(234 nm)single barrier

Out

put p

ower

(mW

)


EL

inte

nsity

(a.u

.)

Figure 17.31 Current versus output power (IeL) for 237 nm AlGaN multiquantum well(MQW) light-emitting diodes (LEDs) with an multiquantum barrier (MQB) and with a singleelectron blocking layers (EBL).


introducing the MQB are approximately 10, 4, and 3 times for 235, 250, and 270 nmAlGaN LEDs, respectively. Fig. 17.33 shows the current versus output power (IeL)and the EQE (hext) for a high-output-power 270 nm AlGaN MQW LED with anMQB, as measured under RT CW operation. An output power of 33 mWwas obtainedfor a bare-chip sample, which is the highest reported value to date. A much higheroutput power can be obtained with heat dissipation through a flip-chip geometry.The highest value of the EQE for a 270 nm AlGaN DUV-LED was 3.8% using asimilar LED structure.22

The improvement in the LEE of a DUV-LED is particularly important for the nextresearch subject, because LEE is less than 10% for conventional AlGaN DUV-LEDs.The low LEE for AlGaN LEDs is caused by strong DUV absorption in the p-GaN con-tact layers and the p-side electrodes. A p-type AlGaN layer, which is expected to be atransparent contact layer, is not yet useful because the hole concentration ofmagnesium-doped AlGaN is quite low. The reflectance of Ni/Au p-electrodes isalso low (approximately 25%). The reflectance of DUV by aluminum is 92%, butohmic contacts are hard to obtain. Additionally, a transparent p-type electrode forAlGaN is yet to be developed. For these reasons, improvements in LEE for AlGaNDUV-LEDs are relatively difficult to obtain, although they have already been achievedfor InGaN/GaN-based blue LEDs.

RT CWλ = 270 nm

00

100 200 300 400 500

5

10

15

20

25

30

Out

put p

ower

(mW

)

Current (mA)

2.2

2

1.8

1.6

1.4

Ext

erna

l qua

ntum

effi

cien

cy (%

)

Figure 17.33 Current versus output power (IeL) and external quantum efficiency (EQE) (hext)for a high output power 270 nm AlGaN multiquantum well (MQW) light-emitting diode(LED) with a multiquantum barrier (MQB), measured under room temperature continuouswave (CW) operation.


The LEE can be improved by using a thin p-GaN contact layer, which absorbs lessDUV, as well as by using highly reflective aluminum-based electrodes. The LEE canalso be enhanced by using photonic nanostructures, fabricated on the backside of thesapphire substrate or on the interface of the substrate and the AlN buffer layer.

Fig. 17.34 is a schematic image of the improvement in the LEE of an AlGaN DUV-LED by using a thin p-GaN contact layer and an aluminum-based highly reflective p-electrode. The LEE can be improved from 8% up to 20%e30% by thinning the p-GaNcontact layer to approximately 10 nm and by fabricating a highly reflective Ni/Al p–electrode. Ohmic contacts can be obtained by inserting a very thin (<1 nm) nickellayer between the aluminum and the p-AlGaN. In the actual experiment, the reflec-tance of the p-type electrode was increased from 25% to 64% by replacing the usualNi/Au by a Ni (1 nm)/Al (150 nm) electrode. Light absorption was reduced to about50% by using a thin (about 30 nm) p-GaN contact layer. We obtained an ohmic contacton a p-electrode by using 33 nm of p-GaN and inserting about 1 nm of nickel.Fig. 17.35 shows the enhancement of EQE for a 270 nm AlGaN LED by using thinp-GaN and a Ni/Al electrode. The EQE was approximately 1.3 times higher due tothe enhancement of the LEE.

(20%–30%)

LEE = 30%

LEE: 8%

LEE = 8% Sapphire

AI N

n-AIGaN

p-AIGaN

Active layer

p-GaN

EmissionEmission

Ni/Au electrodereflectance <30%

High absorptionin p-GaN

AI electrodereflectance >80%

Low absorption(thin p-GaN, 10 nm)

Figure 17.34 Improvement of the light extraction efficiency (LEE) in an AlGaN deepultraviolet (DUV) light-emitting diodes (LED) using a thin p-GaN contact layer and analuminum-based highly reflective p-electrode.


17.6 Conclusions and future trends

Fig. 17.36 summarizes the maximum output power of AlGaN- and InAlGaN-basedDUV-LEDs fabricated on low TDD ML-AlN templates by the RIKEN group16e22

in the past 5 years. Between 2007 and 2011 we achieved significant increases in theEQE and output power of AlGaN and InAlGaN DUV-LEDs with emission wave-lengths ranging from 222 to 351 nm by introducing low TDDAlN templates, InAlGaNemitting and p-type layers, and MQB electron-blocking layers. The maximum outputpower increased to 15e33 mW for 245e270 nm LEDs. We also obtained approxi-mately 5 mW CW power for a 237 nm short wavelength LED. These achievementswill contribute to accelerating the practical application of DUV-LEDs and to expand-ing them for a wide range of applications.

The EQE of AlGaN DUV-LEDs will be significantly increased by improving theIQE, EIE and LEE in the near future. The IQE for AlGaN QW DUV emission is ex-pected to increase from 50% to more than 90% by reducing the TDD of the AlN tem-plates from the current 3 � 108 to 1 � 107 cm�2 by using an ELO method or by usingAlN single-crystal wafers. Quaternary InAlGaN emitting layers are also effective forincreasing the IQE, as mentioned in Section 17.4.

A high EIE can be obtained by using an optimized MQB structure in AlGaN DUV-LEDs, as mentioned in Section 17.5. Further improvements in EIE (>90%) could berealized by increasing the hole concentration of p-type AlGaN, which is considered tobe relatively difficult by the usual method.

The LEE could be significantly increased by replacing the p-GaN contacts with atransparent p-AlGaN layer, as well as by introducing a transparent p-type electrode.

Ni/Au1.3 times

RT CW

EQE: 2.75%

Al-electrode

00

50 100Current (mA)

1

2

3

EQ

E (%

)

Reflectivity of AI-electrode: 64%Absorption through p-GaN: 35%

Figure 17.35 Enhancement of external quantum efficiency (EQE) for a 270 nm AlGaN light-emitting diode (LED) by using thin p-GaN and a Ni/Al electrode.


The achievement of a high hole concentration for AlGaN is difficult because themagnesium-acceptor level is very deep in AlGaN. This difficulty could be solvedby introducing a codoping technique. Codoping of MgeO, MgeSi or CeO is consid-ered to be effective for obtaining shallow acceptor levels, even in AlGaN semiconduc-tors. If shallow acceptor levels are realized by using a codoping technique, then theLEE could be improved to more than 60%, which has already been achieved forblue LEDs. Through these developments, the EQE of AlGaN DUV-LEDs is expectedto improve by several tens of percent in the near future.

References

1. Zukauskas A, Shue MS, Gaska R. Introduction to solid-state lighting. New York: Wileyand Sons Publishers; 2002.

2. Hirayama H. Quaternary InAlGaN-based high-efficiency ultraviolet light-emitting diodes.J Appl Phys 2005;97. 091101 1e19.

3. Han J, Crawford MH, Shul RJ, Figiel JJ, Banas M, et al. AlGaN/GaN quantum well ul-traviolet light emitting diodes. Appl Phys Lett 1998;73:1688e90.

4. Kinoshita A, Hirayama H, Ainoya M, Hirata A, Aoyagi Y. Room-temperature operation at333 nm of Al0.03Ga0.97N/Al0.25Ga0.75N quantum-well light emitting diodes with Mg-dopedsuperlattice layers. Appl Phys Lett 2000;77:175e7.

220 240 260 280 30020010–5

10–4

10–3

10–2

102

10–1

1

10

Max

imum

out

put p

ower

(mW

)

NTT210 nm0.02 μW

TDD > 2 × 1010 cm–2 (2006)

TDD: 3 × 109 cm–2 (2007)

TDD 3–7 × 108 cm–2

Emission is week,no single peak

Sterilizationwavelength

(2010-11)Using MQB

(2008)Using high EBL

Wavelength (nm)

Figure 17.36 Maximum output power of AlGaN- and InAlGaN-based deep ultraviolet (DUV)light-emitting diode (LED) fabricated on low threading dislocation density (TDD) multilayer(ML)-AlN templates by the RIKEN group16e22 in the last 5 years.


5. Nishida T, Saito H, Kobayashi N. Efficient and high-power AlGaN-based ultraviolet light-emitting diode grown on bulk GaN. Appl Phys Lett 2001;78:711e3.

6. Sun WH, Adivarahan V, Shatalov M, Lee Y, Wu S, Yang JW, et al. Continuous wavemilliwatt power AlGaN light emitting diodes as 280 nm. Jpn J Appl Phys 2004;43:L1419e21.

7. Adivarahan V, Wu S, Zhang JP, Chitnis A, Shatalov M, et al. High-efficiency 269 nmemission deep ultraviolet light-emitting diodes. Appl Phys Lett 2004;84:4762e4.

8. Khan MA. MOVPE of nitride UV emitters and detectors. In: 13th International Conferenceon metal organic vapor phase epitaxy (ICMOVPE-XIII), vol. 3; 2006.

9. Taniyasu Y, Kasu M, Makimoto T. An aluminum nitride light-emitting diode with awavelength of 210 nanometers. Nature 2006;444:325e8.

10. Hirayama H, Enomoto Y, Kinoshita A, Hirata A, Aoyagi Y. Efficient 230e280 nm emis-sion from high-Al-content AlGaN-based multi-quantum wells. Appl Phys Lett 2002;80:37e9.

11. Hirayama H, Kinoshita A, Yamabi T, Enomoto Y, Hirata A, et al. Marked enhancement of320e360 nm UV emission in quaternary InxAlyGa1-x-yN with In-segregation effect. ApplPhys Lett 2002;80:207e9.

12. Hirayama H, Enomoto Y, Kinoshita A, Hirata A, Aoyagi Y. Room-temperature intense 320nm-band UV emission from quaternary InAlGaN-based multi-quantum wells. Appl PhysLett 2002;80:1589e91.

13. Hirayama H, Akita K, Kyono T, Nakamura T, Ishibashi K. High-efficiency 352 nm qua-ternary InAlGaN-based ultraviolet light-emitting diodes grown on GaN substrates. Jpn JAppl Phys 2004;43:L1241e3.

14. Fujikawa S, Takano T, Kondo Y, Hirayama H. Realization of 340-nm-band high-output-power (7 mW) InAlGaN quantum well ultraviolet light-emitting diode with p-type InAl-GaN. Jpn J Appl Phys 2008;47:2941e4.

15. Fujikawa S, Takano T, Kondo Y, Hirayama H. 340 nm-band high-power InAlGaN quantumwell ultraviolet light-emitting diode using p-type InAlGaN layers. Phys Stat Sol (C) 2008;5:2280e2.

16. Hirayama H, Yatabe T, Noguchi N, Ohashi T, Kamata N. 231e261 nm AlGaN deep-ultraviolet light-emitting diodes fabricated on AlN multilayer buffers grown by ammoniapulse-flow method on sapphire. Appl Phys Lett 2007;91. 071901 1e3.

17. Hirayama H, Yatabe T, Ohashi T, Kamata N. Remarkable enhancement of 254e280 nmdeep ultraviolet emission from AlGaN quantum wells by using high-quality AlN buffer onsapphire. Phys Stat Sol (C) 2008;5:2283e5.

18. Hirayama H, Tsukada Y, Maeda T, Kamata N. Marked enhancement in the efficiency ofdeep-ultraviolet AlGaN light-emitting diodes by using a multiquantum-barrier electronblocking layer. Appl Phys Express 2010;3:031002.

19. Hirayama H, Noguchi N, Yatabe T, Kamata N. 227 nm AlGaN light-emitting diode with0.15 mW output power realized using thin quantum well and AlN buffer with reducedthreading dislocation density. Appl Phys Express 2008;1(051101):1e3.

20. Hirayama H, Noguchi N, Kamata N. 222 nm deep-ultraviolet AlGaN quantum well light-emitting diode with vertical emission properties. Appl Phys Express 2010;3:032102.

21. Hirayama H, Noguchi N, Fujikawa S, Norimatsu J, Takano T, et al. 222e282 nm AlGaNand InAlGaN based high-efficiency deep-UV-LEDs fabricated on high-quality AlN onsapphire. Phys Status Solidi (A) 2009;206:1176e82.

22. Fujikawa S, Hirayama H, Maeda N. High-efficiency AlGaN deep-UV LEDs fabricated ona- and m-axis oriented c-plane sapphire substrates. Phys Status Solidi (C) 2012;9:3e4.790e793.


23. Shatalov M, SunW, Bilenko Y, Sattu A, Hu X, et al. Large chip high power deep ultravioletlight-emitting diodes. Appl Phys Express 2010;3:062101.

24. Grandusky JR, Gibb SR, Mendrick MC, Moe C, Wraback M, et al. High output power from260 nm pseudomorphic ultraviolet light emitting diodes with improved thermal perfor-mance. Appl Phys Express 2011;4:082101.

25. Adivarahan V, Heidari A, Zhang B, Fareed Q, Hwang S, et al. 280 nm deep ultraviolet lightemitting diode lamp with an AlGaN multiple quantum well active region. Appl Phys Ex-press 2009;2:102101.

26. Hwang S, Morgan D, Keslar A, Lachab M, Zhang B, et al. 276 nm substrate-free flip-chipAlGaN light-emitting diodes. Appl Phys Express 2011;4:032102.

27. Pernot C, Kim M, Fukahori S, Inazu T, Fujita T, et al. Improved efficiency of 255e280 nmAlGaN-based light-emitting diodes. Appl Phys Express 2010;3:061004.

28. Fujioka A, Misaki T, Murayama T, Narukawa Y, Mukai T. Improvement in output power of280-nm deep ultraviolet light-emitting diode by using AlGaN multi quantum wells. ApplPhys Express 2010;3:041001.

29. Yoshida H, Yamashita Y, Kuwabara M, Kan H. Demonstration of an ultraviolet 336 nmAlGaN multiple-quantum-well laser diode. Appl Phys Lett 2008;93. 241106 1e3.

30. Iida K, Kawashima T, Miyazaki A, Kasugai H, Mishima A, et al. 350.9 nm UV laser diodegrown on low-dislocation-density AlGaN. Jpn J Appl Phys 2004;43(4A):L451e99.

31. Takano T, Narita Y, Horiuchi A, Kawanishi H. Room-temperature deep-ultraviolet lasing at241.5 nm of AlGaN multi-quantum-well laser. Appl Phys Lett 2004;84:3567e9.

32. Banal RG, Funato M, Kawakami Y. Optical anisotropy in [0001]-oriented AlGaN/AlNquantum well (x > 0.69). Phys Rev B 2009;79. 121308(R).

33. Kawanishi H, Senuma M, Yamamoto M, Niikura E, Nukui T. Extremely weak surfaceemission from (0001) c-plane AlGaN multiple quantum well structure in deep-ultravioletspectral region. Appl Phys Lett 2006;89:081121.

34. Iga K, Uenohara H, Koyama F. Electron reflectance of multiquantum barrier (MQB).Electron Lett 1986;22:1008.

35. Kishino K, Kikuchi A, Kaneko Y, Nomura I. Enhanced carrier confinement effect by themultiquantum barrier in 660 nm GaInP/AlInP visible lasers. Appl Phys Lett 1991;58:1822.


Infrared emitters using III-nitridesemiconductors 18Akhil Ajay, Yulia Kotsar, Eva MonroyCEA-Grenoble, INAC-PHELIQS, Grenoble, France

18.1 Introduction

Rapid progress achieved in light-emitting diodes (LED) based on the III-nitride mate-rial system have permitted the accessibility to a large spectral range from green toultraviolet using InGaN/GaN and AlGaN/GaN quantum wells (QWs) as active media.Much less effort has been devoted to InN and In-rich alloys, which appeared as thenatural nitride-based choice for red optoelectronics. A major breakthrough in 2002,stemming from much improved quality of InN films grown using molecular beamepitaxy (MBE), resulted in the band gap of InN being revised from 1.9 eV to amuch narrower value of 0.64 eV (z1.9 mm wavelength).1 This finding extended theinterband optoelectronic capabilities of the III-nitride family into the infrared (IR)spectrum. However, this IR technology is penalized by the high residual doping,poor thermal stability, and surface effects characteristic of InN.2 Thus, alternativeapproaches have been proposed to develop a III-nitride IR technology. First, a numberof research groups have developed light emitting devices by incorporating rare-earth(RE) elements such as Erbium (Er), Thulium (Tm), and Praseodymium (Pr) in GaNin order to achieve luminescence via intra-atomic transitions in the 4f electronic shellof these lanthanide ions.3e5 Another approach is the implementation of an intersub-band (ISB) technology, where the IR transition occurs between confined electronicstates in the conduction or in the valence band of semiconductor nanostructures.6e8

In this chapter, we summarize the state of the art of the various III-nitride IR technol-ogies, namely high-In-content interband optoelectronics, rare earth doping, and ISBtechnology.

18.2 High-indium-content alloys for IR emitters

The external quantum efficiency of InGaN LEDs, as high as 75% for blue devices,drops below 25% in the green spectral range (around 560 nm) and even lower forlonger wavelengths.9,10 This efficiency drop stems from the combination of multipledegradation mechanisms including spinodal decomposition of InGaN at high In con-tents, stress relaxation, nonradiative recombination associated with native defects, andpiezoelectricity-induced quantum confined Stark effect (QCSE), which reduces theelectronehole wavefunction overlap.11 Although intense InGaN photoluminescence(PL) is achieved over the entire spectral range from the band gap of InN to that of


http://dx.doi.org/10.1016/B978-0-08-101942-9.00018-6

GaN,12,13 electrically driven InGaN LEDs require the formation of a p-n junction,which at long wavelengths implies either facing the challenge of p-type doping andcontacting high-In-content InGaN layers or dealing with the huge lattice mismatchbetween the p-GaN contact layer and the high-In-content active region (11% in planelattice mismatch between GaN and InN). LED emission up to z740 nm has beenreported,14 although there is no information on the external quantum efficiency atsuch long wavelengths.

An approach to circumvent these material issues consists in nanostructuring theactive region, either incorporating InGaN quantum dots (QDs), which can be synthe-sized defect free in a GaN matrix,15 or by growth of nanocolumn arrays, which can beeither self-assembled16 or spatially localized via selective area growth.17 The elasticrelaxation of misfit strain in these nanostructures enables the fabrication of InGaN-based red LEDs, although their efficiency and reliability remains still far from com-mercial values. Emission up to z740 nm has been demonstrated using InGaN/GaNQDs as active region,18 and up to 690 nm using coreeshell InGaN/GaN nanowires.19

Farther into near-IR spectral range, 1.55 mm (0.8 eV) light emission can be readilyachieved by quantum confining InN or from InGaN with very low indiumcontent,20e22 but the challenge for device fabrication lies again in the electron injec-tion. A demonstration of IR (1.46 mm wavelength) LEDs has been reported with anactive region consisting of a GaN:Si/In0.87Ga0.13N/In0.3Ga0.7N:Mg nanocolumn array,as illustrated in Fig. 18.1.23

The above-described results are based on InGaN structures grown on the Ga-polar(0001) plane, which is the crystallographic orientation of commercial InGaN LEDs.Alternative orientations have been studied to enhance the indium incorporation athigh temperature, and hence improve the InGaN crystalline quality at high In concen-trations. Research in N-polar (000�1) InGaN was particularly motivated by the higherthermal stability of this crystallographic plane, in comparison to (0001).24,25 As aresult, at the same growth temperature, it is possible to incorporate more indium in(000�1)-oriented InGaN than in (0001)-oriented InGaN, and this is valid both forMBE and metal-organic vapor phase epitaxy (MOVPE).26,27 This research has ledto the fabrication of N-polar red (633 nm) LEDs.28

The use of nonpolar or semipolar crystallographic orientations has also been inves-tigated as an approach to enhance indium incorporation, and at the same time addressthe polarization-induced charge separation issue.29 The indium incorporation limit at acertain growth temperature depends on the crystallographic orientation. There is acertain disagreement in the literature about the behavior of various crystallographicplanes, maybe due to the different growth conditions, but it is generally acceptedthat MOVPE growth on (11�22) favors indium incorporation,30e33 which has encour-aged research efforts on that orientation. As a result, (11�22)-oriented yellow-amber(525e575 nm) LEDs with a performance comparable to those of phosphide alloyshave been demonstrated.34e36 However, the reduced critical thickness associated tothe growth on (11�22) has shifted the attention to alternative planes, like (20�21),which has allowed the fabrication of red (624e660 nm) semipolar LEDs.37

Looking at the other edge of the IR spectrum, the terahertz domain, the quest forbrighter far-IR sources has placed InN into the focus of investigation for THz


emission. Irradiated with ultrafast laser pulses, semiconductor surfaces can be used toefficiently generate broad THz pulses.38 InN is considered as a promising candidatefor this application due to its high mobility, the large difference between thediffusion length of electrons and holes, and the very low probability of intervalleyscattering. Several groups have reported THz generation from InN films andnanostructures39e51; however, the underlying physical mechanisms are still underdebate. THz generation in InN was primarily attributed to optical rectification42 orto the photo-Dember effect.41 Later, Xu et al. explained the THz generation fromInN by the destructive interference between optical rectification and photocurrentsurge.50 The role of the polarization fields in InN to enhance the THz emission isalso under discussion.44

p-electrode

In0.3Ga0.7N: Mg

In0.87Ga0.13N

GaN: Si

N-electrode

(0001) Sapphire

0.8

0.6

0.4

0.2

0.0

–8 –6 –4 –2 0 2 4Voltage (V)

0.8 1.2 1.6 2.0 2.4Wavelength (μm)

EL

inte

nsity

(arb

.uni

t)

Cur

rent

(mA

)

turn-onvoltage:~ 1 V

λ = 1.46μm

17.8 mA11.3 mA5.91 mA2.94 mA1.46 mA

(DC current)R.T.

MOCVD-GaN

SOG

Ti

φ = 65 μm

ITO

1.0 μm

(a) (b)

(d)(c)

Figure 18.1 (a) Cross-sectional schematic structure of near infrared nanocolumn light-emittingdiode (LED). (b) Scanning electron microscopy image of the LED crystal taken at an angle of55 degrees from normal. (c) Current versus voltage characteristics under direct current (DC)current injection. (d) Electroluminescence spectra under DC current injection.Reprinted, with permission, from Kishino K, Kamimura J, Kamiyama K. Near-infrared InGaNnanocolumn light-emitting diodes operated at 1.46 mm. Appl Phys Express 2012;5:031001. ©2012 The Japan Society of Applied Physics.

Infrared emitters using III-nitride semiconductors 589

18.3 RE-doped GaN emitters

Luminescence from RE ions in solid hosts involves transitions within the 4f electronicshell that are parity forbidden in the free ion by the Laporte selection rule, but becomeallowed through the admixture of states of opposite parity induced by the crystal fieldin solids. These intra-4f transitions are still not fully allowed, resulting in excited statelifetimes of z1 ms to 1 ms for RE-doped systems. The energy of the emission lines isrelatively independent of the host material; however, the crystal has a very strongeffect on the radiative transition probability, and hence on the photoemission intensity.The study of luminescence from RE ions embedded in semiconductors has quite a longhistory, particularly the 1.54 mm IR band emitted by Er3þ,52 which is suitable for trans-mission through silica fiber over long distances. Favennec et al.53 established that thethermal quenching in Er-doped semiconductors decreases with increasing band gap,which impelled research on RE-doped GaN.

The RE doping of GaN can be accomplished by ion implantation54,55 or by in situdoping during growth.56e58 Ion implantation has the advantage of simplicity and itgrants a good control of the dopant location and density independent of the growthconditions. However, implantation induces structural damage which cannot becompletely removed by annealing.59 On the other hand, in situ doping requires agood understanding and control of the growth process, since the RE flux interactswith the other species and modifies the GaN growth kinetics. In situ RE-doped layersdo not suffer from the damage effects of ion implantation and therefore present asignificantly higher radiative efficiency.

RE atoms that incorporate the III-N lattice get located in the III substitutional site,SIII.

60 Only in the case of Eu, a second incorporation site has been reported, with the Euion shifted away from SIII along the [0001] direction.61,62 It should be noted, though,that the RE atoms do not occupy the perfect substitutional position, but they arerandomly displaced around this location. The major limitation to incorporating REson SIII sites is the creation of implantation-related defects in implanted layers, orclustering and formation of RE-nitrides in the case of in situ doping.

Light emission from GaN:RE has been demonstrated via PL, cathodoluminescence,and electroluminescence. The energy from excited carriers is transferred to the REdopants by impact excitation of hot carriers or as a result of nearby electroneholerecombination.63 The RE then relaxes either radiatively or nonradiatively (multipho-non emission or Auger mechanism). Radiative emission in the near-IR fromRE-doped GaN is obtained at 801 nm using Tm,64 at 1000 and 1540 nm withEr,55,65 at 956, 1303, and 1914 nm with Pr,66 and at 905, 1082, and 1364 nm withNb,67 as summarized in Fig. 18.2.68

It is known that the emission intensity is optimized for a certain RE concentration.As the RE ion density is increased, the average distance between ions is reduced pro-portional to the cube root of the RE concentration. When RE ions are located suffi-ciently close to each other, the excitation residing in one ion can migrate to aneighboring RE ion as a result of resonant energy transfer (cross-relaxation), whichincreases the nonradiative relaxation probability. The GaN lattice allows unusually


high substitutional RE doping incorporation (up to z3e5 at.%) while preserving theoptical activation of RE dopants.60 However, the optimum RE concentration in termsof radiative efficiency is around 0.5%e1%.5

Visible electroluminescent devices consisting of a dielectric/GaN:RE/dielectricstructure have been demonstrated, operated under alternated current (AC) high-voltage (z100 V) conditions. Red, green, and blue emission lines originating fromEu, Er and Tm ions in GaN or AlGaN have been reported.5 Devices incorporatingEr as a dopant exhibit equally intense visible (537/558 nm) and IR (1.54 mm) emission.Near-IR emitting devices with a similar design were reported by Kim and Hollowayusing Tm-, Nd-, and Er-doped GaN to obtain electroluminescence at 800 nm,1.08 mm, and 1.55 mm.67,69 Another approach consists of using a thick (500 nm)GaN:Er layer as a phosphor that is excited by an ultraviolet LED.3 Emission at1.54 mmwas obtained in such devices, with the IR line intensity scaling almost linearlywith the LED driving current.

18.4 III-nitride materials for ISB optoelectronics

An alternative approach of nitride devices for IR optoelectronics consists in using ISBtransitions, i.e., electronic transitions between confined levels in either the conductionband or the valence band of QWs or QDs. The operation wavelength can be tuned bydesign, by varying the dimensions of the active nanostructures. The first observation ofISB transitions in QWs was reported in 1985,70 but it is almost one decade later that wefind the first ISB device: the QW IR photodetector.71 In 1994 Faist et al.72 presented amajor breakthrough in the ISB technology: an alternative to laser diode with a noveloperating principle e the quantum cascade laser (QCL). This was the beginning oftremendous development of the ISB technology that resulted in commercially avail-able devices operating in the mid- and far-IR.73

200 300 400 500 600 800 1000 2000 3000 5000Wavelength λ (nm)

Gd3+ Er3+Eu3+ Pr3+

Tm3+ Tm3+Er3+Pr3+ Tm3+ Er3+Pr3+Tm3+

6 5 4 3 2 1 0.6 0.4 0.2

Bandgap energy (eV)

AIN GaN InNBxAI1–xN AlxGa1–xN lnxGa1–xN

Figure 18.2 Emission wavelengths from selected transitions in rare earth ions and associatedband gap energies of alloys of III-N compound semiconductors.Reprinted, with permission, from Steckl AJ, Park JH, Zavada JM. Prospects for rare earth dopedGaN lasers on Si. Mater Today 2007;10:20e7. © 2017 Elsevier Ltd.


ISB transitions are governed by certain polarization selection rules. Thus, ISBelectronephoton interactions are only possible when the light has an electric fieldcomponent parallel to the direction of carrier confinement. In the case of QWs, thisselection rule forces the use of waveguide configurations or surface gratings toimprove the light coupling. Moreover, due to the inversion symmetry potential transi-tions only with opposite parity of envelope wavefunctions are allowed in symmetricQWs. This later rule can be overthrown by breaking the QW symmetry either bydesign or by an electric field. A comprehensive introduction to ISB physics can befound in Refs. 74 or 75.

The ISB technology has developed in the mid-IR and far-IR spectral ranges usingAs-based systems like GaAs/AlGaAs and InGaAs/AlInAseInP. The ISB shortestoperation wavelength is limited by the conduction band offset, which attainsz1.8 eV for the GaN/AlN system,76 large enough to reach the optical fiber transmis-sion windows at 1.3 and 1.55 mm. In the near-IR, a specific advantage of ISB transi-tions in III-N materials is the ultrashort electron recovery time of about few hundredsfemtoseconds,77,78 which finds application in all-optical switches in the 0.1e1 Tbit/sregime. In the far-IR domain, due to large longitudinal-optical (LO) phonon energy(92 meV) ISB lasers based on III-nitrides have a potential of operation above roomtemperature, overcoming the thermal limitations of As-based THz lasers attributedto thermally activated LO-phonon scattering.

18.4.1 Electronic structure

The optical properties of nitride QWs are strongly affected by the presence of aninternal electric field inherent to wurtzite-phase nitride heterostructures grown alongthe<0001> axis. This field, arises from the piezoelectric and spontaneous polarizationdiscontinuity between the well and barrier materials.79 Modeling of quantum confine-ment in nitride QWs should therefore go beyond the flat-band approximation andaccount for the internal electric field in the QW and in the barriers. As an example,Fig. 18.3(a) presents the band diagram of GaN/AlN (2/3 nm) superlattices calculatedusing the nextnano3 8-band k.p Schr€odingerePoisson solver80 with the materialparameters described in Ref. 81. The structures were considered strained on the AlNsubstrate. The potential takes on a characteristic saw-tooth profile due to the internalelectric field. The electron wavefunctions of the ground hole state, h1, the groundelectron state, e1, and the excited electron states, e2 and e3, are presented. In narrowQWs (z1 nm) the energy difference between e1 and e2 is mostly determined by theconfinement in the QW, whereas for larger QWs (>2 nm) this difference is mostlydetermined by the QCSE, because both electronic levels lay in the triangular part ofthe QW potential profile.

A detailed description of the evolution of the ISB transitions e2�e1 and e3�e1 withthe QW thickness and strain state is presented in Fig. 18.3(b) and compared with therespective experimental data from GaN/AlN superlattices strained on AlN. Theincrease in the e2�e1 ISB energy difference in the superlattice with a larger in-planelattice parameter is related to the enhancement of the electric field in the QW,


due to the larger piezoelectric coefficients of the AlN barrier in comparison to theGaN QW.81

Another consequence of the polarization-related internal electric field is that thecharge distribution in polar heterostructures depends not only on the dopants butalso on the carrier redistribution due to the electric field. The polarization discontinuitybetween the active region (in general, a QW or QD superlattice) and the contact orcladding layers leads to the formation of a depletion layer on one side of the activeregion, and an electron accumulation layer on the other side. Therefore, a reliableview of the charge distribution in a device is only achieved by extending the electronicmodeling effort to the whole structure.

18.4.2 Growth and structural properties

A main requirement for the growth of the III-nitride nanostructures required for ISBdevices is a precise control of the thickness and interfaces. Plasma-assistedmolecular-beam epitaxy (PAMBE) seems the most suitable technique for this applica-tion thanks to its low growth temperature that hinders GaN-AlN interdiffusion.82 Thegrowth of GaN (0001) by PAMBE is extensively discussed in the literature.83e85

Deposition of two-dimensional (2D) GaN layers requires a precise control over III/V ratio during the growth, particularly it demands Ga-rich conditions, and hencegrowth optimization translates into the determination of the adequate metal excess

4

2

0

–2

–4

–60 4 8 12 16

0.4

0.6

1.0 1.5 2.0 2.5QW thinkness (nm)

Strained on GaNStrained on AIN

e2 –e1

e3 –e1

T = 300K

1.6

1.4

1.2

1.0

0.8Ene

rgy

(eV

)

Ene

rgy

(eV

)

Conduction band

e3e2e1

[0001]

h1

Valence band

Depth (nm)

(a) (b)

Figure 18.3 (a) Band diagram of GaN/AlN quantum wells (QWs) in a superlattice with 3-nm-thick AlN barriers and 2-nm-thick GaN QWs. (b) Variation of the e1�e2 and e1�e3 inter-subband (ISB) transition energy as a function of the QW thickness. Triangles indicateexperimental data and solid and dashed lines correspond to theoretical calculations assumingthe structure fully strained on AlN and on GaN, respectively.Reprinted, with permission, from Kandaswamy PK, Guillot F, Bellet-Amalric E, Monroy E,Nevou L, Tchernycheva M, Michon A, Julien FH, Baumann E, Giorgetta FR, Hofstetter D,Remmele T, Albrecht M, Birner S, Dang LS. GaN/AlN short-period superlattices forintersubband optoelectronics: a systematic study of their epitaxial growth, design, andperformance. J Appl Phys 2008;104:093501. © 2008 American Institute of Physics.


and growth temperature. At a substrate temperature higher than 700�C and for a certainrange of Ga fluxes corresponding to slightly Ga-rich conditions, the Ga excess remainson the growing surface in a situation of dynamical equilibrium, i.e., the Ga coverage isindependent of the Ga exposure time. It is possible to stabilize a Ga amount frombelow one monolayer (ML) up to 2.5 ML. However, smooth surfaces can only beachieved with a Ga coverage of 2.5 � 0.1 ML, when the Ga excess arranges into aso-called “laterally contracted Ga bilayer,” which consists of two Ga layers adsorbedon the Ga-terminated (0001) GaN surface.83,85 In the case of AlN, the deposition oflayers with atomically flat surface morphology also requires metal-rich conditions.However, Al does not desorb from the surface at the standard growth temperaturefor GaN. Therefore, to eliminate the Al excess at the surface, it is necessary to performperiodic growth interruptions under nitrogen. An alternative approach to achieve 2Dgrowth of AlN and low Al content (<50%) Al(Ga)N layers is to use Ga as a surfactant,with the Al flux corresponding to the required Al mole fraction.81,86,87

GaN/AlN is a lattice mismatched system (2.5% in-plane lattice mismatch), whichresults in the presence of strain. The mechanisms of strain relaxation can be elastic(undulation of the surface) or plastic (introduction of misfit dislocations or stackingfaults, crack propagation, or decohesion of the layer). These defects can affect thedevice properties causing nonradiative recombination, carrier scattering, and enhanceddiffusion of dopants and impurities. In the case of nitride heterostructures grown alongthe [0001] axis, which is the predominant growth orientation in commercial devices,the formation of regular networks of misfit dislocations is hindered since the most crys-tallographically favorable slip system, the (0001) basal plane with <11�20> {0002}slip directions, lies parallel to the heterointerfaces. In PAMBE growth, the metal-to-Nratio and the growth temperature are key parameters that define the mechanisms ofstrain relaxation during growth. Indeed, Ga-rich conditions hinder crack formationand minimize strain relaxation.87 In the case of GaN/AlN superlattices, the periodicmisfit relaxation is associated to the formation of stacking fault loops that initiate atthe beginning of the AlN deposition, propagate through the barrier and close withinthe following QW.87 In contrast, transmission electron microscopy (TEM) imagesfrom GaN/AlGaN superlattices reveal sharp interfaces free of stacking faults or otherperiodic defect.88 Fig. 18.4(a) and (b) present high-resolution TEM view graphs ofGaN/AlN and GaN/Al0.44Ga0.56N superlattices, respectively, showing abrupt inter-faces at the atomic layer scale. The misfit stress between the superlattice and the under-layer is relaxed mostly by generation of 60 degrees 1

3 ð11� 20Þ dislocations,87e89

which fold toward the growth direction giving rise to edge-type threading dislocations.The density of edge-type dislocations should be kept as reduced as possible since theycause loss of transverse magnetic (TM) polarization that affect adversely the perfor-mance of ISB devices.90,91

GaN/AlN QWs displaying ISB transitions in the near-IR can also be synthesized byMOVPE.92,93 In this case, a critical parameter is the reduction of the growth temper-ature from the 1050e1100�C required for 2D GaN layers to 900e950�C, in order tominimize the GaN-AlN interdiffusion. Furthermore, deposition under compressivestrain (e.g., using AlN substrates) is recommended at these growth temperatures toprevent instabilities of the GaN/AlN interface.94


18.4.3 Optical characterization

Fig. 18.5(a) shows the room temperature photo-induced IR absorption spectra of GaN/Al(Ga)N superlattices with different dimensions and Al contents in the barriers. Thesamples show a pronounced TM-polarized absorption, attributed to the transitionfrom the first to the second electronic level in the QW (e1 / e2), while no absorptionis observed for transverse electric (TE) polarized light within experimental sensitivity.By changing the geometry and composition, the ISB absorption can be tailored tocover the near-IR range from 1.0 mm and mid-IR region up to 10 mm.76,81,93,95,97e101

In large GaN/AlGaN QWs, the first two confined electron levels get trapped in thetriangular section of the QW, which results in a saturation of the ISB energy. To shiftthe ISB absorption toward longer wavelengths, it is necessary to engineer the bands tocompensate the internal electric field in the wells. As a first approach, Machhadaniet al.102 proposed a three-layer structure (step-QW), composed of a 3 nm GaN step-barrier, a 10 nm Al0.05Ga0.95N well and a 3 nm Al0.1Ga0.9N high-Al-content barrier.The design creates a flat band in the “well” by having the “barrier” þ “step barrier”ensemble balanced at the same average Al mole fraction (i.e., same average polariza-tion) as the “well”. Transmission measurements performed at 4K reveal TM-polarizedISB absorption at 4.2 THz, respectively, in good agreement with simulations, asillustrated in Fig. 18.6.

The weakness of the step-QW design lies in the fact that any deviation in the Alcontent or thickness of the layers has a drastic effect on the ISB transition energy.

GaN AIN

10 nm(0001) (1–100)

10 nm

(0002)

GaN

AIGaN

(11–20)

(a) (b)

Figure 18.4 High-resolution cross-sectional transmission electron microscope (TEM) images of(a) GaN/AlN (1.5/3 nm) superlattice taken along the < 1120 > zone axis, and (b) GaN/Al0.44Ga0.56N superlattice viewed along the <1e100> zone axis.(a) Reprinted, with permission, from Kandaswamy PK, Bougerol C, Jalabert D, Ruterana P,Monroy E. Strain relaxation in short-period polar GaN/AlN superlattices. J Appl Phys 2009;106:013526. © 2009 American Institute of Physics. (b) Reprinted, with permission, from Kotsar Y,Doisneau B, Bellet-Amalric E, Das A, Sarigiannidou E, Monroy E. Strain relaxation in GaN/AlxGa1-xN superlattices grown by plasma-assisted molecular-beam epitaxy. J Appl Phys 2011;110:033501. © 2011 American Institute of Physics.


Sapphire

1 2 3 4 5 6 7 8 9 10 11 12Wavelength (μm)

Nor

mal

ized

abs

orpt

ion

Si(111)

p-po

lariz

ed a

bsor

ptio

n (a

rb.u

nit)

200 300 400Photon energy (meV)

4.0×1012

2.0×1012

6.6×1011

Theory Nd=6.0×1012cm–2

(a) (b)

Figure 18.5 (a) Room temperature transverse magnetic (TM)-polarized infrared (IR) photo-induced absorption spectra measured in GaN/AlGaN superlattices with different barrier Alcontents and quantum well (QW) width, grown either on sapphire or on Si(111) templates. (b)Blue shift and broadening of the absorption spectra of the mid-IR intersubband (ISB) transitionin nonpolar m-plane 10-period Al0.485Ga0.515N/GaN (3.20/2.85 nm) superlattices withincreasing doping density. Spectra are vertically shifted for clarity.(a) Reprinted, with permission, from Kandaswamy PK, Machhadani H, Bougerol C, Sakr S,Tchernycheva M, Julien FH, Monroy E. Midinfrared intersubband absorption in GaN/AlGaNsuperlattices on Si(111) templates. Appl Phys Lett 2009;95:141911. © 2009 American Instituteof Physics. (b) Reprinted, with permission, from Kotani T, Arita M, Arakawa Y. Dopingdependent blue shift and linewidth broadening of intersubband absorption in non-polar m-planeAlGaN/GaN multiple quantum wells. Appl Phys Lett 2015;107:112107. © 2015 AmericanInstitute of Physics.

0.2

0.1

0

0 5 10 15–0.1

Growth axis (nm)

Tran

smis

sion

(nor

m.u

)

1

0.9

0.8

0.73 4 5 6 7 8 9 10

Frequency (THz)

ARSi

Si

10 20 30 40Energy (meV)

Ene

rgy

(eV

)

e2e1

(a) (b)

Figure 18.6 (a) Conduction band profile and squared envelope functions of first two electroniclevels (e1, e2) for a step-quantum well (QW) sample with 15 nm thick step barrier. (b)Transmission spectra for transverse magnetic (TM)-(square) and TE-(circle) polarized lightmeasured at T ¼ 4.7K. The inset describes the sample configuration during the measurements:Two pieces of each sample, were polished at 30 degrees to form multipass waveguides. Thetwo pieces were placed face-to-face on the cold finger of a liquid helium-cooled cryostat.

In search for improved robustness, alternative architectures consisting of four layers ofGaN/AlGaN have been proposed,103 with demonstration of ISB absorption tunable inthe 150 to 50 mm range.104

18.4.4 The AlInN/(In)GaN system

The lattice mismatch between GaN and AlN can lead to high defect densities and riskof cracking in GaN/AlN superlattices. An alternative material approach to overcomethis problem is the use of AlInN alloys. AlInN with an In composition around17%e18% is lattice matched to GaN and presents a refractive index contrast equiva-lent to AlGaN with 46% Al content (6% contrast with GaN at 1.55 mm wavelength).Therefore, AlInN is a promising material to form distributed Bragg reflectors and thickwaveguide layers. However, we must keep in mind that lattice-matched AlInN/GaNheterostructures exhibit an electric field as large as 3 MV/cm, solely generated dueto the spontaneous polarization discontinuity.

The potential of AlInN/GaN lattice-matched systems for application in ISB technol-ogy has been explored.105e109 However, this material system is not adapted for ISBoptoelectronics for telecommunication applications since the conduction band offsetis in the range of z1 eV.110 ISB absorption in the near-IR spectral region has beenreported at 2.3e2.9 mm in GaN/Al0.85In0.15N superlattices,107 despite the presenceof significant alloy inhomogeneities.106

An alternative approach to manage the strain in the structure while retaining accessto shorter wavelengths is possible by adding small concentrations of In (below 5%)both in the barrier and in the QW, forming an AlInN/GaInN superlattice.111 This ma-terial combination reduces the probability of crack propagation in comparison to GaN/AlN, although it maintains a certain degree of strain. However, it is difficult to controlprecisely the In mole fraction, and the simulation of the electronic structure remains achallenging task.

18.4.5 Alternative crystallographic orientations

The already high design complexity in terms of modeling IR QCLs further increases inmaterials with internal electric field like polar III-nitrides. A simple solution to thisproblem consists in using nonpolar crystallographic orientations like m-plane{1�100} or a-plane {11�20}, or semipolar planes, which are those (hkil) planeswith at least two nonzero h, k, or i Miller indices and a nonzero l Miller index. Adetailed analysis of the strain and piezoelectric effects in nonpolar and semipolarplanar layers can be found in Ref. 112.

Near-IR ISB absorption has been reported on semipolar (11-22)-oriented GaN/AlNsuperlattices.113,114 In comparison to polar QWs, semipolar structures exhibit quasi-square potential band profiles with symmetric wavefunctions due to reduced electricfield of 0.5e0.6 MV/cm for QWs. The evolution of ISB transition energy with theQW thickness is represented in Fig. 18.7. In semipolar structures, the reduction inthe internal electric field results in a red shift of the ISB energy. The experimentaldata was obtained from identical polar and semipolar samples consisting of 40 periods


of GaN/AlN with 3 nm AlN barriers. The absorption full width at half maximum(FWHM ¼ 80e190 meV) is comparable to the one measured in polar structures.81

Regarding nonpolar orientations, first reports on ISB characterization referred toa-plane GaN/AlN superlattice grown on r-sapphire, which presented ISB absorptionat 2.1 mm, with FWHM ¼ 120 meV.115 Activities on nonpolar orientations increasedwhen bulk GaN substrates became commercially available. Comparative analysis of a-versus m-plane GaN/AlN superlattices grown on bulk GaN has favored research on them-plane, easier to grow with better structural properties.116 ISB absorption in GaN/AlGaN superlattices is demonstrated in the near- and mid-IR,116e118 as well as theTHz spectral region.119,120,121

18.4.6 Cubic III-nitrides

Another approach to eliminate the internal electrical field in III-nitride heterostruc-tures is the use of III-nitride semiconductors crystallize in the zinc-blend

1.0

0.8

0.6

0.41.0 1.5 2.0 2.5

QW thickness (nm)

Semipolar

e 1–e

2 (eV

)4

2

0

–2

–4

Strained on AIN Strained on GaN Strained on AIN Strained on GaN

e3e2e1

h1

2

0

–2

–4

Strained on GaNStrained on AIN

┴(11–22)

Ene

rgy

(eV

)

Ene

rgy

(eV

)

(0001)

Polar

Figure 18.7 Top: band diagram of (0001)- and (112�2)-oriented GaN/AlN (2.5/5 nm)superlattices assuming the structure is fully strained on AlN and on GaN. Down: Variation ofe2ee1 energy as a function of well width in polar and semipolar QWs strained on GaN andAlN.


structure. Furthermore, the LO phonon energy in cubic GaN is almost the same asin wurtzite GaN (92.7 meV122), whereas the effective mass is much smaller(m* ¼ 0.11 � 0.17m0

123,124) than in wurtzite GaN (m* ¼ 0.2m0), which shouldresult in higher gain and lower threshold current in QCLs. The cubic orientationcan be selected by PAMBE using 3C-SiC substrates. However, due to their thermo-dynamically unstable nature, cubic films present low structural quality with a highdensity of stacking faults. ISB absorption in the 1.40e4.0 mm spectral range hasbeen reported in cubic GaN/AlN QWs,125,126 in agreement with theoretical calcula-tions assuming a conduction band offset of 1.2 eV and an effective massm* ¼ 0.11m0. In cubic GaN/AlN MQWs exhibiting near-IR absorption centered at1.77 mm, ultrafast carrier relaxation (<100 fs) have been measured using pump-probe spectroscopy, showing third order nonlinear optical susceptibility.127,128 ISBTHz absorption at 4.7 THz has also been observed in cubic GaN/Al0.05Ga0.95N(12/15 nm) QWs.126

18.4.7 Quantum dots and nanowire heterostructures

An alternative approach to QWs for the fabrication of intraband devices is the use ofQD superlattices. Polar GaN QDs are hexagonal truncated pyramids with {10�13}facets, which form on top of a 2-ML-thick GaN wetting layer.129 In the case of polarGaN, QDs can be synthesized by PAMBE by deposition of GaN under compressivestrain and N-rich conditions.130,131 They can also be induced by a growth interruptionin vacuum when using ammonia-MBE,132 or generated by MOVPE under very lowV/III ratios.133 (Al)GaN/AlN QD superlattices with intraband absorption in the1.38e2.0 mm range (FWHM ¼ 80e100 meV) have been reported.131,134 This absorp-tion is attributed to electronic transitions from the ground state of the conduction band,s, to the first excited state confined along the growth axis, pz. Transitions associated tothe lateral confinement, s-px,y, have been observed in photocurrent experiments at lowtemperature.135

QDs can also be synthesized within nanowires, which allows better control of thedot height. Self-assembled GaN nanowires can be grown almost defect-free on siliconsubstrates with the large surface-to-volume ratio allowing misfit strain to be elasticallyreleased (see Fig. 18.8(a)e(c)). This expands the possibility of active region design interms of size and composition. Intraband absorption in the range of 1.4e1.95 mm,assigned to s-px, has been observed in GaN/AlN QDs inserted in GaN nanowiresgrown on Si(111) by PAMBE,136e138 as illustrated in Fig. 18.8(d) for nanowire heter-ostructures with different doping levels in the GaN disks.

18.4.8 Effect of doping

In order to observe ISB absorption, it is necessary to control the carrier concentrationin the QWs to guarantee that the first electronic level is populated. Doping is acritical parameter to reach the targeted operating wavelength, since the ISB absorptionenergy blue shifts markedly and broadens with increasing doping level due tomany-body effects, particularly to the exchange interaction and depolarization


shift,96,104,121,137,139e141 as illustrated in Fig. 18.5(b). In mid- and far-IR, the value ofthe spectral shift induced by dopants can be comparable to the nominal e2-e1 transitionenergy.

Delta doping is a technique that has been applied to reduce electron-impurity ionscattering and improve interface roughness, thereby reducing ISB line width.142 Inthe case of step-QWs, theoretical calculations predict that delta doping at the GaNwell/AlGaN step-well interface should cause a blue shift of the ISB transition, whereasdelta doping in the barrier or near the barrier/GaN well and barrier/step-well interfacesshould cause first a red shift first and then a blue shift with increasing dopingdensity.143

Although most of the studies on ISB transitions in III-nitrides have been performedusing silicon as n-type dopant, there has been a tremendous increase in the publications

T= 300K

TE TM BEP1.5×10–9 mbar1.0×10–9 mbar5.0×10–10 mbar

Ge

1

IR tr

ansm

issi

on

2 3 4Wavelength (μμm)

(b) (d)

(c)

(a)

300 nm

Figure 18.8 (a) Cross-section scanning electron microscopy and (b) high-angle annulardark-field scanning transmission electron microscopy images of Ge-doped GaN/AlN nanowireheterostructures. (c) High-resolution transmission electron microscopy image of several GaNnanodisks and AlN barriers in a nanowire. The presence of an AlN shell is indicated in theimage. The AlN barriers present {1�102} facets. (d) Room-temperature infrared (IR)transmission spectra for TE-(dashed lines) and transverse magnetic (TM)-polarized (solidlines) light measured for Ge-doped GaN/AlN (4/4 nm) nanowire heterostructures withdifferent doping levels in the nanodisks. The legend indicates the beam equivalent pressure ofGe (BEPGe) used for doping, which corresponds to nominal germanium concentrations of9 � 1019, 1.7 � 1020 and 3.1 � 1020 cm�3.Reprinted, with permission, from Beeler M, Hille P, Sch€ormann J, Teubert J, de la Mata M,Arbiol J, Eickhoff M, Monroy E. Intraband absorption in self-assembled Ge-Doped GaN/AlNnanowire heterostructures. Nano Lett 2014;14:1665e73. © 2014 American Chemical Society.


concerning germanium doped GaN.144,145 However, for structures with near-IR ISBabsorption, a comparison of germanium and silicon in planar layers and nanowiresshows no significant difference in the optical behavior of the structures as a functionof the nature of the dopant.137

18.5 ISB devices

18.5.1 All-optical switches

Thanks to the ultrafast ISB recovery time (in the 150e400 fs range78,146) associated tothe strong interaction of electrons with LO phonons, GaN/AlN QWs or QDs have beenproposed as active medium for all-optical switches (saturable absorbers) operating atTbit/s data rate at telecommunication wavelengths. These ultrafast all-optical devicesare of great interest for optical time division multiplexed systems. The switching isbased on the bleaching of the intersubband absorption by an intense control pulse.All-optical switching at 1.55 mmwith sub-ps commutation time has been demonstratedby several groups.90,91,147e151 As a best result, control switching energies of z8 pJwith 150 fs pulses for 10 dB modulation depth have been demonstrated.150 This resultis achieved in a GaN-on-AlN waveguide structure containing three periods of GaN/AlN QDs.

18.5.2 IR photodetectors

Room-temperature operation of near-IR photoconductive QW IR photodetectors(QWIPs) has been demonstrated.152 However, these devices present a low yield dueto the large dark current originated by the high density of threading dislocations in het-eroepitaxial III-nitrides (w109 cm�2). An alternative to bypass the leakage problem isthe fabrication of photovoltaic devices, where zero-bias operation guarantees a mini-mum dark current.92,153,154 The operation principle of photovoltaic ISB detectors isbased on a nonlinear optical rectification processes in asymmetric QWs.154 A strongperformance enhancement (responsivity increase by a factor of 60) of these detectorshas been achieved by using QDs instead of QWs in the active region.155 The improve-ment is attributed to the longer electron lifetime in the upper QD states and theincreased lateral electron displacement.

Lateral quantum dot IR photoconductors (QDIPs) have been fabricated in samplesconsisting of 20 periods of Si-doped GaN/AlN QDs. The devices exhibit near-IRphotocurrent only observed for TM-polarized light, following the intraband s-pz selec-tions rules.156,157 The appearance of photocurrent due to bound-to-bound intrabandtransitions within the QDs is attributed to lateral hopping conductivity.135

Finally, IR photodetectors based on cubic GaN/AlN QW superlattices have alsobeen reported.158 These devices exhibit photovoltaic effect that is overtaken by thedark current for temperatures above 215K. The photoresponse is consistent withISB transition phenomena, but the mechanism behind the photovoltaic behaviorremains unknown.


In the mid- and far-IR, the reduction of the lattice mismatch in the superlatticesmakes it more accessible to develop photoconductive QWIPs. GaN/AlGaN devicesoperating in the 3e5 mm window159 and in the THz domain160,161 have been reported.InGaN/(Al)GaN QWIPs have also been fabricated using free-standing nonpolarm-GaN substrates.162 These devices, consisting of In0.095Ga0.905N/Al0.07Ga0.93N andIn0.1Ga0.9N/GaN superlattices, displays photocurrent peaks at 7.5 and 9.3 mm,respectively.

18.5.3 Quantum cascade detectors

Quantum cascade detectors (QCDs) are photovoltaic devices consisting of severalperiods of an active QW coupled to a short-period superlattice which serves asextractor.163 Under illumination, electrons from the ground state are excited to theupper state of the active QW and then transferred to the extractor region where theyexperience multiple relaxations toward the next active QW. This results in a macro-scopic photovoltage in an open circuit configuration, or in a photocurrent if the deviceis loaded on a resistor. As major advantage, the dark current is extremely low and thecapacitance can be reduced by increasing the number of periods, which enables highfrequency response.

GaN/AlGaN QCDs operating in the near-IR have been reported,164,165 with thestructure illustrated in Fig. 18.9(a)e(c). These devices take advantage of thepolarization-induced internal electric field in the heterostructure to design an efficientAlGaN/AlN electron extractor where the energy levels are separated by approximately90 meV forming a phonon ladder. The peak responsivity of GaN/AlGaN quantumcascade detector at room temperature is z10 mA/W (z1000 V/W). Detectorscontaining 40 periods of active region with the size 17 � 17 mm2 exhibit the �3 dBcut-off frequency at 11.4 GHz.166 However, the speed of these quantum cascadedetectors is governed by the RC constant of the device and not by an intrinsic mech-anism. Pump and probe measurements of these devices showed relaxation times in therange of z1 ps, as shown in Fig. 18.9(d), which points to an available bandwidthexceeding 200 GHz.167 The response of such detectors at normal incidence can beenhanced by using a nanohole array integrated on top of the detector.168 GaN/AlGaNQCDs have also been demonstrated in the mid-IR, with a peak responsivity of 100 mA/W at 4 mm when operated at 140K.169

18.5.4 Light emitters

ISB luminescence has been observed both in GaN/AlN QWs and QDs under opticalpumping.170e173 Fig. 18.10 shows the emission at 2.1 mm wavelength obtainedfrom GaN/AlN (2/3 nm) QWs. It is important to remind that ISB PL is a very ineffi-cient process because of the very short nonradiative ISB relaxation lifetime. However,this does not prevent the realization of high performance ISB lasers because large stim-ulated gains can be achieved thanks to the high oscillator strength associated with ISBtransitions. The observation of ISB luminescence proves the feasibility of opticallypumped ISB emitting devices in the near-IR. However, in order to develop quantum


fountain lasers, further work is required in terms of growth optimization, processingand dedicated laser active region and cavity design.

Due to the large lattice mismatch between GaN and AlN, the fabrication of a GaN-based near-IR electrically pumped QCL does not appear feasible, despite severaltheoretical calculations.7,174,175 However, using a GaN/AlGaN quantum cascadestructure, Sont et al. managed to produced electroluminescence in the mid-IR (peakemission wavelength at 4.9 mm, with FWHM ¼ 110 meV) under pulsed operation at80K.176

There is an interest to push the operation of ISB nitride devices to longer wave-lengths, particularly to the THz frequency range. The THz spectral region is subjectto intensive research in view of its potential in a number of application domains

200 nm

×40

AIGaNAIN

GaN550 nm

1 μm

AIGaN:Si

AIGaN:Si

AIN

Sapphire

2 nm

Pho

tocu

rren

t (ar

b. u

nit) 0.8

0.6

0.4

0.2

0 –1 –0.5 0 0.5 1ΔT (psec)

1

0 5 10 15Growth axis (nm)

1.5

0.5

–0.5

1

0

(0001)

Ene

rgy

(eV

)

(a) (b) (c)

(d)

Figure 18.9 (a) Schematic description of a GaN/AlN/AlGaN quantum cascade detector (QCD).(b) HRTEM image of a period of the structure (active GaN quantum well [QW] followed byfive-period AlGaN/AlN extractor), viewed along the <11�20> axis. (c) Band diagram andenergy levels in one stage of the structure. (d) QCD photocurrent as a function of pump-probedelay at room temperature under zero bias conditions. Full line: simulation fit based on rateequations and phonon scattering theory.(b) and (c) Reprinted, with permission, from Vardi A, Bahir G, Guillot F, Bougerol C, MonroyE, Schacham SE, Tchernycheva M, Julien FH. Near infrared quantum cascade detector in GaN/AlGaN/AlN heterostructures. Appl Phys Lett 2008;92:011112. © 2008 American Institute ofPhysics and (d) Reprinted, with permission, from Vardi A, Sakr S, Mangeney J, KandaswamyPK, Monroy E, Tchernycheva M, Schacham SE, Julien FH, Bahir G. Femto-second electrontransit time characterization in GaN/AlGaN quantum cascade detector at 1.5 micron. Appl PhysLett 2011;99:202111. © 2011 American Institute of Physics.


such as medical diagnostics, security screening or quality control. In terms of sources,QCLs based on GaAs/AlGaAs QWs have emerged as excellent candidates for appli-cations requiring a few tens-of-mW power in the 1.2 to 5 THz spectral range.177

Although much progress has been accomplished in terms of QCL performance, themaximum operating temperatures reported so far (z199.5 and 129K for pulsed andcontinuous wave operation178,179) are still too low for widespread applications. Oneintrinsic reason limiting the temperature is the small energy of the LO phonon inGaAs (36 meV, 8.2 THz), which hinders laser action close to room temperaturebecause of thermally activated LO-phonon scattering. Wide band gap semiconductormaterials such as GaN, with an LO-phonon energy of 92 meV (22.3 THz), should pavethe way for THz QCLs operating above room temperature. Furthermore, the large GaNLO-phonon energy opens prospects for QCLs at wavelengths inaccessible to other III-V semiconductors due to Reststrahlen absorption.

There are a number of theoretical proposals for nitride devices operating in the far-IR region,180e190 all focusing on the resonant-phonon architecture,191 as illustrated inFig. 18.11. However, the demonstration of a nitride-based THz QCL must still facetwo major challenges: lack of a more robust design of the active QWs to obtain ISBtransitions in the far-IR, and the presence of leakage currents associated to threadingdislocations arising from the heteroepitaxial growth of GaN. So far, only Terashimaand Hirayama have reported THz electroluminescence from QCL structures.192,193

The availability of single-crystal bulk GaN substrates with various crystallographicorientations has brought new hopes for the development of a nitride-based QCL inthe near future.

2 2.2 2.4 2.6 2.8Wavelength (μm)

e 3 –

e 2 p

hoto

lum

ines

cenc

e (a

rb.u

nit)

Figure 18.10 Intersubband (ISB) photoluminescence (PL) from 2-nm-thick GaN/AlN quantumwells (QWs). The inset illustrates the conduction band energy levels.Reprinted, with permission, from Nevou L, Tchernycheva M, Julien FH, Guillot F, Monroy E.Short wavelength (l ¼ 2.13 mm) intersubband luminescence from GaN∕AlN quantum wells atroom temperature. Appl Phys Lett 2007;90:121106. © 2017 American Institute of Physics.


18.6 Conclusions

In this chapter, we have reviewed research on III-nitride IR technologies, namely highIn-content InGaN optoelectronics, RE incorporation and GaN/Al(Ga)N ISB approach.Driving the InGaN LED technology from blue toward longer wavelengths results in asignificant drop in internal quantum efficiency due to problems associated to the largelattice mismatch between InN and GaN, and the challenges associated to p-type dopingand contacting high-In-content InGaN layers. Approaches to circumvent these prob-lems lead toward nanostructuration of the active region via introduction of QD ornanocolumn arrays. Devices operating in the near IR, around 1.5 mm wavelength,have been demonstrated, but their reliability and performance is not yet competitivewith other IR technologies. Nonetheless, research on InN has revealed the potentialof this semiconductor as for the development of far-IR sources. Broad THz pulsesare efficiently generated by irradiation of the InN surface with femtosecond laserpulses.

Several groups have demonstrated IR light emitting devices by incorporating REelements in GaN in order to achieve luminescence via 4f intra-atomic transitions inthese ions. Electroluminescent devices operating at the technologically important1.55 mm wavelength have been demonstrated, but they require high bias voltage(>100 V) and present relatively low efficiency.

Finally, ISB optoelectronics emerged as a potential application field for III-nitridematerials. III-nitride heterostructures are excellent candidates for high-speed ISBdevices in the near-IR thanks to their large conduction band offset (z1.8 eV for theGaN/AlN system) and subpicosecond ISB scattering rates. However, band gap

40

35

30

25

20

15

100 50 100 150 200 250 300 350 400

Temperature (K)

GaN/AlGaNZnO/MgZnOGaAs/AIGaAs

Pop

ulat

ion

inve

rsio

n Δn

(%)

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0–400 –350 –300 –250 –200 –150 –100 –50 0

Growth axis (A)

Ene

rgy

(eV

)

1 period

Ι3>Ι2>Ι1>

Ι4>

(a) (b)

Figure 18.11 (a) Conduction-band profile and squared envelope functions of a GaN/Al0.15Ga0.85N THz quantum cascade laser (QCL) design. (b) Calculated fractional populationinversion of THz QCL structures based on various material families, as a function oftemperature.Reprinted, with permission, from Bellotti E, Driscoll K, Moustakas TD, Paiella R. Monte Carlosimulation of terahertz quantum cascade laser structures based on wide-bandgapsemiconductors. J Appl Phys 2009;105:113103. © 2009 American Institute of Physics.


engineering requires an exquisite control of material growth and modeling that arenotoriously difficult in GaN/AlGaN. First prototypes of nitride-based ISB deviceswere room-temperature multi-Tbit/s all-optical switches operating at 1.5 mm,photovoltaic and photoconductive quantum well IR photodetectors, quantum dot IRphotodetectors, and ISB electro-optical modulators. The concept of quantum cascadeapplied to III-nitrides has been demonstrated by the development of QCDs operating inthe 1.5e2.0 mm spectral range. Near-IR ISB luminescence from GaN/AlN QWs andQDs has been reported, and ISB electroluminescence from a quantum cascade struc-ture has been observed in the mid-IR (peak emission wavelength at 4.9 mm).

An emerging field for GaN-based ISB devices is the extension toward the far-IRspectral range, with several theoretical designs of a GaN-based THz QCL have beenreported. At far-IR wavelengths, the large GaN LO-phonon energy (92 meV) becomesa valuable property to achieve ISB operation at relatively high temperatures, and alsoto cover IR wavelengths that are not accessible by other IIIeV semiconductors due toReststrahlen absorption. However, the extension of this ISB technology toward longerwavelengths requires a reduction of the polarization-induced internal electric field,which sets new material challenges.

References

1. Davydov VY, Klochikhin AA, Seisyan RP, Emtsev VV, Ivanov SV, Bechstedt F,Furthmuller J, Harima H, Mudryi AV, Aderhold J, Semchinova O, Graul J. Absorption andemission of hexagonal InN. Evid Narrow Fundam Band Gap Phys Status Solidi B 2002;229:r1e3.

2. Wu J. When group-III nitrides go infrared: new properties and perspectives. J Appl Phys2009;106:011101.

3. Dahal R, Ugolini C, Lin JY, Jiang HX, Zavada JM. 1.54 mm emitters based on erbiumdoped InGaN p-i-n junctions. Appl Phys Lett 2010;97:141109.

4. O’Donnell K, Dierolf V, editors. Rare earth doped III-nitrides for optoelectronic andspintronic applications, topics in applied physics. Dordrecht, the Netherlands, New York;Bristol, UK: Springer; In Association with Canopus Academic Pub; 2010.

5. Steckl AJ, Heikenfeld JC, Lee D-S, Garter MJ, Baker CC, Yongqiang Wang, Jones R.Rare-earth-doped GaN: growth, properties, and fabrication of electroluminescent devices.IEEE J Sel Top Quan Electron 2002;8:749e66.

6. Beeler M, Trichas E, Monroy E. III-nitride semiconductors for intersubband optoelec-tronics: a review. Semicond Sci Technol 2013;28:074022.

7. Hofstetter D, Baumann E, Giorgetta FR, Théron R, Wu H, Schaff WJ, Dawlaty J,George PA, Eastman LF, Rana F, Kandaswamy PK, Guillot F, Monroy E. Intersubbandtransition-based processes and devices in AlN/GaN-based heterostructures. Proc IEEE2010;98:1234e48.

8. Machhadani H, Kandaswamy P, Sakr S, Vardi A, Wirtm€uller A, Nevou L, Guillot F,Pozzovivo G, Tchernycheva M, Lupu A, Vivien L, Crozat P, Warde E, Bougerol C,Schacham S, Strasser G, Bahir G, Monroy E, Julien FH. GaN/AlGaN intersubbandoptoelectronic devices. New J Phys 2009;11:125023.

9. Damilano B, Gil B. Yellowered emission from (Ga,In)N heterostructures. J Phys ApplPhys 2015;48:403001.


10. Saito S, Hashimoto R, Hwang J, Nunoue S. InGaN light-emitting diodes on c-face sapphiresubstrates in green gap spectral range. Appl Phys Express 2013;6:111004.

11. Barletta PT, Acar Berkman E, Moody BF, El-Masry NA, Emara AM, Reed MJ,Bedair SM. Development of green, yellow, and amber light emitting diodes using InGaNmultiple quantum well structures. Appl Phys Lett 2007;90:151109.

12. Franssen G, Gorczyca I, Suski T, Kami�nska A, Pereiro J, Mu~noz E, Iliopoulos E,Georgakilas A, Che SB, Ishitani Y, Yoshikawa A, Christensen NE, Svane A. Bowing ofthe band gap pressure coefficient in InxGa1�xN alloys. J Appl Phys 2008;103:033514.

13. Valdueza-Felip S, Bellet-Amalric E, N�u~nez-Cascajero A, Wang Y, Chauvat M-P,Ruterana P, Pouget S, Lorenz K, Alves E, Monroy E. High In-content InGaN layerssynthesized by plasma-assisted molecular-beam epitaxy: growth conditions, strain relax-ation, and in incorporation kinetics. J Appl Phys 2014;116:233504.

14. Ohkawa K,Watanabe T, Sakamoto M, Hirako A, Deura M. 740-nm emission from InGaN-based LEDs on c-plane sapphire substrates by MOVPE. J Cryst Growth 2012;343:13e6.

15. Xu T, Nikiforov AY, France R, Thomidis C, Williams A, Moustakas TD. Blue-green-redLEDs based on InGaN quantum dots grown by plasma-assisted molecular beam epitaxy.Phys Status Solidi A 2007;204:2098e102.

16. Kikuchi A, Kawai M, Tada M, Kishino K. InGaN/GaN multiple quantum disk nano-column light-emitting diodes grown on (111) Si substrate. Jpn J Appl Phys 2004;43:L1524e6.

17. Sekiguchi H, Kishino K, Kikuchi A. Emission color control from blue to red with nano-column diameter of InGaN/GaN nanocolumn arrays grown on same substrate. Appl PhysLett 2010;96:231104.

18. Lv W, Wang L, Wang J, Xing Y, Zheng J, Yang D, Hao Z, Luo Y. Green and red light-emitting diodes based on multilayer InGaN/GaN dots grown by growth interruptionmethod. Jpn J Appl Phys 2013;52:08JG13.

19. Hong YJ, Lee C-H, Yoon A, Kim M, Seong H-K, Chung HJ, Sone C, Park YJ, Yi G-C.Visible-Color-tunable light-emitting diodes. Adv Mater 2011;23:3284e8.

20. Che S-B, Mizuno T, Wang X, Ishitani Y, Yoshikawa A. Fabrication and properties ofcoherent-structure In-polarity InN/In0.7Ga0.3N multiquantum wells emitting at around1.55 mm. J Appl Phys 2007;102:083539.

21. Grandal J, Pereiro J, Bengoechea-Encabo A, Fern�andez-Garrido S, S�anchez-García MA,Mu~noz E, Calleja E, Luna E, Trampert A. InN/InGaN multiple quantum wells emitting at1.5 mm grown by molecular beam epitaxy. Appl Phys Lett 2011;98:061901.

22. Naranjo FB, Kandaswamy PK, Valdueza-Felip S, Calvo V, Gonz�alez-Herr�aez M, Martín-L�opez S, Corredera P, Méndez JA, Mutta GR, Lacroix B, Ruterana P, Monroy E.Nonlinear absorption of InN/InGaN multiple-quantum-well structures at optical telecom-munication wavelengths. Appl Phys Lett 2011;98:031902.

23. Kishino K, Kamimura J, Kamiyama K. Near-infrared InGaN nanocolumn light-emittingdiodes operated at 1.46 mm. Appl Phys Express 2012;5:031001.

24. Koblm€uller G, Gallinat CS, Speck JS. Surface kinetics and thermal instability of N-faceInN grown by plasma-assisted molecular beam epitaxy. J Appl Phys 2007;101:083516.

25. Xu K, Yoshikawa A. Effects of film polarities on InN growth by molecular-beam epitaxy.Appl Phys Lett 2003;83:251e3.

26. Keller S, Fichtenbaum NA, FurukawaM, Speck JS, DenBaars SP, Mishra UK. Growth andcharacterization of N-polar InGaN/GaN multiquantum wells. Appl Phys Lett 2007;90:191908.

27. Nath DN, G€ur E, Ringel SA, Rajan S. Molecular beam epitaxy of N-polar InGaN. ApplPhys Lett 2010;97:071903.


28. Shojiki K, Tanikawa T, Choi J-H, Kuboya S, Hanada T, Katayama R, Matsuoka T. Red toblue wavelength emission of N-polar (0001) InGaN light-emitting diodes grown bymetalorganic vapor phase epitaxy. Appl Phys Express 2015;8:061005.

29. Speck JS, Chichibu SF. Nonpolar and semipolar group III nitride-based materials. MRSBull 2009;34:304e12.

30. Bhat R, Guryanov GM. Experimental study of the orientation dependence of indiumincorporation in GaInN. J Cryst Growth 2016;433:7e12.

31. De Mierry P, Guehne T, Nemoz M, Chenot S, Beraudo E, Nataf G. Comparison betweenpolar (0001) and semipolar (11\bar22) nitride blueegreen light-emitting diodes grown onc- and m-plane sapphire substrates. Jpn J Appl Phys 2009;48:031002.

32. Wang Y, Shimma R, Yamamoto T, Hayashi H, Shiohama K, Kurihara K, Hasegawa R,Ohkawa K. The effect of plane orientation on indium incorporation into InGaN/GaNquantum wells fabricated by MOVPE. J Cryst Growth 2015;416:164e8.

33. Wernicke T, Schade L, Netzel C, Rass J, Hoffmann V, Ploch S, Knauer A, Weyers M,Schwarz U, Kneissl M. Indium incorporation and emission wavelength of polar, nonpolarand semipolar InGaN quantum wells. Semicond Sci Technol 2012;27:024014.

34. Bai J, Xu B, Guzman FG, Xing K, Gong Y, Hou Y, Wang T. (11�22) semipolar InGaNemitters from green to amber on overgrown GaN on micro-rod templates. Appl Phys Lett2015;107:261103.

35. Koslow IL, Hardy MT, Shan Hsu P, Dang P-Y, Wu F, Romanov A, Wu Y-R, Young EC,Nakamura S, Speck JS, DenBaars SP. Performance and polarization effects in (112) longwavelength light emitting diodes grown on stress relaxed InGaN buffer layers. Appl PhysLett 2012;101:121106.

36. Sato H, Chung RB, Hirasawa H, Fellows N, Masui H, Wu F, Saito M, Fujito K, Speck JS,DenBaars SP, Nakamura S. Optical properties of yellow light-emitting diodes grown onsemipolar (112) bulk GaN substrates. Appl Phys Lett 2008;92:221110.

37. Kawaguchi Y, Huang C-Y, Wu Y-R, Zhao Y, DenBaars SP, Nakamura S. Semipolar(20-21) single-quantum-well red light-emitting diodes with a low forward voltage. Jpn JAppl Phys 2013;52:08JC08.

38. Yang KH, Richards PL, Shen YR. Generation of far-infrared radiation by picosecond lightpulses in LiNbO3. Appl Phys Lett 1971;19:320e3.

39. Asc�azubi R, Wilke I, Denniston K, Lu H, Schaff WJ. Terahertz emission by InN. ApplPhys Lett 2004;84:4810e2.

40. Chern GD, Readinger ED, Shen H, Wraback M, Gallinat CS, Koblm€uller G, Speck JS.Excitation wavelength dependence of terahertz emission from InN and InAs. Appl PhysLett 2006;89:141115.

41. Matth€aus G, Cimalla V, Pradarutti B, Riehemann S, Notni G, Lebedev V, Ambacher O,Nolte S, T€unnermann A. Highly efficient THz emission from differently grown InN at800 nm and 1060 nm excitation. Opt Commun 2008;281:3776e80.

42. Mu X, Ding YJ, Wang K, Jena D, Zotova YB. Resonant terahertz generation from InN thinfilms. Opt Lett 2007;32:1423.

43. Shubina TV, Andrianov AV, Zakhar’in AO, Jmerik VN, Soshnikov IP, Komissarova TA,Usikova AA, Kop’ev PS, Ivanov SV, Shalygin VA, Sofronov AN, Firsov DA, Vorob’ev LE,Gippius NA, Leymarie J, Wang X, Yoshikawa A. Terahertz electroluminescence of surfaceplasmons from nanostructured InN layers. Appl Phys Lett 2010;96:183106.

44. Sun G, Xu G, Ding YJ, Zhao H, Liu G, Zhang J, Tansu N. Efficient terahertz generationwithin InGaN/GaN multiple quantum wells. IEEE J Sel Top Quan Electron 2011;17:48e53.


45. Sun G, Chen R, Ding YJ, Nguyen HPT, Mi Z. InGaN/GaN dot-in-a-wire: ultimateterahertz nanostructure. Laser Photon Rev 2015;9:105e13.

46. Turchinovich D, Uhd Jepsen P, Monozon BS, Koch M, Lahmann S, Rossow U,Hangleiter A. Ultrafast polarization dynamics in biased quantum wells under strongfemtosecond optical excitation. Phys Rev B 2003;68.

47. Turchinovich D, Monozon BS, Jepsen PU. Role of dynamical screening in excitationkinetics of biased quantum wells: nonlinear absorption and ultrabroadband terahertzemission. J Appl Phys 2006;99:013510.

48. Wang XQ, Zhao GZ, Zhang Q, Ishitani Y, Yoshikawa A, Shen B. Effect of Mg doping onenhancement of terahertz emission from InN with different lattice polarities. Appl Phys Lett2010;96:061907.

49. Wilke I, Ding YJ, Shubina TV. Optically- and electrically-stimulated terahertz radiationemission from indium nitride. J Infrared Millim Terahertz Waves 2012;33:559e92.

50. Xu G, Ding YJ, Zhao H, Liu G, Jamil M, Tansu N, Zotova IB, Stutz CE, Diggs DE,Fernelius N, Hopkins FK, Gallinat CS, Koblm€uller G, Speck JS. THz generation from InNfilms due to destructive interference between optical rectification and photocurrent surge.Semicond Sci Technol 2010;25:015004.

51. Xu G, Sun G, Ding YJ, Zotova IB, Jamil M, Ferguson IT. Mechanism for THz generationfrom InN micropyramid emitters. J Appl Phys 2011;109:093111.

52. Ennen H, Schneider J, Pomrenke G, Axmann A. 1.54-mm luminescence of erbium-implanted III-V semiconductors and silicon. Appl Phys Lett 1983;43:943e5.

53. Favennec PN, L’Haridon H, Salvi M, Moutonnet D, Le Guillou Y. Luminescence oferbium implanted in various semiconductors: IV, III-V and II-VI materials. Electron Lett1989;25:718e9.

54. Ruterana P, Lacroix B, Lorenz K. A mechanism for damage formation in GaN during rareearth ion implantation at medium range energy and room temperature. J Appl Phys 2011;109:013506.

55. Wilson RG, Schwartz RN, Abernathy CR, Pearton SJ, Newman N, Rubin M, Fu T,Zavada JM. 1.54-mm photoluminescence from Er-implanted GaN and AlN. Appl Phys Lett1994;65:992e4.

56. Hori Y, Biquard X, Monroy E, Jalabert D, Enjalbert F, Dang LS, Tanaka M, Oda O,Daudin B. GaN quantum dots doped with Eu. Appl Phys Lett 2004;84:206e8.

57. Hori Y, Andreev T, Jalabert D, Monroy E, Dang LS, Daudin B, Tanaka M, Oda O. GaNquantum dots doped with Tb. Appl Phys Lett 2006;88:053102.

58. MacKenzie JD, Abernathy CR, Pearton SJ, H€ommerich U, Wu X, Schwartz RN,Wilson RG, Zavada JM. Er doping of III-nitrides during growth by metalorganic molecularbeam epitaxy. J Cryst Growth 1997;175e176:84e8.

59. Lorenz K, Barradas NP, Alves E, Roqan IS, Nogales E, Martin RW, O’Donnell KP,Gloux F, Ruterana P. Structural and optical characterization of Eu-implanted GaN. J PhysAppl Phys 2009;42:165103.

60. Birkhahn R, Hudgins R, Lee D, Steckl AJ, Molnar RJ, Saleh A, Zavada JM. Growth andmorphology of Er-doped GaN on sapphire and hydride vapor phase epitaxy substrates.J Vac Sci Technol B Microelectron Nanometer Struct 1999;17:1195.

61. Bang H, Morishima S, Li Z, Akimoto K, Nomura M, Yagi E. MBE growth of Eu- orTb-doped GaN and its optical properties. J Cryst Growth 2002;237e239:1027e31.

62. Monteiro T, Boemare C, Soares MJ, S�a Ferreira RA, Carlos LD, Lorenz K, Vianden R,Alves E. Photoluminescence and lattice location of Eu and Pr implanted GaN samples.Phys B Condens Matter 2001;308e310:22e5.


63. George DK, Hawkins MD, McLaren M, Jiang HX, Lin JY, Zavada JM, Vinh NQ.Excitation mechanisms of Er optical centers in GaN epilayers. Appl Phys Lett 2015;107:171105.

64. Steckl AJ, Garter M, Lee DS, Heikenfeld J, Birkhahn R. Blue emission from Tm-dopedGaN electroluminescent devices. Appl Phys Lett 1999;75:2184e6.

65. Garter M, Scofield J, Birkhahn R, Steckl AJ. Visible and infrared rare-earth-activatedelectroluminescence from indium tin oxide Schottky diodes to GaN: Er on Si. Appl PhysLett 1999;74:182e4.

66. Birkhahn R, Garter M, Steckl AJ. Red light emission by photoluminescence and elec-troluminescence from Pr-doped GaN on Si substrates. Appl Phys Lett 1999;74:2161e3.

67. Kim JH, Holloway PH. Near-infrared electroluminescence at room temperature fromneodymium-doped gallium nitride thin films. Appl Phys Lett 2004;85:1689e91.

68. Steckl AJ, Park JH, Zavada JM. Prospects for rare earth doped GaN lasers on Si. MaterToday 2007;10:20e7.

69. Kim JH, Holloway PH. Near-infrared-electroluminescent light-emitting planar opticalsources based on gallium nitride doped with rare earths. Adv Mater 2005;17:91e6.

70. West LC, Eglash SJ. First observation of an extremely large-dipole infrared transitionwithin the conduction band of a GaAs quantum well. Appl Phys Lett 1985;46:1156e8.

71. Levine BF. Quantum-well infrared photodetectors. J Appl Phys 1993;74:R1e81.72. Faist J, Capasso F, Sivco DL, Sirtori C, Hutchinson AL, Cho AY. Quantum cascade laser.

Science 1994;264:553e6.73. Vitiello MS, Scalari G, Williams B, De Natale P. Quantum cascade lasers: 20 years of

challenges. Opt Express 2015;23:5167.74. Bastard G. Wave mechanics applied to semiconductor heterostructures. Les Editions de

Physique. Les Ulis Cedex, France; New York, N.Y: Halsted Press; 1988.75. Helm M. The basic physics of intersubband transitions. In: Liu HC, Capasso F, editors.

Intersubband transitions in quantum wells: physics and device applications I. San Diego:Acad. Press; 2000.

76. Tchernycheva M, Nevou L, Doyennette L, Julien F, Warde E, Guillot F, Monroy E, Bellet-Amalric E, Remmele T, Albrecht M. Systematic experimental and theoretical investigationof intersubband absorption in GaN∕AlN quantum wells. Phys Rev B 2006;73:125347.

77. Hamazaki J, Matsui S, Kunugita H, Ema K, Kanazawa H, Tachibana T, Kikuchi A,Kishino K. Ultrafast intersubband relaxation and nonlinear susceptibility at 1.55 mm inGaN/AlN multiple-quantum wells. Appl Phys Lett 2004;84:1102e4.

78. Iizuka N, Kaneko K, Suzuki N, Asano T, Noda S, Wada O. Ultrafast intersubband relax-ation (�150 fs) in AlGaN/GaN multiple quantum wells. Appl Phys Lett 2000;77:648e50.

79. Bernardini F, Fiorentini V, Vanderbilt D. Spontaneous polarization and piezoelectricconstants of III-V nitrides. Phys Rev B 1997;56:R10024e7.

80. Birner S, Zibold T, Andlauer T, Kubis T, Sabathil M, Trellakis A, Vogl P. Nextnano:general purpose 3-d simulations. IEEE Trans Electron Devices 2007;54:2137e42.

81. Kandaswamy PK, Guillot F, Bellet-Amalric E, Monroy E, Nevou L, Tchernycheva M,Michon A, Julien FH, Baumann E, Giorgetta FR, Hofstetter D, Remmele T, Albrecht M,Birner S, Dang LS. GaN/AlN short-period superlattices for intersubband optoelectronics: asystematic study of their epitaxial growth, design, and performance. J Appl Phys 2008;104:093501.

82. Sarigiannidou E, Monroy E, Gogneau N, Radtke G, Bayle-Guillemaud P, Bellet-Amalric E, Daudin B, Rouvi�ere JL. Comparison of the structural quality in Ga-face andN-face polarity GaN/AlN multiple-quantum-well structures. Semicond Sci Technol 2006;21:612e8.


83. Adelmann C, Brault J, Mula G, Daudin B, Lymperakis L, Neugebauer J. Galliumadsorption on (0001) GaN surfaces. Phys Rev B 2003;67:165419.

84. Heying B, Averbeck R, Chen LF, Haus E, Riechert H, Speck JS. Control of GaN surfacemorphologies using plasma-assisted molecular beam epitaxy. J Appl Phys 2000;88:1855e60.

85. Neugebauer J, Zywietz T, Scheffler M, Northrup J, Chen H, Feenstra R. Adatom kineticson and below the surface: the existence of a new diffusion channel. Phys Rev Lett 2003;90:056101.

86. Iliopoulos E, Moustakas TD. Growth kinetics of AlGaN films by plasma-assistedmolecular-beam epitaxy. Appl Phys Lett 2002;81:295e7.

87. Kandaswamy PK, Bougerol C, Jalabert D, Ruterana P, Monroy E. Strain relaxation inshort-period polar GaN/AlN superlattices. J Appl Phys 2009;106:013526.

88. Kotsar Y, Doisneau B, Bellet-Amalric E, Das A, Sarigiannidou E, Monroy E. Strainrelaxation in GaN/AlxGa1�xN superlattices grown by plasma-assisted molecular-beamepitaxy. J Appl Phys 2011;110:033501.

89. Sugiura L. Dislocation motion in GaN light-emitting devices and its effect on devicelifetime. J Appl Phys 1997;81:1633e938.

90. Iizuka N, Kaneko K, Suzuki N. Polarization dependent loss in III-nitride optical wave-guides for telecommunication devices. J Appl Phys 2006;99:093107.

91. Iizuka N, Kaneko K, Suzuki N. All-optical switch utilizing intersubband transition in GaNquantum wells. IEEE J Quan Electron 2006;42:765e71.

92. Baumann E, Giorgetta FR, Hofstetter D, Golka S, Schrenk W, Strasser G, Kirste L,Nicolay S, Feltin E, Carlin JF, Grandjean N. Near infrared absorption and room temper-ature photovoltaic response in AlN/GaN superlattices grown by metal-organic vapor-phaseepitaxy. Appl Phys Lett 2006;89:041106.

93. Bayram C. High-quality AlGaN/GaN superlattices for near- and mid-infrared intersubbandtransitions. J Appl Phys 2012;111:013514.

94. Nicolay S, Feltin E, Carlin J-F, Grandjean N, Nevou L, Julien FH, Schmidbauer M,Remmele T, Albrecht M. Strain-induced interface instability in GaN∕AlN multiplequantum wells. Appl Phys Lett 2007;91:061927.

95. Kandaswamy PK, Machhadani H, Bougerol C, Sakr S, Tchernycheva M, Julien FH,Monroy E. Midinfrared intersubband absorption in GaN/AlGaN superlattices on Si(111)templates. Appl Phys Lett 2009;95:141911.

96. Kotani T, Arita M, Arakawa Y. Doping dependent blue shift and linewidth broadening ofintersubband absorption in non-polar m-plane AlGaN/GaN multiple quantum wells. ApplPhys Lett 2015;107:112107.

97. Bayram C, Péré-laperne N, Razeghi M. Effects of well width and growth temperature onoptical and structural characteristics of AlN/GaN superlattices grown by metal-organicchemical vapor deposition. Appl Phys Lett 2009;95:201906.

98. Gmachl C, Ng HM, Cho AY. Intersubband absorption in GaN/AlGaN multiple quantumwells in the wavelength range of l w 1.75e4.2 mm. Appl Phys Lett 2000;77:334e6.

99. Gmachl C, Ng HM, Cho AY. Intersubband absorption in degenerately doped GaN/Alx-Ga1�xN coupled double quantum wells. Appl Phys Lett 2001;79:1590e2.

100. Péré-Laperne N, Bayram C, Nguyen-The L, McClintock R, Razeghi M. Tunability ofintersubband absorption from 4.5 to 5.3 mm in a GaN/Al0.2Ga0.8N superlattices grown bymetalorganic chemical vapor deposition. Appl Phys Lett 2009;95:131109.

101. Sherliker B, Halsall M, Kasalynas I, Seliuta D, Valusis G, Vengris M, Barkauskas M,Sirutkaitis V, Harrison P, Jovanovic VD, Indjin D, Ikonic Z, Parbrook PJ, Wang T,Buckle PD. Room temperature operation of AlGaN/GaN quantum well infrared photo-detectors at a 3e4 mm wavelength range. Semicond Sci Technol 2007;22:1240e4.


102. Machhadani H, Kotsar Y, Sakr S, Tchernycheva M, Colombelli R, Mangeney J, Bellet-Amalric E, Sarigiannidou E, Monroy E, Julien FH. Terahertz intersubband absorptionin GaN/AlGaN step quantum wells. Appl Phys Lett 2010;97:191101.

103. Beeler M, Bougerol C, Bellet-Amalric E, Monroy E. Terahertz absorbing AlGaN/GaNmulti-quantum-wells: demonstration of a robust 4-layer design. Appl Phys Lett 2013;103:091108.

104. Beeler M, Bougerol C, Bellet-Amalric E, Monroy E. Pseudo-square AlGaN/GaN quantumwells for terahertz absorption. Appl Phys Lett 2014;105:131106.

105. Edmunds C, Tang L, Li D, Cervantes M, Gardner G, Paskova T, Manfra MJ, Malis O.Near-infrared absorption in lattice-matched AlInN/GaN and strained AlGaN/GaN hetero-structures grown byMBE on low-defect GaN substrates. J Electron Mater 2012;41:881e6.

106. Edmunds C, Tang L, Cervantes M, Shirazi-HD M, Shao J, Grier A, Valavanis A,Cooper JD, Li D, Gardner G, Zakharov DN, Ikoni�c Z, Indjin D, Harrison P, Manfra MJ,Malis O. Comparative study of intersubband absorption in AlGaN/GaN and AlInN/GaNsuperlattices: impact of material inhomogeneities. Phys Rev B 2013;88:235306.

107. Malis O, Edmunds C, Manfra MJ, Sivco DL. Near-infrared intersubband absorption inmolecular-beam epitaxy-grown lattice-matched InAlN/GaN superlattices. Appl Phys Lett2009;94:161111.

108. Nicolay S, Carlin J-F, Feltin E, Butté R, Mosca M, Grandjean N, Ilegems M,Tchernycheva M, Nevou L, Julien FH. Midinfrared intersubband absorption in lattice-matched AlInN/GaN multiple quantum wells. Appl Phys Lett 2005;87:111106.

109. Nicolay S, Feltin E, Carlin J-F, Mosca M, Nevou L, Tchernycheva M, Julien FH,IlegemsM, Grandjean N. Indium surfactant effect on AlN∕GaN heterostructures grown bymetal-organic vapor-phase epitaxy: applications to intersubband transitions. Appl PhysLett 2006;88:151902.

110. Gonschorek M, Carlin J-F, Feltin E, Py MA, Grandjean N, Darakchieva V, Monemar B,Lorenz M, Ramm G. Two-dimensional electron gas density in Al1�xInxN/AlN/GaN het-erostructures (0.03�x�0.23). J Appl Phys 2008;103:093714.

111. Cywi�nski G, Skierbiszewski C, Fedunieiwcz- _Zmuda A, Siekacz M, Nevou L,Doyennette L, Tchernycheva M, Julien FH, Prystawko P, Kry�sko M, Grzanka S,Grzegory I, Presz A, Domagała JZ, Smalc J, Albrecht M, Remmele T, Porowski S. Growthof thin AlInN/GaInN quantum wells for applications to high-speed intersubband devices attelecommunication wavelengths. J Vac Sci Technol B 2006;24:1505e9.

112. Romanov AE, Baker TJ, Nakamura S, Speck JS, ERATO/JST UCSB Group. Strain-induced polarization in wurtzite III-nitride semipolar layers. J Appl Phys 2006;100:023522.

113. Lahourcade L, Kandaswamy PK, Renard J, Ruterana P, Machhadani H, Tchernycheva M,Julien FH, Gayral B, Monroy E. Interband and intersubband optical characterization ofsemipolar (11-22)-oriented GaN/AlN multiple-quantum-well structures. Appl Phys Lett2008;93:111906.

114. Machhadani H, Beeler M, Sakr S, Warde E, Kotsar Y, Tchernycheva M, Chauvat MP,Ruterana P, Nataf G, De Mierry P, Monroy E, Julien FH. Systematic study of near-infraredintersubband absorption of polar and semipolar GaN/AlN quantum wells. J Appl Phys2013;113:143109.

115. Gmachl C, Ng HM. Intersubband absorption at w2.1 mm in A-plane GaN∕AlN multiplequantum wells. Electron Lett 2003;39:567e9.

116. Lim CB, Beeler M, Ajay A, L€ahnemann J, Bellet-Amalric E, Bougerol C, Monroy E.Intersubband transitions in nonpolar GaN/Al(Ga)N heterostructures in the short and mid-wavelength infrared regions. J Appl Phys 2015;118:014309.


117. Kotani T, Arita M, Arakawa Y. Observation of mid-infrared intersubband absorption innon-polar m-plane AlGaN/GaN multiple quantum wells. Appl Phys Lett 2014;105:261108.

118. Kotani T, Arita M, Hoshino K, Arakawa Y. Temperature dependence of mid-infraredintersubband absorption in AlGaN/GaN multiple quantum wells. Appl Phys Lett 2016;108:052102.

119. Lim CB, Ajay A, Bougerol C, Haas B, Sch€ormann J, Beeler M, L€ahnemann J, Eickhoff M,Monroy E. Nonpolar m-plane GaN/AlGaN heterostructures with intersubband transitionsin the 5e10 THz band. Nanotechnology 2015;26:435201.

120. Edmunds C, Shao J, Shirazi-HD M, Manfra MJ, Malis O. Terahertz intersubbandabsorption in non-polar m-plane AlGaN/GaN quantum wells. Appl Phys Lett 2014;105:021109.

121. Lim CB, Ajay A, Bougerol C, L€ahnemann J, Donatini F, Sch€ormann J, Bellet-Amalric E,Browne DA, Jiménez-Rodríguez M, Monroy E. Effect of doping on the far-infraredintersubband transitions in nonpolar m-plane GaN/AlGaN heterostructures. Nanotech-nology 2016;27:145201.

122. Brazis R, Raguotis R. Monte Carlo modeling of phonon-assisted carrier transport in cubicand hexagonal gallium nitride. Opt Quan Electron 2006;38:339e47.

123. Pugh SK, Dugdale DJ, Brand S, Abram RA. Electronic structure calculations on nitridesemiconductors. Semicond Sci Technol 1999;14:23e31.

124. Suzuki M, Uenoyama T. Optical gain and crystal symmetry in IIIeV nitride lasers. ApplPhys Lett 1996;69:3378e80.

125. DeCuir EA, Fred E, Manasreh MO, Sch€ormann J, As DJ, Lischka K. Near-infraredintersubband absorption in nonpolar cubic GaN∕AlN superlattices. Appl Phys Lett2007;91:041911.

126. Machhadani H, Tchernycheva M, Sakr S, Rigutti L, Colombelli R, Warde E, Mietze C,As DJ, Julien FH. Intersubband absorption of cubic GaN/Al(Ga)N quantum wells in thenear-infrared to terahertz spectral range. Phys Rev B 2011;83:075313.

127. Jostmeier T, Wecker T, Reuter D, As DJ, Betz M. Ultrafast carrier dynamics and resonantinter-miniband nonlinearity of a cubic GaN/AlN superlattice. Appl Phys Lett 2015;107:211101.

128. Wecker T, Jostmeier T, Rieger T, Neumann E, Pawlis A, Betz M, Reuter D, As DJ. Linearand nonlinear behaviour of near-IR intersubband transitions of cubic GaN/AlN multiquantum well structures. J Cryst Growth 2017.

129. Chamard V, Sch€ulli T, Sztucki M, Metzger T, Sarigiannidou E, Rouvi�ere J-L, Tolan M,Adelmann C. Strain distribution in nitride quantum dot multilayers. Phys Rev B 2004;69:125327.

130. Daudin B, Widmann F, Feuillet G, Samson Y, Arlery M, Rouvi�ere J. Stranski-Krastanovgrowth mode during the molecular beam epitaxy of highly strained GaN. Phys Rev B 1997;56:R7069e72.

131. Guillot F, Bellet-Amalric E, Monroy E, Tchernycheva M, Nevou L, Doyennette L,Julien FH, Dang LS, Remmele T, Albrecht M, Shibata T, Tanaka M. Si-doped GaN∕AlNquantum dot superlattices for optoelectronics at telecommunication wavelengths. J ApplPhys 2006;100:044326.

132. Damilano B, Grandjean N, Semond F, Massies J, Leroux M. From visible to white lightemission by GaN quantum dots on Si(111) substrate. Appl Phys Lett 1999;75:962e4.

133. Miyamura M, Tachibana K, Arakawa Y. High-density and size-controlled GaN self-assembled quantum dots grown by metalorganic chemical vapor deposition. Appl PhysLett 2002;80:3937e9.


134. Himwas C, Songmuang R, Le Si Dang, Bleuse J, Rapenne L, Sarigiannidou E, Monroy E.Thermal stability of the deep ultraviolet emission from AlGaN/AlN Stranski-Krastanovquantum dots. Appl Phys Lett 2012;101:241914.

135. Vardi A, Bahir G, Schacham SE, Kandaswamy PK, Monroy E. Photocurrent spectroscopyof bound-to-bound intraband transitions in GaN/AlN quantum dots. Phys Rev B 2009;80:155439.

136. Beeler M, Hille P, Sch€ormann J, Teubert J, de la Mata M, Arbiol J, Eickhoff M, Monroy E.Intraband absorption in self-assembled Ge-Doped GaN/AlN nanowire heterostructures.Nano Lett 2014;14:1665e73.

137. Ajay A, Lim CB, Browne DA, Polaczynski J, Bellet-Amalric E, den Hertog MI,Monroy E. Intersubband absorption in Si- and Ge-doped GaN/AlN heterostructures in self-assembled nanowire and 2D layers. Phys Status Solidi B 2017.

138. Tanaka K, Ikuno K, Kasai Y, Fukunaga K, Kunugita H, Ema K, Kikuchi A, Kishino K.Ultrafast intersubband relaxation dynamics at 1.55 mm in GaN/AlN multiple quantum disknanocolumns. J Lumin 2008;128:1084e6.

139. Helman A, Tchernycheva M, Lusson A, Warde E, Julien FH, Moumanis K, Fishman G,Monroy E, Daudin B, Le Si Dang D, Bellet-Amalric E, Jalabert D. Intersubband spec-troscopy of doped and undoped GaN/AlN quantum wells grown by molecular-beamepitaxy. Appl Phys Lett 2003;83:5196e8.

140. Kandaswamy PK, Machhadani H, Kotsar Y, Sakr S, Das A, Tchernycheva M, Rapenne L,Sarigiannidou E, Julien FH, Monroy E. Effect of doping on the mid-infrared intersubbandabsorption in GaN/AlGaN superlattices grown on Si(111) templates. Appl Phys Lett 2010;96:141903.

141. Liu D-F, Jiang J-G, Cheng Y, He J-F. Effect of delta doping on mid-infrared intersubbandabsorption in AlGaN/GaN step quantum well structures. Phys E 2013;54:253e6.

142. Edmunds C, Tang L, Shao J, Li D, Cervantes M, Gardner G, Zakharov DN, Manfra MJ,Malis O. Improvement of near-infrared absorption linewidth in AlGaN/GaN superlatticesby optimization of delta-doping location. Appl Phys Lett 2012;101:102104.

143. Tang C, Shi J. Influence of delta doping on intersubband transition and absorption inAlGaN/GaN step quantum wells for terahertz applications. Phys E 2015.

144. Ajay A, Sch€ormann J, Jiménez-Rodriguez M, Lim CB, Walther F, Rohnke M, Mouton I,Amichi L, Bougerol C, Den Hertog MI, Eickhoff M, Monroy E. Ge doping of GaN beyondthe Mott transition. J Phys Appl Phys 2016;49:445301.

145. Nenstiel C, B€ugler M, Callsen G, Nippert F, Kure T, Fritze S, Dadgar A, Witte H,Bl€asing J, Krost A, Hoffmann A. Germanium e the superior dopant in n-type GaN. PhysStatus Solidi RRL e Rapid Res Lett 2015;9:716e21.

146. Heber JD, Gmachl C, Ng HM, Cho AY. Comparative study of ultrafast intersubbandelectron scattering times at w1.55 mm wavelength in GaN/AlGaN heterostructures. ApplPhys Lett 2002;81:1237e9.

147. Iizuka N, Kaneko K, Suzuki N. Sub-picosecond modulation by intersubband transition inridge waveguide with GaN/AlN quantum wells. Electron Lett 2004;40:962963.

148. Iizuka N, Kaneko K, Suzuki N. Sub-picosecond all-optical gate utilizing GaN intersub-band transition. Opt Express 2005;13:3835e40.

149. Li Y, Bhattacharyya A, Thomidis C, Moustakas TD, Paiella R. Ultrafast all-opticalswitching with low saturation energy via intersubband transitions in GaN/AlN quantum-well waveguides. Opt Express 2007;15:17922e7.

150. Monteagudo-Lerma L, Valdueza-Felip S, Naranjo FB, Corredera P, Rapenne L,Sarigiannidou E, Strasser G, Monroy E, Gonz�alez-Herr�aez M. Waveguide saturable


absorbers at 155 mm based on intraband transitions in GaN/AlN QDs. Opt Express 2013;21:27578.

151. Nevou L, Mangeney J, Tchernycheva M, Julien FH, Guillot F, Monroy E. Ultrafastrelaxation and optical saturation of intraband absorption of GaN/AlN quantum dots. ApplPhys Lett 2009;94:132104.

152. Hofstetter D, Schad S-S, Wu H, Schaff WJ, Eastman LF. GaN/AlN-based quantum-wellinfrared photodetector for 1.55 mm. Appl Phys Lett 2003;83:572e4.

153. Hofstetter D, Baumann E, Giorgetta FR, Graf M, Maier M, Guillot F, Bellet-Amalric E,Monroy E. High-quality AlN/GaN-superlattice structures for the fabrication of narrow-band 1.4 mm photovoltaic intersubband detectors. Appl Phys Lett 2006;88:121112.

154. Hofstetter D, Baumann E, Giorgetta FR, Guillot F, Leconte S, Monroy E. Opticallynonlinear effects in intersubband transitions of GaN/AlN-based superlattice structures.Appl Phys Lett 2007;91:131115.

155. Hofstetter D, Di Francesco J, Kandaswamy PK, Das A, Valdueza-Felip S, Monroy E.Performance improvement of AlN/GaN-based intersubband detectors by using quantumdots. IEEE Photon Technol Lett 2010;22:1087e9.

156. Doyennette L, Nevou L, Tchernycheva M, Lupu A, Guillot F, Monroy E, Colombelli R,Julien FH. GaN-based quantum dot infrared photodetector operating at 1.38 mm. ElectronLett 2005;41:1077e8.

157. Vardi A, Akopian N, Bahir G, Doyennette L, Tchernycheva M, Nevou L, Julien FH,Guillot F, Monroy E. Room temperature demonstration of GaN/AlN quantum dot intra-band infrared photodetector at fiber-optics communication wavelength. Appl Phys Lett2006;88:143101.

158. DeCuir EA, Manasreh MO, Tschumak E, Sch€ormann J, As DJ, Lischka K. Cubic GaN/AlN multiple quantum well photodetector. Appl Phys Lett 2008;92:201910.

159. Rong X, Wang XQ, Chen G, Zheng XT, Wang P, Xu FJ, Qin ZX, Tang N, Chen YH,Sang LW, Sumiya M, GeWK, Shen B. Mid-infrared photoconductive response in AlGaN/GaN step quantum wells. Sci Rep 2015;5:14386.

160. Durmaz H, Nothern D, Brummer G, Moustakas TD, Paiella R. Terahertz intersubbandphotodetectors based on semi-polar GaN/AlGaN heterostructures. Appl Phys Lett 2016;108:201102.

161. Sudradjat FF, Zhang W, Woodward J, Durmaz H, Moustakas TD, Paiella R. Far-infraredintersubband photodetectors based on double-step III-nitride quantum wells. Appl PhysLett 2012;100:241113.

162. Pesach A, Gross E, Huang C-Y, Lin Y-D, Vardi A, Schacham SE, Nakamura S, Bahir G.Non-polar m-plane intersubband based InGaN/(Al)GaN quantum well infrared photode-tectors. Appl Phys Lett 2013;103:022110.

163. Giorgetta FR, Baumann E, Graf M, Yang Q, Manz C, Kohler K, Beere HE, Ritchie DA,Linfield E, Davies AG, Fedoryshyn Y, Jackel H, Fischer M, Faist J, Hofstetter D. Quantumcascade detectors. IEEE J Quan Electron 2009;45:1039e52.

164. Sakr S, Kotsar Y, Haddadi S, Tchernycheva M, Vivien L, Sarigiannidou I, Isac N,Monroy E, Julien FH. GaN-based quantum cascade photodetector with 1.5 mm peakdetection wavelength. Electron Lett 2010;46:1685e6.

165. Vardi A, Bahir G, Guillot F, Bougerol C, Monroy E, Schacham SE, Tchernycheva M,Julien FH. Near infrared quantum cascade detector in GaN/AlGaN/AlN heterostructures.Appl Phys Lett 2008;92:011112.

166. Vardi A, Kheirodin N, Nevou L, Machhadani H, Vivien L, Crozat P, Tchernycheva M,Colombelli R, Julien FH, Guillot F, Bougerol C, Monroy E, Schacham S, Bahir G.


High-speed operation of GaN/AlGaN quantum cascade detectors at l z 1.55 mm. ApplPhys Lett 2008;93:193509.

167. Vardi A, Sakr S, Mangeney J, Kandaswamy PK, Monroy E, Tchernycheva M,Schacham SE, Julien FH, Bahir G. Femto-second electron transit time characterization inGaN/AlGaN quantum cascade detector at 1.5 micron. Appl Phys Lett 2011;99:202111.

168. Pesach A, Sakr S, Giraud E, Sorias O, Gal L, Tchernycheva M, Orenstein M, Grandjean N,Julien FH, Bahir G. First demonstration of plasmonic GaN quantum cascade detectors withenhanced efficiency at normal incidence. Opt Express 2014;22:21069e78.

169. Song Y, Bhat R, Huang T-Y, Badami P, Zah C-E, Gmachl C. III-nitride quantum cascadedetector grown by metal organic chemical vapor deposition. Appl Phys Lett 2014;105:182104.

170. Driscoll K, Liao Y, Bhattacharyya A, Zhou L, Smith DJ, Moustakas TD, Paiella R.Optically pumped intersubband emission of short-wave infrared radiation with GaN/AlNquantum wells. Appl Phys Lett 2009;94:081120.

171. Nevou L, Julien FH, Colombelli R, Guillot F, Monroy E. Room-temperature intersubbandemission of GaN/AlN quantum wells at l ¼ 2.3 mm. Electron Lett 2006;42:1308e9.

172. Nevou L, Julien FH, Tchernycheva M, Guillot F, Monroy E, Sarigiannidou E. Intrabandemission at lz 1.48 mm from GaN∕AlN quantum dots at room temperature. Appl PhysLett 2008;92:161105.

173. Nevou L, Tchernycheva M, Julien FH, Guillot F, Monroy E. Short wavelength(l ¼ 2.13 mm) intersubband luminescence from GaN∕AlN quantum wells at room tem-perature. Appl Phys Lett 2007;90:121106.

174. Jovanovi�c VD, Ikoni�c Z, Indjin D, Harrison P, Milanovi�c V, Soref RA. Designing strain-balanced GaN/AlGaN quantum well structures: application to intersubband devices at 1.3and 1.55 mm wavelengths. J Appl Phys 2003;93:3194e7.

175. Stattin M, Berland K, Hyldgaard P, Larsson A, Andersson TG. Waveguides for nitridebased quantum cascade lasers. Phys Status Solidi C 2011;8:2357e9.

176. Song AY, Bhat R, Allerman AA, Wang J, Huang T-Y, Zah C-E, Gmachl CF. Quantumcascade emission in the III-nitride material system designed with effective interfacegrading. Appl Phys Lett 2015;107:132104.

177. K€ohler R, Tredicucci A, Beltram F, Beere HE, Linfield EH, Davies AG, Ritchie DA,Iotti RC, Rossi F. Terahertz semiconductor-heterostructure laser. Nature 2002;417:156e9.

178. Fathololoumi S, Dupont E, Chan CWI, Wasilewski ZR, Laframboise SR, Ban D,M�aty�as A, Jirauschek C, Hu Q, Liu HC. Terahertz quantum cascade lasers operating up tow200K with optimized oscillator strength and improved injection tunneling. Opt Express2012;20:3866e76.

179. Wienold M, R€oben B, Schrottke L, Sharma R, Tahraoui A, Biermann K, Grahn HT. High-temperature, continuous-wave operation of terahertz quantum-cascade lasers with metal-metal waveguides and third-order distributed feedback. Opt Express 2014;22:3334.

180. Bellotti E, Driscoll K, Moustakas TD, Paiella R. Monte Carlo study of GaN versus GaAsterahertz quantum cascade structures. Appl Phys Lett 2008;92:101112.

181. Bellotti E, Driscoll K, Moustakas TD, Paiella R. Monte Carlo simulation of terahertzquantum cascade laser structures based on wide-bandgap semiconductors. J Appl Phys2009;105:113103.

182. Chou H, Anwar M, Manzur T. Active layer design and power calculation of nitride-basedTHz quantum cascade lasers. Proc SPIE 2012;8268:82680O.

183. Jovanovi�c VD, Indjin D, Ikoni�c Z, Harrison P. Simulation and design of GaN/AlGaNfar-infrared (l w 34 mm) quantum-cascade laser. Appl Phys Lett 2004;84:2995e7.


184. Mirzaei B, Rostami A, Baghban H. Terahertz dual-wavelength quantum cascade laserbased on GaN active region. Opt Laser Technol 2012;44:378e83.

185. Shishehchi S, Paiella R, Bellotti E. Numerical simulation of III-nitride lattice-matchedstructures for quantum cascade lasers. Proc SPIE 2014;8980:89800T.

186. Sun G, Soref RA, Khurgin JB. Active region design of a terahertz GaN/Al0.15Ga0.85Nquantum cascade laser. Superlattices Microstruct 2005;37:107e13.

187. Sun G, Khurgin JB, Tsai DP. Spoof plasmon waveguide enabled ultrathin room temper-ature THz GaN quantum cascade laser: a feasibility study. Opt Express 2013;21:28054.

188. Terashima W, Hirayama H. Design and fabrication of terahertz quantum cascade laserstructure based on III-nitride semiconductors. Phys Status Solidi C 2009;6:S615e8.

189. Vukmirovi�c N, Jovanovi�c VD, Indjin D, Ikoni�c Z, Harrison P, Milanovi�c V. Opticallypumped terahertz laser based on intersubband transitions in a GaN/AlGaN double quan-tum well. J Appl Phys 2005;97:103106.

190. Yasuda H, Hosako I, Hirakawa K. Designs of GaN-based terahertz quantum cascade lasersfor higher temperature operations. In: Conference on lasers and electro-optics (CLEO).Presented at the conference on lasers and electro-optics (CLEO); 2012.

191. Williams BS, Callebaut H, Kumar S, Hu Q, Reno JL. 3.4-THz quantum cascade laserbased on longitudinal-optical-phonon scattering for depopulation. Appl Phys Lett 2003;82:1015e7.

192. Terashima W, Hirayama H. Spontaneous emission from GaN/AlGaN terahertz quantumcascade laser grown on GaN substrate. Phys Status Solidi C 2011;8:2302e4.

193. TerashimaW, Hirayama H. GaN-based terahertz quantum cascade lasers. Proc SPIE 2015.948304.



LEDs for liquid crystal display(LCD) backlighting 19Chi-Feng ChenNational Central University, Taoyuan City, TaiwanRevised by JianJang Huang

19.1 Introduction

With the rapid increase of modern multimedia requirements, a wide range of displaydevices have become available. Thin-film-transistor liquid crystal displays (TFT-LCDs) are one of the most popular display devices. They range from small to large1e3

devices such as mobile phones, notebooks, and netbooks; car navigation systems, andtelevisions. Because an LCD cannot radiate by itself, it has to rely on an externalsource to provide the illumination. Generally, a backlight unit (BLU) positioned onthe back of an LCD cell is used to supply sufficient and uniform brightness and lightfor transmissive and transflective LCDs. Around 2004, LCD technology became suit-able for televisions. Due to the strong market demand for thin flat-panel televisions andthe limits on other display technology from technological bottlenecks, LCD televisionsgradually penetrated the television market. The penetration rate of LCD televisionsexceeded 50% for the first time in 2008, and they substantially replaced the cathoderay tube (CRT). However, typical LCD televisions have many intrinsic shortcomings,such as poor viewing angles, response times, contrast ratios, and color gamutcompared to plasma televisions or even traditional CRTs. LEDs can effectively over-come the shortcomings of LCDs and enhance their quality. Total global large-area TFTLCD panel shipments from 2009 to 2015 is in the range of 600e700 million units,which is a large proportion using LED backlight module (Fig. 19.1).

19.2 Types of LED LCD backlighting units

19.2.1 Technical considerations for the light source

The most significant part of a BLU is the light source. The best light source for aspecific BLU is determined by factors such as spectral content, luminous flux andefficiency, operating temperature range and stability over that range, and dimmability.The types of light sources used in BLUs include: LEDs, CCFLs, hot cathode fluores-cent lamps (HCFLs), external electrode fluorescent lamps (EEFLs), flat fluorescent


http://dx.doi.org/10.1016/B978-0-08-101942-9.00019-8

lamps (FFLs), and electroluminescent (EL) devices.3 The general properties of theselight sources for BLUs are listed in Table 19.1.5 CCFLs were once considered thebest light sources for LCDs, even though they have many shortcomings. The advan-tageous properties of LEDs and FFLs make them more suitable than CCFLs for usein a BLU.

Nowadays, with the rapid development of LED technology, especially the advancesin luminous flux and efficiency, LEDs have gradually replaced CCFLs, and they areused in all LCD products. Further, the rate of incorporation of LEDs into each type

100

80

60

40

20

02010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4

Quarter

CCFLLED

Shi

pmen

t pro

porti

on (%

)

Figure 19.1 Shipment proportion of LED and CCFL LCD televisions.4

Table 19.1 General characteristics of light sources for BLUs5

Characteristics CCFLWhiteLED

RGBLED FFL EEFL HCFL

Typicalefficiency(lm/W)

80 >100 >60 30 80 65

Color gamut(% of NTSCspecification)

72%e80% >65% >100% 80% 72% 92%

Mercury w4 mg 0 mg 0 mg 0 mg <4 mg >5 mg

Produced heat Moderate Moderate Moderate High Low Moderate

Gascomposition

Hg, Ar, Ne None None Xe, Ar,Ne

Hg Ar


of LCD product has begun to gradually increase. Their use in LCD backlights has beenthe main driving force behind the LED market.

The applications of LED BLUs to LCDs roughly fall into three categories depend-ing on how the white light is generated:

• The multicolor type of LED is composed of multiple-colored LEDs or LED chips and can becontrolled to produce different colors at different temperatures. The most common are RGBLEDs consisting of red (R), green (G), and blue (B) LEDs. The cell of an RGB LED has onered LED, one green LED, and one blue LED or one red LED, two green LEDs, and one blueLED. Now, two kinds of six-color LED solutions have been proposed: either red, green,blue, cyan (C), yellow (Y), and magenta (M) or red, green, blue, cyan, yellow, and purple (P).

• A white LED is composed of a blue LED chip closely packaged with a yellow phosphor orred and green phosphors. Though white LED can be realized by UV LEDs and phosphors, itspenetration is still limited due to poor UV LED efficiency. Alternatively, white LED can berealized by placing blue LED chip on a homoepitaxially grown zinc selenide (ZnSe)substrate. The function of the ZnSe substrate is for color conversion.

• Remote phosphor is also used by coating the phosphor on a separate substrate.6

Within the above categories, the main types of LEDs used include: side-view whiteLEDs, top-view white LEDs, RGB multi-chip LEDs, and single-color LEDs.

19.2.2 BLU classification

In general, there are four LED BLU structures for LCDs differentiated by the positionof the light source and by their structural characteristics: edge-type, direct-type,hollow-type, and folded-mixing-light guide plate (LGP)-type. Schematic diagramsof each type are shown in Fig. 19.2(a)e(d).

19.2.2.1 Edge-type structure

An edge-type structure has at least one LED light bar located at an edge of the LGP ofthe LCD device, as shown in Fig. 19.2(a). Light is transmitted through a light guide bymeans of total internal reflection. The key components of an edge-type structureinclude: LEDs, an LGP with a microstructure or dots, a back reflector, and diffusers.7

To meet the requirement for low power consumption, a brightness enhancement film(BEF) or dual brightness enhancement film (DBEF) needs to be used in the BLU.8

19.2.2.2 Direct-type structure

A direct-type structure has LEDs positioned below the LCD panel, as shown inFig. 19.2(b). A typical direct-type backlight device includes: LEDs placed above ametal core printed circuit board (MCPCB), a reflector, a diffusion plate, and someoptical function films.9 This structure has the advantages of a large backlight, highbrightness, and light weight and is easier for local dimming. However, to maintain auniform brightness, a light-mixed cavity between the light sources and the diffusionplate is necessary. For this reason, a direct-type BLU is thicker than an edge-type BLU.

LEDs for liquid crystal display (LCD) backlighting 621

Optical function films

LED light barReflector

LGP

Optical cavity

Reflector and MCPCB


Hollow cavity

ReflectorPatterned reflector

Directional LED light bar

Structural diffusion plate



Diffusion plate

LED

Main LGP180 degrees couplingreflective mirror

ReflectorMixing LGPReflector

90 degrees coupling reflective mirror

Multiple-color LEDs

(a)

(b)

(c)

(d)

Figure 19.2 Schematic diagrams of (a) edge-type, (b) direct-type, (c) hollow-type and (d)folded-mixing-LGP-type LED backlight structures for LCDs.


19.2.2.3 Hollow-type structure

Hollow-type structures have two directional LED light bars located at the two edges ofthe hollow cavity of the BLU, as shown in Fig. 19.2(c). A hollow-type structure BLUis composed of at least two LED light bars, a hollow cavity, a patterned reflector placedat the bottom of the cavity, a structural diffusion plate, and some optical films.10 Thepoor brightness uniformity is generally a serious drawback of this type.

19.2.2.4 Folded-mixing-LGP-type structure

A folded-mixing-LGP-type structure has at least one folded mixing color LGP andseveral multiple-color LEDs, typically RGB LEDs, positioned below the main LGP,as shown in Fig. 19.2(d). Its main components comprise: high-power multiple-colorLEDs, a folded mixing LGP, a main LGP placed behind the LCD panel, a 90 degreescoupling reflective mirror, and a 180 degrees coupling reflective mirror.11,12 PhilipsLumileds Lighting Company has presented a typical structure for a BLU assembly.This uses a Luxeon DCC as a light source, which is based on RGB LEDs with aLambertian radiation pattern, as shown in Fig. 19.3.12 The multicolored light emittedfrom the multiple-color LEDs is coupled into the mixing light guide with a 90 degreesmirror. This light is propagated, diffused, and mixed in the mixing light guide.Uniform white light can be obtained from the folded mixing LGP. The 180 degreesmirror then directs this white light into the main light guide. This type of structurehas the advantages of high brightness, wider color gamut, compact shape, and goodthermal dissipation.

Mixing light guide

Light guide

PCB with electronics Coupling mirror

Luxeon light source for backlight

180 degreesmirror

Figure 19.3 BLU assembly using a Luxeon DCC as a light source of Ref. 12.


19.3 Technical considerations for optical films andplates

The optical qualities of a BLU depend on the optical films or plates used, such as anLGP, a diffusion film, a prismatic BEF, a micro-lens BEF, a reflective polarizer BEF,and a diffusion plate. LGPs are usually made of optical grade materials such aspolymethyl-methacrylate (PMMA), ZEONOR, or polycarbonate (PC). Table 19.2 liststhe general characteristics of these three materials. The materials used for LGPs in thecurrent market are still based on PMMA. An LGP can be either wedge-shaped or flat.In general, because of space considerations, mobile phones, car satellite navigationdevices, notebooks, and small- and medium-sized products commonly use a wedge-shaped LGP such as that shown in Fig. 19.2(a).

Structured BEFs and reflective polarizer BEFs can recycle some of the wasted lightenergy, increasing the effective area of the luminance. The 3M Company manufacturesBEFs, named the Vikuiti, BEF II,13 and BEF III,14 which can direct diffused light intothe backlight and through the LCD. This increases the brightness for an on-axisviewer. Typically, two orthogonally aligned BEFs are used in the BLU of mobileproducts while the BLU of monitors and televisions uses a single BEF. In comparison,the Vikuiti DBEF, a very common reflective polarizer BEF, uses 3M’s multilayeroptical film technology.15 The DBEF increases the amount of light available for illu-minating LCD displays by making use of light that would normally be absorbed by therear polarizer of the LCD panel. The backlight efficiency is increased while maintain-ing the viewing angle. It can increase on-axis brightness by up to 60% in notebookdisplays with a slab LGP and up to 97% in notebook displays with a wedge LGP.The brightness gains of the Vikuiti BEF II and BEF III are listed in Table 19.3.

Table 19.2 General characteristics of three materials used for LGPs

Characteristics PMMA ZEONOR PC

Proportion 1.2 1.01 1.20

Water absorption (%) 0.3 <0.01 0.2

Transmittance at 3 mm 92 92 88

Index of refraction 1.49 1.53 1.59

Hardness 2H H B

Glass transition temperature, Tg (�C) 105 140 145

Table 19.3 Brightness gains of Vikuiti BEF II13 and BEF III14

Single sheet Two sheets crossed at 90 degrees

BEF II 60% 120%

BEF III 59% 111%


The main aim in the development of optical films and plates is multifunctional inte-gration, to produce a so-called integrated optical film or plate. Examples are a BEFwith a diffusion function, a diffusion film with a brightness enhancement function(usually called a gain diffusion film), and an LGP with a diffusion and/or brightnessenhancement function.

19.4 Requirements for LCD BLUs

19.4.1 Features of LCD BLUs

An ideal BLU needs to have features such as good uniformity of brightness, ultra-slim,light weight, ultra-narrow bezel, low power consumption, long life, wide color gamut(good white spectrum), short response time, large brightness adjustment range, fastmodulation, temperature insensitivity, color that is adjustable according to the temper-ature, support for field sequential color technology, flexibility, support for two-dimensional (2D) and three-dimensional (3D) convertible displays, user-friendly,environmentally friendly, and low cost. Therefore, the light source itself should beslim, light, and user- and environmentally friendly and have low power consumption,a long life, quick response time, fast modulation, wide color gamut, and a color adjust-able according to temperature.

19.4.2 Environmental requirements

There are both legal environmental requirements and optional energy saving programsthat can affect the development and use of electronic equipment. In July 2006, theEuropean Union (EU) began the formal implementation of the RoHS Directive: therestriction of the use of certain substances in electrical and electronic equipment thatare potentially hazardous to us. One of these controlled substances is mercury.However, as there is an exclusion clause for it, CCFLs can still continue to use mercuryfor now.

The Official Journal of the EU published Commission Regulation (EC) No.642/2009 on 22 July 2009. This implemented Directive 2005/32/EC of theEuropean Parliament and of the Council with regard to ecodesign requirementsfor televisions. The first phase of the regulations was implemented from 20 August2010.

Energy Star is an international standard for energy-saving consumer productsand programs. The project was initiated in the United States in the 1990s andhas become multinational. Manufacturers can choose of their own accord to affixthe Energy Star label to qualified products. The first products included in theproject were mainly computers and other information appliances. It was thengradually extended to motors, office equipment, lighting, appliances, and so on.Table 19.4 lists the calculations for the maximum power requirements in the onmode.16


19.5 Advantages and history of LED BLUs

19.5.1 Advantages of LEDs for LCD BLUs

LEDs provide numerous options and advantages for LCD BLUs. Generally the advan-tages of an LED BLU are the following.

19.5.1.1 Low operating DC voltage

LEDs use a low-voltage power-driven supply, unlike the high voltage power neededfor a CCFL. Therefore, they do not need an inverter. This significantly helps tomake an LED safer and reduces electronic noise. Not only that, an LED circuit canuse less space cost less, and its energy consumption is reduced and the heat that wouldbe generated by the inverter is not lost. The design of the power supply module is alsorelatively simple.

19.5.1.2 Wide operating temperature range

LEDs can work instantly at all temperatures without the need for heaters. The oper-ating temperature range of an LED is between about �40 to þ85�C. LEDs can startpromptly at �40�C unlike CCFLs, which do not work properly in such environments.As LEDs are functional over a wide temperature range, they are favored by the militaryand in aviation, exploration, and similar fields.

19.5.1.3 High luminous efficiency (low power consumption)

At present, the luminous efficiency of a white LED for use in a BLU is over about130 lm/W. This is nearly twice as efficient as a CCFL. As LED technology progressesfurther, this luminous efficiency will continue to improve.

Table 19.4 Calculation of maximum on mode power requirements(PON_MAX)

16

Product type PON_MAX (watts)

diagonal screen size,d (in.)

(r is the screen resolution in megapixels and A is the viewable screenarea, rounded to the nearest 0.1 in2)

d < 12.0 (6.0 � r) þ (0.05 � A) þ 3.0

12.0 < d < 25.0 (6.0 � r) þ (0.0145 � A) þ 4.0

25.0 < d < 30.0 (6.0 � r) þ (0.18 � A) � 40.0

30.0 < d < 60.0 (0.27 � A) þ 8.0


19.5.1.4 Package size and chroma selection flexibility

The optical design and use of LEDs is flexible. They are scalable and the chromaselection is flexible.

19.5.1.5 Wide color gamut

For the RGB LEDs used in LCDs, the National Television System Committee (NTSC)color gamut can be over 100% and can even achieve 150%.

19.5.1.6 Longer operating life

An LED has a specific lifespan. The lifetime is 60,000 to 100,000 h, for a suitablecurrent and voltage, which is far longer than a CCFL. Using LEDs can greatly extendthe life of an LCD television and has overwhelming advantages compared with plasmatechnology.

19.5.1.7 Rapid switching speed

The response time of an LED is as short as a nanosecond, which is about one millionththat of a CCFL. The screen on an LCD device can appear blurry because of the slowresponse time of the liquid crystal. This is caused by screen persistence of fast movingobjects. This drawback can be solved to some extent by using an LED BLU. An LEDcan support instant backlight blinking technology and dynamic scanning backlighttechnology. This technique effectively reduces motion blur and the display qualitywill be significantly improved. An LED can support the field sequential color technol-ogy. The color filter, which accounts for 30% of the cost of an LCD device, can bereplaced by quick scanning RGB LEDs. An LED can also support local dimmingtechnology. This technique can achieve high contrast and enhance the color saturationfor low power consumption.

19.5.1.8 Wide adjustment range for brightness, contrast, andchromaticity

LED power control is simple and the brightness adjustment range is large, unlikeCCFLs, which have a threshold for minimum brightness. Therefore, in the brightoutdoors or dark indoors, it is simple for users to adjust the brightness of the displaydevice for ease of viewing. This is particularly useful in automotive, avionic, andmarine electronics, where the product display must be able to deal with lighting con-ditions ranging from bright sunlight to the moonlight at night. In addition, when thevideo display source switches between computers and DVD players, it can be easilyadjusted between a 9600 and 6500K white balance, without sacrificing the brightnessand contrast.


19.5.1.9 Environmentally friendly

LEDs are made of nontoxic materials. Unlike incandescent lamps, they cause nomercury pollution, and there is no UV or IR radiation in their spectra.

19.5.1.10 Quicker illumination to stable brightness

LEDs require no warm-up time or general heating.

19.5.1.11 Robustness

LEDs have a high resistance to mechanical shock as they do not have glass tubes.LEDs are slim, light, safe, and quiet: they are a solid-state solution without an inverter.

The above advantages significantly promote the use of LEDs in BLUs. Moreover,for specialized products such as military, avionics, marine and automotive displays,LCDs based on CCFLs are unable to meet user requirements.

However, there are still challenges in using LEDs. In particular, these include cost,system design complexity, and performance with temperature. For a high-power LEDBLU, it is very difficult to use the edge-type structure with a narrow bezel and thedirect-type with high brightness uniformity and slim body. For a high-power RGBLED model, the optical design makes it difficult to obtain good color control andhigh color uniformity. Some of the time there will be an uneven color because thedecay rate of each one of the LED colors is inconsistent. The heat produced by theLED will lead to LED color variation, LED brightness variation, shortened lifetime,deformation, and LGP aging.

19.5.2 History of LED BLU development

LED BLUs were first used in mobile phones. This application of LEDs was also thefastest to penetrate the market for LCD products. Since edge-type BLUs with aside-view-type white LED do not require much packaging or high operating power,they have proven to be the best choice. From about 2006, the market began to useLEDS in the BLUs of notebooks. In 2006, computer manufacturers produced only11.3ʺ and 12.1ʺ notebooks using LEDs. At the beginning of 2007, several computermanufacturers launched a 13.3ʺ and even a larger 15.4ʺ product using an LED back-light. The most common choice is a wedge-shaped LGP with a white top-view-typeLED bar.

The LED industry and LCD industry intended to use LEDs in televisions. An RGBLED BLU was first used by Sony Corporation in August 2004 for its 40ʺ and 46ʺQUALIA 005 LCD televisions. The power consumption for the 40ʺ and 46ʺ modelsis 470 and 550 W, respectively. To solve the problem of how hot these become, themodels use fans, heat pipes, and heat sinks. The heat pipes are horizontally arrangedand there are large heat sinks on both sides of the back of the BLU. The heat passesthrough the heat pipes to the heat sinks on both sides, which are cooled by fans. Usingthis particular design, the color gamut can be extended up to 105%. This elaboratedesign is expensive and uses a thick backlight unit (up to 10 cm). South Korea’s


Samsung Electronics demonstrated 46ʺ and 40ʺ LCD televisions without additionalthermal design at the 2005 CES show. They used direct-type BLUs with a medium-power (0.3e0.5 W) LED and color sensor.

In 2006 there were crucial developments for LCD televisions. There were severaltechnological advances and solutions to problems. Larger LCDs were successfullydeveloped, such as LG Display’s 100ʺ model and Samsung Electronics’ 82ʺ model.A double frame rate (or higher) became possible, which can effectively solve thedynamic image blur. Local dimming (also known as high dynamic range) technologyusing an LED backlight can make the contrast quality up to 10,000 or more, andgreatly reduce energy consumption. Further, eight-domain multi-domain verticalalignment (MVA) technology can reduce the color washout problem of the previousMVA technology, solving the problem of the LCD viewing angle.

The main focus of research for LCD LED BLUs is still to achieve high color satu-ration (or wide color gamut). Ultra-slim LCD televisions are realized by using a direct-type BLU with low-power LEDs or an edge-type BLU with medium-power LEDs. Thefield sequential color method works without a color filter. It uses direct fast-switchingR-, G-, and B-LEDs to produce the respective R-, G-, and B-display pictures, throughthe persistence of vision to create full-color display effects.

The first large-screen LCD television with an edge-type LED BLU, the KLV-40ZX1M, was launched by Sony in September 2008. It was Sony’s thinnest 40ʺLCD monitor measuring a mere 9.9 mm in width. However, it was expensive; sodemand for the product was poor. Samsung Electronics then launched a mid-pricedLCD television with an edge-type BLU using medium-power LEDs. Due to a salesstrategy highlighting the slim body, price, and quality, Samsung Electronics success-fully created the LED LCD television market. Thus, LED LCD televisions were asought after commodity in 2009 and the market for these televisions was created.

According to a report by NPD DisplaySearch,17 due to lower-than-expected con-sumer adoption of LED LCD televisions at the end of 2011, television makers changedtheir strategy for direct-type LED BLU televisions by developing products that use lesspower and cost less by reducing the number of LEDs per television set. Their aim is todevelop an adaptation of the original LED BLU with a slim design and better picturequality.

19.6 Market trends and technological developments

19.6.1 Market trends

Over the past decade, LEDs have slowly been incorporated into various LCD devices,initially in mobile phones and recently in LCD televisions. Due to a combination offactors, including the fact that OLED development has made less progress thanexpected, that there has been no significant breakthrough in the technology of otherlight sources used in LCDs, and that the quality of LEDs has been continuouslyimproved, LCDs with an LED BLU have gradually become the best choice fordisplays. For LCD devices, the quality and features of the light source affect or


even directly determine the optical quality and market competitiveness of the LCDproducts. Several advantages of LEDs solve a few of the outstanding problems withLCDs. It is clear that using an LED inside an LCD has increased the mainstream statusof LCDs. A market forecast for large BLUs based on a survey by LEDinside publishedin August 2011 is shown in Fig. 19.4. A cost forecast based on a survey by NPDDisplaySearch for different BLU types for 32ʺ high-definition (HD) 60-Hz LCDtelevisions is shown in Fig. 19.5.4 The high cost of LED BLUs obviously affectsthe commercial market penetration of LCDs based on LEDs. To reduce costs, televi-sion manufacturers have been adopting two-chip LEDs to reduce the number of LEDs.

LCD monitor2000

1500

1000

500

02009 2010 2011 2012

Year

NB/media tablets LCD TV

Mar

ket s

ize

(mill

ions

of U

S$)

Figure 19.4 Market forecast for large BLUs based on a survey by LEDinside published inAugust 2011.

10

15

20

25

30

35

40

45

Edge-type

Direct-type

CCFL-type

Cos

t (U

S$)

Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q42014

Q1 Q2 Q3 Q420152011 2012 2013

Figure 19.5 Cost forecast for different BLU types for 32ʺ HD 60 Hz LCD televisions.4


The number of LEDs used per set with a direct-type BLU is expected to be less than ina set with an edge-type BLU.4 Fig. 19.6 shows a forecast for the number of LED pack-ages per television set.17

Overall, the focus of technological development is to produce devices with aslimmer body; higher quality (in terms of better brightness uniformity, higher bright-ness, less unevenness, lower color washout, and higher color saturation); lighterweight; lower cost; larger size; narrower bezel (for aesthetic reasons and ease ofapplication, e.g., for an LCD video wall); better environmental factors (lower carbonemissions, lower power consumption, and less use of nontoxic materials); rapidswitching or rapid scanning; wide brightness and contrast adjustment ranges; finerlocal dimming, and more smart functions, for example, auto-adjusting the brightness.However, many of these technologies are conflicting, for example, display size andpower consumption or thickness and brightness uniformity for direct-type BLUs.Based on these desired qualities, an ideal BLU can almost be realized with LEDsand LEDs are able to meet the requirements of any application.

For different products and different product positioning, there are different trends inthe technological developments. Tables 19.5 and 19.6 list these trends for the differentproducts and product positioning, respectively.

The current status and the trends in development of LED LCD televisions are dis-cussed below. Table 19.7 shows the status of the technology for LED LCD televisions.Nowadays, technological developments for LED LCD televisions using an edge-typestructure have included:

• reducing the number of LED bars so that they are on one side instead of two sides to lowerthe cost

• reducing the thickness of the LGPs to reduce the cost, weight, and module thickness

300

250

200

150

100

50

02010 2011

Year2012

Effect of 2-chip package

Effect oflow-costdirect-type62

46

136112

Effect of 2-chip package32-in HD direct-type

32-in HD edge-type

46-inch full HD edge-type

Num

ber o

f LE

Ds

per T

V s

et

Figure 19.6 Forecast for number of LED packages per television set.17


• reducing the chromaticity deviation of the LGPs• using microstructures with higher efficiency for the LGPs• using local dimming technology• changing the packaging from the 5630 type to the longer 7030 type.

19.6.2 Advantages and key technologies of LED LCD televisions

Compared with CCFL-type LCDs, LED-type LCDs18

• produce images with greater dynamic contrast,• can be extremely slim with some screens less than half an inch (0.92 cm) thick,19

Table 19.5 Technological development trends for different products

ProductCommon LEDtype

CommonBLU type

Developmenttrends

Mobile phoneNetbook

Notebook

B-LEDchip þ Yphosphor

B-LEDchip þ RGphosphors

Low-powerchip

Medium-powerchip

Edge-type SlimmerLighter weightHigher qualityNarrower bezel-Lower powerconsumption

Wide brightnessadjustmentrange

Auto-adjust thebrightness

Environmentallyfriendly

Lower cost

Monitor B-LEDchip þ Yphosphor

RGB chipsLED

RGB LED

Medium-powerchip

High-powerchip

Edge-typeDirect-type

Higher qualityNarrower bezel-Lower powerconsumption

Auto-adjust thebrightness

Environmentallyfriendly

Lower cost

TelevisionLargeannouncementdisplay

B-LEDchip þ Yphosphor

RGB LED

Low-powerchip

Medium-powerchip

High-powerchip

Edge-typeDirect-type

All


Table 19.6 Technological development trends for different productpositioning

Productposition

Common LEDtype

CommonBLU type

Developmenttrend

High-gradetelevision

RGB LEDB-LEDchip þ RGphosphors

Low-powerchip

Medium-powerchip

Direct-type All

Medium-gradetelevision

B-LED chip þ Yphosphor

Medium-powerchip

High-powerchip

Edge-type SlimmerHigher qualityEnvironmentallyfriendly

Lower cost

Low-gradetelevision

B-LED chip þ Yphosphor

Medium-powerchip

High-powerchip

Direct-type Environmentallyfriendly

Lower cost

Table 19.7 Status of technology for LED LCD televisions

Direct-type Edge-type

LED chip type RGB LED White LEDa White LED White LED

Low-powerchip

Medium-powerchip

High-powerchip

Low-powerchip

Medium-powerchip

High-powerchip

Medium-powerchip

High-powerchip

Color gamut >100% 70% 70% 70%

Local dimming Yes Yes No No

Thickness Thicker Slimmer Thicker Slimmer

Cost Highest Higher Lowest Lower

Powerconsumption

Lower Lower Higher Higher

Chromaticitydeviation

Larger Smaller Smaller Larger

Image quality Highest Higher Lower Lower

For edge-type structures, the technological development of LED LCD televisions mainly focuses on cost and quality.aB-LED chip þ Y phosphor.


• offer a wider color gamut when an RGB-LED BLU is used,20,21

• produce less environmental pollution on disposal,• have typically 20%e30% less power consumption,• are more robust and reliable,• have a nonlinear, wider dimming range,• give full and flicker-free dimming at all temperatures down to 5% or lower,• can have a higher image quality,• can realize programmable chromaticity adjustments.

The key technologies used in LED LCD televisions that make them so advanta-geous include:

• LED spectral and LED light bar design,• good design potential,• thermal design,• optical design of modules,• integration and efficiency of the drive circuit,• local dimming technology.

19.6.3 New display technologies using LEDs: crystal LEDdisplays

In early 2012, Sony Corporation announced that it had developed a next-generationself-luminous display technology called the crystal LED display, and unveiled a55ʺ crystal LED prototype at CES 2012.22 Each pixel of an ultrafine RGB colorLED chip is directly connected to a light emitter. For 1080 full HD resolution,the total number of LED chips is about 6 million. So far, according to informationfrom Sony Corporation, compared with existing LCDs, the prototype has about 3.5times the contrast ratio, 1.4 times the color gamut, and 10 times faster responsetime.

19.7 Optical design

19.7.1 Design factors

According to the specific application requirements, some of the design factors to beconsidered are:23

• diagonal display size,• panel thickness,• luminance,• color gamut,• thermal environment and associated constraints,• power limits,• dynamic contrast,• BLU cost.


19.7.2 Key design considerations

Compared to a CCFL-based BLU, the key design considerations include:

• designing the LED light guide bar to form line-shaped light, as in a CCFL,• for an edge-type structure, designing the microstructure of the light injection surface of the

LGP to diffuse light emitted from the LED chips fully,• for edge-type and direct-type structures, designing the secondary optical parts to diffuse light

emitted from a middle- or high-LED chip array fully.

The effect of the microstructure of the light injection surface of an LGP is shown inFig. 19.7.

19.7.3 Edge-type BLUs

The optical design considerations for edge-type LED BLUs are listed in Table 19.8.When the white LED light is just coupled into the LGP, the light is more concentratedin front of the LED and then slowly spreads out. This unequal distribution of the light

LEDα

w w

p

L

LGP

AirLEDLEDα

LED

LGP

Air

w

L

p

w

(a)

(b)

Figure 19.7 Light propagation from the light injection surface of an LGP (a) without and(b) with a microstructure.


Table 19.8 Optical design considerations for edge-type LED BLUs

Item Description

Selection andconfiguration of LEDtype

1. Select the type of LEDs, such as side-view type or top-viewtype; white LED or multicolor LEDs.

2. Based on the features of the selected LED type, configureand design the relative position of all LEDs.

Diffusing light emittedfrom LED chips

1. To obtain a narrow bezel, select the microstructure type anddesign the structure parameters of the light injection surfaceof the LGP.

2. Because of the cost, some commercial products still do notuse a microstructure on the light injection surface of theLGP.

LGP geometry, physicalproperties, andprocessing

1. Select a wedge-shaped or flat-shaped LGP.2. Select the LGP material.3. Select the processing method such as printing or injection

molding.

Selection of theextraction patterns anddesign of theextraction patterndensity on the bottomsurface of the LGP

1. Select the extraction patterns for the up- and down-surfacesof the LGP. The light escape probability and escape angleare related to the geometry and optical features of theextraction patterns.

2. Consider manufacturing issues.3. Based on the features of the selected LED type, design the

extraction pattern (microstructure) density on the bottomsurface of the LGP to meet the requirements for uniformity,efficiency, and manufacturability.

4. Based on cost, the extraction pattern is fabricated by etchingand dot printing for most commercial products.

Injection molding or flatpanel cutting

1. The LGP is fabricated by injection molding or flat panelcutting combining with dot printing or hot embossing.

Selection and dispositionof optical films

1. The selection and disposition of the optical films will affectthe viewing angle, efficiency, optical quality, and cost.

Chromaticity uniformityand chromaticityshifting

1. The optical films, LGP material and manufacturing methodwill affect the chromaticity deviation.

2. Using laser direct processing and hot embossing to fabricatethe extraction patterns can reduce the chromaticitydeviation.

3. The LED chromaticity must be uniform.

With or without localdimming

1. LGP design is specific to the 1D local dimming technologyused.

2. Based on cost and technology maturity, most commercialproducts do not use local dimming.


creates a hot spot. Using the estimated length of the color mixing area as a model,24 therough length of the hot spot area L for an injection surface without a microstructure canbe written as:

L ¼ ð p� wÞ=2

tan

"

sin�1

�sin an

�#; (19.1)

where p is the LED pitch, w is the width of the LED emitting area, a is half of the half-intensity angle, and n is the LGP index. Schematic diagrams of these parameters andlight propagation are shown in Fig. 19.7. This estimated result is very rough. If theselected LEDs are multicolor, the estimated length of the color mixing area is the sameas in Eq. (19.1), except that p is modified to represent the maximum pitch betweenequal color LEDs.24

To meet the requirements for a narrow bezel and uniform brightness, the LED hot-spot problem needs to be eliminated, which can be achieved using microstructuressuch as prisms, pyramids, cylindrical lenses, or lenses. These are used on the lightinjection surface of an LGP to couple and diffuse the light efficiently from theLEDs into the LGP.7 The secondary optical element is also used to diffuse the lightfully to form an approximately linear light source injecting into the LGP.25 A totalinternal reflection lens has been designed and used to improve the brightness anduniformity of the backlight. The brightness and uniformity were improved by 40%and 83%, respectively, compared with a conventional BLU. Furthermore, the technol-ogies used for the LGP and optical films are the same as for a CCFL-type device.

Listed below are some examples of applications of LED edge-type BLUs. A 19ʺLCD monitor with a six-lead MULTILED, the LRTB G6SG, was developed byOSRAM Opto Semiconductors.26 It uses a light bar with 77 LEDs and an LED pitchof 5 mm instead of some of the CCFLs. Only two of the CCFLs were replaced and therest of the design (housing, light guide, optical films, etc.) remained unchanged. It onlyrequires a passive cooling system of ventilation slots in the housing and the MCPCBswere mounted on thin heat sinks. Due to the continuous increasing LED brightness, asmaller number of LEDs are needed and the heat generated is relatively reduced.

Using high-power LEDs for larger LCD BLUs has clear advantages over smallLEDs. Philips Lumileds Lighting Company developed an edge-type LED BLU withhigh-power side-emitting white Luxeon LEDs.27,28 The schematic construction isshown in Fig. 19.8. The light from the side-emitting LEDs is coupled into an LGPwith an optical incoupling efficiency of 82%. This BLU has the advantages of athin design and high coupling efficiency.

Consider a commercial 7ʺ BLU in which the lighting area is 145.8 mm � 82.2 mmand the LGP thickness is 0.6 mm. The optical specifications and materials used arelisted in Tables 19.9 and 19.10, respectively. Here Lmin and Lmax are the measured min-imum and maximum luminance for the nine points shown in Fig. 19.9. H and W are


measured in the vertical and horizontal directions, respectively. To avoid the hot spotarea, the lighting area begins at 3 mm away from the incident surface. The opticaldesign considerations for this example are shown in Table 19.11.

Consider a commercial example of a 46ʺ BLU. The requirements are specified inTable 19.12. The optical measured positions are shown in Fig. 19.10.

19.7.4 Direct-type BLUs

Compared with a CCFL-based BLU, the most important point of optical design for anLED-based BLU is designing the secondary optical parts to diffuse the light emittedfrom the medium- or high-power LED chip arrays fully.

If low-power LEDs are adopted, the structure is generally simple and does not havesecondary optical parts. The estimated thickness of the light-mixed cavity H can bewritten as:

H ¼ ðp� wÞ=2tan a

; (19.2)

where p is the maximum pitch between equal color LEDs, w is the width of the LEDemitting area, and a is a half of the half-intensity angle. For the optical design of direct-type BLUs with medium- or high-power LEDs, the main considerations are listed inTable 19.13.

Some application examples of specific LED direct-type BLUs now follows.Consider a 23-inch direct-type BLU based on 72 high-power side-emitting RGBLuxeon LEDs consisting of two strips of 36 LEDs each.28 The LED pitch is12 mm. The variance of the brightness profile of the resulting backlight as a functionof the spacing between the two strips is shown in Fig. 19.11(b) with the spacingranging from 50 to 120 mm. For RGB LED BLUs, the color uniformity is a relevantperformance parameter. The measured color uniformity for this BLU as a function ofLED pitch is shown in Fig. 19.12. The upper and lower curves show results for arandom placement of LEDs from R, G, and B batches of LEDs (called unselectedLEDs) and individual LEDs selected based on optimizing the flux for color unifor-mity (called selected LEDs), respectively. Comparing with the unselected LEDs,the color nonuniformity of the selected LEDs was reduced by approximately half.

LCD panel

Optical films

LGPReflector films

Heat sink

LCD controller

Figure 19.8 Edge-type LED BLU with high-power side-emitting white Luxeon LEDs.27


Table 19.9 Optical specifications for a commercial example of a 7-in. BLU

Item Symbol Unit Condition Minimum Typical Maximum Remark

Center luminance L cd/m2 25�C 5500 6400 e Measured at the center of the lighting area

Luminanceuniformity

DL % 25�C 80 85 e Lmin/Lmax � 100%

Chromaticity x e e 25�C 0.280 0.305 0.330 LEDs from different color bins are not allowedon the same MCPCB

Chromaticity y e e 25�C 0.280 0.305 0.330

LEDsfor

liquidcrystal

display(LCD)backlighting

639

By using LED selection, RGB LED light sources can easily be used for a BLU withhigher color uniformity.

A 32ʺ LCD television based on high-power Golden DRAGON ARGUS LEDs,which are a Golden DRAGON combined with a wide radiating ARGUS lens, hasbeen designed and manufactured.9 Fig. 19.13 shows the arrangement of the RGGBLED clusters with a reflector cover foil. The ARGUS lens deflects the light emittedfrom the chip to give a flat homogeneous distribution. For optimum color homogene-ity, a compact cluster arrangement is adopted. This BLU, used with a reflector boxwith an inner height of 35 mm, consists of 41 RGGB LED clusters mounted on anMCPCB. A hexagonal arrangement of the clusters with a pitch of 82 mm betweencluster centers was used. For the thermal design, the RGGB LED clusters weremounted on a 2-mm-thick metal plate, without active cooling. The NTSC color gamutof the complete LCD television was up to 105%.

Table 19.10 Materials used in a commercial example of a 7-in. BLU

Part Quantity Part model (supplier)

LEDs 7 pieces � 2 strings NSSW206 (NICHIA)

MCPCB 1 AL-5052 (CS Aluminium)

Reflector sheet 1 RW188 (Kimoto)

Light guide plate 1 Idemitsu LC-1500 (PC material)(Green Point)

Diffuser 1 BS-04(188) (Keiwa)

Prism sheet (horizontal) 1 BEF III-T 90/50 (3M)

Reflective polar 1 DBEF-D400 (3M)

1

4

7

W/10W/2

9H/2

6

3 H/102

5

8

Figure 19.9 Measured positions for the 7ʺ BLU.


Table 19.11 Optical design considerations for a commercial example ofa 7-in. BLU

Item Description

Selection and configuration of LEDtype

White top-view-type LED, NFSW036C, is used.Two strings of LED strips are arranged on thelong side.

Diffusing light emitted from LEDchips

There is no microstructure on the light injectionsurface of the LGP.

LGP geometry, physical properties,and process

A flat-shaped PMMA LGP is used and fabricatedby injection molding.

Selection of the extraction patterns anddesign of the extraction patterndensity on the bottom surface of theLGP

The extraction pattern structure is fabricated byetching.

It is designed according to previous experienceand software tools.

Injection molding or flat panel cutting The LGP is fabricated by injection molding.

Selection and disposition of opticalfilms

Selection and disposition of optical films willaffect the viewing angle, efficiency, opticalquality, and cost.

Chromaticity uniformity LEDs from different color bins cannot be used onthe same MCPCB.

With or without local dimming Local dimming is not used.

Table 19.12 Requirements specification for a commercial example of a46ʺ BLU

Part Description

LED LED bar arrangement Bottom (single side)

Package type top-view, 5630

LED phosphor RG or YAG

LGP Thickness &3 mm

Optical films Reflector sheet 1 piece

Diffuser plate 1 piece

Prism sheet 1 piece

Optical specification Brightness 5900 cd/m2

Luminance uniformity Minimum 75%, typical 80%

Chromaticity uniformity Dx& 0.015, Dy & 0.015

Chromaticity shifting Dx& 0.003, Dy & 0.003

Power LED electric power &63 W


1

4

7

W/6

8

5

2 3

6

9

W/2

H/2

H/6

Figure 19.10 Measured positions for the 46ʺ BLU.

Table 19.13 Optical design considerations for a direct-type LED BLU

Item Description

Selection andconfiguration of LEDtype

Select the type of LEDs, such as RGB LEDs or a white LED.Based on the features of the selected LED type, configure anddesign the relative position of all LEDs.

Diffusing light emittedfrom LED chips with aside-emitting lens29e32

or wide-distribution-radiating lens28

Several secondary optical elements are used to diffuse lightemitted from LED chips.

Side-emitting and wide-distribution-radiating lenses arecommonly used. These lenses can reduce the thickness andincrease the optical quality of the device.

If a side-emitting lens is adopted, use local dimming.

Designing the shape of thelight-mixing cavity

Design the inclination angle and shape of the four sides.Design the shape of the reflective bottom surface.Select and design the structure of the reflective bottom surface.

Selecting the diffusionplate

Select a suitable diffusion plate.

Selection and dispositionof the optical films

Selection and disposition of the optical films will affect theviewing angle, efficiency, optical quality, and cost.

Chromaticity uniformity Selection of the optical films, LGP material, andmanufacturing methods will affect the chromaticitydeviation.

The LED chromaticity must be uniform.

With or without localdimming

LGP design is specific to the local dimming technology (0D,1D, or 2D) used.

NichiaT ¼ 0.8 mm:NINSW208B, C.T ¼ 0.6 mm:NSSW206B, C.T ¼ 0.4 mm:NSSW206B, C.


Spacing

Pitch

0.31

m

Spacing50 mm80 mm100 mm120 mm

G R B G

0.51 m

10, 000

9000

8000

7000

6000

5000

4000–0.15 –0.1 –0.05 0 0.05 0.1 0.15

Position of vertical direction (m)

Lum

inan

ce (n

its)

(a)

(b)

Figure 19.11 (a) Basic design parameters and (b) measured luminance profile as a function oflight source spacing for a 23ʺ BLU with side-emitting RGB Luxeon LEDs.28

16

12

8

4

08 10 12 14 16 18

LED pitch (mm)

Unselected LEDs

Selected LEDs

Du′

v′ (

××100

0)

Figure 19.12 Measured color uniformity for a 23ʺ BLU as a function of LED pitch.28 (Du0v0 isthe deviation of chromaticity on the CIE 1976 (u0, v0) diagram.)

References

1. Hunt R. Displays: past, present, and future. In: 11th Int. Display Workshop 04; 2004.p. 3e6.

2. Boer W. Active matrix liquid crystal displays: fundamentals and applications. London:Newnes; 2005.

3. Insight Media. LCD backlight report. Insight Media; May 2008.4. Total global large-area TFT LCD panel shipments from 2009 to 2015 (in million units).

Available at: https://www.statista.com/statistics/221650/global-large-area-tft-lcd-shipments-since-2009/.

5. Delphi Display Systems, Inc. Comparison of LCD backlight technology for outdoorapplications. Delphi display systems whitepapers. 2008.

6. Ito Y, Tsukahara T, Masuda S, Yoshida T, Nada N, et al. Optical design of phosphor sheetstructure in LED backlight system. SID Symp Dig Tech Pap 2008;39:866e9.

7. K€al€ant€ar K, Okada M, Ishiko H. Monolithic block-wised light guide with controlled opticalcrosstalk for field-sequential color/scanning LCD. SID Symp Dig Tech Pap 2009;40:1038e41.

8. Anderson J, Schardt C, Yang J, Koehler B, Ostlie B, et al. New back reflector and front filmfor improved efficiency of direct-lit LED backlights for LCD TV. SID Symp Dig Tech Pap2007;38:1236e9.

9. Ploetz L, Stich A. LED display backlighting e large screen and TV application usingGolden DRAGON® ARGUS®. In: Osram Opto semiconductors application note, March11, 2005; 2005. p. 1e8.

Figure 19.13 Arrangement of RGGB LED clusters with reflector cover foil.9


https://www.statista.com/statistics/221650/global-large-area-tft-lcd-shipments-since-2009/

https://www.statista.com/statistics/221650/global-large-area-tft-lcd-shipments-since-2009/

10. Tsuchiya R, Kawasaki Y, Ikebe S, Shiba T, Kinoshita J. Thin side-lit, hollow-cavity flatLED lighting panel for ultra-uniform LCD backlight applications. SID Symp Dig Tech Pap2008;39:874e7.

11. Martynov Y, Huub Konijn H, Nicola Pfeffer N, Simon Kuppens S, Wim Timmers W. High-efficiency slim LED backlight system with mixing light guide. SID Symp Dig Tech Pap2003;34:1259e61.

12. Philips Lumileds Lighting. SnapLED and SuperFlux categories and labels: Luxeon DCCfor LCD backlighting, application brief AB27. 2005. Available at: http://www.philipslumileds.com/support/documentation/application-briefsb.

13. 3M Corporation. Vikuiti™ brightness enhancement film II (BEF II). 2012. Available at:http://products3.3m.com/catalog/us/en001/electronics_mfg/vikuiti/node_QVCDZB50FVbe/root_GST1T4S9TCgv/vroot_S6Q2FD9X0Jge/gvel_GD378D0HGJgl/theme_ us_vikuiti_3_0/command_AbcPageHandler/output_html.

14. 3M Corporation. Vikuiti™ brightness enhancement film III (BEF III). 2012. Available at:http://products3.3m.com/catalog/us/en001/electronics_mfg/vikuiti/node_Q2ZGN85GDRbe/root_GST1T4S9TCgv/vroot_S6Q2FD9X0Jge/gvel_GD378D0HGJgl/theme_ us_vikuiti_3_0/command_AbcPageHandler/output_html.

15. 3M Corporation. Vikuiti™ dual brightness enhancement multi-functional film (DBEF).2012. Available at: http://products3.3m.com/catalog/us/en001/electronics_mfg/vikuiti/node_DPGRHQ6199be/root_GST1T4S9TCgv/vroot_S6Q2FD9X0Jge/gvel_K6PCGS29T6gl/theme_us_vikuiti_3_0/command_AbcPageHandler/output_html.

16. Eligibility Criteria. ENERGY STAR® program requirements product specification fordisplays. Draft 2, version 6.0. 2012.

17. Uno T, Liu L. Quarterly LED backlight report. Santa Clara: NPD Display Search; 2012.18. LED TELE. LED vs LCD TV comparison. 2012. Available at: http://www.ledtele.co.uk/

ledvslcd.html. [email protected]. AOC Corporate Website. 23-inch Aire Pro i2353Ph monitor. 2012. Available at: http://us.

aoc.com/monitor_display/i2353ph.20. West R, Konijn H, Sillevis-Smitt W, Kuppens S, Pfeffer N, et al. High brightness direct

LED backlight for LCD-TV. SID Symp Dig Tech Pap 2003;34:1262e5.21. Hsieh P, Lin H. Colorimetric transformation with two sets of LED backlight for multi-

primary LCD. SID Symp Dig Tech Pap 2009;40:649e52.22. Semenza P. Sony unveils new TV display technology: direct-emission LED, display search.

2012. Available at: http://www.displaysearchblog.com/2012/01/sony-unveils-new-tv-display-technology-direct-emission-led/.

23. Smith-Gillespie R. Advanced flat panel display backlighting techniques. Proc SPIE 2008;6956:1e11. 69560K.

24. Stich A. LEDs, new light sources for display backlighting. In: Osram Opto semiconductorsapplication note, February 2, 2004; 2004. p. 1e9.

25. Hwang J, Shin D, Gong D, Park C, Kang W, et al. Enhancement of brightness anduniformity by LED backlight using a total internal reflection (TIR) lens. SID Symp Dig TechPap 2008;39:1645e8.

26. H€uttner J. LED display backlighting-monitor applications using 6-lead MULTILED®. In:Osram Opto semiconductors application note, September 9, 2003; 2003. p. 1e7.

27. Folkerts W, West R. LED backlight designs using Luxeon high flux light source solutions.SID 2004 seattle, oral presentation file. 2004.

28. Folkerts W. LED backlighting concepts with high flux LEDs. SID Symp Dig Tech Pap2004;35:1226e9.


http://www.philipslumileds.com/support/documentation/application-briefsb

http://www.philipslumileds.com/support/documentation/application-briefsb

http://products3.3m.com/catalog/us/en001/electronics_mfg/vikuiti/node_QVCDZB50FVbe/root_GST1T4S9TCgv/vroot_S6Q2FD9X0Jge/gvel_GD378D0HGJgl/theme_%20us_vikuiti_3_0/command_AbcPageHandler/output_html



http://products3.3m.com/catalog/us/en001/electronics_mfg/vikuiti/node_Q2ZGN85GDRbe/root_GST1T4S9TCgv/vroot_S6Q2FD9X0Jge/gvel_GD378D0HGJgl/theme_%20us_vikuiti_3_0/command_AbcPageHandler/output_html



http://products3.3m.com/catalog/us/en001/electronics_mfg/vikuiti/node_DPGRHQ6199be/root_GST1T4S9TCgv/vroot_S6Q2FD9X0Jge/gvel_K6PCGS29T6gl/theme_us_vikuiti_3_0/command_AbcPageHandler/output_html



http://www.ledtele.co.uk/ledvslcd.html

http://www.ledtele.co.uk/ledvslcd.html

mailto:[email protected]

http://us.aoc.com/monitor_display/i2353ph

http://us.aoc.com/monitor_display/i2353ph

http://www.displaysearchblog.com/2012/01/sony-unveils-new-tv-display-technology-direct-emission-led/

http://www.displaysearchblog.com/2012/01/sony-unveils-new-tv-display-technology-direct-emission-led/

29. West R, Konijn H, Kuppens S, Pfeffer N, Vader Q, et al. LED backlight for large area LCDTV’s. In: Proceedings of the 10th International Display Workshops; 2003. p. 657e60.

30. West R, Sasser G, Stewart J. Side emitting light emitting device. US Patent 6598998.Lumileds Lighting US, LLC; July 29, 2003.

31. West R, Sasser G, Stewart J. Side emitting LED and lens. US Patent 6679621. LumiledsLighting US, LLC; January 20, 2004.

32. Noh J, Park J, Kim D, Wang J, Hwang J. Side emitting device, backlight unit using the sameas light source and liquid crystal display employing the backlight unit. US Patent 7322721.Samsung Electronics Co, Ltd; January 29, 2008.


LEDs and automotive lightingapplications 20John D. BulloughRensselaer Polytechnic Institute, Troy, NY, United States

20.1 Introduction

The majority of roads in North America and much of the rest of the world are notilluminated by fixed pole-mounted roadway lighting systems.1 Because of this, auto-motive lighting is a key component for driving safely at night. The performancerequirements for vehicle headlamps, such as those published in the United States asFederal Motor Vehicle Safety Standard (FMVSS) 108, are based on standards and rec-ommendations published by the Society of Automotive Engineers (SAE) and similarindustry organizations. These requirements stipulate certain minimum or maximumluminous intensities toward different directions from the center of the vehicle lightingsystem. A similar set of photometric performance requirements exists for countriesoutside North America; these differ in particulars but have the same objectives ofspecifying luminous intensities to ensure vehicle lighting systems provide sufficientlight for drivers to see at night while minimizing glare to other drivers, and to ensurethat vehicle signal lights can be detected promptly and without ambiguity.

This chapter summarizes some of the performance requirements for vehicle lightingsystems and includes a discussion of the impact of light emitting diode (LED) sourceson driver visual responses, compared to filament sources (such as incandescent andtungsten-halogen lamps), the traditional light source used in most automotive lightingapplications.

20.2 Forward lighting

For automotive headlamps that provide illumination ahead of the vehicle, two head-lamps are required to be mounted as far apart as practical. Each headlamp mustmeet the same performance requirements. There are two primary types of beampatterns (beam patterns are the resulting distributions of luminous intensity producedby vehicle headlamps): the high (or driving) beam and the low (or passing) beam.North American requirements for several angular locations for high- and low-beamheadlamps are given in Tables 20.1 and 20.2.

As expected, requirements for high beams have higher intensities and fewermaximum intensity values than low beams. Additionally, the high beam has a symmet-rical beam pattern. In contrast, the low beam has an asymmetrical beam pattern, with


http://dx.doi.org/10.1016/B978-0-08-101942-9.00020-4

more stringent maxima toward the left side (where oncoming traffic in North Americais more likely to be found; beam patterns in countries with left-side traffic are reversedleft-to-right). Fig. 20.1 shows the intensity requirements for a low-beam headlamppattern overlaid onto the angular locations of a straight, two-lane road. Using theinverse-square law, it is possible to convert the angular luminous intensity values toilluminances on the roadway and on objects located ahead of the vehicle; illuminancesfrom each headlamp should be added together to obtain the total.

Fig. 20.2 is a photograph of a low-beam headlamp pattern projected onto a wall infront of the headlamp. To reduce glare on oncoming and preceding drivers, low-beampatterns usually have the sharp vertical gradients shown in Fig. 20.2. Above theso-called cutoff boundary between the light and dark portions of Fig. 20.2, intensitiesare low and there is little light. The cutoff boundary makes it possible to check andadjust the vertical aim of the headlamp. Most North American headlamps requirethe location of the right-side cutoff boundary to be at the same height as the headlamp.2

The left-side cutoff boundary is usually lower than the right-side boundary to reducethe amount of light entering oncoming drivers’ eyes. The sharp cutoff boundary oflow-beam headlamp patterns restricts the visibility of drivers ahead. When drivingspeeds exceed 60e65 km/h, it can be difficult for a driver to detect and stop in timeto see some potential hazards3 when using low-beam headlamps. High beams are

Table 20.1 Selected high beam headlamp photometric requirements inthe United States

Angular location (degrees left/right, up/down)

Maximum luminousintensity (cd)

Minimum luminousintensity (cd)

(0� right, 0� up) 75,000 40,000

(3� left, 1� up) and (3� right, 1� up) e 5,000

(0� right, 2� up) e 1,500




(12� left, 0� up) and (12� right,0� up)

e 1,500

(0� right, 1.5� down) e 5,000

(0� right, 2.5� down) e 2,500

(9� left, 1.5� down) and (9� right,1.5� down)

e 2,000

(12� left, 2.5� down) and (12�right, 2.5� down)

e 1,000

(0� right, 4� down) 12,000 e


Table 20.2 Selected low-beam headlamp photometric requirements inthe United States

Angular location (degrees)Maximum luminousintensity (cd)

Minimum luminousintensity (cd)

(8� left, 0� up) and (8� right, 0� up) e 64




(1.5� right, 0.5� down) 20,000 8,000

(6� left, 1� down) e 750

(2� right, 1.5� down) e 15,000

(9� left, 1.5� down) and(9� right, 1.5� down)

e 750

(15� left, 2� down) and(15� right, 2� down)

e 700

(1.5� left, 1� up) 700 e

(1.5� left, 0.5� up) 1,000 e

(1.5� left, 0.5� down) 3,000 e

(1� right, 1.5� up) 1,400 e

(1� right, 0.5� up), (2� right, 0.5� up)and (3� right, 0.5� up)

2,700 e

(4� right, 4� down) 8,000 e

15

10

5

0

–5–30 –25 –20 –15 –10 –5 0 5 10 15 20 25 30

Maxima

Minima

Road edge

Center line

Horizontal angle (degrees)

Verti

cal a

ngle

(deg

rees

)

SAE low beam requirements

Figure 20.1 Photometric requirements for low-beam headlamp patterns in the United Statessuperimposed onto an image of a two-lane roadway.

LEDs and automotive lighting applications 649

warranted for such conditions, except if approaching traffic is within 100 m or so.However, most drivers underutilize their high-beam headlamps.4

As a result of the sharp cutoff boundaries of low-beam headlamp patterns, verticalaim is a very important factor in their proper performance. In the United States, forexample, most states do not require headlamps to be properly aimed as part of a safetyinspection.1 Studies of vertical aim among vehicles5,6 found that most vehicles had atleast one poorly aimed headlamp. When aim is too high, headlamps can contribute todisability glare and discomfort glare,7,8 resulting in poor ratings from organizationsthat test headlamps for safety.9 When aim is too low, drivers’ forward visibility canbe compromised because of the sharp vertical cutoff; this can also negatively impactsafety ratings.9

Vehicle headlamp systems that can change or adapt in response to different drivingconditions, called adaptive forward-lighting systems (AFS), are starting to causereevaluation of the fixed high- and low-beam headlamp patterns that have been usedfor many decades.10 Cornering and bending lights are being used on some vehicles;bending lights sometimes use mechanical elements to swivel one or both headlampstoward roadway curves. Some European vehicles are equipped with a “town” head-lamp beam pattern that has lower maximum luminous intensities and a broader distri-bution than most low beams to help detect pedestrians while driving at low speeds inurban locations. AFS requirements for most nations are promulgated by the EconomicCommission on Europe (ECE) in Vehicle Regulation No. 123. The FMVSS 108 in theUnited States is presently silent with respect to AFS, although swiveling beams that donot change in their overall intensity distribution are permitted. Adaptive driving beam(ADB) headlamp systems using cameras to identify oncoming headlamps and preced-ing tail lights can dim headlamp intensity specifically in the direction of those sources.As a result, they allow drivers to use high-beam headlamp functionality continuously,while reducing glare for other drivers.11 LED light sources which can dim or switch onand off easily within a “matrix” configuration are ideal for ADB headlamp systems.The SAE12 has developed a standard for testing ADB headlamps, and the inclusionof ADB systems into FMVSS 108 appears likely in the near future.

Presently, most automotive headlamps use filament sources (tungsten-halogen, ormore simply, halogen) with reflector or projector optical systems to produce the

Figure 20.2 Photograph of a low-beam headlamp beam pattern projected onto a wall.


necessary beam pattern. A relatively small proportion of headlamps use high-intensitydischarge (HID) lamps, metal halide types containing xenon, to allow the lamps to beswitched on immediately. Headlamps using LEDs are now available on several vehiclemodels. Regardless of the light source used, all headlamps are required to meet thesame photometric requirements.

20.3 Signal lighting

Vehicles need to have signal lights to allow drivers to communicate about their actionswith respect to braking and turning during both day and night. An increasing propor-tion of vehicle signals uses LEDs for signal lighting purposes. Different signal lightshave different requirements for both color and luminous intensity. Federal require-ments for vehicle signals in the United States are based on SAE standards and recom-mendations. In Table 20.3, the color and permissible luminous intensity values forseveral vehicle signal light types are listed.

Performance requirements for vehicle signal lights in Europe do not differ muchfrom those in North America regarding color and luminous intensity,13 with an impor-tant exception. Turn signal lights at the back of a vehicle can be red or yellow in theUnited States with different intensity requirements depending upon which color theyare. In most of the rest of the world rear turn signals must be yellow. Allen14 reportsthat yellow rear turn signals tend to result in fewer crashes, possibly because of theirhigher luminous intensities than red rear turn signals, or because the yellow colormakes it easier to tell apart rear turn signals from brake or tail lights. The NationalHighway Traffic Safety Administration (NHTSA) is considering whether yellowshould be required for all rear-mounted automotive turn signals.

Dynamic or sweeping turn signals, which provide indication to other drivers aboutthe intended direction of turning, have begun to appear on vehicles outside North

Table 20.3 Photometric and color requirements for vehicle signallights in the United States

Signal functionRequiredcolor

Minimum-maximum luminousintensity (cd)

Tail (presence) light Red 2e18

Stop light Red 80e300

Center high-mounted stop light Red 25e130

Rear turn light Red or yellow 80e300 (if red), 130e750 (if yellow)

Front turn light Yellow 130e750

Backup light White 80e300 (if two), 80e500 (if one)


America. They are presently not allowed for use in the United States under FMVSS108. Research studies have demonstrated that such dynamic signal lights can allowdrivers to identify the turning direction of other vehicles from larger peripheral anglesthan conventional turn signals.15

20.4 Human factor issues with LEDs

LED sources are substantially different from filament lamps used in most present-dayautomotive lighting applications in a number of important ways:

• LEDs have higher luminous efficacies (in lm/W) than filament sources, meaning they canproduce higher intensities or broader beam patterns for the same amount of energy, or similarlight output with lower energy requirements

• The narrowband spectral output of colored LEDs produces a highly saturated color appear-ance, in contrast to broadband sources such as filament lamps, which require filters in orderto produce colored illumination (Fig. 20.3)

• White phosphor-converted LEDs can be produced with higher correlated color temperature(CCT) than filament lamps, which results in a more bluish color appearance

• LEDs have very rapid onset and offset times: 10e20 ns, including the decay time of yttrium,aluminum, and garnet phosphors, compared to about 80e250 ms for filament lamps

The photometric, colorimetric and temporal properties of LED sources can also in-fluence drivers’ ability to see and respond to potential hazards in and along the roadway.For vehicle headlamp systems, the spectral distribution of typical phosphor-convertedwhite LEDs, based on blue InGaN devices in combination with YAG phosphors, has

1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0400 420 440 460 480 500 520 540 560 580 600 620 640 660 680 700

Wavelength (nm)

Rel

ativ

e po

wer

Yellow incandescent

Red incandescent

LED dominant wavelength (nm) 596 616 637

Figure 20.3 Spectral distributions of yellow and red LED and (filtered) filament sources.


a larger proportion of short-wavelength (blue) light than the spectral distribution offilament sources like incandescent and halogen lamps (Fig. 20.4). This difference isrelevant to visual performance while driving, because at light levels commonly experi-enced while driving at night, resulting in asphalt pavement luminances between 0.1 and1 cd/m2,16 visual detection of hazards is supported by a combination of cone and rodvisual receptors in the eye.

However, photometric quantities such as illuminance (in lx), luminance (in cd/m2),luminous intensity (in cd), and luminous flux (in lm) are entirely based on the spectralresponse of the cone receptors in the eye. Cone receptors are used exclusively forseeing at light levels typically experienced outdoors and indoors during the daytime(usually between 10 and 1000 cd/m2). This apparent discrepancy between the waylight is measured and how we see matters because collectively rod receptors aremore sensitive to short visible wavelengths (such as blue and green light) than conereceptors.17 Thus, the usual photometric quantities (lx, cd/m2, cd, lm) can underesti-mate a driver’s ability to see under LED sources at night, relative to his or her abilityto see under filament lamps.

A unified photometric system has been published by the Commission Internationalede l’�Eclairage (CIE) to quantify the relative role of rods and cones18 in seeing at night.As a consequence, it could be possible to obtain equivalent nighttime visual perfor-mance using LED sources that produce light levels that are 20%e30% lower thanthose produced by filament lamps.19 Another visual response that may favor LEDsover filament sources is the perception of roadway scene brightness, according to abrightness model developed by Rea et al.20 This response appears to have increasedshort-wavelength sensitivity. Fig. 20.5 shows the predicted roadway scene brightnessunder headlamps using filament, HID, and LED sources.

The relatively high amount of short-wavelength spectral power in white LEDillumination might also have some possible negative impacts for vehicle lighting,however. When headlamps of different colors produce equivalent conventional

1

0.8

0.6

0.4

0.2

0400 500 600 700

Wavelength (nm)

FilamentLEDR

elat

ive

pow

er

Figure 20.4 Spectral distributions of white LED and (unfiltered) filament sources.


photometric quantities, disability glare (a reduction in visual performance that iscaused by scatter in the eyes from a bright light) is not influenced by the spectral con-tent of the headlamp illumination.21 This is not the case for discomfort glare, which isdefined as an annoying or painful sensation that is experienced when viewing a brightlight in the visual scene of interest. Like the perception of roadway scene brightness,discomfort glare also exhibits increased sensitivity to short-wavelength light.22 It is notfully understood whether, or to what extent, increased discomfort glare affects drivingsafety. There is some evidence that shows that when drivers experience discomfortglare from oncoming headlamps, they are more likely to exhibit driving behaviorssuch as increases in head movements and increased throttle variability, which inturn have been found to be correlated with increased crash risk.23

Regarding the visual detection of vehicle signal lights, because LEDs have substan-tially shorter onset times than filament lamps, they can have some advantages, espe-cially for vehicle brake lamps. Bullough24 demonstrated that visual reaction times tothe onset of a colored light signal, such as a brake light or turn signal, could be pre-dicted by a threshold quantity light energy (in cd s) received at drivers’ eyes. A tung-sten filament lamp when first switched on results in a relatively gradual increase inillumination from the filament, which can take up to 250 ms to reach full brightness.LEDs have practically instantaneous rise times and can produce the threshold quantityof light energy more quickly. As a result, LEDs elicit shorter visual reaction times thanfilament sources of the same nominal color and peak luminous intensity.25

Importantly, because the rate of deceleration of a braking vehicle is linked to thesame action that turns on the brake light itself, pressing the brake pedal and shorterlight source rise times can provide a stopping distance benefit of nearly 7 m for adriver following a braking vehicle,26 a small but sometimes practically significantincrease.

140%

120%

100%

80%

60%

40%

20%

0%

Rel

ativ

e br

ight

ness

1 lx 3 lx 10 lxIlluminance on pavement

HalogenHIDLED

Figure 20.5 Relative brightness of roadway pavement surfaces illuminated by photometricallyequated light sources (halogen, HID, and LED).


20.5 Energy and environmental issues

Because they have higher luminous efficacies compared to filament sources, automo-tive lighting systems using LED sources can have substantially reduced power require-ments. In separate studies, Hamm27 and Schoettle et al.28 estimated the typicalwattages for conventional filament source-based vehicle lighting systems and forLED lighting systems. The average of their estimates for different lighting andsignaling functions are summarized in Table 20.4.

Also listed in Table 20.4 are estimated values for the total annual hours of use foreach type of lighting system, based on driving patterns in United States.29 Table 20.4also includes the resulting total annual lighting energy use for filament- and LED-based automotive lighting systems. Under the assumption that each kilowatt hour oflighting energy use on a vehicle powered by gasoline corresponds to CO2 emissionsof 1.29 kg,28 the total reduction in annual energy use that would be expected to accom-pany a shift from filament lamps to LEDs for automotive lighting would be 27.4 kWh/year, and would correspond to an annual reduction of CO2 emissions of about 35 kg/year for each automobile.

Table 20.4 Estimated power and energy use of filament lamp and LEDautomotive lighting systems

Power per vehicle(W/vehicle)

Annual energy use(kWh/year)

FunctionFilamentsource

LEDsource

Annual use(h/year)

Filamentsource

LEDsource

Low beam headlamp 124 87 97 12.08 8.47

High beam headlamp 132 64 10 1.29 0.63

Daytime running lamp 48 18 382 18.30 7.03

Position lamp 14 3 107 1.54 0.29

Front turn signal 52 14 22 1.15 0.31

Rear turn signal 52 10 22 1.15 0.22

License plate lamp 17 2 107 1.80 0.16

Reverse lamp 43 7 4 0.16 0.03

CHMSL 34 4 81 2.73 0.28

Brake signal 52 11 81 4.16 0.86

Tail lamp 14 2 107 1.52 0.26

Total annual energy use (kWh/year): 45.9 18.5


20.6 Future outlook

LED automotive lighting systems are already common for signal lighting applications,and have been introduced for forward headlamp systems. The rapid advances in lumi-nous efficacy will continue to make them increasingly attractive for automotive use.The solid-state construction of LED systems, modular configurations, and relativeease of intensity control through current modulation or pulse width modulation pro-vides significant promise for energy-saving vehicle lighting systems that can adaptin real time to changing roadway traffic and weather patterns. LEDs are particularlysuited for adaptive headlamp systems, where the intensity can be modulated inresponse to oncoming or preceding vehicles. The advantages of LEDs are also makingdynamic turn signal lighting systems practical.

20.7 For further information

For additional information about automotive lighting in general, including the growinguse of LED sources, consult W€ordenweber et al.10 An overview of the components ofthe roadway transportation lighting system, including automotive lighting, roadwaylighting, and traffic signals, is provided by Bullough.30 For an extensive discussionof the human factors aspects of lighting for transportation, Boyce31 is an excellentresource. Research from the National Highway Traffic Safety Administration on vehiclelighting systems can be found online at https://www.nhtsa.gov/human-factors/human-factor-program-areas#human-factor-program-areas-visibility-and-lighting, and reportsfrom the Lighting Research Center at Rensselaer Polytechnic Institute are availableonline at http://www.lrc.rpi.edu/programs/transportation/TLA/PublicInformation.asp.

References

1. National Highway Traffic Safety Administration. Nighttime glare and driving performance:report to Congress. Washington: U.S. Department of Transportation; 2007.

2. Schoettle B, Sivak M, Takenobu N. Market-weighted trends in the design attributes ofheadlamps in the U.S. In: Automotive lighting technology. Warrendale: Society ofAutomotive Engineers; 2008. p. 85e93.

3. Andre J, Owens DA. The twilight envelope: a user-centered approach to describing roadwayillumination at night. Hum Factors 2001;43:620e30.

4. Sullivan JM, Adachi G, Mefford ML, Flannagan MJ. High-beam headlamp usage onunlighted rural roadways. Lighting Res Technol 2004;36:59e65.

5. Bullough JD, Skinner NP, Pysar RP, et al. Nighttime glare and driving performance:research findings. Washington: National Highway Traffic Safety Administration; 2008.

6. Skinner NP, Bullough JD, Smith AM. Survey of the present state of headlamp aim. In:Transportation Research board 89th annual meeting. Washington: Transportation ResearchBoard; 2010.


https://www.nhtsa.gov/human-factors/human-factor-program-areas#human-factor-program-areas-visibility-and-lighting

https://www.nhtsa.gov/human-factors/human-factor-program-areas#human-factor-program-areas-visibility-and-lighting

http://www.lrc.rpi.edu/programs/transportation/TLA/PublicInformation.asp

7. Perel M. Evaluation of headlamp beam patterns using the Ford CHESS program. In:Gaudaen G, editor. Motor vehicle lighting. Warrendale: Society of Automotive Engineers;1996. p. 153e7.

8. Sivak M, Flannagan MJ, Miyokawa T. Quantitative comparisons of factors influencing theperformance of low-beam headlamps. Ann Arbor: University of Michigan; 1998.

9. Bullough JD. Vehicle headlights: aiming for better driving safety. Leukos 2016;12:183e4.10. W€ordenweber B, Wallaschek J, Boyce P, Hoffman DD. Automotive lighting and human

vision. New York: Springer; 2007.11. Bullough JD, Skinner NP, Plummer TT. Assessment of an adaptive driving beam

headlighting system: visibility and glare. Transp Res Rec 2016;2555:81e5.12. Society of Automotive Engineers. Adaptive driving beam. J3069. Warrendale (PA): Society

of Automotive Engineers; 2016.13. Bullough JD, Van Derlofske J, Kleinkes M. Rear signal lighting: from research to standards,

now and in the future. In: Automotive lighting technology and human factors in drivervision and lighting. Warrendale: Society of Automotive Engineers; 2007. p. 157e66.

14. Allen K. The effectiveness of amber rear turn signals for reducing rear impacts.Washington: U.S. Department of Transportation; 2009.

15. Bullough JD, Skinner NP. Dynamic signal lighting: off-axis detection and directional signalidentification. In: Societe des Ingenieurs de l’Automobile Vehicle and Infrastructure SafetyImprovement in Adverse Conditions and Night Driving Congress Proceedings, Paris,France, October 13e14; 2016. 5 p.

16. He Y, Rea MS, Bierman A, Bullough JD. Evaluating light source efficacy under mesopicconditions using reaction times. J Illum Eng Soc 1997;26:125e38.

17. Rea MS, Bullough JD, Freyssinier JP, Bierman A. A proposed unified system ofphotometry. Lighting Res Technol 2004;36:85e111.

18. Commission Internationale de l’�Eclairage. Recommended system for mesopic photometrybased on visual performance. Vienna: Commission Internationale de l’�Eclairage; 2010.

19. Van Derlofske J, Bullough JD. Spectral effects of LED forward lighting: visibility and glare.In: Automotive lighting technology and human factors in driver vision and lighting. War-rendale: Society of Automotive Engineers; 2006. p. 11e8.

20. Rea MS, Radetsky LC, Bullough JD. Toward a model of outdoor lighting scene brightness.Lighting Res Technol 2011;43:7e30.

21. Schreuder DA.White or yellow lights for vehicle head-lamps?. Voorburg: Institute for RoadSafety Research; 1976.

22. Bullough JD. Spectral sensitivity for extrafoveal discomfort glare. J Mod Opt 2009;56:1518e22.

23. Bullough JD. Investigation of the influence of headlight glare and aim on risk-relateddriving behavior. In: Society of automotive Engineers world Congress experience,Detroit, MI, April 4e6; 2017. 11 p.

24. Bullough JD. Onset times and the detection of colored signal lights. Transp Res Rec 2005;1918:123e7.

25. Bullough JD, Yan H, Van Derlofske J. Effects of sweeping, color and luminance distri-bution on response to automotive stop lamps. In: Advanced lighting technology for vehicles.Warrendale: Society of Automotive Engineers; 2002. p. 179e83.

26. Sivak M, Flannagan M, Sato T, et al. Reaction times to neon, LED, and fast incandescentbrake lamps. Ergonomics 1994;37:989e94.

27. Hamm M. Green lighting: analysing the potential for reduction of CO2 emissions in full-LED headlamps. In: Automotive lighting technology. Warrendale: Society of AutomotiveEngineers; 2009. p. 9e14.


28. Schoettle B, Sivak M, Fujiyama Y. LEDs and power consumption of exterior automotivelighting: implications for gasoline and electric vehicles. In: 8th International symposium onautomotive lighting. Munich: Herbert Utz Verlag; 2009. p. 11e20.

29. Buonarosa ML, Sayer JR, Flannagan MJ. Real-world frequency of use of lightingequipment. Ann Arbor: University of Michigan; 2008.

30. Bullough JD. Roadway transportation lighting. In: Kutz M, editor. Handbook oftransportation engineering, vol. II. New York: McGraw-Hill; 2011. 8.1e8.24.

31. Boyce PR. Lighting for driving. New York: CRC Press; 2009.


LEDs for large displays 21Linas SvilainisKaunas University of Technology, Kaunas, Lithuania

21.1 Introduction

Large-scale LED displays have proved to be valuable information presentation devicefor decades. The paper by R. Haitz1 dated as far back as 1974 provides trends on LEDdisplay technology. First LED display prototype is attributed to J.P. Mitchell, who pre-sented such in 1977.2 First patent on such a display was filled in 1979 by Y. Okuno.3

The boost to LED displays started at the end of 1990,4e6 changing the Times Square inNew York and the Piccadilly Circus in London. Other markets were ready for LEDvideo displays considerably later because LEDs were so expensive that it was unthink-able they could be used in quantities required for displays. LED produces light by theluminescence7 generated in p-n junction diode that gives light output after a biasvoltage is applied to its terminals. Large amounts of carriers are injected into the deple-tion region thanks to the forward bias current. Recombination of these carriers emitsenergy. Direct recombination results in a photon emission. With appropriate bandgapsize, this radiated energy may be of wavelengths in the visible spectrum. Thoughsingle-element semiconductors are not suitable for LEDs’ production, there aremany binary or ternary compounds that can be used. Energy gaps can be tuned to adesired emission wavelength by a semiconductor composition. Boost of LED displayshappened in 1980s, thanks to the development of GaAlAs.8 This technology providedsuperior performance over previous devices: 10 times greater brightness and loweroperation voltage. It is essential for displays that LEDs can be easily pulsed or multi-plexed at high speed. This enabled their application for message and outdoor signs.Appearance of InGaAlP LEDs in 1990s gave the mechanism for LED output colordesign: if color can be adjusted by the energy bandgap, then same technology canbe used to manufacture green, yellow, orange, and red LEDs with sufficient efficiency.Moreover, the LED light output degradation of InGaAlP material was less affected bytemperature and humidity. Display color rendering is the same RGB tristimulusapproach as in the CRT. Blue was missing. Introduction of InGaN in 1994 completedthe required set of colors. The only problem was how to regulate the pixel intensitywith display resolution and size growing: both price and efficiency are necessary. Inthe CRT, this is done by an electron beam deflected by a scanning system. In theLCD, intensity is controlled by shading out the unwanted light portion, so image for-mation is relatively simple and large number of pixels can be controlled sequentially.In LED displays, the pixel control is the major problem.5,9,10 Much development hadtaken place until LED control settled to some driving circuitry standard. When semi-conductor manufacturers rushed into this market, this resulted in dedicated control ICs


https://doi.org/10.1016/B978-0-08-101942-9.00021-6

and a significant price drop. Today a dozen manufacturers offer LED drivers that arededicated for displays. Reduction of LED price and introduction of blue LED7,8,11 wasa new push and worldwide development started around 2000.10,12e18 The worldwidemarket for these displays in 1999 was already 300 million USD.19 According to aglobal market study20 provided by Persistence Market Research, the global outdoorLED displays market was 5 billion USD in 2014, and it is anticipated to grow to 15billion USD by 2021. Same study20 indicated that the LED billboard segment wasvalued at 2.5 billion USD in 2014, the LED video walls segment at 807 millionUSD in 2014, rental market for outdoor LED billboard displays market at 2.2 billionUSD, and rental outdoor LED displays market at 1.5 billion USD. Indoor surfacemounted rental market was valued at 2.6 billion USD in 2014. All markets were fore-cast to grow at 15%e18% compound annual growth rate (CAGR). Increasing demandfor advertisement and use of LED displays at stage, shopping malls, and conferencerooms is driving the market growth. Demand for better image quality (resolutionand contrast) for large audience and display price reduction are the primary factorsincreasing the market adoption of LED displays. The luminous performance ofLEDs can now satisfy the most demanding LED screens. Achievements in blue andgreen LED technology no longer object the color gamut of LED displays. Color gamutof LED screen is larger than conventional TV standards or CRT phosphor. Althoughbeing capable of displaying many colors, LED displays are mainly intended for largeaudience imaging due to their price. LED displays have a long life; can provide theunbeatable luminance allowing for clear visibility outdoor even under the direct sunlight; can work under harsh environmental conditions. LED displays can satisfy anyneed for large-scale text, graphics, or video information; are used in street billboards,traffic signage, stage decorative illumination, and imaging.

This chapter discusses LED video display types, structure, operation, quality pa-rameters, and issues encountered.

21.2 LED display types

By operation conditions (mainly ingress protection, defined by IEC standard 6052921)displays are classified as outdoor and indoor. Outdoor LED displays require higher de-gree of protection22 against dust and humidity (usually must meet IP65 class). Outdoordisplays are exposed to sun radiation. Therefore, LEDs and outer surfaces have to beUV resistant, have special louvers, and its front face design should be used to increasethe contrast. In order to ensure the high contrast, recommended luminance (brightness)of outdoor display is 5000 nt (nit, nonstandard name, 1 nt ¼ 1 cd/m2). Such brightnessis roughly 15,000 lux, which compares to full daylight, not direct sun conditions(10,000e25,000 lux), overcast day is 1000 lux. Fortunately, required viewing anglefor outdoor is not wide,� 35 to�50 degrees,23 so larger louvers and higher directivityLEDs can be used.24,25 Heat drainage from electronics and proper ventilation areessential here. Components used must account the operating temperature range. IndoorLED displays are usually operated in room conditions; therefore, some parameters can


be relaxed: IP20 class for ingress protection, brightness 1000 nt. Yet, since usuallyviewed at closer distances, indoor displays (Fig. 21.1) require better image quality,higher resolution (pitch), and wider viewing angle.

By construction LED displays can be divided into modular, mesh, flexible, rotary,or mobile design. Modular design allows for rigid assembly of large display using rela-tively small-size modules (Fig. 21.2).

Module size and shape varies from few meters to few decimeters. Modules can bededicated for stationary or temporary installation. In case of temporary installation, fastassembly and dismantling is required. Therefore easy to install yet strong enough tohold large structure and accurate interlocking mechanisms are included in moduledesign (see Fig. 21.3).

Such temporary assemblies are usually used for concerts and political and socialevents. A lot of care is paid for the ease and accuracy of assembly since even a fractionof millimeter deviation is noticeable on displayed image. The weight of the moduleand durability of the design are important factors. All this reflects on the cost ofsuch a display.

Figure 21.1 TWA Series LED video wall display with the 0.9-mm pixel pitch from Leyard.

LED lmap LED array(LED tile unit)

LED tile(LED module) LED screen

Figure 21.2 Concept of the modular display design.15

LEDs for large displays 661

In case of stationary installation, holding structure is usually produced locally, sim-ple fixation of the modules is used, and sometimes even the air conditioning on thebackside is made available from the supporting infrastructure. Therefore the cost forthis type of displays is lower and impressive size can be achieved (Fig. 21.4).

Modular construction can be used to create images on a floor either for stage deco-rative lighting or for entertainment: a variety of visual effects can be created dependingon the movement of the objects on it (Fig. 21.5).

Figure 21.3 Rear view of outdoor RENTAL 768-3 module from RGG LED.Copyright RGG LED 2017.

Figure 21.4 Large stationary outdoor LED display at Piccadily Circus, London.


Mesh design allows covering large areas using semitransparent construction of theLED display (Fig. 21.6). Sometimes mesh is arranged into the module, which providespower supply and controls signals distribution.

If made for outdoor use, LED strips or individually controllable LED pixels can beinstalled on a building façade (Fig. 21.7 left). View from inside the building is notobstructed when mesh-type display modules are used. Strip-type LED display modulescan be installed in between the window frames (Fig. 21.7 right). In such case light is radi-ated outwards the building interior so illumination inside the building is not disturbed.

If pixel strips or individual pixels are the building blocks, any shape of LED displaycan be produced. Most popular form is the pixel strip, where pixels are mounted on somethin supporting material with possibility to supply power and serial data through someconnector or solderable contacts. Flexible forms can be obtained using novel designsoffered by many manufacturers. Individual pixels give higher flexibility; yet, data andespecially power delivery can be a challenge. Pixel spacing can be controlled by placingthem randomly or in a raster fashion. Large areas can be easily covered if pixels are

Figure 21.5 LED floor display with interactive function by EKTA.Copyright © 2017 EKTA.

Data and powerconnector

Linear LED modules

Assembly hoist

Figure 21.6 Transparent LED display modules construction example.


arranged in a curtain. Power supply modules and signal distribution electronics are hid-den either in the upper part of truss, on the top, or on the ground, in the lower part of thedisplay. Mesh display does not obstruct the view behind, and is almost transparent. LEDscan be directly deposited on glass or plastic substrate, tiny or transparent conductors canbe used, printed circuit boards (PCBs) for electronics can be made very thin or orientedalong the viewing direction. Transparency beyond 90% can be attained. Such transparentdisplays are also used to install in building windows. Impressive effects can be obtained,when the image appears from thin air on a stage.

Rotary or spin design is rendering the image using just 1-dimensional LED array:another dimension is obtained by scanning LED array in space, usually using rotarymotion (Fig. 21.8).

Window

Linear LEDdisplay modules

Figure 21.7 LED display installation on a building façade of the Torre Agbar in Barcelona (left)and construction details (right) .

Motor

Rotationaxis

Rotation

LinearLEDpixelsarray

Figure 21.8 Rotary LED display principle (left), photo of image rendering (center) and finaldesign (right) .


Such an approach allows to render the image using much smaller amount of LEDdisplays; yet, the image obtained is well blurred along mechanical scanning directionand usually is 360 degrees.

Mobile LED screens are LED displays mounted on the side or rear portion of thevehicle, can be installed into truck walls, or extractable. Such type of LED displaysare popular for their extreme mobility and are used at outdoor events for promotion,bar or restaurant openings, exhibiting shows, political campaigns, sports events, con-certs, and festivals. They can be quickly put into operation or moved to a new placeand so are an ideal choice for short-time rentals.

21.3 Display parameters

A display is the final device in the video information transmission system. The end-useris a human vision that has to be deceived in order to fulfill the actual function of displays.Therefore, human visual sensing and cognition must be accounted by the display designof quality evaluation. The International Committee for Display Metrology, part of Def-initions and Standards Committee of Society for Information Display, is trying to set thestandards for display metrology.26 ANSI/INFOCOMM standard 3M-201127 definescontrast definition. VESA Flat-Panel-Display Measurements (FPDM) standard28 is anattempt for conventional displays. ISO 9241 standard series29 cover essential aspectsof display measurements. See publication30 on standards analysis. No dedicated standardexists that can be applied to LED display quality evaluation. Most authors agree thatdisplay parameters have to be related to human vision.

The human eye can be compared to a camera: light is refracted by an adjustable lensand brought into focus at retina where neural sensing is performed.31 Two different typesof photoreceptors at the retina are responsible for sensing: cones and rods. There are just6 to 7 million cones, most of them are concentrated at the focal spot fovea centralis about0.3 mm diameter area (2 degrees of the visual field). Amount of rods is much larger,about 120 million, but distributed evenly, though are absent at fovea spot. Cones havea greater resolving power than rods thanks to one-to-one mapping onto visual nervesand are sensitive to wavelength, that is, can resolve hue. Rods are more sensitive to lightthan cones, have many-to-one connections with the visual nerves, thus providing sum-mation of low light signals. Rods are also better sensitive to motion.

21.3.1 Radiant intensity terms

Radiant flux or radiant power (W) is a measure of the total power of electromagneticradiation.32,33 Radiant intensity is a measure of the intensity of radiation. It is definedas the amount of power per unit solid angle (W/sr). Discrimination of differences inlevels of brightness, saturation, or hue is governed by a psychophysical functionknown as Weber’s law.31 Photopic vision occurs at approximately 2 lx level whereboth rods and cones are sensitive so color can be perceived. Scotopic vision takes placeat lower illumination (2$10�7 to 2 lx) but color perception is gradually reduced. Two


luminance functions describe the human eye sensitivity to light at different wave-lengths. For high illumination, the photopic luminance function is used (Fig. 21.9).It is considered pure photopic above 3.4 cd/m2. For low light levels, the scotopic func-tion is relevant. Pure scotopic vision occurs below 0.034 cd/m2. Mesopic vision occursbetween photopic and scotopic vision in low lighting situations. A mesopic curve hasbeen proposed in CIE Technical Report 191:2010 by the Technical Committee TC1-58. The photopic luminance function is the CIE standard curve used in the CIE 1931color space.34,35 A scotopic curve, though available from CIE, is not of concern forlarge-scale video displays.

As can be seen from Fig. 21.9, luminous flux differs from radiant flux: luminousflux is adjusted to the varying sensitivity of the human eye to different wavelengths.It represents the rate at which light energy is emitted from a source; expressed in lu-mens (lm). Luminous intensity is the luminous flux per unit solid angle, measuredin candelas (cd).7,16,31,36 Luminance is a measure of brightness, that is, the amountof light per unit area, either emitted by or reflected from a surface, measured in can-delas per square meter (cd/m2) or nits (nt, not officially recognized unit). The optimalbrightness of the display is defined by the current level of the ambient lighting. There-fore display brightness should be adjusted, especially for outdoor display, dependingon the daylight conditions. This requirement will limit the performance of otherdisplay parameters that are discussed further.

Contrast is the measure of the display to reproduce the various levels of brightness.Different procedures have been proposed in the literature.37,38 Contrast ratio orSequential-contrast28 of LED display is measured as full On/Off contrast:

Cseq ¼ Lmax

Lmin; (21.1)

where Lmin is the display brightness (luminance) at black and Lmax is the luminance atmax white. This type is the most favored by manufacturers as it yields a larger number

400 450 500 550 600 650 700 7500.0

0.2

0.4

0.6

0.8

1.0

470

520

625

Nor

mal

ized

lum

inos

ity, a

.u.

wavelength, nm

Figure 21.9 Photopic luminance function with most common RGB LED wavelengthsindicated.


for the contrast ratio. The reason is that even at black level there could be a leakage inelectronics or scatter in the light path, projection lens, or LED encapsulation. Whendisplaying just a black image, there is no leak to compromise the black level.

Weber Contrast or Weber fraction, ANSI contrast measurement27 uses a checker-board pattern of 16 rectangles, 8 white and 8 black. Brightness values of all the whitesquares are measured and averaged, and brightness of the black squares is measuredand averaged as well. The ratio of the averaged white readings to the black readingsis the ANSI contrast ratio.

CW ¼ Lmax � Lmin

Lmin; (21.2)

With the checkerboard, part of light can be scattered from LED tinting or dust.There are a lot of critics on such measurement.39 Measurements are carried out indark room or using a straylight-elimination tube which has little relation to real displayoperation conditions: both outdoor and indoor displays are operated under highambient light levels. Daylight contrast (ambient contrast) proposed in Ref. 40 is thecontrast under outdoor illumination conditions. Ambient illuminance can be veryintense when dark areas of a display are corrupted by reflected light. Anotherapproach, proposed in Ref. 41, is called contrast rating for high ambient light. It isbased on brightness and reflectance measurements to indicate how well a display per-forms in high ambient lighting conditions. Reflectance is expressed as a fraction ofincident light that is reflected at the surface. High display surface reflectance can createglare, dramatically reducing the display contrast, causing annoyance and discomfortwhile watching such a display. Minimizing or eliminating reflectivity and glare isessential for display performance.24 Lower luminance displays require greater contrastto achieve the same visibility of objects.31 Contrast-enhancing filters are effectivelyused in conventional displays. Especially attractive are micromesh filters: placed onthe display surface, they limit light penetration insomuch as only rays falling perpen-dicular to the mesh can penetrate; this stops both specular and diffuse reflections andincreases contrast. However, such filters are not applicable and too expensive for largesurfaces; they also reduce the amount of the display light output. Therefore LED dis-plays usually use exposed LEDs and etching or frosting of the display surface aroundthe LEDs is used to break up and scatter the specular reflections. Quarter-wave coat-ings can be used to reduce the reflections from the material around the LEDs. Opticalcross talk can occur between neighboring LEDs which further reduces the contrast.Contrast enhancement is also achieved by using hooded shades, louvers, or just tiltingthe display away from the offending light.24,25

As pointed in Ref. 42 an absolute luminance of the original scene is rarely importantin image capture and presentation. Human vision lightness sensation L* response tonormalized luminance Y is nonlinear43: two neighboring grayscale patches cannotbe distinguished if their luminance differs less than 1%.42 In other words, vision dis-criminates more shades in the dark areas of the image. Eye can resolve 1:1,000,000 oreven 1:10,000,000 contrast range if allowed to scan (saccade) the image for a long timeand up to 1:1000 in transient mode.44,45 Such wide dynamic range would have been


impossible to code if linear image transfer from camera to display is used. Fortunately,at the beginning of TV broadcasting it turned out that CRT response is nonlinear.Luminance L of CRT is roughly a power g function of cathode voltage V 46:

L ¼ Vg: (21.3)

Then it turned enough to have an inverse (Eq. 21.3) at the camera to efficientlyencode the image. The constant g, that is, “gamma” gave a name for gamma correctionin video and computer graphics. Coding the luminance to lightness by the use of a po-wer law between 1/3 and 0.45 matches the perceptual performance. Coding of theincoming image intensity into a gamma-corrected signal makes maximum perceptualuse of the channel capacity, allowing significant data steam compression. Thenonlinear signal is transformed back to the linear intensity at the display using the in-verse. Sensitivity at black level is infinity if coding is done using (Eq. 21.3), whichwould create some noise at black. CIE publication 1547,48 recommends the approxima-tion where linear segment at low levels minimizes the effect of noise:

L� ¼

8>><

>>:

116$Y13n � 16; Yn > 0:008856

903:3$Yn; Yn < 0:008856; (21.4)

where Yn is the luminance normalized by the white balance reference. Roughly(Eq. 21.4) corresponds to “pure” gamma 0.42 in Eq. (21.3). Such gamma correction isapplied (Fig. 21.10) at the camera end.

The resulting luminance at the system output should be the same as at the input.With CRT displays it was easy: CRT had the inherent gamma correction so decom-pression at the display end was natural. LED display, as well as OLED49

pixel luminance has linear response; therefore, decompressing gamma correctionhas to be implemented to adapt the standardized image compression used in camera.

Use of (Eq. 21.4) creates an excessive compression at low intensities.42 ITU recom-mendation BT202050 offers slightly different response to mimic lightness L2020

perception at the camera end balancing the coding space better:

Y 02020 ¼

(4:5$Yn; Yn � b

aY0:45n � ð1� aÞ; Yn > b

; (21.5)

Eye emulation, compressionusing γ=1/2.4=0.42

Decompressionusing γ=2.4

Yin YoutL LStorage or

transmission

Linear Yout< > Yin relation

Figure 21.10 Path of image compression using the gamma function.


where a ¼ 1.099 and b ¼ 0.018 for 10-bit systems a ¼ 1.0993 and b ¼ 0.0181 for12-bit systems. Extensive analysis of gamma correction influence on image codingquality can be found in Ref. 42. Refer to Fig. 21.11 for graphical comparison of all therecommendations and “pure” gamma 2.5.

Decompressing gamma at LED display uses power of 2.4 in Eq. (21.3). Followingthe ITU recommendation 2020 following approximation can be derived:

Yn ¼

8>>>>>><

>>>>>>:

Y 0

4:5; Y 0 � 4:5b

�Y 0 þ ð1� aÞ

a

� 10:45

; Y 0 > 4:5b

; (21.6)

where Yn is the decompressed, linear RGB tristimulus value of R, G, or B, and the Y0 isthe gamma-compressed RGB values. Publication51 offers the approximation using twopower functions as better approximation of the CRT response. The idea mainly is thesame: as linearity of the last few codes has almost no effect on the image, use the imagecoding with peak codes going beyond the screen capability and then simply clip theresponse at the maximum. In such a way, the total image appears brighter giving slightadvance over the competitors. The coding table is also exploited better. The number ofthe reproducible gray shades is useful for evaluating the capability of the display torender pictorial information or the range of luminance levels that can be used forcoding. The highest luminance level is determined by display capabilities, but thelowest level is determined by the luminance of the display surface when no signal ispresent. A bright light incident on the display can elevate this minimum level andreduce the number of usable gray shades.

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00

10

20

30

40

50

60

70

80

90

100

"Pure" γ=2.5L*=Y1/γ

CIE 15 ITU rec.2020Ligh

tnes

s L*

,a.u

.

Normalised luminance Yn, a.u.

Figure 21.11 Nonlinear human vision response approximations.


Gamma correction also serves to improve the image acceptance and is widely usedin image and video editing (Fig. 21.12). This is equivalent of inserting additionalgamma correction in data transmission chain in Fig. 21.10. For instance: if displayis viewed in a dim environment, the image will lack a contrast.43 If additional end-to-end 1.1e1.2 gamma correction is applied, this effect can be overcome. Viewingin bright environments, such image will appear as high contrast. In such case no addi-tional gamma correction is required.

Actually, this applies to the whole scenery (display plus ambient), only assumingthat human vision cone is relatively narrow and display is viewed at standard condi-tions,52 just image on the display can be analyzed.

21.3.2 Color sensing-related terms

Cones are sensitive to wavelength of the light and are used to derive the hue.53 Conesare mainly of three types, each sensitive to a different portion of the visible light spec-trum (Fig. 21.13 for standard CIE 31 observer sensitivities). Peak sensitivities are atyellow (570 nm), green (540 nm), and violet (440 nm). Thanks to the three types ofphotoreceptors, three numerical components are sufficient to describe a color if

Figure 21.12 Additional gamma correction effect on image: original image (center), g ¼ 0.3(left) g ¼ 3 (right).

350 400 450 500 550 600 650 700 750 8000.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

sens

itivi

lty, a

.u.

wavelength, nm

R,xλ

G,yλ

B,zλ

Figure 21.13 Color matching functions xl; yl; zl34 for standard CIE31 observer.


appropriate spectral weighting functions are used. In 1931, CIE adopted standardcurves for a Standard Observer. The CIE system is widely used on illumination sourcesand displays.

To calculate CIE X, Y, and Z values for light with spectrum P(l), one must sum theproducts of the color matching function weights and intensity recorded (Fig. 21.13) ateach wavelength:

X ¼ DlXlmax

lmin

xlPðlÞ;

Y ¼ DlXlmax

lmin

ylPðlÞ;

Z ¼ DlXlmax

lmin

zlPðlÞ;

Dl ¼ lmax � lmin:

(21.7)

In the CIE31 system, the intensities of red, green, and blue are transformed intowhat are called the tristimulus values, which are represented by the capital letters X,Y, and Z. These values represent the relative quantities of the primary colors. The co-ordinate axes of chromaticity diagram can be derived from Eq. (21.8) values:

x ¼ X

X þ Y þ Z;

y ¼ Y

X þ Y þ Z;

z ¼ Z

X þ Y þ Z:

(21.8)

The coordinates x, y, and z are called chromaticity coordinates, and they always addup to 1. Then z can always be expressed in terms of x and y, so only x and y are requiredto specify any color. By assuming monochrome emitter of P(l), a range of colors eye isable to sense is enclosed in a horseshoe diagram (Fig. 21.14).

In addition, a source or display usually specifies the (x, y) coordinate of the whitecolor used as pure white is not usually captured or reproduced.16 White is defined asthe color captured or produced when all three primary signals are equal; it has a subtleshade of color to it.

To specify the white screen, color temperature of a white light source, given in units ofKelvin, is used. That is, chromaticity coordinates of the Planckian black-body radiator atthis temperature are the same. Planckian locus onCIE31 chromaticity coordinates is givenin Fig. 21.14. The correlated color temperature (CCT) is used if the color of a white lightsource is not exactly on the Planckian locus. The CCT of a white light source is defined asthe temperature of the Planckian black-body radiator which color is the closest.7


CIE 1976 (L*, u*, v*) color space is more perceptually uniform: distance betweenany two points in this color space represents same color difference perception of thehuman vision. The CIE 1976 L*a*b* color space is a color-opponent space: L forlightness, a and b for the color-opponent dimensions. Chromaticity spaces are obtainedby nonlinear transforms of CIE XYZ coordinates. CIE31 is preferred space because ofsimplicity and wide acceptance. The CIE 1931 xy chromaticity diagram has severallimitations. Area of green shades is exaggerated, compared to human vision. If intwo dimensions, then there is no information whether color is dark or light. Luminanceinformation is not included in the (x, y) chromaticity diagram, it is on the orthogonalaxis. The lighter the color, the more restricted is the chromaticity range. The triangleformed by the three (x, y) coordinates of display primary colors encloses the gamut ofcolors the source or display can reproduce.41 Fig. 21.15 compares the color gamutsused.

Color spaces can fall into three categories according to the area covered by colorgamut: small, medium, and wide gamut. Refer Table 21.1 for chromaticities of pri-maries and the reference white54e56 of most common chromaticity standards.

Small-gamut spaces are comparable to CRT monitors. sRGB (gamma ¼ 2.2, whitepoint ¼ 6500K) is the default color space for Windows and the World Wide Web.57

sRGB is weak in green and cyan (not a big problem if analyzed in (u, v) or L*a*b*perceptually uniform space). It covers just 35.9% of the CIE31 color space. PAL/SECAM and SMPTE-C are appropriate for video output.

Medium-gamut spaces cover slightly larger than CRT area.54 NTSC 1953 is obso-lete. Adobe RGB 1998 is the best known and most widely recommended(Gamma ¼ 2.2; white point ¼ 6500K). A standard was born as a typing mistake: inan attempt to include SMPTE 240M. Adobe attempt to include SMPTE 240M

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Planckian locus

D65, 6500K

A, 2856K

104 K

C, 6770K

B, 4870K E

y, a

.u.

x, a.u

380nm

470nm, blue

505nm, cyan

520nm, green

555nm, "pure" green

580nm, yellow

605nm, orange

625nm, red

Virtual purple

490nm

Figure 21.14 CIE1931 XYZ chromaticity diagram with color regions indicated.


primaries green coordinates turned very saturated. Fortunately, Adobe RGB 1998 satu-rated green primary was practical and useful; it covers 52.1% of the CIE31.

Wide-gamut spaces (Wide gamut RGB, Universal RGB, CIE RGB, Chrome 2000D65, and Kodak ProPhoto RGB) outperform CRTs or inkjet printers, sometimes even

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

R2020White

D50

D65

PAL,SECAM,NTSC,HDTV Adobe RGB 1998 Adobe Wide RGB ITU R2020

y, a

.u.

x, a.u

Figure 21.15 Evolution of color spaces.

Table 21.1 Coordinates of primaries of chromaticity standards inCIE31 space

Color space

Red Green Blue White

x y x y x y x y

1953 NTSC 0.67 0.33 0.21 0.71 0.14 0.08 0.3101 0.3162

Modern NTSC 0.63 0.34 0.31 0.595 0.155 0.07 0.3127 0.329

PAL, SECAM 0.64 0.33 0.29 0.6 0.15 0.06 0.3127 0.329

HDTV(ITU-709)

0.64 0.33 0.3 0.6 0.15 0.06 0.3127 0.329

Adobe RGB1998

0.64 0.33 0.21 0.71 0.15 0.06 0.3127 0.329

Adobe widegamut

0.735 0.265 0.115 0.826 0.157 0.018 0.3457 0.3585

ITU Rec.BT2020

0.708 0.292 0.170 0.797 0.131 0.046 0.3127 0.3290


including the colors that eye cannot see; therefore, color shifts or clipping can occurwhen put on the monitor or printer.

ITU provided Recommendation BT.202050 in 2012, final version is 2015. It covers75.8% of the CIE31 color space and is a compromise between Adobe 1998 and widegamut. Colors are real, RGB primaries corresponding to monochromatic light source(630, 532, and 467 nm for RGB correspondingly).

An RGBW pixel configuration was proposed in Ref. 58: RGB is complemented bywhite LED. In such way hues that have not pure color can be modeled as the additionof a certain amount of white and some intensities of two of the three colors. Such pro-posal has an advantage for LCD display, because R, G, and B subpixels are producedby filtering the white, so color filters remove significant portion of the light. RGBWproduces a brighter white because the white subpixel is not filtered so display powerconsumption can be reduced. OLED display usually also use color filters.59 Idea wasquickly picked up by industry though it receives a lot of critics on image quality.60,61

It must be noted that tristimulus set of colors cannot generate all possible colors, so im-age representation on RGB display is never completely accurate.43 Furthermore, imageproduced using one set of primaries (gamut) should be accordingly transformed beforerepresentation on display containing different gamut triangle coordinates.8 According toRef. 41 it is amisconception that thewider the color gamut, the better. Such display cannotshowcolors that are not in the original content: colors are distorted. Visually, smaller colorgamut is better than too large.MostLCDshavea smaller andmostOLEDshave larger thanthe standard color gamut. Nowadays, the reproducible color range of displays has beenexpanded in the aspect of natural color reproduction.54e56,62,63 Multi-primary displaytechniques or displays with highly saturated pure primary colors (LED, Laser) are typicalexamples of a wide color gamut systems (refer Fig. 21.16 for Nichia oval LEDs64e66 co-ordinates comparison with standard gamuts).

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9 PAL,SECAM,NTSC,HDTV Adobe RGB 1998 ITU R2020 Nichia NSPx346

y, a

.u.

x, a.u

Figure 21.16 Some representative color spaces versus Nichia LEDs gamut.


Note that green is not located on the horseshoe curve, while red and blue are almoston this curve. The reason is that green LED is not purely monochromatic and has acertain bandwidth. Because coordinate is defined by weighted sum of all wavelengths,such color is no longer pure and is located away from the horseshoe curve. Yet modernLEDs and especially lasers are approaching the highest boundaries.63 It should benoted that such wide color gamut displays have a different color gamut compared tothe standard. In particular, the display that uses the LED or laser light source has amuch wider color gamut than Rec.709 (sRGB) standard67 as shown in Fig. 21.16.Generally, a mismatch of color gamuts between signal definition and display systemcauses color distortion on reproduced images.8,68e70 For proper color reproductionregarding definitions and display systems, it generally needs color transformationbased on a colorimetric display model.56 Paper55 introduces a simple optical processto build individual standard color gamuts. It composes new primary colors by an op-tical additive mixture of original primary colors in a time-sequential display system.Paper56 indicates that an increased need for color management of LCDs that wouldenable accurate control of color in the displayed images is also observed. The authorsclaim that AMLCD color displays can be color-calibrated with good accuracy by usinga simple offset, matrix, and tone-response correction for the conversion of tristimulusvalues into display RGB. If the LUT is to be used for color correction, accuracy of thecharacterization can be increased without requiring additional measurements throughusage of an alternate model that does not assume constancy of chromaticity for eachchannel. Calibration accuracy is lower than one achieved by CRTs, but it is acceptablefor most applications. Despite all the shortcomings, indicated tristimulus color codinghas proved an acceptable color reconstruction technique.

Though single wavelength is expected LED emission is not exactly monochromaticand is often approximated by Gaussian spectral shape.71 The LED is often specified bya single wavelength, but different descriptions are in general use: (1) peak wavelength,lp, the wavelength of the spectral peak; (2) center wavelength, l0.5m, average betweenthe two 50% of spectral density points; (3) centroid wavelength, lc, or weighted meanor (4) the dominant wavelength, ld. Though widely used, peak wavelength has littlesignificance for display design purposes as two LEDs may have the same peak wave-length but different color perception.33,69 The dominant wavelength is defined as themonochromatic color of particular wavelength located on the perimeter of the chroma-ticity diagram (horseshoe curve) that appears to be closest in color (Fig. 21.17).

It is determined by drawing a straight line from a white reference point through themeasured (x, y) chromaticity coordinate of the LED to the perimeter of the chroma-ticity diagram. The intersection point is the dominant wavelength. CIE has definedthe color coordinates of a few different white illuminants, but several manufacturers(Lumileds, Cree) apply CIE Illuminant E.72,73

Color purity or color saturation of the LED. The color purity is obtained as

Color Purity ¼ a

aþ b¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðx� xref Þ2 þ ðy� yref Þ2

q

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðxd � xref Þ2 þ ðyd � yref Þ2

q ; (21.9)


where a and b are shown in Fig. 21.17, and (x, y), (xref, yref), and (xd, yd) are thechromaticity coordinates of the light source under test, of the reference white point E,and of the dominant color correspondingly. Color purity of 100% means that the lightsource is a pure color (monochromatic). Color purity of 0% means that the light sourceis white with the same color as the reference illuminant used. Full Width Half Max, thespectral bandwidth at half of the peak power, can be used for monochromaticityevaluation. It is calculated from the difference of the two wavelengths l00.5 and l

〞0.5 on

left and right sides of the peak. Chromaticity coordinates are the best estimate of colorgamut and LED suitability for a particular display since color differences will benoticeable between two neighboring modules. In order to control the spectral emis-sions of every display pixel, LED manufacturers are offering color binning.74

The difference between two colors can be expressed by distance in color spaceDE.75e77 In CIE76, Delta E (or CIE dE 76) is obtained by using coordinates L�1, a

�1,

b�1 and L�2, a�2, b

�2 of the two colors in question:

DE ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�L�2 � L�1

�2 þ �a�2 � a�1

�2 þ �b�2 � b�1

�2q

: (21.10)

CIE 1994 (CIE dE 94) modified this calculation by introducing the L*C*h colorspace. Last modification can be named CIEDE2000 (or CIE dE 00) equation for colordifference calculation.

Just using proper wavelength LED is not sufficient to have color reproduction:white balance should be maintained. Approximate balance for white can be derivedby analyzing Fig. 21.9: rule of 3:6:1 RGB components luminance should be

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

b

a

Tested LED

E

DW=529nm

y

x

Figure 21.17 Determination of the dominant wavelength and color purity.


used.78 Actual percentage depends on chromaticity coordinates of LEDs and white.Furthermore, it is complicated to keep such match just by LED intensity binning.Therefore usually final match is achieved by adjusting (usually intensities are low-ered) RGB LEDs’ intensities. Because of such balancing, display luminance de-creases; therefore, more accurate way of display luminance specification is atwhite balance.

One least representative but most frequently stated LED display parameter is num-ber of colors. Number of colors does not mean a larger color gamut: it’s just the totalnumber of possible combinations of the RGB primaries’ intensities.41 Assuming eachprimary can take 256 (color coding standard) possible intensity levels results in16.7 million combinations, which are not real colors taking CIE31 nonlinearitiesinto account. Furthermore, larger than 256 pixel intensity levels are needed to codethe lightness perception nonuniformity as it was indicated in Chapter 21.3.1. Then,either smaller amount of levels has to be used which should end up with smaller num-ber of combinations; or larger amount of intensity levels is needed to keep 256 light-ness perception nonuniformity-corrected levels. Then some manufacturers calculatethe number of possible combinations based on these larger numbers resulting innonsense trillions of colors.

According to Ref. 41 smaller number of intensity levels is larger problem, whichintroduces visually noticeable discrete steps into images with smooth intensityvariationdfalse contouring. The effect can be masked by spatial or temporaldithering, but either degrades the resolution or introduces content-dependentflicker.

21.3.3 Spatial distribution terms

Display size is specified by the width and height of the screen. Pixel pitch describes thephysical distance between pixels on screen. Usually it’s a physical distance betweencenters of LED clusters assigned as single RGB pixel. Some proposals exist wheretemporal scanning is used to form virtual pixels79,80: if RGB LEDs are placed at equaldistances, each image frame can use different LEDs to form a pixel (Fig. 21.18).Required size of display and pixel pitch can be derived by analyzing human cognitiveabilities.

R B R B

G R G R

R B R B

G R G R

Pixel

Pixel pitch

Pixe

l pitc

h

Pixel

R B R B

G R G R

R B R B

G R G R

1-st frame

Virtual pixel pitch

Virt

ual p

ixel

pitc

h

2-nd frame

3-rd frame

4-th frame

Figure 21.18 Real (left) and virtual (right) pixels explanation.


Visual acuity is a measure to see the smallest features. Minimum separable acuity isimportant in pixel pitch determination. It is expressed in cycles per degree (CPD), theangle at which eye can differentiate an object in arc minutes31:

VA ¼ arctg

�H

D

�z

3438$HD

; (21.11)

where H is the height of the smallest detectable object and D is the observationdistance.

Contrast sensitivity testing complements and extends the acuity tests. It is assessedby presenting the observer with a target of sine-wave grating of given spatial frequency(the number of luminance cycles per degree of a visual angle) modulated by Gaussianenvelope. The contrast of the target grating (carrier frequency amplitude) is then varied0.5 to 32 cycles per degree of a visual angle until the detection threshold of theobserver is determined. The obtained thresholds are converted to the contrast sensi-tivity score (1/contrast) and are plotted versus target spatial frequency yielding thecontrast sensitivity function (CSF). Limited number of eye receptors is the causewhy detection of a high frequency pattern is more difficult. Most commonly acceptedMannos and Sakrison model (Fig. 21.19) of the CSF82 is:

CSF ¼ 0:04992ð1þ 5:9375$f Þ$e�0:114f 1:1 : (21.12)

It can be seen that the model underestimates the spatial response for higher fre-quencies.83 The CSF approximation is used for image quality assessment in imagingand compression.84e86

A viewing distance is determined primarily by the minimum size (i.e., visual angle)for the objects that a user must see. The minimum viewing distance is defined bydisplay resolution or a pixel pitch. The viewing angle at the eye is measured from a

Spatial frequency (cycles/degree)

Sen

sitiv

ity

Figure 21.19 Mannos and Sakrison model of the contrast sensitivity function overlaid onCampbell-Robson test image for human visual system response evaluation.81


line through the visual axis to the point being viewed, and determines where an objectwill register on the retina. The best image resolution occurs at the fovea, directly on theline of gaze, and visual acuity degrades with an increasing angle away from this axis.87

According to the CSF for a human with excidealellent acuity, the maximum theoreticalresolution would be 50 cycles per degree (CPD).88 For the sake of simplicity, it isassumed that 60 CPD is the limit for image pixels to be rendered in order to createa comfortable image (1 arc minute per pixel52). Then using (Eq. 21.12) the minimumviewing distance for display with pixel pith P can be calculated as

D ¼ P$3438: (21.13)

However, the eye can only resolve contrast of 5%. Taking this fact into account theeye can resolve the maximum resolution of 37 CPD. Therefore, the equation can besimplified to

D � P� 1000; (21.14)

that is, the minimum viewing distance in meters corresponds to pixel pith P inmillimeters.

The maximum viewing distance defines the required display size.89,90 Many sour-ces suggest that maximum recommended viewing distance should be three to sixscreen widths for video. At this distance most people will begin having trouble pickingout details and reading the screen. For instance, for a 10-m wide screen, the last viewershould be located 60 m away. Maximum recommended viewing distance according toSMPTE standard EG-18-199489 corresponds to 30 degrees viewing angle minimum.Such distance results in a more immersive experience and also reduces the eye strainand is recommended for home theater. For the same example (10-m wide screen): thelast viewer should sit closer than 17 m. For the advertising displays placed on a street,the maximum distance can be increased beyond the aforementioned standards, butcognitive ability will be reduced.

Pixels blend into a complete image when LED display is viewed at the distancewhere closest pixels are less than one inch apart. For instance, the 10-mm pitch displayshould be viewed at a distance of at least 34 m in order to see a smooth image. Size andamount of the LED define a fill factor17: the ratio of area occupied by the pixel LEDsand the total display area assigned for a pixel (Fig. 21.20).

Figure 21.20 Comparison of a high (left) and low (right) fill factor.


In order to get the high contrast, the area around LEDs is filled with a nonreflecting,light absorbing material. Unfortunately, this black area significantly degrades the qual-ity of the LED screen image when observed at a normal viewing distance, causing theLED flaring, color blending problems, and other difficulties. According to Ref. 17, fillfactor should not be lower than 0.5, whereas the majority of large displays exhibitmuch lower values. Publication17 suggests increasing the fill factor by using a signif-icant reflection chamber inside the LED (Fig. 21.21).

The display viewing angle is the angle, in degrees, between a line normal to thedisplay surface and the user’s visual axis where threshold is established.28 Usuallyit is defined as the maximum angle at which a display luminance falls below 50%of the frontal value. This angle can vary depending on the LED and the technical fea-tures of the display: if LEDs are with lenses, those usually compress the directivity invertical direction and expand in horizontal direction; lenses and louvers can produceshadowing effect. Contrast ratio can be used for viewing angle evaluation, which isusually the case in TV displays, but rarely used in LED displays evaluation. Standard28

defines 1:10 contrast drop at maximal viewing angle. Also the color shift, measured asthe D(u0v0) coordinates (CIE1976) can be used for directional color performanceassessment of display:

Dðu0; v0Þq ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�u00 � u0q

�2 þ �v00 � v0q

�2q

; (21.15)

Chromaticity uniformity28 is evaluated displaying monochrome test pattern, usingu0, v0 derived from colorimetry of five or nine distinct points on display or individualLEDs. Difference between any two points is:

Dch ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�u01 � u02

�2 þ �v01 � v02

�2q

; (21.16)

Luminance uniformity is carried out using photometer. Nonuniformity is expressedas

Dch ¼�1� Lmin

Lmax

�100%; (21.17)

While display chromaticity or luminance nonuniformity may not be noticeable athigh intensities, it can appear at low intensities. Pixel intensity can be controlled by

Reflecting cup

Encapsulating epoxy

Terminal pins

LED chipReflecting cup

Figure 21.21 Conventional LED construction does not allow for large reflecting cup (left) whiledifferent construction allows for a large reflector (right).


same technological processes in conventional displays. Color uniformity in LED dis-plays can only be controlled by binning the LEDs. Luminance uniformity can beaffected by directionality of individual LEDs, intensity bins, driving current accuracy,or even current distribution on PCB.91 Deviation of driving current can influence theintensity of whole tile (PCB) or just a LED cluster that is driven by same driver IC(Fig. 21.22).

Final uniformity tuning for individual LEDs can be accomplished using so-calleddot correction (Fig. 21.23).

Lens application on LED allows for more efficient use of the light produced by thecrystal: usually there is no need for wide vertical viewing angle so intensity can beredistributed horizontally. Unfortunately, lenses of the LEDs can produce the shadow-ing effect, reducing the viewing angles; cross talk can occur in lenses, degrading thedisplay contrast. Surface mounted LEDs are free from this type of defects but Lamber-tian directivity wastes a portion of light distributed vertically.

Figure 21.22 Luminance nonuniformity at low intensities: either whole tile (large rectangles)or driver (smaller squares) influence can be seen.

Figure 21.23 Individual pixels’ luminance nonuniformity at low intensities: before (left) andafter (right) dot correction.


21.3.4 Temporal performance terms

Flicker is a term for detectable changes in display luminance, and it occurs when fre-quency of these changes is below the integrating capability of the eye. Flicker occurswhen changes in display luminance occur at frequency below the integrating capabilityof the eye. The minimum frequency this occurs at is the critical flicker fusion fre-quency (CFF)31

CFF ¼ a$logðLaÞ þ b; (21.18)

where a ¼ 12.5 (for high ambient light level) or 1.5 (for low), La is the averageluminance of an image in cd/m2 or nits, and b ¼ 37. Taking the LED screen luminance10,000 cd/m2 and assuming a high ambient light level, one gets CFF ¼ 87 Hz or12 ms.92

A refresh rate is the frequency at which the display pixels are re-illuminated.Refresh rates below 50e80 Hz may induce perceptible flicker. The update rate isthe frequency at which the information content of the display is changed. In practice,image refresh frequencies can reach 100 Hz or even 240 Hz. Paper93 indicated thatincreasing spatial resolution to improve visual perception alone is not sufficient andit must be combined with higher frame rates. In professional LED screen applications,not a human vision is the factor determining image refresh frequency. The reason isthat usually almost every event is captured on a camera for clip production or TVbroadcasting. Until the scene background is relatively dim, camera sensitivity canbe controlled by adjusting the aperture size. When the scene gets very bright, aperturesize reduction is not sufficient and exposure time is varied then. In case the LED screenrefresh rate is too low compared to shooting equipment speed, the camera will captureonly a part of the PWM cycle dimming the LED. Therefore, the captured intensity willnot correspond to the set value. This can create the moiré pattern, image blinking,banding, patching, or even complete disappearance of the image at some frames. Syn-chronization of video shooting equipment frequency is not sufficient. Therefore, therefresh rate needs to be much higher than the requirement for a human spectator. Usu-ally, 400 Hz refresh frequency is used. Some manufacturers claim success at 240 Hz.94

Frame or update rate is the frequency at which display video content is updated.While mostly defined by video content parameter, it can be the cause of unexpectedeffects. Interlaced black-and-white frames video test is the best demonstration ofLED video display performance. If video frame is not simultaneously undated on alltiles of the display, image banding occurs.

21.4 Technology in detail

A raster principle is used to form the image on display. In most common case raster isformed as a set of equally spaced pixels, though there are designs where pixels areplaced at random positions. Pixels are usually formed from RGB LED cluster andembedded massively on a larger tile (Fig. 21.24 left). Surface mount LEDs are used


in case of smaller pixel pitch. RGB chips are placed into single LED (Fig. 21.24 right)to obtain better color mixing and denser pitch.

Individual pixels can be used, as separately encapsulated enclosure, if larger pixelpitch is required or display has to be transparent and covers large area. Individualpixels allow for any display shape. Mounting has to be designed to ease the massiveinstallation procedure. Ingress protection is required if used outdoors. Multiple LEDsor power LEDs are used to achieve the high pixel brightness.

21.4.1 Display electronics structure

A tile (can be addressed as “module” or “panel”) is a major constructive element incase of the modular display design. The tile is an encapsulated PCB responsible fordriving a relatively small number of LEDs. One side of PCB holds large number ofLEDs; 256 or 512 are the most frequent case (Fig. 21.25). Opposite side either holdsdriving electronics or has a connector; in such case add-on PCB is used to carry thedriving electronics part.

Figure 21.24 Closer look at LED display surface can reveal the pixel structure which can beformed by DIP (left) or SMD (right) LEDs.

Figure 21.25 LED tile PCB front (left) and rear (right) view before encapsulation.


Such arrangement allows using fewer ICs to control the individual LEDs: usually 8or even 64 LEDs are controlled by one driver chip; just one IC for input and outputdata buffering. Tile PCB is encapsulated into plastic enclosure. The outer face ofthe tile contains plastic louvers for contrast enhancement and mechanical protectionof the LEDs.24 In case of floor-mount, for example, baseball perimeter displays, lou-vers are made of rubber to stand the damage because of possible mechanical impactand avoid injuries. Humidity protection of outer face is accomplished by potting thetile with a special compound.95 A compound should have sufficient electrical resistiv-ity; should have low uncured viscosity in order to penetrate the inner space of theenclosure; have good adhesion to both plastic and PCB components; have sufficientthermal conductivity to take out the heat from the encapsulated electronics; have adark and matt surface to keep a good contrast ratio of a display image; be self-curable to self-seal occurring mechanical damages; be soft enough to allow easyLED dismantling in case of failure. Compound should not degrade display contrast;therefore, surface is usually frosted. Room temperature vulcanization (RTV) siliconcompounds are the good candidates. Tile is usually mounted on the front surface ofa larger module (Fig. 21.26). Sealing gaskets protect the inner module space againstdust and humidity penetration.

Louvers are typically used to improve the contrast in case of intense ambient light.Louvers act as a hood, cancelling some of ambient light coming from the top(Fig. 21.27). The larger the louver’s length, the better is the display shading, as wellas the contrast ratio and image visibility.

LED Plastic case with louvers Compound Outer wall ofthe module

Driver IC'sVentilation openingFixing bolt

Figure 21.26 Tile structure and placement on the module wall.

LEDLouver Louver

SunLED

Shaded area

Viewer

Area with blocked view

Figure 21.27 Shading provided by louvers increase the display contrast (left), but louvers arealso blocking the view (right).


Though useful for contrast enhancement, louvers also reduce the vertical displayviewing angle. Refer Fig. 21.28 for display directivity comparison with and without lou-vers reported inRef. 25. Pixel pitch is 10 mm, louvers thickness is 2 mm, length is 10 mm.Oval 5.1 � 3.8 mmLEDwas used; vertical gap between LED and top louver is 0.86 mm.

If size and position of the louvers regarding the LED pixels is disregarded, signif-icant display directivity performance degradation can occur. Publication24 proposed acontrast sensitivity function for outdoor visibility optimization.

The inner side of the tile PCB (oriented toward a macroblock) is covered by a thinconformal coating layer: it is exposed to the air flow circulating inside the macroblockand should have good heat conduction. A special opening is designed on the inner sideof the encapsulating box (Fig. 21.26) in order to allow air circulation to the driver ICssurface. Despite increasing efficiency of the LED drivers,6,9 heat dissipation on driverICs is significant: for the 16 � 16 pixel board, an average of 4 W power is dissipatedon driver ICs at a full white level. The heat source is also in the LED itself: while in-candescent lamps emit the majority of losses as infrared radiation, LED losses are leftinside the chip, raising the junction temperature. Introduction of louvers and pottingprevent the front face ventilation. Therefore, heat extraction techniques should beused. In case of outdoor displays, this task has to be accomplished by not hamperingthe ingress protection (usually IP65) requirements. The power and control signals forthe tile are delivered through the dedicated connectors. Usually, the LED is consideredto serve 100,000 h.96e98 For normally operating, manufactured and assembled LED,this can hold true. However, the LED video display is operated under harsh conditions:direct sun exposure results in UV degradation and elevated operation temperatures;direct rain; accidental mechanical impacts during the display installation.98e100 There-fore, some LEDs can fail.101,102 LED encapsulant failure can occur due to the stress onLED pins in a tile manufacturing phase.99 Hence, the LED shall be easily replaceable.Usually whole tile is replaced and individual LED replacement is accomplished later,under laboratory conditions in order to ensure the same sealing quality and directivity.Therefore easy replacement of the tile should be designed: fixing bolts (Fig. 21.28) areeasy to detach; size and shape of the opening in a module should allow removing thetile inwards; tile should have dedicated handle.

–60 –40 –20 0 20 40 600.0

0.2

0.4

0.6

0.8

1.0

Screen directivity after louvers application

Screen directivity iforiginal LED used

Iv /

Ivm

ax, a

.u.

Angle, deg100 200 300 400 500

50

100

150

200

250

300

350

400

450

500

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

Figure 21.28 Louvers influence on the vertical directivity of the display (left) and displaybrightness modulation for viewer at 10 m away (right).25


Tiles are assembled in a module (Fig. 21.29). Modules usually are termed “macro-blocks” to indicate that those are construction parts. If tile usually contains 64 (8 � 8monochrome image resolution) to 4096 LEDs (32 � 16 pixel in color), a macroblockcan contain 3072 (64 � 48 resolution) to 49,152 pixels (256 � 192 resolution).103e106

Modular LED display construction12,18,103,104 gives the flexibility to final display size:modules can be arranged in any fashion. Modules are easy to transport and can behandled manually. Macroblocks assembly and interlocking mechanisms have to bedesigned for ease of assembly and also provide the invisible seams from the front. Spe-cial locks are used to hold both vertical and horizontal seals in position (Fig. 21.29). Arange of innovations is required to match the carrying construction and the spacerequired for electronics.

The module can be easily identified looking at the LED display from the rear (referto Fig. 21.30).

Module contains AC/DC power converters. Each module usually requires500e1000 W peak power. Power is needed for communication, control, and LEDdrive electronics of the module. Fans are usually applied to extract the heat generatedinside the module, while other manufacturers rely on heat transport to rear aluminumsurface and convection cooling.

21.4.2 Data production and distribution

The incoming video stream can be produced in many ways. Simple LED display(Fig. 21.31) may only contain only a local controller which is in charge of delivering

Communication electronics

LED tiles

Rear access door

Stiffening plate

Power supply Load carying andinterlocking mechanisms

Next module One module (macroblock)

Figure 21.29 Top view on module construction.

Figure 21.30 Rear view of the assembled LED display.


the control signals to several LED tiles (Fig. 21.25). Communication part can be absentif remote control or configuration are not needed.

Message display, signs,107,108 or indicators use such structure. Simple microcon-troller can do the control over signals required for display operation. Pixel decoding,refresh cycles can be implemented in software, though some automation can be off-loaded to CPLD or FPGA.5,10,15,109e111

Most common is to use a modular structure. Tiles are assembled in a module(Fig. 21.29 or Fig. 21.30). Modules usually are termed “macroblocks” to indicatethat those are construction parts. If tile usually contains 64 (8 � 8 monochrome imageresolution) to 4096 LEDs (32 � 16 pixel in color), a macroblock can contain 3072(64 � 48 resolution) to 49,152 pixels (256 � 192 resolution).103,104 Modular LEDdisplay construction12,18,103,104 gives the flexibility to final display size: modulescan be arranged in any fashion. Modules are easy to transport and can be handledmanually. Modular construction is used to distribute the pixel control tasks betweenlighter, smaller size modules (macromodules, macroblocks).10,18 Module(Fig. 21.32) incorporates the (1) communication interface; (2) module controller; (3)arbitrary number of LED tiles; and (4) other supporting equipment like power supply,ventilation fans, indicators, and mechanical infrastructure. Macromodule (Fig. 21.32)also contains AC/DC power converters. Each module usually requires 500e1000 Wpeak power. Power is needed for communication, control, and LED drive electronicsof the module. Fans are usually applied to extract the heat generated inside the module,while other manufacturers rely on heat transport to rear aluminum surface and convec-tion cooling.

Remote control(optional) Controller

LED tile /pixel

LED tile /pixel

LED tile /pixel

Figure 21.31 Simple LED board structure.

Incomingdata Communication

interface

Data relay tonext module

Mod

ule

cont

rolle

r

LEDtile

LEDtile

LEDtile

LEDtile

LEDtile

LEDtile

LEDtile

LEDtile

LEDtile

LEDtile

LEDtile

LEDtile

Figure 21.32 LED module (macroblock/macromodule) structure.


Communication interface is responsible for incoming data and control commandsreception and data routing to next module (optional). Usually it is an addressableunit with programmable address. In a simplest case communication interface just stripsout certain portion of the incoming data stream conveying the rest of stream to nextmodule.

Controller can be responsible for tile status (LED failure, temperature, current con-sumption) monitoring, temperature management (cooling fans, heat pipes, or powerreduction), power sequencing and module address or status presentation on theexternal indicator. Main task of module controller is to interpret the data receivedfrom communication interface and convert that data to serial data stream for LED tiles.

LED tile incorporates several LED drivers (Fig. 21.33). Each driver can control 8 to64 LEDs in individual drive configuration and up to 256 LEDs in case of multiplexeddrive. LEDs placement depends on PCB routing: LEDs can be grouped into clusterclose to driver or evenly distributed over tile area. Pixel decoding is required fromdisplay controller in order to render the image according to the pixels configuration.

Data is delivered to the tile or pixel modules using serial protocol. It is similar to SPIwith only exception that number of data lines can be larger and signals for multiplexedcontrol can be added.1,10,112e114 Simple LED display (Fig. 21.31) may only containonly a local controller which is in charge of delivering the control signals to severalLED tiles. Communication part can be absent if remote control or configuration arenot needed.

Data delivery to the LED drivers has settled to some standard circuit which is basedon 74HC595 functionality (Fig. 21.34).

Base of the data delivery is the shift register: note the lines D, SDO, CLK, LTD,OE, and the double-layer structure. Lower latch is needed to hide the data shifting pro-cess so it is not visible on the screen. The upper shift register is responsible for serialdata communication. It uses a serial data input (D), data output (SDO) for cascading,and a clock (CLK) for shifting. The lower latch is used to hold the previous data. Thislayer is controlled by a separate control line (LTD). The output stage is using the datafrom the latch to control the LED. All driver outputs can be turned off simultaneously

Drv

Drv

Drv

Drv

Drv

Drv

...

...

...

Power supply

Inpu

t buf

fers

Out

put b

uffe

rs

Figure 21.33 LED drivers connection on the tile.


by using an output-enabled signal/OE. The driver shall be loaded with the current LEDstate (on/off) of the time tick scale.114 The driver loading speed is important to ensurethe quality of the image on screen.

LED control configurations within a tile can be addressed to as individual and mul-tiplexed drive. The multiplexed drive (Fig. 21.35, left) can reduce the number ofdriving ICs.

LED lines are driven (lit-up) in sequence.10,113 The resulting brightness is dividedbetween the scanned lines. A 20-mA LED is usually allowed for a 100e200 mA peakcurrent in case of pulsed drive. Then, multiplexing of more than 4e8 lines in sequencewill reduce brightness. Lines are driven in sequence, so increase of the number of thelines results in refresh frequency reduction. The individual drive10 uses individualdrivers for every LED (refer to Fig. 21.35, right). This topology is prevailing in allmodern designs thanks to the LED driver IC price reduction.

Data into macroblock is supplied by video processor. The processor decodes theincoming video, which can be analog or digital, and converts them into resolutionand format agreed with macromodules, according to pixel arrangement within themacromodule and arrangement of the macromodules in display (Fig. 21.36).

Shift registerSDI

SDOCLK

Latch

Output stage

LTD

OE

Serial data shift clockSerial data in

Latch signal

Output enable strobe

Serial data out

To the LEDs

Figure 21.34 LED driver data delivery is based on shift register in pair with the latch.

VDCVDC

Figure 21.35 Multiplexed drive (left) configuration uses sequential drive of LED groups;individual drive (right) configuration uses individual drive of every LED.


Raster size of incoming video can be different from display resolution. Low reso-lution displays are used at the parts of the stage that attract less attention from theviewers or at the large viewing distances. Control and manufacturing of every pixelcosts some fraction, so most economical solution usually is sought. Some types of dis-plays, like curtains, floor, wall, façade displays are designed in such a way that imagebehind them is obstructed the least, so their resolution usually is low. Display resolu-tion increases with reduced viewing distance and the attention it receives from specta-tors. In case of stage display, video capturing equipment should be considered too: itusually captures the artists on the stage, but display is in the background; if displayresolution is low, interference pattern is produced. Once display resolution has beendecided, display video processor is informed of the resolution and display arrange-ment. Video processor has to capture incoming video (e.g., DVI) frame and convertit into resolution required. In a simplest case resolution conversion may not berequired: incoming video resolution can be adjusted. Such solution is most commonbecause there is no need for conversion and there are no resampling distortions ifnative resolution is maintained. Usually DVI signal is taken directly from computerdisplay port.115

Required video can be supplied from local computer which can be single-boardcomputer or standard PC if display is used as standalone video source. Computercan have a network connection, and then such display can be a part of street advertisingnetwork. Sometimes data for LED modules is produced directly, overriding the com-puter’s video card, from incoming stream or file stored in computer memory.10,109 Yet,direct DVI signal stripping is the most popular.

Modular construction assumes far more complex display structure when it comes toprofessional LED display use. General complex video display structure is presented inFig. 21.37.

Videoprocessor

HDMI,VGA,AV,TV,Ethernet,AVB

LED display

Macromodules

Figure 21.36 Most common LED display structure.

Videoprocessor

HDMI,VGA,AV,TV,Ethernet,AVB

Video splicer(optional)

Monitor

Remote control,ethernet

LED display1

LED display

LED display 3

2

Figure 21.37 General structure of complex LED video display.


Any type of video information can be supplied to video processor, if it is supportedby a proper connection. Display resolution usually differs from the incoming videostream.16,17 Video processor is responsible for rendering the video information intodata stream dedicated to LED display. In order to accomplish this task, video processorshould know how display has been arranged and what type of modules have been used.This information is either entered manually or in a semi-automated or complete way,using video camera, that presents the display image to the video processor. Operationcontrol can be accomplished locally, using the monitor and keyboard or remotely, us-ing the Ethernet connection. In general, several LED displays can be serviced by singlevideo processor. Video splicer may be required if video processor cannot support suf-ficient number of data lines.

It is worth noting that all circulating data is digital: LED intensity control, refresh,correction parameters, diagnostics, data routing, addressing, and macroblock health re-ports. This creates significant data streams that have to be delivered down to singleLEDs in display. Display stream size is mainly defined by resolution, size, video framerate, refresh rate of the frame, and gamma correction requirements.

Stream is delivered serially in order to reduce the cable size and avoid the delayskew problems. Still, at high display resolution and size, simple FTP cable becomesinsufficient. Problem partially is solved by providing several serial channels. Opticalfiber communication can be used for longer distances (beyond 100 m) and largerstreams. Another solution is to use the local gamma correction (at macroblock) or com-pressed transmission with decompression at macroblock.

A rounding error is introduced114 if incoming 8-bit (24-bit pixel color coding) gray-scale Cin is converted to gamma-corrected grayscale Cgwith resolution of N bits. ReferFig. 21.38 for gamma-corrected grayscale resolution (bits N) impact on conversionresult at low intensities if g ¼ 2.5 is used in Eq. (21.3).

0 5 10 15 200

1

2

3

4

5

6

7

8

9

10

10bit

16bit

14bit 12bit

8bit

19bit

Gam

ma-

corr

ecte

d gr

aysc

ale,

Cγ

Original grayscale, Cin

Figure 21.38 Grayscale resolution impact on gamma correction.


It can be seen that gamma correction will exhibit some rounding error. It is notablethat less than 5% variation of the LED peak intensity is hardly achieved by sorting, noris it practical. In addition, the LED driver accuracy has some limit, driving currentmismatch to 3%.116 Therefore, it makes no sense in reaching better than 5% error ingamma correction coding. Taking 5% as lower bound will result in 25 lower levelsin case of 8-bit grayscale coding, 14 levels in case of 10-bit coding, 7 levels for 12-bit, 5 levels for 14-bit, 3 levels for 16-bit, and 1 level for 19-bit coding. The errorwill be large just at low intensities (refer Fig. 21.39 for 10-bit output coding errors).

Fig. 21.38 also indicates that only 19-bit resolution output coding is capable ofmonotonic variance when change of the input code leads to the change in an output.Problem is usually solved by using very small intensity instead of zero at the lowestend. At larger intensities, many of the codes will be of no use.117 Even applicationof slightly different correction10,51,118,119 cannot solve the problem completely. Thefloating point coding can be a solution. It is also worth pointing out that Rec.2020does not use 0 as black level. In 10-bit system black is 64 and in 12-bit it is 256.Same applies to nominal peak: 940 in case of 10-bit and 3760 for 12-bit. Evenmore intensity levels resolution is required if dot correction of individual display pixelshas to be done. Furthermore, tile correction has to be envisaged, because tile LEDs areaging, and in case of replacement their intensity might have to be tuned down. Addi-tional dimming is required according to the ambient, surrounding light conditions:high brightness is needed during a sunny day, but image intensity has to be signifi-cantly reduced during the night.120

21.4.3 Driving the LED: pixel control

LED driving requirements significantly differ from other applications: (1) intensity hasto be regulated in very wide range to account the nonlinearity of the human vision;

13 20 30 40 50-30

-20

-10

0

10

20

30

10% limit,rorregnidnuor

noitcerroc-am

maG

apro

x (%)

Original grayscale, Cin

5% limit

δ

Figure 21.39 Approximation error for 10-bit output coding in case of gamma correction.


(2) radiation stability has to be maintained over wide range of operation conditions;(3) driving circuits must provide a luminance adjustment for white balance, dot correc-tion, tile correction, nighttime dimming purposes120; (4) spectral stability has to bemaintained; (5) control has to be accomplished in a cost-efficient way and (6) energyefficiency is demanded.

The output stage structure divides the drivers into four categories: (1) switch outputwith external current limiting resistors; (2) constant current output; (3) constant currentplus PWM output (programmable) and (4) complex drivers (using DC/DC converterplus current feedback and PWM).

LED light is produced by the luminescence in a solid-state p-n junction diode whenforward bias voltage is applied.7,8 Fig. 21.40 represents a typical voltage-current rela-tionship of the LEDs used in LED video displays.

Forward current is small before forward voltage VF exceeds the internal barriervoltage of LED. Increasing the forward voltage further, curve follows the shape of aknee and rises rapidly at a linear rate. Forward voltage and other parameters are spec-ified at IF 20 mA; it is usually a nominal current for low-power diodes. Exceeding themaximum specification of the manufacturer can seriously reduce useful life of theLED. Forward current can be increased up to 100e200 mA if LED is operated inpulsed mode. Few essential drawbacks can be seen from this graph: (1) LED forwardvoltage is different for red (1.9e2.1 V depending on technology) and green and blue(3.4e3.6 V); (2) power delivery to LED is highly nonlinear at low currents. Higherdrive currents (350 mAe1 A) are needed if power LEDs are used. Further problemscan be seen by analyzing the luminous intensity variation with forward current(Fig. 21.41).

LED light output is proportional to a forward current, but it is not exactly linear. If aforward current is not controlled appropriately, it can result in an unacceptable

0

5

10

15

20

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Blue

Green

Red

VF, V

I F, mA

Figure 21.40 Voltage-current relationship of the JUNDUOLI SJ-E54DxR24VU-S LEDs.


variation of light output.111 A challenge in designing the LED driver is to create a sta-ble, programmable, constant current source possessing high efficiency.110,121

Unfortunately, LED pixel intensity cannot be controlled using forward currentbecause of several reasons: (1) color shift with forward current for green and blue;(2) nonlinearity of human vision; (3) instability of LED emission and driver feedbackproblems at very low currents; and (4) driving every LED by the analog source iscomplicated.

Forward current affects the emission wavelength, influencing the chromaticity co-ordinates of the display gamut.8 This is especially valid for InGaN, which is used inblue and green LED production. Note the color triangle shift in Fig. 21.42 whenLED forward current is changed from 1 to 100 mA. Data was taken fromRefs. 122e124. Data for red uses the temperature effects in �30 to 85�C range.

It can be seen that green has the largest variation, while red is stable with currentand temperature. If current is varied only slightly, gamut can be considered stable.Constant current adjustment can be used to provide a luminance adjustment, white bal-ance, dot correction or even to compensate for temperature effects.

Light output has to be varied over a very wide dynamic range. Nonlinearity of thehuman sense of light43 requires gamma correction to decompress the image color cod-ing. 16-bit LED intensity coding (more than 80 dB) is required to avoid image bandingafter gamma correction. LED output with current is nonlinear, as it can be seen fromFig. 21.41. Furthermore, LED emission is not stable at low currents. It is very compli-cated for the driver to provide the wide range of constant current. Electromagneticinterference from neighboring LEDs driven at high current will distort this very lowcurrent even if two aforementioned problems can be solved. Driving the LED bythe Pulse Width Modulation (PWM) of constant current solves these problems: thelinearity of LED output modulation can be maintained.117,119 Average power of the

0 5 10 15 200.0

0.2

0.4

0.6

0.8

1.0

Linear

Green

Blue

RedIv

, a.u

.

IF, mA

Figure 21.41 Luminous intensity variation with forward current of the same LEDs.


PWM is linearly proportional to the duration of the pulse (refer to Fig. 21.43). Just theamount of the available duration steps limits the dimming resolution. PWM is easy toimplement with digital circuits; a steady drive over a wide range of intensities can beensured. Display image has to be blinking anyway to avoid the flickering when motionpicture is rendered on screen,94 so periodicity of PWM is in synergy with thisrequirement.

The PWM drive has its own shortcomings:

1. The number of available steps is limited by the switching speed of the LED or driver elec-tronics and the required refresh frequency;

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1, 5, 20, 50, 100mA

Adobe RGB 1998 @ IF=1mA @ IF=100mA

y, a

.u.

x, a.u

Figure 21.42 Color gamut shift when LED current is changed from 1 to 100 mA.

1 765432 8 9 10 11 12 13 14 15 1 765432

7%13%

20%27%

67%100%

Figure 21.43 PWM dimming time diagram.


2. Constant current PWM has to be supplied to every individual LED on screen. This both com-plicates the design and increases the cost of the display;

3. The time required to deliver this data to LED driver output limits the minimal PWM pulseduration;

4. There is a moment when all LEDs are switched on (Fig. 21.42: the start of the period) nomatter what pixel intensities are; this will produce a large current spike, rising an EMI issue;

5. Refresh rate has to be synchronized with the frame data, otherwise different frame fragmentswill be displayed on screen when refresh cycle changes; a good test for this phenomenon isthe movie with interlaced black-and-white frames;

6. Color shift occurs even in case of PWM drive125 due to temperature rise.

A range of PWMmethods have been suggested for LED control (Fig. 21.44). Pulsefrequency modulation (PFM), Bit-angle modulation (BAM),126 binary PWM(BPWM), gated PWM (GPWM),114 pulse frequency modulation, and pulse amplitudemodulation,67,111 probability-based PWM127 and scrambled PWM (S-PWM)116 arejust a few candidates. PFM is basically same PWM, only in case of PWM the refreshperiod is fixed and the pulse duration is varied; while PFM is the opposite: pulse dura-tion is fixed but refresh period is varied. PFM is rarely used in video displays becauseimage blinking at stable period is required to maintain the smooth motion on screen.Binary-weighted PWM (BPWM), discrete PWM,119 or BAM126 dimming techniqueuses different weight for the pulses in a period to form the required duty. ReferFig. 21.43 for conventional PWM and BPWM comparison. In BPWM, width ofeach separate pulse is proportional to the weight of the bit in the corresponding binarycode, but the total pulse width is the same as for conventional PWM. Data delivery toLED driver is serial (Fig. 21.34): the amount of the communication lines isreduced18,113,114; every display controller can control large amount of LEDs5;complexity and price of control electronics is reduced. Data is serially shifted out ofthe controller in order to reach the LED. BPWM driving further simplifies the systemand reduces the amount of data upload cycles thanks to bitplanes (Fig. 21.44) decoding(binary code of pixel grayscale can be used directly).

It is essential that BPWM also reduces the data flow: every time tick (1,2,3,4, and soon), indicated in Fig. 21.45 requires an upload of new data for PWM driver; whileGPWM requires an upload of corresponding bitplanes, that is, on time tick 1,2,4,8,

Bit no

Bit 0Bit 1

Bit 2Bit n

Bitplane rowDrv Drv

Figure 21.44 Transferring the image bitplanes to driver is direct with BPWM dimming.


and so on. BPWM (BAM) dimming will also reduce EMI by spreading current surgespikes in time (compare PWM and BPWM waveforms in Fig. 21.45).

Unfortunately, both PWM and BPWM have the same limitation: the number of theachievable grayscale levels, the refresh frequency, and number of serially controlledLEDs are interrelated and limit each other. The problem lies in limited speed of shiftregisters of LED drivers: clock speed is limited by technology and currently is about25e30 MHz. PWM refresh cycles can be subdivided into subframes (ticks). This sub-frame data (“1” or “0”) shall be delivered to every LED driver by shifting serial datathrough driver registers. The subframe loading time sload required to shift-load NREG

channels of the LED driver at shifting clock frequency fCLK can be calculated as:

sload ¼ 1fCLK

� NREG. (21.19)

1 765432 8 9 10 11 12 13 14 15 1 432

1

3

2

Conventional P

WM

Binary P

WM

Gated P

WM

1 2 4 8

1

3=1+2

2

1 2 4 8 3216

1

3=1+2

2

Bit plane no

Bit plane no

Time tick no

PFM

1

3

2

Figure 21.45 Waveforms of the most common PWM dimming techniques. PWM, Pulse WidthModulation.


For instance, the shortest subframe duration will be 1280 ns for the 32-channel LEDdriver, 3840 ns for the 96-channel LED driver, etc. Subframe loading time and arequired number of ticks in PWM period limits the minimum attainable period ofthe PWM signal TPWM. Flicker occurs when changes in the display luminance occurat the frequency below the integrating capability of the eye. For professional LEDscreen application, 1000e400 Hz refresh frequency9 is used in order to match the cam-era exposure time as it was indicated before. Then the number of the available levelsnPWM is defined by required PWM, image refresh period:

nPWM ¼ TPWM

sload. (21.20)

Data in Table 21.2 lists the attainable grayscale resolution in bits versus the numberof LEDs driven serially and PWM or image refresh frequency.

It must be noted that modern LED drivers have the OE signal which can be used foroutput blanking. And the switching speed of the LED driver output is comparable tothe shifting speed: duration of the OE signal can be considerably shorter than the dura-tion required for loading the entire subframe data. GPWM technique proposed inRef. 114, explores this fact to significantly increase the number of gray levels and im-age refresh frequency. Resolution is achieved without compromising the number ofserially controlled LED drivers. In GPWM LED is lit on just for a short durationsmin (note the GPWM pulse duration in Fig. 21.45 and 21.46) during the data loadingoperation instead of the shortest PWM duration defined by the frame data load timesload. This would correspond to replacing sload used Eq. (21.20) by smin. This way,almost unlimited number of PWM levels can be achieved. Further performance gainis achieved when bitplane loading as per BPWM is used.

Table 21.2 Attainable PWM grayscale resolution versus the refreshfrequency of the frame and serially controlled LEDs in case of30 MHz serial data clock frequency

PWM/image refresh frequency, Hz

50 100 240 400 1000

LEDs Grayscale resolution, bits

1 19 18 17 16 15

8 16 15 14 13 12

16 15 14 13 12 11

32 14 13 12 11 10

96 13 12 10 10 8

256 11 10 9 8 7


Actually there is a limit in GPWM speed. It is related to the shortest achievable LEDflash duration, which depends on LED response time and LED driver used. As noted inTable 21.1, the best propagation delay for high and low level is 100 ns. Then 200 nsshould be used as the minimum GPWM duration smin. Studies in Refs. 9117 illustratedthat the LEDs commonly used for video screens have response time below 100 ns. Theamount of available bits NG,thanks to additional gating, is:

NG ¼ log2

�sloadsmin

�: (21.21)

Remaining number of bits NM corresponds to what would be attainable if PWM orBPWM techniques are used:

NM ¼ log2

�TPWM � NG$sload

sload

�. (21.22)

The total available bits are the sum of NG and NM. The attainable grayscale resolu-tion of GPWM versus the number of LEDs driven serially and image refresh frequencyis presented in Table 21.3.

GPWM not only offers significant speed performance improvement as can be seenby comparing Tables 21.2 and 21.3, but also has more flexibility in rounding of PWMperiod, since better resolution of PWM cycle allows for two options. Because of addi-tional gating in GPWM the maximum available LED brightness is reduced. This effectbecomes significant only at high numbers: at 1000 Hz refresh frequency and 256 seri-ally driven LEDs; 5% reduction in brightness is expected. Another GPWM disadvan-tage is related to the fact that more on/off transitions of the LED are implemented withGPWM (as well as with BPWM); therefore LED pulse driving skew will affect itstronger than conventional PWM as indicated in Ref. 117.

It is worth reminding here that despite PWM drive, LED has to be driven by con-stant current during on cycle. The simplest way of setting a current is to use a resistor inseries (Fig. 21.47).

TPWM

min

One level (bit plane) loading cycle

Bit 0 Bit 1 Bit 2 Bit 3 Bit 4

OE

τ

Figure 21.46 Time diagrams of the GPWM dimming technique. GPWM, gated binary-weighted PWM.


Current is set by Rset value and depends on supply voltage VDC and LED forwardvoltage VF:

IF ¼ VDC � VF

Rset. (21.23)

Circuit is simple, low cost, and produces no noise. LED can be turned on and off ifswitch is added in the series. A switch driver is the simplest type: an output stage usu-ally is a saturated switch (typically, an open drain) connecting the output to the ground.One external resistor per LED is added in order to set the LED current. Refer toFig. 21.48 for a simplified topology drawing.

High LED current can be used because switch operates in saturated mode. This to-pology is quite old; application of a saturated switch does not allow for fast switchingtimes. It does not offer current accuracy and efficiency. Forward voltage varies fromLED to LED: �0.3 to þ0.5 V for red; and �0.4 þ 0.6 V for green and blue, resultingin �20% to þ10% current variation in case VDC is 5 V and �38% to þ23% currentvariation in case of 3.3 V power supply for red. Accuracy is low if resistor voltage dropis low. Variation decreases if power supply voltage is larger, but then large portion ofenergy is wasted as a heat on resistor Rset. More than 60% of energy is dissipated onRset for red and 30% for green and blue in case of 5 V power supply. Closer examina-tion of Fig. 21.40 can reveal that forward voltage VF is changing with current. In

Table 21.3 Attainable GPWM grayscale resolution versus the refreshfrequency and serially controlled LEDs in case of 30 MHz serialdata clock frequency

PWM/image refresh frequency, Hz

50 100 240 400 1000

LEDs Grayscale resolution, bits

1 19 18 17 16 15

8 17 16 15 14 13

16 17 16 15 14 13

32 17 16 15 14 13

96 17 16 15 14 13

256 17 16 15 14 13

4096 17 16 14 13 10

VDC Rset

Figure 21.47 Series resistor application for LED current control.


addition to that, it also varies with temperature. Due to the large amount9,128 of theLEDs powered, a power supply voltage will fluctuate as well. Furthermore, theLED forward current is a function of temperature Tj, diode forward voltage VF, andpower supply voltage VDC.

A constant current (usually a current sink) output driver (Fig. 21.49) offers moreadvantages.

Constant current output driver (Fig. 21.50) topology is the most popular in LEDdisplays.129e133 The constant current is provided by individual regulated current sinks.It allows for faster switching times as the output switch is not saturated.

D SDO

CLK

OE

Pow+

LTD

Figure 21.48 Switch output driver topology.

0.6V<

Figure 21.49 LED current control using a constant current sink.

Pow+

SDISDO

CLK

OE

I REF

LTD

+–

VR

EF

16:1

Figure 21.50 Constant current output driver topology uses current mirror to set the current.CLK, clock, LTD, latch signal, OE, output enable strobe, SDI, serial data in, SDO, serial data out


The output current is usually set by a single external resistor for all outputs simul-taneously. The resistor is connected to the driver’s internal reference voltage source(usually 1.2 V). Value of this resistor defines the reference current flowing from thisreference voltage source. This current is copied by a current mirror and multipliedby a certain coefficient (usually 16), setting the driver’s output current.

The major advantage of this driver type is that only 0.2e0.7 V driver dropoutvoltage (depending on technology used) is usually sufficient. The required drivervoltage drop-out relation to the operating current of the LED driver IC can be rectifiedby analyzing Fig. 21.51. Data was taken from MBI5045 datasheet (MacroblockInc.).129

It can be seen that 0.2 V dropout voltage is sufficient for stable constant current sinkoperation for 20 mA and below output currents, while higher currents might need0.4 V output voltage. This means that 2.7 V power supply is sufficient for red LED(2 þ 0.5 V VF) resulting in 75% LED drive efficiency. Best AC/DC power supply ef-ficiency for such low output voltage is about 75%,134 resulting in 56% wall-plug driveefficiency. Efficiency would have been 48% in case of 3.3 V (79.5% for AC/DC and61% for LED drive).

LED drivers with internal PWM in addition to constant current drive (Fig. 21.52)represents a new generation of intelligent drivers used to address the problem ofdata delivery bottleneck. This type of driver combines the constant current outputwith the individual PWM generators for every channel.135e140 PWM can be complextype, like enhanced-spectrum PWM (ES-PWM)135 or S-PWM.116

Driver is quite efficient since usually 0.4e0.7 V output dropout voltage is sufficienttomaintain the sink current at the programmed level. The output current is set by a singleexternal resistor but the gain of the current mirror can be programmed using same serialinterface. This type of driver allows for significant reduction of the data flow over driverIC. The current PWM dimming code can be loaded only once per video frame. Internal

0.0 0.2 0.4 0.6 0.8 1.00

5

10

15

20

25

30

35

Out

put c

urre

nt, m

A

Output voltage, V

IF2mAIF5mAIF10mAIF20mAIF30mA

Figure 21.51 Driver output current versus drop-out voltage for MBI5045 from Macroblock.


PWMgenerator of the IC takes care of the refresh, dimming, multiplexing (if any) func-tions. Larger number of internal registers is used here different from previously dis-cussed drivers. Internal registers are not only used to store the PWM value but also todo the current correction, to adjust the tile brightness to neighboring tiles, or to dimthe whole display during the nighttime or to provide the dot correction. Staggered delayof PWMoutput, scrambled PWM116 can be implemented to address the problemof largecurrent pulse at the beginning of the PWM cycle.

Even application of the low output voltage constant current output and internalPWM does not solve the driving efficiency problem. LED display consumes signifi-cant amounts of power. It is essential that LED is a low voltage device: the forwardvoltage is 2 or 3.6 V depending on color and technology. Moderate 640 � 480 reso-lution and three LEDs per pixel display with a 20 mA current per LED requires20 kA of total current. A macroblock of 64 � 48 pixels will need 180 A. Losses in po-wer distribution wires will occur when such currents have to be distributed. Further-more, efficiency of the AC/DC power supply decreases if the output voltage is lowdue to rectification losses (0.5e1.5 V forward voltage drop of the rectifier). Synchro-nous rectification reduces the problem but low efficiency for lower output voltageSMPS remains.134 Higher LED driver power supply voltage would ease the problems.

Several solutions for power LED driving have already been proposed in light-ing.141,142 Some automotive ICs dedicated are close to the required solution.142,143

Up to eight channels buck is offered by Linear Technology.144 However, these driversare dedicated for high power LEDs, require external PWM source, require the externalanalog voltage source for the LED current setting, and are complex and expensive.Interesting LED driver solution (Fig. 21.53) was proposed in Ref. 9.

It was proposed to incorporate floating point dimming in the driver using localPWM controller with required gamma correction already be stored on the chip. The

Pow+

DSDO

CLKLTD

PWM

contr

PWM

contr

P WM

con tr

P WM

contr

P WM

con tr

PWM

contr

PWM

contr

P WM

contr

I RE

F

+–

VR

EF

K:1

Configuration

Figure 21.52 Programmable LED constant current driver with the internal PWM. PWM, PulseWidth Modulation.


majority of PWM control codes are thrown away in order to get an appropriate gammacorrection curve. Floating point coding would exploit the coding space better. Incor-porated buck converters solve the current stability problem thanks to the current feed-back. The cost of converter switches should be low thanks to low output current. Sucha driver could preserve the same serial control interface for backward compatibilitywith older generation drivers. It is interesting to point out that current delivery tracefrom the driver to LED is usually long and exhibits some inductance. Usually thisinductance creates unwanted oscillations. Floating buck converter topology can incor-porate this inductance into energy storage inductor path. Efficiency is further increasedif synchronous rectification and smart current sensing technologies are used. Proposedtopology also eases problem with large current distribution: higher voltage can be usedin distribution (in case of aforementioned 180 A at 5 V distributed current can go downto 18 A at 50 V). Higher efficiency mains input AC/DC converter can be used.

Color rendering quality depends on white balance: red, green, and blue luminanceratios have to be 3:6:1.78 Chromaticity coordinates of white and RGB primaries definethe actual percentage. LED intensity binning can be used, but better idea is to matchthe RGB intensities approximately and then adjust the driving current. Current adjust-ment is not possible if limiting resistors are used for setting LED current, but use of theinternal reference voltage and current mirror in constant current output driver allowscurrent regulation. Refer Fig. 21.54 for possible topologies for current adjustment.

The constant reference current is usually set by a single external resistor. Theresistor is connected to the driver’s internal reference voltage source (usually1.2 V). Value of this resistor defines amount of the current flowing from this referencevoltage source (Fig. 21.54). The flowing current is copied by a current mirror andmultiplied by a certain coefficient K. This amplified current is used to set the driver’s

DC/DCcontroller

Pow+

Feedback

PWMcontrollerD

CLKLTD

SDO

Figure 21.53 LED driver incorporating serial digital communication, local PWM controller,and buck converter.9 PWM, Pulse Width Modulation.

+–

K:1

IREF=Vref/RREF

V REF

RR

EF

ILED=IREF x K

+–

K:1

I RE

FV

RE

F

+–V

AD

J

RR

EF

IREF=(Vref -VADJ)/RREF

ILED=IREF x K

Figure 21.54 Output current regulation using external resistor (left) or external voltage (right).


output current. It must be pointed out that current is regulated in inverse proportion toRREF value so RREF and Iout relation is nonlinear.

Problems occur when total output current adjustment145e149 of all the driverslocated on the tile PCB is required (Fig. 21.55). Usually, the tile incorporates 16 to128 drivers. Drivers are grouped into groups for red, green, and blue LED. It ismore convenient when such R, G, or B group of 4e64 drivers has a common outputcurrent adjustment. Reference resistors RREF can be grounded through a single addi-tional potentiometer, or they should be grounded into a regulated voltage source. Incase of a common resistor (Fig. 21.55), the reference current can be varied by regu-lating the common resistor value.145,147,148

This method is low-priced, easily operated by an unskilled operator, but not suitablefor remote control by module controller. Reference current regulation149 by varyingthe voltage difference between the internal reference and external regulated sourceis suitable for remote control. DAC (Fig. 21.56) or filtered PWM output can serveas the external regulated voltage source. Additional control circuit increases thecomplexity; nonlinearity and current mismatch between drivers are growing whenvoltage levels of external source is close to internal reference voltage.

Though more advanced techniques of internal current gain compensation have beenenrolling,136,150 simple LED display manufacturers still prefer to use the commonresistor method. Such simple control is not free of induction of errors that havebeen reported in Ref. 114. The LED tile with 16 � 32 LEDs was examined for the

RREF1 RREF2 RREFn

Drv1 Drv2 Drvn

RADJ

Figure 21.55 Common resistor application for simultaneous drivers’ current adjustment.

RREF1 RREF2 RREFn

Drv1 Drv2 Drvn

DAC

Figure 21.56 External voltage source application for simultaneous drivers’ current adjustment.


LED current errors induced by unaccounted currents on the PCB. Usually the tile has asingle potentiometer for red, green, and blue LEDs’ intensity adjustment. It was foundthat placement of the common resistors involves the additional voltage drop along theground traces of the PCB. This influences LED forward current set for driver. It wasdemonstrated that distance from the driver to the common control potentiometer andcurrent errors are correlated, the normalized cross-correlation of 0.94 was obtainedfor red LED drivers, 0.96 and 0.7 for green and blue correspondingly. Errors are rela-tively small: maximum 4.7% current deviation was found. Programmable LEDdrivers116,136,150 have the ability to adjust the gain of the reference current, so asoftware-adjustable driver is more immune to such type of errors.

21.4.4 LED dynamics-related issues

It has been presented that LED has to be driven by constant current source, but its in-tensity must be controlled by high-resolution PWM. High PWM resolution is neededto cancel the gamma correction which was used for image compression. LED intensityadjustment to match the neighboring pixels (dot correction, see Fig. 21.23), tile inten-sity matching (see Fig. 21.22), total display intensity adjustment to match the ambientlighting conditions increase PWM resolution requirements further. Resolution of19 bits is the best option,114 while 12-bit PWM resolution is the lowest requirement.Furthermore, refresh frequencies (PWM repetition frequency) beyond 240 Hz(400 Hz or even 1000 Hz) are required to avoid the image flicker.117 Shortest lightpulse in case of the 12 bit and 240 Hz PWM is around 1000 ns, but in case of400 Hz and 16 bit PWM the shortest pulse is less than 40 ns. Two factors limit theshortest PWM pulse duration, that is, the driver response time and LED response time.

Driver manufacturers129e133 mention few parameters that relate to the driverresponse speed. One is the level propagation time. It is measured as the time from inputsignal transition (crossing 50% of signal swing) to LED control output transition (50%crossing). For high to low transition it is tPHL and tPLH for low to high transition. Thespecified values are symmetrical, that is, tPHL ¼ tPLH. For Allegro’s A6278, the typicalvalue is 75 ns, while for the latest MBI5043 driver from Macroblock it is 20 ns.Another related parameter is the output rise and fall time. The rise time usually refersto the time it takes for a signal to rise from 10% to 90% of its peak value, and the oppo-site for the fall time. For A6278 it is 75 ns and 15 ns for MBI5043. Such symmetry ofdriver response times suggests that the LED driver will not introduce nonlinearity tothe PWM waveform. In addition, it might be expected that with IC manufacturingtechnology improvement of the switching times will go even faster.

LED manufacturers generally are not aware of the need for faster LED responsetimes. Normally, it is considered to be fast enough to satisfy dimming requirements.LED manufacturers do not specify the response time. Literature151e160 analysisrevealed that LED speed could get few to hundreds of ns. LED transient response de-pends on the LED chip structure, doping level, technology applied, and drivingconditions.

Investigation presented in Ref. 117 revealed that asymmetry in LED response time(skew) introduces nonlinearity in PWM dimming. Response skew reported for green


LEDs was 91.7 ns (standard deviation was 8.7 ns) and 86.2 ns for blue (7.6 ns standarddeviation). Red LEDs (AlGaInP) have smallest skew, 74.7 ns (1.1 ns standarddeviation).

The nonlinearity due to LED turn-on time tR and turn-off time tF as a function ofexpected PWM pulse duration son can be calculated as:

dnonlðsonÞ ¼ tF � tR2son

$100% ¼ Dt

2son$100%. (21.24)

Fig. 21.57 is used to demonstrate the skew influence on blue LED PWM nonline-arity at various refresh frequencies.

Then it’s the response skew that has to be used for PWM resolution limitation.Table 21.4 is used to present the achievable PWM resolution for most common refreshfrequencies.

0 20 40 60 80 100–15

–10

–5

00.125MHz PWM

0.25MHz PWM

0.65MHz PWM

1.25MHz PWM

I v e

rr, %

Duty, %

Figure 21.57 LED PWM nonlinearity due to skew in LED response time.117

Table 21.4 Available PWM grayscale resolution limitation by LEDnonlinearity117

Refresh frequency, Hz

100 240 500 1000

dskew, % Bits

20 15 14 13 12

10 14 13 12 11

5 13 12 11 10


Results in Ref. 117 demonstrate that the LED response time skew introduces aserious limitation on linearity of PWM, thus for the available resolution. Therefore,the LED light pulse appears to be a limiting factor for the PWM dimming resolution.

It is interesting to point out that skew is quite stable (note the standard deviationvalues). The nonlinearity introduced can be compensated using a look-up table orsome analytical approximation. But in case of more complicated PWM techniqueslike BPWM/BAM,126 GPWM,114 pulse frequency modulation, and pulse amplitudemodulation67,111 or S-PWM116 the nonlinearity will have more complex response;therefore, it should be studied for the artifacts indicated in Ref. 117. Fig. 21.58 isused to represent the relative nonlinearity error for the red LED operated at 1 kHzPWM frequency for more complicated PWM dimming techniques.

Analysis of data in Fig. 21.58 can be confusing since the relative error is more than5% for the duties below 0.4%. It should be kept in mind that the large number of PWMcodes or PWM resolution is required for gamma correction.43,46 It is roughly a powerfunction of luminance Y. Coding a luminance signal to a signal by the use of a powerlaw with an exponent of between 1/3 and 0.45 has excellent perceptual performance.

Experimental investigation has been carried out using the setup of Ref. 117. Theconventional PWM, BPWM, and GPWM dimming was applied. Results for BPWMare presented in Fig. 21.59. It can be seen that skew influence is even larger becauseone BPWM period contains more than one pulse. Eq. (21.24) has been used to calcu-late the theoretical output deviation from linear law.

Results show good agreement between the experiment and theory. Slightly largerskew at short pulse durations is because of the crowding effect. Skew is the samefor PWM and BPWM where the pulse durations are the same. Results for theGPWM are presented in Fig. 21.60.

It is worth to mention that LED driver also influences LED response skew. Whilefew nanoseconds short response times can be achieved when special carrier sweep-out

0.0 0.2 0.4 0.6 0.8 1.0–15

–10

–5

0

BPWM

GPWM

PWM

Rel

ativ

e er

r, no

nl (%

)

Duty (%)

δ

Figure 21.58 Nonlinearity due to skew in LED response time for complex dimming techniques.BPWM, binary-weighted PWM; GPWM, gated binary-weighted PWM.


technologies are used,161 in reality situation is different since LED display drivers seekboth efficiency and economy. Investigation presented in Ref. 162 was carried out toevaluate the response time of visible light LEDs used in video displays when drivenby different driver topologies. Namely, four topologies were analyzed (Fig. 21.61):(1) constant current with disconnect (commercial MBI5026 driver by Macroblock);(2) passive driver (current limiting resistor); (3) Transconductance amplifier; and (4)constant current driver with LED tamper for fast carriers sweep-out.

Commercial driver is a representative of actual situation in LED display. It containsa constant current sink. Turn-off is provided by disconnecting the output. Passive

0 200 400 600 800 1000 1200 1400 1600

–120

–100

–80

–60

–40

BPWM: Experiment Theory

PWMS

kew

, t (

ns)

Pulse duration, on

(ns)

Δ

τ

Figure 21.59 Experimentally measured nonlinearity due to skew in LED response time forBPWM and PWM dimming techniques.

0 200 400 600 800 1000 1200 1400 1600

–200

–180

–160

–140

–120

–100

–80

–60

–40

GPWM: Experiment Theory

PWM

Skew

, t (

ns)

Pulse duration, on

(ns)

Δ

τ

Figure 21.60 Experimentally measured nonlinearity due to skew in LED response time forGPWM and PWM dimming techniques.


Cμ1

0.1 F+5V

Test patternupload

OE

OUT

LED

+5V

Rin56

X1BNC

GND

PWR

Rprog

Iset

Rin56Ω

LEDRs

51ΩX1

BNC

Rin51

+– U1

AD8001

+5V

-Pow

Rfb50

X1BNC

M1FDV301N

Rin56

LED

+50V Rs2k5X1

BNC

Ω

Ω

Ω

Ω

Figure 21.61 Driver topologies used in Ref. 162 experiment for LED response times’measurement (from left to right): (i) commercial LED driver, (ii)passive current limit, (iii) transconductance amplifier, and (iv) constant current with tamper turn-off.

710Nitride

Sem

iconductorLight-E

mitting

Diodes

(LEDs)

driver uses current limiting resistor Rs to set the LED driving current. Transconduc-tance amplifier was using current feedback amplifier AD8001, the output currentwas proportional to the input voltage because of transconductance topology. It was ex-pected that such driver should deliver shortest rise and fall fronts, approximately1.5 ns. Constant current driver with tamper turn-off was using passive current source(thanks to high voltage and large Rs value). LED turn-off was carried out by tamperingN-channel MOSFET FDV301N. Such topology ensured that current flowing throughLED or tampering FET was relatively stable (less than 5%). Expected turn-on delaywas 3.5 ns and turn-off delay 6 ns.

Measured rise-and-fall front durations are presented in Fig. 21.62. Mean values forthe measurement are complemented by 3s (standard deviation) to indicate the 99.9%probability range for the measurement results.

It can be seen that LED drivers exhibit different operation speed when in connectionwith LED. Commercial driver did not produce the response speed specified by manu-facturer: instead of specified typical 40 and 70 ns for tR and tF, measurements indicatethat 283 and 111 ns mean values, which are far away even from specified maximum(120 and 200 ns accordingly). Longer response times could be attributed to LEDresponse influence, but measurements on tampered topology (rise time) should besimilar. Therefore conclusion was drawn that commercial driver did not producetypical response performance. Passive driver performance was expected to matchthe transconductance amplifier, but it did not perform better than transconductanceand tampered constant current drivers. Measured rise front mean value was 62 nsand turn-off duration was 30 ns. Such asymmetry was expected because LED turn-off time could be faster because carriers sweep-out prevails here. As expected, tamperturn-off topology exhibited shortest response times: tampering the LED ensured fastestlight cut-off: 4.2 nsdclose to receiver speed (4 ns) used in measurements. Though

0

100

200

300

Rise time

TransconductancePassiveTampered

Tim

e, n

s

Commercial

Fall time

Driver topology type

Figure 21.62 LED photoluminescence pulse rise and fall front measurement results for differentLED driver topologies.162


such a setup seems attractive, it exhibits an asymmetry in pulse response (skew) whichin turn will degrade the dimming linearity. Yet, this driver topology indicates the limitfor rising front duration (56 ns mean) which can be achieved with constant currentsource. Transconductance topology driver performance was the best: minimal skewfor tR and tF (54 and 55 ns). This driver should have exhibited some skew in drivingresponse since LED is driven by 20 mA/0 mA (on/off) current pulses and this shouldcause the skew in the light pulse (turn-off is not promoted by additional current). Mostprobably, this was counterweighted by the phenomena that LED turn-off is muchfaster due to carrier sweep-out process prevalence in photoluminescence cut-off.This topology will produce lowest nonlinearity in case of PWM dimming.

This investigation demonstrates that the LED response time skew is significant:LED response times can reach 300 ns and skew can become almost 200 ns if commer-cial LED drivers are used. Since this is the most usual case in LED displays, noticeablenonlinearity can be introduced if high refresh frequency and high resolution PWM isused for LED pixel grayscaling. Situation can become even worse if complex PWMdimming techniques are used. However, errors have relatively low variance which al-lows for look-up table compensation.

21.4.5 LED directivity influence

Directional (angular) dependence of the luminous radiation pattern measured at the far-field of the LED emitter is called the LED far-field pattern.7,32,33 Some recommenda-tions for measurement conditions are defined by the CIE.36 Definitions of the axesproposed by the CIE Technical Committee 2e46 are:

• LED front tip: is the center point of the LED outer surface of the emitter.• Optical axis: is the axis in the direction of centroid of the FFP.• Peak intensity axis: is the axis through the front tip in the direction of a maximum intensity.• Mechanical axis: the axis through the tip in the direction of the axis of symmetry of the body.

Refer to Fig. 21.63 for the graphical explanation of above. Any of the three axes canbe chosen as a reference axis for intensity measurements.

0

30

6090

120

150

180

–150

–120–90

–60

–30

peak

Mechanical axis

Peak intensity axis

Optical axis2 0.5

LED

Θ

Θ

Figure 21.63 LED axis definitions.


Polar angle Q is the inclination angle in vertical direction measured from the LEDmechanical axis,163 according to Ref. 164 it is a line through the tip in the direction ofthe axis of symmetry of the body. Azimuth angle f is the angle in horizontal direction,measured in plane perpendicular to mechanical axis with zero aligned with line alongLED pins. FFP is obtained as the spatial intensity I distribution over the observationangles Q and f, resulting in I(Q) and I(f) functions.165,166 LEDs used in video dis-plays can have different directivities along vertical and horizontal axis in order tomatch the display viewing angles and get the best use of the energy emitted. ThenFFP is measured both in a horizontal and vertical plane. FFP can be presented in polar(Fig. 21.64 left) or Cartesian (Fig. 21.64 right) coordinates system. Since absolute in-tensity is not important for FFP analysis, normalization to unity is used.

Numerical parameters can be extracted from FFP: (1) peak emission directionQpeak, and (2) half power beam angle 2Q0.5, where the source’s relative intensity isdropping to the half of the peak emission can be obtained from the FFP using measuredI(Q) (Fig. 21.65).

–90

–75

–60

–45

–30–15 0 15

30

45

60

75

90

I(φ)

I(Θ)

–90 –60 –30 0 30 60 900.0

0.2

0.40.50.6

0.8

1.0

I v/I vmax

, a.u

.

Angle , deg

I( )

I( )

Θ

Θ

φ

Figure 21.64 LED far-field pattern in horizontal (thick line) and vertical (thin line) plane can bedifferent.

–60 –50 –40 –30 –20 –10 0 10 20 30 40 50 600.2

0.4

0.6

0.8

1.0

peakAprox

peakMax

peakHalf

2 0.5

0.50.5

I v, no

rm

Angle, deg

Θ

Θ ΘΘ

Θ

Θ

Figure 21.65 LED FFP parameters determination.


Actually several methods for Qpeak location can be used. The most evident way tofind the peak emission angle is to locate the absolute maxima on the FFP (Qpeakmax inFig. 21.65). LED manufacturers usually rank the intensity on mechanical axis (angle0 in Fig. 21.63). Variation ofQpeakmax will cause that the LED display speckle despitetight peak intensity ranking. If FFP is approximated by some analytical function, thenpeak of this function is used (QpeakAprox in Fig. 21.65). Middle of half power beamwidth 2Q0.5 (QpeakHalf) has similar meaning. Same meaning can be assigned to thecentroid position (center of gravity of FFP).

LED tinting improves the fill-factor, smooths the directivity pattern, but increasesthe cross talk between the LEDs reducing the contrast. If LED encapsulation is not per-formed carefully, or tinting materials are not properly distributed, FFP variation is sig-nificant. See example of several LED measurements of the same batch from Brilliancetechnologies LED BTL-55NRDS-O1-K2FFP in Fig. 21.66.

Individual LED FFPs are influencing the display image quality. Display surfaceluminance directivity is differing from individual LED, because when viewed fromdistance, number of LEDs per spectator surface is varying with angle: the sharperthe angle, the more LEDs are per surface perpendicular to viewer (Fig. 21.67).167e169

Then display luminance FFP Bv(Q) can be calculated from LED FFP I(Q) as:

BvðQÞ ¼ IvðQÞcosðQÞ . (21.25)

The measured LED intensity170 FFP and resulting display brightness obtained us-ing Eq. (21.25) are presented in Fig. 21.68.

It can be seen that display viewing angle is wider than the individual LED. Thebroadening of viewing angle depends on the shape of the LED FFP. It should be notedthat the LED FFP used in Fig. 21.68 has been obtained as an average or multiple LEDmeasurements (as per Fig. 21.66) since not the individual but statistical LED response

–60 –50 –40 –30 –20 –10 0 10 20 30 40 50 60

0.5

1.0

I v,nor

m

Angle, deg

Figure 21.66 LED FFP variation within same batch in case of low manufacturing quality.


will form the screen FFP. It should be pointed out that Eq. (21.25) is extreme simpli-fication of actual situation: it does not account the shouldering effect, diffraction, lou-vers, and other obstacles which will further modify the display luminance.

LED directivity and display front face design determine another important param-eter: point spread function (PSF). This issue is related to “spreading” of the individ-ual pixel brightness into surrounding area which can cause blurring of the image.Being a serious issue in LCD and CRT displays, this phenomenon is very weak inLED displays as it is easy to isolate each pixel optically. Sometimes, this phenom-enon is addressed as a modulation transfer function (MTF).171 Square wave gratingsare used to determine the MTF and resulting display image quality.172 Some authorshave introduced such terms as a spatial Nyquist limit (based on a pixel pitch) andvisual resolution limit (obtained by subjective evaluation).173 The visual resolutionlimit and MTF can be considered as a more perceptive estimation of display visual

β2 β1

L1

L2

Observer 2 Observer 1

αα

LEDs

Figure 21.67 Display directivity defined by luminance depends on viewing angle: area coveredby same angle a is different for Observer 1 and Observer 2.

–80 –60 –40 –20 0 20 40 60 800.0

0.2

0.4

0.6

0.8

1.0 Display

LED

).U.A( ytilanoitceri

D

Angle Θ (deg)

Figure 21.68 Display directivity and LED directivity comparison.


performance, while the PSF can be obtained by purely hardware-based measure-ments and a spatial Nyquist limit can be estimated by simple calculation. Authorshave determined that a visual resolution limit is roughly half of that of a spatialNyquist limit.

Study174 showed that visual discomfort glare is produced by flicker frequency,panel luminance, and viewing angle. It was also found that there is an interactionbetween display luminance and the viewing angle. Visual comfort models showthat optimal comfort can be achieved by display operating conditions selection.

Analysis and prediction of display performance are significantly simplified if LEDFFP is approximated by some analytical function.175e177 The need for FFP approxima-tion in LED video displays can be justified by several reasons: (1) when the FFP haslarge variability at peak emission, which might distort the Qpeak value; (2) when themeasured FFP dataset is too sparse; (3) when it is required to evaluate the FFP bysome analytical form, which then is used for calculation and so on.178,179 AnalyticalLED radiation far-field pattern approximations have been used in a wide variety ofapplications.178e182 For example, to determine and optimize the radiation flux reach-ing the projection display screen,182 irradiated premises181 or drastic reduction of mea-surement errors associated with an incorrect rotation axis.180 With no doubt, suchequations are useful in the design of LED video displays.168,178,179,183,184 Image puritycan be described as smoothness of an image. The last is defined by LED brightness anddirectivity pattern repeatability between the LEDs used for screen manufacturing.They define both image purity and screen image quality at various viewing angles.

Several functions can be used for LED directivity approximation: parabolic is theeasiest, well suited for peak approximation; cosg is used by LED society; Gaussianis widely accepted. Simplex algorithm is usually used for fitting with L1 or L2norm as convergence criteria.168,169,179,184

Parabolic function in a form of polynomial:

IvðQÞ ¼ a0 þ a1Qþ a2ðQÞ2; (21.26)

can be used for peak emission direction Qpeak angle estimation:

Qpeak ¼ � a12a2

; (21.27)

and half power beam width:

2Q0:5 ¼ �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2a21 � 8a0a2

q

2a2. (21.28)

Power function of cos32 can be used as:

IðQÞ ¼ Imax cosðQ�QpeakÞg�1. (21.29)


where Qpeak is the peak emission angle and g is a coefficient proportional toviewing angle 2Q0.5:

2Q0:5 ¼ 2 arccos

0

B@e

lnð2Þg�1

1

CA. (21.30)

Gaussian approximation185 does not have side lobes:

IðQÞ ¼ Imax$e

��lnð2Þ ðQ�QpeakÞ2

Q20:5

�

(21.31)

Then Gaussian with DC offset IOff can be used as:

IðQÞ ¼ IOff þ Imax$e

��lnð2Þ ðQ�QpeakÞ2

Q20:5R

�

: (21.32)

The half power beam angle is obtained as:

2Q0:5 ¼ �2Q0:5R

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ln

�Imax � IOff

Imax

�� lnð2Þ

s

lnð2Þ . (21.33)

Moreno et al.179 suggested multiple Gaussian terms function for LED FFPapproximation:

IðQÞ ¼ I1$e

��ðQ�QpeakÞ2

W1

�

þ I2$e

��ðQ�QpeakÞ2

W2

�

; (21.34)

whereW1 and W2 are defining the first and second term width; I1 and I2 are weights ofthe terms.

Analysis presented in Ref. 169 used several LED types with different FFP functionto evaluate the performance of the aforementioned functions. See the approximationexample in Fig. 21.69.

It has been concluded that dual-term Gaussian approximation can produce the re-sults that are close to ideal approximation. Simple, single-term functions can also beused in reduced angular range, corresponding to display half radiance angle. Relativeapproximation error is 5%e8% for these functions. Such performance was namedsatisfying for general engineering tasks like illumination rendering186 or display direc-tivity performance prediction.


FFP approximation function can be useful in directivity analysis. LED directivitymeasurement in situ technique is presented in Ref. 183: quick FFP measurement is pro-posed without dismantling the LEDs from the assembled display tile. Positioning hasbeen replaced by switching on a LED from different positions, obtaining the desiredinspection angle this way (Fig. 21.70).

The 32 � 16 LED’s 10-mm pitch tile was investigated. A sensor has been placed ata 146 mm distance from the tile LED tip, which corresponded to a 106 and 66 degreesmaximal viewing angle in a horizontal and vertical direction respectively. Collecteddataset was approximated by Gaussian function. The approximation was treated asthe averaged LED FFP. Several LEDs have been removed from the tile and directivitymeasured by a goniometer. In situ and goniometry measurement results comparison ispresented in Fig. 21.71.

The averaged goniometry FFPs results are presented as thick lines, in situ results aredashed lines. Results confirm that FFP measurement can be done in situ without

–70 –60 –50 –40 –30 –20 –10 0 10 20 30 40 50 60 700.0

0.5

1.0

Low intensity

Ideal Gaus Gaus x 2 Gaus + offset cosg

x2

x2+x4

Highintensity

I (A

.U.)

Angle (deg)Θ

Figure 21.69 Measured Z2BH LED FFP approximation by candidate functions.169

LED 1

Photosensor

LED 4

LED 3

LED 2

234Θ Θ Θ

Figure 21.70 LED FFP can be measured in situ: different inspection angle is obtained bymeasuring the radiation from sequentially lit LEDs.183


dismantling the LEDs from the tile. The obtainable accuracy and suitable approxima-tion function analysis are beyond the scope of this paper. Such technique could be veryhandy when a fast and portable measurement is needed to evaluate or verify the LEDtile quality.

The viewing angle of display may also consider color shift, which occurs due toencapsulant, tinting, shouldering, diffraction, or multiple reflection effects. If thereis a significant color shift angular dependence before the LED display luminance fallsto 50%, then viewing angle should be limited by required color stability. Louvers areessential part of contrast enhancement, but, as it was mentioned above, they can alsoalter the vertical viewing angle (mostly).

21.4.6 Radiation stability

As stated in Ref. 187 LED failure rate is high at the beginning (early failure) and end(product wear-out) of the product cycle, spontaneous failure (period in between theearly and wear-out) rate is very low and flat over time. Failure rate over time curve(Fig. 21.72) is in a shape of a “bathtub.”

–50 –40 –30 –20 –10 0 10 20 30 40 50

0.5

1.0

Blue,gonio meas

Blue, in situ measGreen,

in situ meas

Green, gonio meas

Red, in situ meas

Red,gonio meas

I v/I vmax

, a.u

.

Angle, deg

Figure 21.71 In situ LED FFP measurement results (thin dashed lines) comparison withgoniometry (thick lines).

Normal life

Time

Failu

re ra

te Earlyfailures,startup

Wearoutperiod,

end-of-life

Figure 21.72 Electronics failure rate over time follows Weibull curve or “bathtub” shape.


Extrinsic failures (usually appear early and spontaneously) are generated by defec-tive materials, deviations in the manufacturing, or by incorrect handling and operationby the customer.187 More than 99% of these failures are observed during product as-sembly or in the first hours of operation. The intrinsic failures are the reason for the so-called wear-out period of the component at the end of the product cycle. It occursbecause of accumulation of the wear and aging of the materials used (chip and encap-sulant188). This change over time can be measured as parameters degradation: changesin brightness or color coordinates. Variability in light output of nominally identicalLEDs can differ by a factor of two.189 Wavelength shift by several nanometers canoccur.190 The main failure reason of AlInGaP LEDs is the degradation.188 Degradationcan be divided into two periods: luminosity increases during the first hours of opera-tion and after that decay follows an exponential trend. The exponential decay rate sparameter follows an Arrhenius equation188:

s ¼ C$eEAkBTj : (21.35)

where EA is the activation energy, kB is the Boltzmann constant, Tj is the junctiontemperature, and C is a constant determined by LED operating conditions (intensityand drive current). Decrease in the light output for every 10K increase in temperature is10% for AlInGaP red LED, 5% for InGaN green LED, and 2% for InGaN blue LEDwas reported in Ref. 191. Lumen maintenance describes the degradation of the LEDluminous flux over time and is measured as percentage of the initial LED output.187,192

A commonly accepted threshold is 70% for LEDs used in video displays.193,194

Study195 revealed that half-life of blue InGaN LEDs is approximately 2$104e4$104 hin case pulsed operation at 40 mA and 40�C. Lifetime dependency on pulse durationwas noted: longer lifetime was found for shorter pulses and 2e4 times longer forpulsed operation than continuous. Results of Ref. 196 suggest that temperature has animpact on the electroluminescent intensity, color, and reliability of the LEDs. Carrierfreezing and shallow defects (nitrogen vacancies or oxygen in nitrogen sites) can trapthe injected carriers and reduce the electroluminescent intensity at temperatures below150K. Deep traps (structure dislocations at the interfaces) significantly reduce theefficiency and trap-assisted tunneling current is increased (causes heat and results inredshift of the emission peak) at temperatures above 300K.196

A mathematical model for the emission spectrum of LEDs at different drive cur-rents was proposed in Ref. 197 Authors developed a simulation program that can pre-dict the spectral power distribution and color rendering index (CRI).

The thermal resistance of the LED defines the heat dissipation speed and efficiency:lower thermal resistance leads to the reduction of the junction temperature.198 Authorsof survey199 found that LED packaging with thermal resistance 5K/W have loweroperating junction temperature and longer lifetime and spectral stability. Epoxybrowning for LEDs at increased temperature tests and drive current influence onchip degradation (increase of non-radiative recombination current and series resis-tance) have been observed in Ref. 188.

LED lifetime can also be based on a statistical analysis of failed components.187

Thermo-mechanical stress on a component is induced by temperature cycles: variation


in pixel intensity, transportation, and ambient conditions. It can cause lens cracking orinterconnects failure. Humidity, pollution, UV radiation, mechanical stresses, anddents further derate the lifetime. Mean time to failure of 3.106 h at 65�C and 20 mAconditions was found for AlInGaP red LEDs in Ref. 188.

Paper194 claims that the typical life of LEDs can be expected from 3000 h (harshoperating conditions: high current, high temperature, and high humidity) to50,000 h in benign environments. It was suggested that LED die cracking can becontrolled by fine-tuning the thermal expansion coefficients between the substrateand epitaxial layers and that optimal medium layer between the substrate and theepitaxial part is a key. The bonding should be optimized by wire type, pad metalliza-tion, and device configurations control. Authors state that UV transparent or silicone-based encapsulants should prevent lens yellowing due to UV radiation. Silicone ofmodified epoxy encapsulants and low thermal resistance substrates are stated usefulfor minimizing the thermal degradation. One more proposal suggested light extractionefficiency improvement thanks to refractive index of LED die and encapsulant match-ing or high refractive index encapsulants and efficient case and reflecting cup design.

LEDs are used in large numbers in large-scale video displays. Therefore the price ofan individual LED has to be low. This fact sets certain design restrictions: no complexconstruction for thermal management is possible, encapsulants, LED chips, and evenpin cost is essential.200

Quoted201 life figures for the LEDs range from 20,000 to 100,000 h. It is obviousthat actual lifetime will differ depending on actual operation conditions. Sometimeshigh LED currents are used to achieve high screen brightness. High drive currentscause temperature rise. Furthermore, ambient temperature of display exploitationcan be high or sun radiation can rise the surface temperature, especially for outdoorapplications. Elevated temperature lead to faster degradation of the LEDs’ perfor-mance: color and light output are changing, and can cause a drop in image uniformityor total luminosity (if much shorter than specified time).

Failure or light output degradation of the several LEDs in display significantly de-grades the image quality. That is why LED tile replacement in the module has to beenvisaged. Individual LED replacement in a tile can be done only under workshopconditions. Even in such case the integrity and performance of the tile cannot be main-tained. Intensity of new module or LED in aforementioned cases differs: electrolumi-nescent output of the whole display has already degraded. The tile or dot correction isnecessary in order to blend-in the new components. Usually it is implemented insoftware.

21.4.7 EMC issues

LED display pixels are operated at constant current and PWM dimming is used forgrayscale production.119 Current spikes are generated on power supply lines whenthe large number of LEDs is switched by PWM. For instance, if 512 LEDs (20 mAcurrent each) on tile are turned on during the PWM cycle, this results in a 10 A currentspike. Such pulses create electromagnetic interference (EMI). Voltage dip, currentsurge, and ringing are induced on power lines which can be a source of EMI, if not


addressed in a corresponding way. Usually the equipment closest to display is affected.Most frequent issue is the clutter for radio microphones.202 Though, the latest caseaccounted was when all TV broadcasts in few kilometers radius were affected by afresh LED display installed on a shopping mall wall. Another problem encounteredwas a curtain display, causing FM interference.

There are several reasons for high EMI level. It could be already mentioned PWM,which is inherent for LED displays. Switch mode power supplies are usually used toget the efficiency and reduce the size and weight. Sometimes LED and driver PCBsare separate to allow for future driver upgrade, repair, or to increase the LEDs density,if through-hole LEDs are used. Such solution leads to the increased length of LEDdriving traces.9,203 Increased path of pulsed LED current leads to larger radiating surfaceof the parasitic magnetic loop antenna. Data transmission between pixel drivers, betweenthe tiles, and betweenmodules is synchronized to clock signal. This clock signal is routedthrough thousands of nodes. Single node communication could be the EMI source, butwith larger numbers this phenomenon multiplies into an unresolvable problem.

Countries acknowledge the electromagnetic compatibility (EMC) problem anddedicated regulatory frameworks have been established since radio communicationsintroduction to allow the variety of electronic devices to coexist. Governments are ex-pected to harmonize the allocation of radio bands and impose their protection. Radiofrequencies use is strictly regulated by national laws and coordinated internationally bythe International Telecommunication Union (ITU). ITU-R develops regulations and ismanaged mainly by national administrations; the ITU-T prepares telecommunicationstandards and is maintained by the industry. When it comes to EMC, CISPR (Interna-tional Special Committee on Radio Interference, founded in 1934, part of the Interna-tional Electrotechnical Commission, IEC) is responsible for standards controlling theEMI in electrical and electronic devices. In the United States, the FCC (Federal Com-munications Commission) regulates and implements standardization functions. InEurope, CENELEC (European Committee for Electrotechnical Standardization) andETSI (European Telecommunications Standards Institute) are responsible for EMCstandards while regulation is by directives 2014/53/EU (Radio Equipment Directive)and 2014/30/EU (Electromagnetic Compatibility Directive) produced by EuropeanParliament and European Council and then enforced by national governments. Inmost cases CISPR is the origin of the standards used.

Electromagnetic compatibility (EMC) of the device means that it is compatible withits electromagnetic (EM) environment and does not emit EM energy beyond the levelswhich can cause electromagnetic interference (EMI) to other devices. European coun-tries not only limit the levels of emission but also demand the immunity of the deviceagainst possible EMI level. LED displays in general fall under European normEN55032 (or CISPR32,204 Electromagnetic compatibility of multimedia equipmentdemission requirements) and EN55035 (or CISPR35,205 Electromagnetic compatibilityof multimedia equipmentdimmunity requirements). Both emission and immunity re-quirements are further subdivided into conducted and radiated emissions for the sameof measurement and interpretation simplification.

It is easy to assume the current or voltage signal spectrum as piecewise linearapproximation, presented in Fig. 21.73.


Yet, the EM radiation of such signal is affected by antenna. Antenna in a sense ofEMI is either a current loop (magnetic antenna) or open-ended conductor (electricalantenna). Those are usually smaller than emitted wavelength. In such case electricalantenna radiation is directly proportional to frequency; electrical field strength at10m distance is206:

E ¼ 1:26$10�7$f $L$ICM; (21.36)

where f is the radiation frequency, Ldantenna length, ICMdantenna current.Radiation of magnetic antenna is proportional to square of the frequency; electrical

field strength at 10m distance is206:

E ¼ 26:3$10�16$f 2$A$IDM; (21.37)

where A is the loop area, IDMdloop current.This means that EM radiation can reach levels (refer Fig. 21.74) not permitted even

at very high harmonics, where radiation was not expected according to Fig. 21.73.Image refresh frequencies are relatively low: can get up to 400 Hz to 1 kHz. But if

16-bit PWM resolution is used, shortest PWM pulse can correspond to 32 MHz. Suchshort pulses shift the first cut-off frequency presented in Fig. 21.73. This can create sig-nificant EMI. Shorter pulse duration also means lower spectral level, thanks to 1/sterm. Simple example demonstrates the application of findings presented above.

20log(2Uτ/T) f1=1/( )

f2=1/( tr)

Ampl

itude

,dB

Log frequency

τπ

π

–40dB/dec

–20dB/dec

Figure 21.73 Simplified frequency spectrum of pulse signal.

20log(2.52.10-7ILτ/T2)

–20dB/dec

f1~fundamental

f2

EM fi

eld

stre

ngth

,dBu

V/m

Log frequency

20log(52.6.10-16IAτ/T3)

+20dB/dec

f1~fundamental

f2

EM fi

eld

stre

ngth

,dBu

V/m

Log frequency

Figure 21.74 Simplified frequency EM field strength spectrum of pulse signal, radiated byelectrical antenna (left) and magnetic antenna (right).


Assuming 20 mA LED current and MBI5043 LED driver (33 MHz grayscalefrequency or 30 ns shortest pulse, 15 ns rise/fall time), one can derive the current spec-trum using Fig. 21.73 (refer Fig. 21.75 for 30 and 300 ns pulse spectrum at 1 kHzPWM period).

If this data is applied to Eq. (21.37) then emission created by current loop can beestimated. Assuming that 1.5 mm PCB was used and all LEDs were routed using50-mm long traces above the ground, 75 mm2 radiating loop is obtained (Fig. 21.76).

It can be seen that single LED current loop contribution is small: it is 125 dB belowthe 40 dBmV/m limit, defined by CISPR32 for Class A equipment.204 Actually onlypart above 30 MHz is of interest, because radiation below this limit is analyzed as

0.1 1 10 100 1000–70

–60

–50

–40

–30

–20

–10

0

10

20

I LED (d

BuV

)

Frequency (MHz)

ILED30ns ILED300ns

Figure 21.75 Current spectrum envelope of 30 and 300 ns PWM pulses with 1 kHz period.

0.1 1 10 100 1000–140

–130

–120

–110

–100

–90

–80

Em

30ns

(dB

uV/m

)

Frequency (MHz)

Em30nsEm300ns

Figure 21.76 Electrical field strength of 75 mm2 radiating loop of 30 and 300 ns PWM pulses.


conducted. It is also notable that it is the pulse rise time that defines the emission level,not the narrower minimum pulse duration as in Ref. 207. Fig. 21.73 can be misleading,causing assumption that LED current does not create EMI beyond required limit.Assumption above was for continuous ground return, which is not possible in caseof double-sided PCB, where LEDs and drivers have to share ever shrinking area. Incase of separate LED and driver PCBs loop area can increase several times, loweringthe noise margin. LED display may contain millions of LEDs, so radiation will growaccordingly: 64 � 48 pixel macroblock (12,288 LEDs) will have the radiation which is82 dB higher. If dozens of macroblocks are used in LED display, radiation can reachlevels beyond the limit. As pointed in Ref. 208, a single panel may be under therequired limit, but 1000 panels radiating the same give 30 dB EMI increase; so itwas assumed that complete system would never pass the EMC test. But modulesare usually certified individually.

It is worth pointing out that Fig. 21.73 to Fig. 21.76 represent spectral envelope.Actual spectral content of the produced EMI is affected by LED dimming technique.Investigation presented in Ref. 209 analyzed the current spectrum compared the EMIpotential of the three major PWM dimming techniques: PWM, BPWM, and GPWM.Spectral content of 1 A current (all LEDs simultaneously switching) spike of PWMand GPWM with 1 kHz repetition frequency and 3/127 duty cycle (7-bit PWM reso-lution) is presented in Fig. 21.77.

GPWM waveform for the same intensity code 3 resulted in different spectralresponse (Fig. 21.77 black) because GPWM contains two pulses; its period had tobe slightly adjusted114 and those two pulses are a bit narrower.

Current spectrum when uniform pixel intensities distribution was used is presentedin Fig. 21.78. PWM dimming spectrum had exponentially decaying frequency

0 10k 20k 30k 40k 50k 60k 70k0.0

10.0m

20.0m

30.0m

40.0m

50.0m

Cur

rent

spe

ctru

m, A

Frequency, Hz

GPWM

PWM

Figure 21.77 Low frequency part of the current spectrum for grayscale code 3 in case of PWM(red) and GPWM (black) dimming. GPWM, gated binary-weighted PWM; PWM, pulse widthmodulation (PWM).


response, GPWM had similar behavior but contained certain modulation, but BPWMhad significantly smaller emissions.

Lower emissions for BPWM were explained by opposite waveforms presence incase of uniform intensities distribution. Same relation was observed209 when largerdisplay (256 LEDs) was assumed and test image “Lena” was used together withgamma correction of power 2.5.43 Fig. 21.79 represents the power supply current spec-trum for discussed case.

Notable that BPWM spectrum is lower by approximately 20 dB. Low frequenciespart was always higher for the PWM technique. BPWM has slightly better perfor-mance in high frequencies region. However, the BPWM technique no longer holdsthe significant advantage noted in the case of uniform image intensities distributionbecause of disproportion in opposite waveforms which was the case in a uniform dis-tribution image. Hence, it may be concluded that EMC potential depends on the dim-ing technique used.

Local power decoupling capacitors (usually MLCC) are expected to absorb the cur-rent spike, but if electrolytic capacitors are used or MLCC capacitors are improperlyplaced or routed, current spike penetrates the power supply leads. Length, so theloop area, of power leads is large. LED display usually has metallic enclosure for me-chanical reasons. This works well in EMC sense too. EMI leak can occur at LED tilesinterface with macroblock enclosure: usually there is no metal contact here and tile canbe encapsulated into plastic case for the cost reasons (refer Figs. 21.26 and 21.29).

LED display power supply is using AC/DC converters. The reason is obvious: withtypical 5 or 3.3 V power for electronics SMPS is required to keep size and efficiency.Aim for efficiency contradicts with EMC: faster switching ensures efficiency, but in-creases EMI; input filtering reduces EMI, but affects efficiency; inrush current limita-tion also lowers the efficiency.

0 100k 200k 300k 400k 500k100e–21

1a10a

100a1f

10f100f

1p10p

100p1n

10n100n

1μ10μ

100μ1m

10mGPWM

BPWM

PWM

Cur

rent

spe

ctru

m, A

Frequency, Hz

Figure 21.78 Current spectrum for the PWM, BPWM, and GPWM grayscale production incase of uniform code distribution in an image. BPWM, binary-weighted PWM; GPWM, gatedbinary-weighted PWM; PWM, pulse width modulation.


Another EMI source is communication interfaces. Data delivery inside the macro-block is usually by ribbon cables, with no individual ground returns for signals used.Data shifting occurs simultaneously on numerous tiles. Symmetric line communica-tion lines are expected for much lower emissions, but driver, cable, and load symme-tricity is quickly lost if display is assembled and disassembled frequently. Level ofsymmetricity and cable length (actually loop area) defines the amount of radiation;shape of the loop is not important.206 Cable shielding is expected to reduce EMI,but pigtails in cable shield connection, improper connector construction, integritydamages significantly degrade the shielding performance.

As it was found in Ref. 208, usually LED video display interference is low and lookslike noise (no distinct spikes in a spectrum) 100 MHz and beyond. This can be explainedby dense lines in spectrum (refer Fig. 21.74 to Fig. 21.76) where spacing is defined byimage refresh frequency. Furthermore, as Ref. 208 points out, EMI is often present evenwith no image on the screen. As can be predicted from Figure 21.74,207,209 EMI signif-icantly depends on the video signal and even the colors or movements of the images. Butfor low quality displays situation is different: large spectral spurs occur which can spanup to 800 MHz, can be 5 MHz wide repeating every 20 MHz.208 If spurs are at fixed po-sition and narrow, radio microphones can be tuned away, but if EMI is wideband, it isimpossible to avoid or filter out in radio microphone channel.

21.5 Summary

LED displays have acquired an unbeatable position in entertainment and advertisingindustry. They are almost the only type of displays that indeed dim the pixel and

1k 10k 100k 1M10

20

30

40

50

60

70

80

PWM BPWM GPWM

BPWM

GPWM

PWMC

urre

nt s

pect

rum

(dBu

A)

Frequency (Hz)

Figure 21.79 Power supply current spectrum for image “Lena.” BPWM, binary-weightedPWM; GPWM, gated binary-weighted PWM; PWM, pulse width modulation.


produce pure and efficient color primaries. LCD technology displays and projectorsblind the light, DLP projectors dump the light. Therefore power consumption is thesame even with the dark screen. Plasma and even some sorts of OLED displays pro-duce color by filtering. Filters remove significant portion of backlight luminance andreduce the color gamut size. LED displays offer highest luminance and contrast, cancover extremely large areas, are irreplaceable for façade and stage decoration.

All above comes at significant cost. LED emission wavelength and lifetime dependson current and temperature. While correlation with current can be solved by usingPWM dimming at constant current, temperature effects can be reduced by proper ther-mal design, but usually have to be compensated in software or tolerated. Demand forconstant current drive requires specific drivers, which have to be fast, efficient, small,and cheap. LEDs used in displays demand much tighter tolerances for luminance andcolor repeatability. Same is applicable to LED directivity: directivity diagram must beof certain shape, different in vertical and horizontal direction and mechanical and op-tical axis match is demanded. Louvers are used in order to achieve contrast under highambient light conditions, which further alters display directivity. Aging, variations inluminous output, directivity, need for LED repair or complete tile replacement call fordot and tile corrections. These, together with unavoidable need for gamma correctiondemand high resolution dimming. Adding large numbers of pixels that have to bedimmed creates a bottleneck in data transmission. High resolution PWM createsanother problem: LED drivers have to be fast. Fast rise and fall fronts together withmillions of LEDs switching simultaneously create significant EMI. New LED driversare being offered by semiconductor manufacturers to distribute the pulse energy andpower demand in time. High luminance comes at expense of high power supply cur-rents. Special mains power sequencing is used in order to cope with large inrush cur-rents. Power supply converters’ efficiency and EMC compliance contradict each other.Enclosure design must account numerous issues: modules have to be light, EMI-tight,ingress-protected, accurate, and easy to assemble. If cost restrains are put on top ofthese, LED display task seems impossible. Yet, LED display designers manage todevelop lighter, more efficient, and elegant products.

References

1. Haitz R. Trends in LED display technology. In: Machine design, Proc. 24th electron.Components Conf.; 1974. p. 2e10.

2. LED TV screen. Brief History of the LED television. 2016. Ledtv.net.3. Light-emitting diode display. November 3, 1981. US patent 4298869 A.4. Takahara K, Yamaguchi T, et al. 16-level gray-scale driver architecture and full-color

driving for TFT-LCD. In: Proc. of Int. Display Res. Conf., Japan; 1991. p. 115e8.5. Gillessen K, Malinowski C, Marshall AJ, Schairer W. Survey of interconnection methods

which reduce the number of external connections for LED displays. In: Sid Int. Symp., vol.22; 1980. p. 181e4.

6. Fletcher R. Visible light-emitting diodes in display applications e an old technology gets anew life. LEOS’98, IEEE 1998;1:185e6.


http://Ledtv.net

7. Shubert E. Light emitting diodes. Cambridge University Press; 2003.8. Zukauskas A, Shur MS, Gaska R. Light-emitting diodes: progress in solid-state lighting.

MRS Bulletin 2001;26(10):764e9.9. Svilainis L. Considerations of the driving electronics of LED video display. In: ITI 2007

IEEE Proc., Croatia; 2007. p. 431e6.10. Kurdthongmee W. Design and implementation of an FPGA-based multiple colour LED

display board. Microprocessors and Microsystems 2005;29:327e36.11. The nobel prize in physics Blue LEDs e filling the world with new light. The Royal

Swedish Academy of Sciences; 2014.12. Putman P. When old is new again. Videosystems. 2002. p. 37e43.13. Displays: Barco, color Web, XL video, Daktronics, Dynamax. LEDs Magazine 2006.14. San Siro stadium installs Odeco LiveAd LED display. LEDs Magazine February 14, 2006.15. Ahn JH. Implementation of an LED tile controller for high-quality image display. Displays

2013;34(1):17e26.16. Karim M. Light-emitting diode displays. electrical measurement, signal processing, and

displays. CRC Press; 2003.17. Chen B, Lo S. Black area and fill factor in LED display systems. LEDs Magazine 2004.18. Abramov VS, Puisha AE, Polyakova IP, Tomilin MG, Chuvashov AI. LED modules for

large screens. J Opt Techn 2003;70(7):492e4.19. Electronic display group, large area electronic display industry Report. 2000.20. Global market study on installed and rental outdoor LED displays: driven by China, Asia

Pacific to witness significant growth by 2021. 2015.21. IEC standard 60529 Degrees of protection provided by enclosures (IP Code). International

Electrotechnical Commission (IEC); 1989.22. Vegas LED Screens What do the IP protection grades for LED screens mean?. 2014.23. What is the relationship between the LED screen viewing angle and the LED screen

brightness? Hot Electronics Co., Ltd.24. Vazquez-Molin D, Alvarez Fernandez-Balbuena A, Bernabeu E, Alda J. Louvers design

for LED displays for sunny days. Displays 2007;28:167e73.25. Svilainis L, Dumbrava V. Estimation of the louvers influence on LED video display image

visibility. In: IEEE Proc. International Conference on Information Technology Interfaceslocation: Cavtat, Croatia; 2009. p. 571e6.

26. Ver. 1.03 Information display measurements standard. Society for Information Display,Definitions and Standards Committee, International Committee for Display Metrology;2012.

27. Standard ANSI/INFOCOMM 3M-2011 Projected image system contrast ratio. 2011.28. VESA standard Flat panel display measurements. 2001. Ver. 2.0.29. Standard ISO 9241-9305 Ergonomics of human-system interaction e 305: optical labo-

ratory test methods for electronic visual displays. 2008. ISO/TC 159/SC 4.30. Downen PA. Closer look at flat-panel-display measurement standards and trends. Inf

Display 2006;1(06):16e21.31. Murray S, Caldwell B. Human factors in displays. The measurement, instrumentation, and

sensors handbook. CRC Press; 1999.32. Instrument systems GmbH Handbook of LED metrology. 2000.33. The radiometry of light emitting diodes. UK: Technical guide, Labsphere Inc.; 2001.34. CIE. Commission internationale de l’Eclairage proceedings. Cambridge: Cambridge

University Press; 1931. p. 1932.35. Broadbent A. A critical review of the development of the CIE1931 RGB color-matching

functions. Color Res Appl 2004;29(4):267e72.


36. Commission Internationale de l’�Eclairage: Measurement of LEDs. CIE 127-2007.37. Rizzi A, Simone G, Cordone R. A modified algorithm for perceived contrast measure in

digital images. In: Proc. CGIV; 2008. p. 249e52.38. Zhao P, Pedersen M, Hardeberg JY, Thomas JB. Measuring the relative image contrast of

projection displays. J Imag Sci Technol 2015;59(3):79e91.39. Powell E. Let’s ignore contrast specs. Projector central. 2008.40. Kelley EF, Lindfors M, Penczek J. Display daylight ambient contrast measurement

methods and daylight readability. JSID 2006;14(11):1019e30.41. Soneira R. 16 misleading display specs and what they really mean. Gizmodo. 2012.42. Poynton C, Funt B. Perceptual uniformity in digital image representation and display.

Color Res Appl 2014;39(1):6e15.43. Poynton C. Gamma and its disguises. J Soc Motion Picture TV Eng 1993;102(12):

1099e108.44. Hood D, Finkelstein M. Sensitivity to light. Handbook Perception Hum Perform 1986;

1(5):1e66.45. Ledda P, Santos LP, Chalmers A. A local model of eye adaptation for high dynamic range

images. In: Proc. Afrigraph; 2004. p. 151e60.46. Poynton C. Frequently asked questions about gamma. 1997.47. Pub. No. 15, colorimetry. 3rd ed. Commission Internationale de l’�Eclairage (CIE); 2004.48. CIE S002 Standard on colorimetric observers. Commission Internationale de l’�Eclairage

(CIE); 1986.49. Kimura M, Suzuki D, Sawamura S, Kato M, Nishinohara D. Gamma correction of pulse

width modulation with current uniformization for AM-OLEDs. J Display Technol 2012;8(5):245e9.

50. Recommendation BT.2020-2: parameter values for ultra-high definition television systemsfor production and international programme exchange. Iternational TelecommunicationUnion, ITU; October 2015. p. 9.

51. Mo X, Zhang Y, Wu C. Grayscale model for video LED display. In: Proc. of 4th AsianSymposium on Information Display; 1997. p. 117e8.

52. Recommendation ITU-R BT.2035 A reference viewing environment for evaluation ofHDTV program material or completed programmes. July 2013.

53. Hunt RWG. The reproduction of colour. 6th ed. Chichester UK: WileyeIS&T Series inImaging Science and Technology; 2004.

54. Fraser B. The full gamut photoshop how-to. 1998. p. 51e5.55. Kim M. Optically adjustable display color gamut in time-sequential displays using LED/

Laser light sources. Displays 2006;27(4e5):137e44.56. Sharma G. LCDs versus CRTs color-calibration and gamut considerations. IEEE Proc

2002;90(4):605e22.57. Anderson M, Motta R, Chandrasekar S, Stokes M. Proposal for a standard default color

space for the Internet: sRGB. In: Proc. IS&T/SID; 1996. p. 238e45.58. Shlayan N, Venkat R, Ginobbi P, Singh AK. Energy efficient RGBW pixel configuration

for light-emitting displays. J Display Technol 2009;5(11):418e24.59. Choi JH, Lee M, Kang K, Kim JO. Adaptive color saturation control for low power RGBW

OLED displays. J Display Technol 2016;12(8):784e90.60. Archer JLG. Hits out at OLED critics, promises HDR 4K OLED TVs by Q3. Forbes. 2015.61. Wang L, Tu Y, Chen L, Teunissen K, Heynderickx I. Trade-off between luminance and

color in RGBW displays for mobile-phone usage. In: SID International Symposium digesttechnical papers, vol. 38; 2007. p. 1142e5.

62. Martindale D, Paeth AW. Television color encoding and “Hot” broadcast colors//graphics gems II. Boston: Academic Press; 1991. p. 147e58.


63. Zukauskas A, Vaicekauskas R, Ivanauskas F, Vaitkevi�cius H, Shur MS. Rendering a colorpalette by light-emitting diodes. Appl Phys Lett 2008;93(2). 021109-1e3.

64. Specifications for Nichia red LED model NSPR346BCS. Nichia Corporation; 2002.65. Specifications for Nichia green LED model NSPG346BS. Nichia Corporation; 2002.66. Specifications for Nichia blue LED model NSPB346AS. Nichia Corporation; 2002.67. Recommendation ITU-R BT.709, basic parameter values for the HDTV standard for the

studio and for international programme exchange. Geneva: ITU; 1990.68. Poynton C. Wide-gamut displays. Information display. SID 2007;7:10e5.69. Shur MS, Zukauskas A. Solid-state lighting: toward superior illumination. IEEE Proc

2005;93(10):1691e703.70. Schanda J. Photometry and colorimetry of LED lightning devices. In: PAL symposium,

Darmstadt; 2003. p. 153e60.71. Zukauskas A, Shur MS, Gaska R. Introduction to solid state lighting. Wiley; 2002.72. Application brief AB08 Optical testing for super flux, SnapLED and Luxeon emitters.

Lumileds Lighting, LLC; 2002.73. RazerThin LEDs CxxxRT290eS0200. Cree, Inc.; 2006.74. TA CLD-AP38 REV 1D. Cree LED color mixing: basics and background. 2010.75. Schanda J, Muray K, Kr�anicz B. LED colorimetry. In: Proc. SPIE 9th Congress of the

International Colour Association, vol. 4421; 2002. p. 781e4.76. Wen S, Wu R. Unique gray method for visually measuring the white point of CRT dis-

plays. Displays 2004;25(4):135e40.77. Sharma G, Wu W, Dalal E. The CIEDE2000 color-difference formula: implementation

notes, supplementary test data and mathematical observations. Color research and app.2004. p. 1e24.

78. Hooi LB. Understand RGB LED mixing ratios to realize optimal color in signs and dis-plays. LEDs Magazine June 2013.

79. US patent. US 7030893 B2 Method for driving full-color LED display board. 2006.80. Real pixel and virtual pixel. RGB Industries Co., Ltd.; 2006.81. Ginesu G, Massidda F, Giusto DD. A multi-factors approach for image quality assessment

based on a human visual system model. Signal Processing: Image Commun 2006;21(4):316e33.

82. Mannos JL, Sakrison DJ. The effects of a visual fidelity criterion on the encoding ofimages. IEEE Trans Inf Theor 1974;20(4):525e35.

83. Thomas CW, Gilmore GC, Royer FL. Models of contrast sensitivity in human vision. IEEETrans Syst Man Cybernetics 1993;23(3):857e64.

84. Zhang XH, LinWS, Xue P. Just-noticeable distortion estimation for image pixels. In: IEEE7th Workshop on Multimedia Sig. Proc.; 2005. p. 1e4.

85. Rushmeier H, Ward G, Piatko S, Sanders P, BL R. Comparing real and synthetic images:some ideas about metrics. Eurographics’95 1995:82e91.

86. Gaddipati A, Machiraju R, Yagel R. Steering image generation with wawelet basedperceptual metric. Eurographics’97 1997;16(3):C241e51.

87. Russ JC. The image processing handbook. CRC Press; 2006.88. Watson A, Barlow HB, Robson JG. What does the eye see best? Nature 1983;302:

419e22.89. Design of effective cine theaters SMPTE engineering guideline EG 18-1994. 1994.90. Recommended guidelines for presentation quality and theater performance for indoor

theaters. Lucasfilm Ltd.; 2000.91. Svilainis L, Dumbrava V. LED forward current errors induced in video display tile’s PCB.

In: Proc ITI; 2009. p. 565e70.


92. Potter MC, Wyble B, Hagmann CE, McCourt ES. Detecting meaning in RSVP at 13 msper picture. Atten Percept Psychophys 2014;76:270e9.

93. Kime S, Galluppi F, Lagorce X, Benosman RB, Lorenceau J. Psychophysical assessmentof perceptual performance with varying display frame rates. J Display Technol 2016;12(11):1372e82.

94. Wendler B. Refresh rate. Technical note 42. Daktronics Inc.; 2002.95. LEDs given a longer life. Sealing Technol 2008;2008(5):1.96. Hong E, Narendran N. A method for projecting useful life of LED lighting systems. SPIE

Proc Int Conf Solid State Lighting 2004;5187:93e9.97. Application Brief I-018 Projected long term HTOL light output degradation of precision

optical performance AlInGaP LEDs. Agilent Technologies, Inc.; 1999.98. Application Brief I-004 Reliability of precision optical performance AlInGaP LED lamps

in traffic signals and variable message signs. Agilent Technologies, Inc.; 1999.99. Auto-insertion of option 002 tape and reel LED lamps AB 74. Agilent Technologies, Inc.;

1999.100. AN 1005 Operational considerations for LED lamps and devices. Agilent Technologies,

Inc.; 1999.101. Hwang N. Failure analysis matrix of light emitting diodes for general lighting applications.

In: Proc. of 15th IPFA Symposium; 2008. p. 1e4.102. Jacob P, Kunz A, Nicoletti G. Reliability and wearout characterisation of LEDs. Micro-

electron Reliab 2006;46:1711e4.103. Outdoor LED panels. LVP2056 model data sheet. Lighthouse Inc.; 2017.104. Outdoor LED displays. NV: Barco; 2017.105. LED video displays. Daktronics, Inc.; 2017.106. Large scale video displays. Diamond vision display systems. Mitsubishi Electric Power

producs Inc.107. Freyssinier JP, Zhou Y, Ramamurthy V, Bierman A, Bullough JD, Narendran N. Eval-

uation of light-emitting diodes for signage applications. In: Proc. Third InternationalConference on solid state lighting, of SPIE, vol. 5187; 2004. p. 309e17.

108. Guan JZ, Zhu X, Lan P, Qian X. The application of LED in the information display oftransportation. Adv Display 2005;53(4):20e3.

109. Svilainis L. Large scale LED video display data flow management. In: Proc ITI; 2010.p. 657e62.

110. Morimasa M, Kazumi T, Shinji F. Large-scale color LED stock-information display board.Natl Tech Rep 1987;33(1):102e7.

111. Pachuca P, Borras R. Microcontroller-based LED drivers: topologies and trade-offs. LEDsMagazine October 2005.

112. AN29503.1 Constant-current LED drivers. Allegro MicroSystems, Inc.; 2007.113. Nguyen F. Challenges in the design of a RGB LED display for indoor applications. Synth

Met 2001;122:215e9.114. Svilainis L. LED brightness control for video display application. Displays 2008;29(5):

506e11.115. Configurable controller for high-resolution LED display systems. J Display Technol 2016;

12(12):1594e601.116. MBI5043 16-channel SPWM constant current LED driver with lower ghosting effect

elimination. Macroblock Inc.; 2015.117. Svilainis L. LED PWM dimming linearity investigation. Displays 2008;29(3):243e9.118. Liu M, Jia C, Chen Z. A single driver chip for gray scale passive matrix OLED. In: Proc. of

4th Int. Conf. on communications, circuits and systems; 2006. p. 2193e6.


119. Mo X, Zhang Y. Consecutive PWM driving video LED display system. In: Proc. of Int.Symposium on circuits and systems; 1997. p. 1437e9.

120. Feng Y, Xu-fei Q, Lin-bo Z. Control system and control method for automatic adjustmentof outdoor LED display brightness. Ubiquitous Comput Appl Wireless Sensor: LectureNotes Electr Eng 2015;331:613e9.

121. Winder S. Power supplies for LED drivers. Elsevier Inc.; 2008.122. Specifications for Nichia blue LED model NSPB546DS. Nichia Corporation; 2008.123. Specifications for Nichia green LED model NSPG546DS. Nichia Corporation; 2008.124. Specifications for Nichia green LED model NSPR546CS. Nichia Corporation; 2008.125. Manninen P, Orrevetelainen P. On spectral and thermal behaviors of AlGaInP light-

emitting diodes under pulse-width modulation. Appl Physics Letters 2007;91(18):181121-1e181121-3.

126. Howell W. An overview of the electronic drive techniques for intensity control and colourmixing of low voltage light sources such as LEDs and LEPs. AN 011. Artistic LicenceLtd.; 2002.

127. Ahn JH, Park S, Son K, Park M, Moon M, Kim H, Hwang E, Kim W, Kim U, Kim K.Probability-based PWM generator for large-size LED display system. In: Proc. IMECS P;2011. p. 480e4.

128. Vivid Effect provides eye-catching LED frontage for Austrian bank. LEDs MagazineSeptember 2006.

129. MBI5168, 8-bit constant current LED sink driver, datasheet VA02. Taiwan: MacroblockInc.; 2005.

130. TB62727FN, 16-bit constant current LED module with built-in dot-current correction foreach output. Japan: Toshiba Corporation; 2005.

131. A6278 and A6279, serial-input constant-current latched LED drivers with open LEDdetection 6278-DS. Allegro MicroSystems Inc.; 2006.

132. Rev 1 MAX6969, 16-port, 5.5V constant-current LED driver 19-3677. CA, USA: MaximIntegrated Products; August 2005.

133. Ref. 5, ST microelectronics STP16C596, 16-Bit, constant current LED sink driver. 2004.134. Mean well RSP-320 series 320W single output with PFC function. 2016.135. Texas instruments TLC5943, 16-channel, 16-bit PWM led driver with 7-bit global

brightness control. 2007.136. MBI5031, 16-channel PWM-embedded LED driver, datasheet V2. Taiwan: Macroblock

Inc.; 2006.137. Data sheet SLVS515 TLC5940, 16-channel LED driver with dot correction and grayscale

PWM control. Texas Instruments Inc.; 2004.138. LDT logic device technology LD71D1032 32-output LED driver/10-bit PWM controller.

2002.139. Maxim integrated products 24-Output PWM LED drivers for message boards. 2006.140. DM163 8x3-channel constant current LED driver. Silicon touch technology Inc.; 2005.141. Hu C, Zhang Y, Wu X, Peng FZ. A multi-channel DC/DC LED driver with inductor-less

series-parallel auto-regulated rectifier. In: Proc APEC; 2013. p. 829e34.142. Data sheet LT3474 step-down 1A LED driver. Linear Technology Corporation; 2005.143. Data sheet DSFPHV9910HV9910 universal high brightness LED driver. Supertex Inc.; 2004.144. LT3476-High current quad output LED driver. Linear Technology Corporation; 2006.

p. 1e14.145. LED driver application note. StarChips Technology Inc.; 2006.146. Application note 1141 Adding global intensity control to the MAX6956 and MAX6957

static LED drivers. 2007. Maxim integrated products.


147. B€ocher A. Constant current LED drivers TB62725A/726A from Toshiba. EBV Elektronik;March 2003. p. 15e6. MIP Edition.

148. Application note Constant current LED driver MBI5001/5016. Macroblock Inc.; 2003.149. Why does LED need current gain adjustment function. Taiwan: Macroblock Inc.; 2007.150. Low dropout, two-bank LED driver with PWM brightness control. USA: Texas Instr. Inc.;

2008.151. Descombres A, Guggenbuhl W. Large signal circuit model for LED’s used in optical

communication. IEEE Trans Electron Devices 1981;ED-28:395e404.152. Hino I, Iwamoto K. LED pulse response analysis considering the distributed CR constant

in the peripheral junction. IEEE Trans Electron Devices 1979;ED-26:1238e42.153. Uhle M. The influence of source impedance on the electrooptical switching behavior of

LED’s. IEEE Trans Electron Devices 1976;ED-23(4):438e41.154. Brailovsky AB, Mitin VV. Fast switching of light-emitting diodes. Solid-state Electronics

2000;44:713e8.155. Descombres A, Guggenbuhl W. Investigation of the influence of several diode parameters

on the light-delay time in large-area SH-junction LED’s. IEEE Trans Electron Devices1978;25(3):379e82.

156. Tessler N, Pinner DJ, Friend RH. Semiconductor device model applied to electricallypulsed polymer LEDs. Synth Met 2000;111e112:269e72.

157. Hassine RL, Fave JL, Bouchriha H. Study of the transient EL slow rise in single layerOLEDs. Org Electron 2006;7:1e7.

158. Veledar O, Danaher S, Allen JIH, Byrne PO, Thompson LF. Review and development ofnanosecond pulse generation for light emitting diodes. In: Proc. 17th International Sci-entific Conference, Mittweida; 2005. p. 3e6.

159. Kapustinsky J, DeVries R, DiGiacomo NJ, Sondheim WE, Sunier JW. A fast timing lightpulser for scintillation detectors. Nucl Instrum Methods Phys Res A 1985;241(2e3):612e3.

160. Veledar O, Byrne PO, Danaher S, Allen JIH, Thompson LF, McMillan JE. Simpletechniques for generating nanosecond blue light pulses from light emitting diodes. MeasSci Technol 2007;18:131e7.

161. Halbritter H, J€ager C, Weber R, Schwind M, M€ollmer F. High-speed LED driver for ns-pulse switching of high-current LEDs. IEEE Photon Technol Lett 2014;26(18):1871e3.

162. Chaziachmetovas A. Driver topology influence on LED luminescence response dynamics.Elektronika Ir Elektrotechnika 2014;20(5):104e7.

163. Young RLED. Measurement instrumentation. CIE advanced LED seminar. 2001.164. Gaertner AA. LED measurement issues. Canada NRC: Institute for National measurement

standards; April 2002.165. Fiorentin P, Scroccaro A. The metrological characterization of a goniometer for the

analysis of surface properties. In: IMTC 2006, Sorrento, Italy; 2006. p. 542e7.166. Burmen M, Pernu�s F, Likar B. Automated optical quality inspection of light emitting

diodes. Meas Sci Technol 2006;17:1372e8.167. Wendler BLED. Viewing angles vs. LED video display viewing angles. Technical note 40.

Daktronics, Inc.; 2002.168. Svilainis L, Dumbrava V. LED far field pattern approximation performance study. In:

Proc. ITI; 2007. p. 645e9.169. Svilainis L, Dumbrava V. Numerical comparison of LED directivity approximation

functions for video displays. Displays 2010;31(4e5):196e204.170. Svilainis L, Dumbrava V. LED goniometry: the feasibility study. Matavimai 2005;35(3):

35e40.


171. Triantaphillidou S, Jacobson RE. Measurements of the modulation transfer function ofimage displays. J Imag Sci Technol 2004;48(1):58e65.

172. Feng Y, €Ostberg O, Lindstr€om B. MTFA as a measure for computer display screen imagequality. Displays 1990;11(4):186e92.

173. Chena LM, Hasegawa S. Visual resolution limits for colour matrix displays. Displays1992;13(4):179e86.

174. Lin MW, Hsieh PH, Chang EC, Chen YC. Flicker-glare and visual-comfort assessments oflight emitting diode billboards II: color display. Appl Optics 2015;54(13):4089e96.

175. Moreno I, Contreras U. Color distribution from multicolor LED arrays. Opt Express 2007;15(6):3607e18.

176. Cassarly WJ. LED modelling: Pros and Cons of common methods, photon, Tech BriefsIIa-2a, special supplement to NASA Tech Briefs. April 2002.

177. Yang H, Bergmans JWM, Schenk TCW, Linnartz JMG, Rietman R. Uniform illuminationrendering using an array of LEDs: a signal processing perspective. IEEE Trans SignalProcess 2009;57(3):1044e57.

178. Yang H, Bergmans J, Schenk T, Linnartz JP, Rietman R. An analytical model for theilluminance distribution of a power LED. Opt Express 2008;16(26):21641e6.

179. Moreno I, Sun C. Modeling the radiation pattern of LEDs. Opt Express 2008;16(3):1808e19.

180. Manninen P, Hovila J, K€arh€a P, Ikonen E. Method for analysing luminous intensity oflight-emitting diodes. Meas Sci Technol 2007;18:223e9.

181. Deller CA, Franklin JB, Smith GB. Lighting simulations using smoothed LED profilescompared with measured profiles. SPIE 2006;6337. 63370X.1e63370X.12.

182. Harbers G, Bierhuizen SJ, Krames MR. Performance of high power light emitting diodesin display illumination applications. J Display Technol 2007;3:98e109.

183. Svilainis L. LED directivity measurement in-situ. Measurement 2008;41(6):647e54.184. Moreno I, Tsai CY, Berm�udez D, Sun CC. Simple function for intensity distribution from

LEDs. In: Proc. Spie, vol. 6670; 2007. 66700H1e7.185. Barton D. Radar technology encyclopedia. Norwood: Artech House Inc.; 1998.186. Moreno I. Image-like illumination with LED arrays: design. Opt Letters 2012;37(5):

839e41.187. Application note. OSRAM opto semiconductors GmbH Reliability and lifetime of LEDs.

2013. p. 14.188. Vazquez M, Nunez N, Nogueira E, Borreguero A. Degradation of AlInGaP red LEDs

under drive current and temperature accelerated life tests. Microelectronics Reliability2010;50(9e11):1559e62.

189. Narendran N, Deng L. Performance characteristics of light emitting diodes. In: Proc.IESNA Annual Conference Technical papers; 2002. p. 157e64.

190. Muthu S, Schuurmans FJ, Pashley MD. Red, green and blue LEDs for white light illu-mination. IEEE J Sel Top Quan Electron 2002;8(2):333e8.

191. Muthu S, Schuurmans FJ, Pashley MD. Red, green, and blue LED based white lightgeneration: issues and control. Proc Ind Appl Conf 2002;1:327e33.

192. Keeping S. Determining LED rated life: a tricky challenge. Electronic products. Digi-KeyArticle Library; 2011.

193. Taylor J. Industry alliance proposes standard definition for LED life. LEDs Magazine2005:9e11.

194. Chang MH, Das D, Varde PV, Pecht M. Light emitting diodes reliability review. Mi-croelectronics Reliability 2012;52:762e82.


195. Yanagisawa T, Kojima T. Degradation of InGaN blue light-emitting diodes undercontinuous and low-speed pulse operations. Microelectron Reliab 2003;43(6):977e80.

196. Liu J, TamWS, Wong H, Filip V. Temperature-dependent light-emitting characteristics ofInGaN/GaN diodes. Microelectron Reliab 2009;49(1):38e41.

197. He GX, Zheng LH. A model for LED spectra at different drive currents. Chin OpticsLetters 2010;8(11):1090e4.

198. Meneghini M, Meneghesso G, Zanoni E. Electrical Properties, Reliability Issues, and ESDRobustness of InGaN-Based LEDs. III-nitride based light emitting diodes and applica-tions. Top Appl Phys 2013;126:197e229.

199. Burmen M, Pernus F, Likar B. LED light sources: a survey of quality-affecting factors andmethods for their assessment. Meas Sci Technol 2008;19(12):122002.

200. DenBaars SP, Feezell D, Kelchner K, Pimputkar S, Pan CC, Yen CC, Tanaka S, Zhao Y,Pfaff N, Farrell R, Iza M, Keller S, Mishra U, Speck JS, Nakamura S. Development ofgallium-nitride-based light-emitting diodes (LEDs) and laser diodes for energy-efficientlighting and displays. Acta Materialia 2013;61(3):945e51.

201. Thilwind D. LED it Be! Read it first before you buy LED screen. Screens Magazine 2007;5.202. Cadena R. Video matters. RF interference from LED screens is a big problem. Julius

Media; 2013.203. Wenyu X, Zhiliang C. A mixed-signal driver chip for 65K-color passive-matrix OLED. In:

Proc. Asicon, vol. 1; 2005. p. 473e7.204. CISPR 32:2015 Electromagnetic compatibility of multimedia equipment e emission re-

quirements. International Electrotechnical Commission; 2015. p. 209.205. CISPR 35:2016 Electromagnetic compatibility of multimedia equipment e immunity re-

quirements. International Electrotechnical Commission; 2016. p. 165.206. Ott HW. Electromagnetic compatibility engineering. Wiley; 2009.207. Song C, Wang RG, Feng YQ. Potential EMI analysis of the LED display panel driving

current. Guangdianzi Jiguang. J Optoelectron Laser 2013;24(6):1059e64.208. Milne A. Unplugging the mystery behind (and inside) video wall RF interference. RF

Venue. 2017.209. Svilainis L. Comparison of the EMI performance of LED PWM dimming techniques for

LED video display application. J Display Technol 2012;8(3):162e5.


LEDs for projectors 22Linas SvilainisKaunas University of Technology, Kaunas, Lithuania

22.1 Introduction

Projector explores the excellent idea: project the image on large surface, using a pointlight source. 1 Small and relatively simple device can be used to provide large-scaleimages. This idea (camera obscura or pinhole camera) dates back through centuries, itis believed that this principle inspired paleolithic cave paintings and could have beenused in building the neolithic structures. Archimedes and Leonardo Da Vinci used theidea in their works. The oldest published drawing is by Gemma Frisius in 1545. Mod-ern principle is pretty the same: a light is projected through electronically modulatedoptical device (best known is liquid crystal display [LCD]) into a lens, which projectsthe image on a large surface. Pattern provided by optical modulator will result in pro-jected image. Most of projectors throw an image on a surface or dedicated projectionscreen2,3 in such way, modulating the light flow by means of LCD, liquid crystal onsilicon (LCoS), silicon X-tal reflective display (SXRD), or digital micromirror device(DMD).4 New types of projectors use the beam steering by means of microelectrome-chanical systems or Bragg cell to construct the vector or raster image directly. Some ofnew projectors can project the image on the windshield of the car or on head-updisplay (HUD)5 or head-mounted display (MHD), overlaying road, or other environ-ment information directly. 6 In a retinal projector, an image is delivered directly on theretina. Xenon and ultrahigh performance (UHP) lamps,7 LEDs, and lasers are used asa light source. LED light source-based projector can deliver 3000 lm, UHP can deliverup to 12,000 lm,8 xenon 33,000 lm, and laser 60,000 lm.9 Color primaries are pro-vided either by a color wheel or by using individual light sources. LED and laserare the winning technologies in color purity (gamut size). These technologies alsooffer lower cost, size, and weight. Projector displays are portable, they offer highlyscalable image size, can provide image on large area on almost any surface, andhave small installation costs.

With all the developments given above for projectors, the future looks bright.With education being the key driver, accounting nearly 37% according to Ref. 10,low-cost, portable projectors find their way into small- and medium-sized enter-prises, government, and become affordable for middle-class consumers. Cinema,stage structured lighting, and entertainment are significant drivers too. New playeris Pico or pocket projector. Although fixed installation projectors dominated themarket in 2015, picoprojector part was expected to take the maximum share bythe end of 2020 in Ref. 10. Key players in this signet are 3M, Microvision, Optoma


http://dx.doi.org/10.1016/B978-0-08-101942-9.00022-8

Technology, Syndiant, and Texas Instruments. According to the study, Marketsand-Markets2 Pico projector market value was USD 1.28 billion in 2015 and expected togrow at a compound annual growth rate (CAGR) of 14.52% between 2016 and 2022.Another newcomer with even more impressive, or 27%, growth is the interactive pro-jector. Total projector markets grow for the period until 2020 was expected to bemore than 8% CAGR in Technavio report.10The key vendors identified are Canon,Epson, JVC, and Sony.

This chapter discusses projection display operation and components.

22.2 Projector technologies

A projector can be separated into three parts: the illumination, spatial modulation, andprojection systems.

Technologies of the projector displays can be subdivided into reflective and trans-missive by a principle how spatial modulation is performed (Fig. 22.1). Although othermodulation types11,12 can be used, today LCD is the main technology used in trans-missive modulation. Reflective technologies can be further subdivided by how imageis formed. It can use (1) one-dimensional pixel in conjunction with mechanical rasteror vector scan using galvanometer, MEMS (microelectromechanical systems) mirror13

or Bragg cell,14 light source usually is a laser; (2) two-dimensional pixel line inconjunction with mechanical scan, laser source can be laser or LED; or (3) three-dimensional modulation, using LCoS, SXRD, D-ILA (direct-drive image light ampli-fier), or digital micromirror device.

Another subdivision of technologies can be based on projector light source type. Itcan be xenon or UHP lamp.7 New sources are LED and laser.

Projector technologies also differ how color primaries are produced: either by colormultiplexing (rotating color filter, see Fig. 22.2) or by individual RGB channels(Fig. 22.3).

One of the most important components is optics. It includes polarizing filters (incase of LCD or LCoS), dichroic mirrors or filters, mirrors, prisms, and projectionlenses.

Reflective modulator (LCOS, DMD)

Polarisers

Projection lens

Screen

Transmissive modulator (LCD)

Light source

Polarisers Projection lens

Screen

Figure 22.1 Reflective (left) ant transmissive (right) projector technologies comparison.


Power supply is not the last important projector component with the constant de-mand for size, weight, and cost reduction. In the mains powered projectors, it is analternating current/direct current converter, which must provide the best efficiencyin order to relax the temperature restrains. In pico projector development, it mustalso include the battery.

Control electronics usually include the image processor and control subsystems. Inthe case of mechanical scanning electronics is responsible for light modulation andmechanical scanning subsystem control and synchronization.

A projector screen is a surface that is dedicated or at least useful for projecting animage. Actually, it is a part of the image quality defining chain. Screen can be operatedin front or rear projection so selection of screen type depends on this choice. Contrastof the rear-projection screen is higher than on a front projection. A perfectly diffusing,100% reflecting, Lambertian distribution screen gain is 1.0. If the screen reflectivitydiffuses, the light into a �60 degrees horizontal angle and �10 degrees vertical angle,such screen will have a gain of 6.1 If multiple projectors are used, then problem in im-ages stitching can occur if high gain screen is used.

DMD

White lightsource

Projection lens

Screen

Rotating filter/color wheel

Red Green BlueMechanical multiplexing:

DMDRed

Projection lens

Screen

Green lightsource

Blue

Dichroic mirror Dichroic mirror

Red Green BlueElectronic multiplexing:

Figure 22.2 Two principles of color multiplexing: color wheel (left) or separate RGB (right).

R

Dichroic mirror

White lightsource

Dichroic mirror

Dichroicmirror

Dichroic mirror

B

G

Dichroicprism

Mirror

Mirror Mirror

LCD

LCD

LCDOutput Output

R

B

G S

White light source

DMD

DMD

DMD

Figure 22.3 Individual RGB channels use individual modulation for each RGB primary: case oftransmissive (left) and reflective (right) modulator.

LEDs for projectors 739

22.2.1 Light source

Xenon and UHP lamp7 are the most common light source in projectors, metal-halideand electrodeless lamps are rarely used.1 A xenon arc lamp (Fig. 22.4) is a gasdischarge lamp. Light is produced by passing the current through the ionized xenongas. Lamp produces quite a uniform spectrum, very near the optimal D65 point(5900e6200K) similar to that of natural light.15 There are few peaks at near infrared(850e900 nm, 10% of the total emission) and substantial UV radiation portion.

Special construction, very high pressure (up 3040 kPa), and metals composition arerequired to improve efficiency and handle high temperatures, pressure, and differentexpansion coefficients. Special power supply is needed: it has to provide high voltageat start-up; later voltage has to be reduced. For instance, Ushio UXW-15KD 15 kWlamp is operated 37.5 V at 400 A and has a 1000 h lifetime. Lamp is water cooledand cooling circuit for each electrode is required. Xenon illuminated projectors canreach 33,000 lm,9 but lamp typical lifetime is only 800 h.16

Because of their short lifetime and replacement complexity, xenon lamps are usedonly in high-end systems. Ultrahigh performance (UHP) lamp7 was introduced by Phi-lips in 1995. It is also an arc lamp, using even higher, beyond 20 MPa pressure, filledwith mercury. Although UHP lamps have a longer lifetime (Philips claim 10,000 h),spectra of UPH lamps in nonuniform (see Fig. 22.4). Both xenon and UHP are large,expensive to replace,17 and large arc length make light projection and collectioninefficient.16

3000

20

40

60

80

100

400 500 600 700 800Wavelength (nm)

UHP

Xenon

Rel

ativ

e in

tens

ity (%

)

Figure 22.4 Ushio UXW-15KD 15 kW water cooled xenon lamp (left) and Xenon lampemission spectrum comparison to ultrahigh performance (UHP) (right).Left: Copyright 2017 Ushio America, Inc.


Lamp is a white color source, dichroic mirrors or color wheel is used for RGB pri-maries production (Fig. 22.5).

Filtering effect of dichroic mirrors is moderate: not a narrowband, but a largeportion of spectrum is passed through (refer Fig. 22.6).

The effect of such filtering on xenon and UHP lamps can be seen in Fig. 22.7. ITURecommendation BT.202018 is a compromise between Adobe 1998 and wide gamut.RGB primaries correspond to monochromatic light source with 630, 532, and 467 nm

Red dichroicmirror

Blue dichroicmirror

LCDLCD

Light source

Dichroic mirror“wavelength selector”

Dichroic combiner cube

MirrorMirror

Lens

Figure 22.5 Use of dichroic mirrors for RGB primaries production from white light source.

Wavelength (nm)

Tran

smitt

ance

(%)

4200

20

40

60

80

100

460 500 540 580 620 660 700

G B

R

RGB

Figure 22.6 Transmittance example of the dichroic mirrors.


wavelength accordingly. RGB sources in Fig. 22.7 are by no means wideband, so colorreproduction performance will suffer.19 Sometimes, in order not to waste the yellowpart of spectrum, RGBY color scheme is used.20

LED is essentially an inexpensive, small, unidirectional, low-cost light source withsignificantly longer life time. It can provide narrowband light emission, resulting inlarger color gamut of the display and higher light source efficiency. Refer Fig. 22.8for LED and lamps color gamut (derived from Ref. 15) comparison.

Xenon (cermax) color spectra at D65 High pressure mercury color spectra at D65

400 450 500 550 600 650 700Wavelength (nm)

400 450 500 550 600 650 700Wavelength (nm)

010002000300040005000600070008000

010002000300040005000600070008000

Inte

nsity

(au)

Inte

nsity

(au)

BlueRedGreenWhite

Blue

RedGreen

White

Figure 22.7 RGB primaries production result for xenon (left) and ultrahigh performance (right)lamps.Reprinted from Yeralan S, Doughty D, Blondia R, Hamburger R. Advantages of using high-pressure, short arc xenon lamps for display systems. Projection displays XI. Proc SPIE 2005;5740:27e35 with permission.

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9 ITU R2020 LED Xenon UHP

y, a

.u.

x, a.u

Figure 22.8 Color gamut of light-emitting diodes, xenon, and ultrahigh performance lampscomparison.


Unfortunately, LED luminance significantly lags the UHP and xenon lamps. Pico-projectors21 are intended for small image size, so even 25 lm (Magnasonic LEDPocket Pico Video Projector) e 50 lm (Lenovo Pocket Projector) brightness couldbe enough. Publication4 presents the development of handheld picoprojector. An illu-mination system was solved with a single freeform lens, using two freeform surfaces,which guarantee small-angle incidence on LCoS. It was demonstrated that single2 mm � 2 mm LED chip was enough to illuminate the system. Over 95% image uni-formity has been achieved with the whole optical engine 60 mm� 28 mm � 16 mm.Compact LED projector presented in Ref. 22 uses RGB LED in the illumination sys-tem, which also has a collimator lens group and a mirror with a color filter and a lensarray integrator instead of an integration rod. Thanks to total internal reflection prismand projection lens, the whole optical engine is smaller (total length of 52 mm), hashigher contrast, and 82% screen uniformity.

Yet, higher luminous output is needed for larger image projectors. This is usuallysolved by using LED array23,24 or even micropixel array.25 Such LEDs summation in-creases the etendue of the system and limits its collecting efficiency.16 Homogeniza-tion is done using integrator, which has to be long (Fig. 22.9 left), yet recentdevelopments propose to use microlens array (Fig. 22.9 right) instead so light enginesize can be reduced.26

Larger LED projectors generate satisfying results: LED projector�1000 offered byCRE can supply 2800 lm at FullHD resolution using LCD engine; Casio XJ-A240 canachieve 2500 lm using laser and LED hybrid. LED does not have gain medium andemission not stimulated by cavity resonance, so only spontaneous emission prevails.Furthermore, light output is not very directional (though better than lap); area of lightgeneration is large so it is difficult to collect the light by the optical system.9 Due toreasons above higher brightness LED projectors are not possible, therefore, LEDsare used only if lower luminance is required (current limit 2800 lm). Although blueand red emissions are better, low green efficacy is limiting LED application in projec-tors. For étendue limitation reasons LED should light the emission surface tomaximum 2 � 2 mm.1 This means that surface area or chips number increase arecomplicated. Color and lumen output repeatability limit LED application in arrays.Furthermore, because of relatively low chip temperatures cooling of LED is less effi-cient as UHP and xenon lamps, but LED color and luminous output are remarkablyaffected by temperature. Still, LED has a lot of attractive features, which push itsdevelopment for projectors: it can be switched very fast, so it is an attractive candidate

Condenser lensLED/Laser

Optical integrator/beam homogeniser

Expansion lens

Input beam profile Output beam profile

Microlens array

Input beam profile Output beam profile

Fourier lens

Homogenisation planeMicrolens array

Figure 22.9 Use of optical integrator in order to achieve the illumination uniformity: using longlight guide (left) or using microlens arrays (right).


in multiplexed color systems. It is relatively cheap and can be outsourced by a thirdparty. It is light and small, relatively simple to drive and to control. It is narrowbandand does not produce glare.

Development of laser technologies allowed this technology for projectors.16,17,27,28

Initially, laser projectors were based on the beam scanning (raster or vector) technol-ogy. With powers of available lasers growing and optics getting better, modern laserprojectors are full frame displays, where no scanning is used or just one-dimensionalscanning is required.29 Laser addresses the request for miniaturization and efficiency.One of the laser benefits is their pure color emissions, which enable very large colorgamut.17 Another benefit is the very low intrinsic étendue of the source.17 Laser pro-jectors are autofocus, that is, no projection lenses are required and so no intermediateimage planes exist. This corresponds to less complex, more efficient, smaller yet morepowerful optical systems. The above is true both for scanning and full frame, two-dimensional light valve projectors.

If direct laser source is used, then RGB primaries are generated by red, green, andblue lasers. Impressive 60,000 lm can be achieved.9 Three-dimensional image can begenerated by using RGB primaries offset by small fraction.30 Using Six Primary 3D(6p 3D, Fig. 22.10) or wavelength multiplexing technology (WMT), there is no

0

0,5

1

Inte

nsity

(a.u

.)

0

0,5

1

Inte

nsity

(a.u

.)In

tens

ity (a

.u.)

0%

50%

100%

400 450 500 550 600 650 700Wavelength (nm)

400 450 500 550 600 650 700Wavelength (nm)

400 450 500 550 600 650 700Wavelength (nm)

Figure 22.10 Thanks to extremely narrowband emission of laser, left and right eye images canbe produced using slightly offset RGB primaries: topdprojector emission, center-googlefilters, bottomduser output; solid linedleft eye, dasheddright eye image.30

By courtesy of INFITEC GmbH.


need for three-dimensional frames multiplexing: these can be separated by wearingglasses, having the corresponding bandpass filters.9 Although used with lamps in Dol-by3D it provides better results when used in conjunction with lasers.

Today, direct laser source projectors are expensive because cost grows proportion-ally with lumens output. In the case of laser phosphor light production (hybrid technol-ogy), only the less expensive blue lasers are used. The red and green primaries arederived by exciting the yellow phosphor (laser wheel) with blue laser.9,31 Broadbandyellow emitted by phosphor is split into red and green by color filters. A phosphorwheel32 includes a rotating disk with a wavelength converting layer. Phosphor wheelconstruction (size) determines max allowable optical pump power.31 Phosphor wheelis rotated to improve the thermal performance. This technique is more cost effective(beyond 4000 lm), though color purity is reduced.

BenQ hybrid laser projectors use a blue laser bank, phosphor wheel, and dual colorwheel to extract red and green components, but also keep yellow, resulting in RGBYcolor. One DMD light valve chip is used for providing 2000 lm from 270 W mains.

Casio laser-hybrid design is adding red LED to the colors produced by phosphorwheel and combine everything into one DMD chip, offering 3000 lm from 165Wmains input in case of slim model and 5000 lm from 4K Ultra HD model.

Sony introduced Z-Phosphor 3 LCD hybrid laser source projectors in 2015. It alsouses 3 LCD light valve chips, VPL-FHZ700L is capable of 7000 lm, WUXGA(1920 � 1200) resolution. Blue laser is used in conjunction with transmissive phos-phor wheel and dichroic mirrors.

Panasonic uses both three (RGB primaries, Fig. 22.11) and one (RGBY colors)DMD chips in their hybrid laser projectors, achieving 65,000 lm WUXGA with1DMD model PT-RZ670 and 31,000 lm SXGAþ with 3DMD model PT-RS30K.

Epson contributed to high brightness projectors with 3LCD laser projector in 2016.It has 25,000 lm 4K resolution and 20,000 h of maintenance free by using inorganic3LCD panels, a laser bank light source and inorganic phosphor wheel (Fig. 22.12).

LensDMD chips Blue laser

3-Chip DLPTM laser projection system

Phosphor wheelColor filter prism

Figure 22.11 3DMD model PT-RS30K from Panasonic achieves 31 000 lm SXGAþ.Courtesy of Panasonic.


Epson LS10000 uses a hybrid design with two blue lasers: one for blue light, the otherto energize a yellow phosphor, which splits the blue light into red and green.

Christie HS Series and Christie DHD850-GS 6900 lm Full HD laser phosphor pro-jector use blue and red laser diodes to boost the amount of red primary and is using oneDMD chip. Christie GS, Captiva, HS Series, and the CP2208-LP cinema use blue laserdiodes and a phosphor wheel to create yellow, whereas blue laser light passes througha diffusion segment in the phosphor wheel (Fig. 22.13).

Direct laser technology is used in Christie D4KLH60, Christie Mirage 4KLH, andChristie CP42LH and the projection head is separate and connects to the laser modulevia a fiber optic link. Up to 60,000 lm, 4K resolution, 120 Hz frame rate, 30 000 h to80% threshold are achieved. Christie 6p 3D laser systems use two projection heads andtwo sets of RGB laser primaries.

Quantum dots (QD) are the promising technology to derive color primaries.33 QDsoffer saturated colors because their structures are dominated by quantum size effect,which provides narrowband emissions. Emission wavelength can be controlled bynanocrystal size or composition.33

Figure 22.12 Epson’s 3LCD laser projector L1000 structure.

Blue laser

Phosphor wheel Color wheel

Figure 22.13 Structure of color primaries derivation in Christie laser phosphor projectors.Copyright of Seiko Epson Corp.


Coherence of the laser photons allows to achieve monochromaticity and to use mul-tiple lasers or even laser arrays to increase the light output.9 Unfortunately, coherencealso creates the laser light interference with a screen surface, creating speckle. Speckleand beam shaping are the major issues with laser projectors. Speckle noise severity ischaracterized by the speckle contrast34:

C ¼ sI

eI; (22.1)

where sI is the standard deviation of the intensity of the uniform image projected, I isthe mean. Speckle is most severe when C ¼ 1 and C ¼ 0 means no speckle. It has beenestablished that speckle is not noticeable if C is below 3%.

Time- and space-varying independent speckle pattern is the most efficient inspeckle reduction. Integration time of the human eye and projector should be consid-ered. Full-frame projector pixel is illuminated for 16.68 ms in the case of 60 Hz framerate. This time it is 8.68 ms for a 1920 � 1080 one-dimensional scan projector, and8.03 ns in the case of two-dimensional scan. In combination with human visionintegration time (30e60 ms), this time has to be multiplied by successive frames aver-aged (two to four).34 Implementation of moving parts is usually used; other solutionslike integrated optical components, hybrid micro-opto-electro-mechanical systemsdevices, and active phase modulators, noncoherent laser sources of different wave-lengths, laser emission chirping, variation of polarization, and illumination beamwobbling have been mentioned in Refs. 34,35. A compact solution for both issuesis presented in Ref. 36 by using a diffractive beam shaper in illumination optics anda vibrating motor attached to the shaper. The beam shaper is a double-sided microlensarray with a lateral shift. The resulting illumination pattern does not have zero-orderdiffraction. Laser speckle is further reduced thanks to vibration induced on thebeam shaper. Measured uniformity and speckle contrast were 78% and 5.5% corre-spondingly. Laser pico projector with a low speckle was proposed in Ref. 22 usingtwo diffusers and a voice coil motor oscillator. Vibrating diffuser concept was pro-posed in Ref. 37. Speckle reduction by laser beam modulation using ultrasonic wavesin a liquid cell was proposed in Ref. 38. Binary micromirror array (BMMA) is pro-posed for speckle reduction in Ref. 29. BMMA mirrors have two states, motionlessmirrors or fully deflected mirrors, giving the phase modulation depth p radians. About5% speckle contrast ratio was achieved. It must be noted that laser phosphor technol-ogy provides despeckling naturally because the phosphor emission is noncoherent andwideband. Another technique is to use angular diversity, mixing the laser light in asmany directions as possible to fill the light cone transmitted through the lens.9 Ifdifferent techniques are combined together larger speckle reduction is achieved.

22.2.2 Image generationdspatial modulation

Spatial modulation is responsible for image formation on screen. It can be provided intwo ways: (1) using mechanical 1D (raster/vector scan) or 2D (line scan); (2) usingfull-frame projection. In both cases, pixels have to be modulated in intensity (dimmed).


Usually it is a laser beam that is scanned in case (1), but focused LED beam can alsobe used. Scanning can be done using galvanometer, MEMS mirror,13 or Bragg cell.14

Pico projectors use a two-dimensional raster scan (flying spot)34,39: a single spot isprojected sequentially on screen in two dimensions with spot intensity modulated byvideo signal to create a picture (Fig. 22.14).

Although being the oldest type of laser projectors, these systems recently got aboost from in the ultracompact image projection systems.39 Appearance of thecompact and competitive cost laser sources provided a breakthrough of this field.Availability of red (640e660 nm) and blue laser diodes (440e470 nm) and recentdevelopments of frequency doubled green offer super video graphics array, 800 x600 resolution.28 This type of display does not require the projection lens and isauto-focused at all distances. Therefore, the system can be very compact and lowcost, which is important in pocket or mobile projector market.

Beam deflection can be accomplished by step motor, piezo-motor,40 galvanometer,polygonal mirrored drum rotating at high speed, a resonant galvanometer mirror, ornonresonant mirrors, MEMS micromiror, and cantilever (Ref. 13; Fig. 22.15).

Some designs use two mirrors or drums, whereas others use a single two-axisdeflection gimbaled MEMS mirror (Fig. 22.16), capable of two-dimensional tilts.

Interesting alternative that can be used to scan the beam is the acousto-opticalmodulator or a so-called Bragg cell.1,14 It uses the acousto-optic effect to diffractthe light using sound waves (usually at hundredths of MHz). A piezoelectric trans-ducer is attached to the optically transparent material or material itself can possessthe piezoelectric properties (e.g., lithium niobate [LiNbO3] quartz). The excitationelectrical signal applied causes the piezoelectric effect and acoustic waves are excitedin transparent material. Standing waves are created. These can be treated as periodiclayers of the material with different index of refraction. Incoming light is scatteredoff such structure causing the interference similar to Bragg diffraction (Fig. 22.17).

Red laser diode

Blue laser diode

Green laser(frequency doubled)

MEMS scanner

Figure 22.14 Two-dimensional raster scan using two galvanometers (left) and micro-electromechanical systems (MEMS) (right).Left: reprinted with permission from krazerlasers.com. Right: reprinted from Chellappan KV,Erden E, Urey H. Laser-based displays: a review. Appl Opt 2010;49(25):F79e98 withpermission from MicroVision.


Laserbeam

Centilever length = CL

DL

y

PSD

X

Z

Nc→

θ

φ

ξ

Figure 22.15 Microcantilever can be used for laser beam deflection.Reprinted with permission from Beaulieu LY, Godin M, Laroche O, Tabard-Cossa V, Grutter P.A complete analysis of the laser beam deflection systems used in cantilever-based systems.Ultramicroscopy 2007;107:422e30.

Aperture angle40°

2D micro mirrorØ 1mmmech. ±10°

PCUSBDigitalcontrolFPGA

2D voltageadaption

Laser driver& modulation



Collimator

Collim

ator

Collimator

Laserdiode (blue)

Laserdiode (green)

Laserdiode (red)

Beamdivider

Beamdivider

Projection module

Figure 22.16 Use of single two-dimensioanl deflection axis gimbaled micromirror.Reprinted with permission from Scholles M, Br€auer A, Frommhagen K, Gerwig C, Lakner H,Schenk H, Schwarzenberg M. Ultracompact laser projection systems based on two-dimensionalresonant microscanning mirrors. J Micro/Nanolith MEMS MOEMS 2008;7(2):021001-1-11.


Diffraction angle q depends on the wavelength of the light l ratio to the wavelengthof the sound L

sinðqÞ ¼ ml

L; (22.2)

in the Bragg regime and

sinðqÞ ¼ m$n$l0L

; (22.3)

where m ¼., �2, �1, �0, þ1, þ2,. is the diffraction order and n is the refractiveindex of the media. Angle of deflection can be varied by changing the excitationfrequency and can reach tens of mrad. Laser beam scanning using an acousto-opticaldeflector has been presented in Ref. 41. It was demonstrated that fast laser beamscanning can be achieved using acoustic waves to deflect the laser beam by varying theacoustic frequency. There are no moving parts and the scan frequency is higher than inthe case of the conventional approaches. Frequency can be varied on the nanoscalescale so high scanning rates can be achieved.

A one-dimensional or line scan display projects a column of pixels simultaneouslyand scans them across the screen to produce the image. Compact and low-cost projec-tion can be obtained using LED array combination with MEMS scanning is displayswith one-dimensional axis scanning. Such devices offer light output sufficient forHUD (head up display) and even HMD (head mounted display) applications. Gratinglight valve-based laser projector is described in Ref. 29. One-dimensional DLP, LCD,or LCoS light valves can be used for light modulation.

Full-frame projectors provide all pixels on-screen simultaneously frame after frameusing arrays of DLP, LCD or LCoS light valves.

( )

~

Transparent media

Ultrasonictransducer

Variable frequencyAC signal source

Deflected beam

Incoming laser beam

Absorber

Λ

Λθ

Figure 22.17 Laser beam can be diffracted in acoustically excited media.


The simplest full-frame projection system is using LCD light valves. Because oflowest cost this type of projectors is the most common and affordable. Its most com-mon issue is pixelation though recent developments have reduced effect significantly.

The most commercially successful micro-light valve is the DMD (Fig. 22.18). Itoffers affordable solution for the large screen displays, providing high-quality images.Device is an addressable array of MEMS micromirrors that can tilt into two distinctpositions þ10 or �10 degrees (new designs �12 degrees) in response to drivingvoltage provided by the memory cell below. Tilt þ12 degrees corresponds to “on”position e incident light is reflected back. Tilt of �12 degrees corresponds to “off”position e incident light is reflected away, into light dump. Position can be switchedin about 15 ms. Grayscale is achieved using PWM or BPWMmodulation.42 Pixel pitchcan be down to 5.4 mm.1

Projector is compact and lightweight, does not require polarization optics that areused in LCD. In its simplest configuration (Fig. 22.19) it uses just one DMD chipand color is provided by color wheel. Note, that in order to reduce the color switchingeffects light is focused on color wheel. However, such setup provides color blurring orrainbow effect when sight is moved across the screen. Light output is reduced becauseof multiplexing. Systems with 3 DMD chips do not have these problems, since RGBprimaries are displayed simultaneously. Also blanking might be required to avoid co-lor transition influence.

DMD light valve chips are preferred in high brightness laser projectors since thesehave better immunity against high power.17 Yet, DLP could not reach the contrast andresolution of state-of-the art LCoS chips. On the other hand, LCoS switching speed issufficient for three chip architectures, but not satisfies single chip projectors. DLP canbe used both in single and three chip setups. Single chip setup is using color multiplex-ing, therefore have lower brightness and color breakup might be perceived due to thecolor sequential approach.

Lens CCDsensor

CCDpixel

“off” state

“static” state

“on” state

Incident lightDMDmicromirror

12°

–12°

–48°–24°24°

Mirror–10 degMirror+10 deg

Hinge

CMPoxide

Metal 3Yoke Spring tip

CMOSsubstrate

Figure 22.18 DMD light valve technology: micromirror construction (left) and on/offreflections (right).Left: reprinted with permission from Hornbeck LJ. From cathode rays to digital micromirrors:a history of electronics projection display technology. Tex Instrum Tech J 1998;15:7e46.Courtesy Texas Instruments. Right: reprinted from Feng W, Zhang FM, Qu XH, Zheng SW.Per-pixel coded exposure for high-speed and high-resolution imaging using a digitalmicromirror device camera. Sensors 2016;16(3):331.


Liquid crystal on silicon (LCoS, Fig. 22.20) is a try to miniaturize the LCD engine:it is a reflective active-matrix LCD on silicon. An LCoS projector combines DLP andLCD features, it was invented by JVC in the late 1990s (as D-ILA, Direct Drive ImageLight Amplifier[57]). The increasing adoption of LCoS technology in the projectorssegment is expected to result in this market’s impressive growth rate of more than25% during the period of 2014 and 2019.

DMD

Color wheel

Light source

Optics

Figure 22.19 Simplest DMD 1 chip and color wheel system structure.Reprinted with permission from The Digital Micromirror Device A Historic MechanicalEngineering Landmark. DLP0350. Texas Instruments Incorporated; 2008, courtesy TexasInstruments.

Cover glassIndium tin oxide

Alignment layer

Liquid-crystal material

Spacers

CMOS substrate

Figure 22.20 Liquid crystal on silicon light valve technology: pixel construction.Reprinted by permission from Macmillan Publishers Ltd: Nature Photonics Jepsen ML. Atechnology rollercoaster: liquid crystal on silicon. Nat Photon 2007;1:276e7, copyright 2007.


LCoS uses similar to LCD units, except that light is reflected back from a mirrorbehind the pixel. Both LCoS and LCD share same advantage over DMD: pixel inten-sity is regulated by LCD cell voltage so no PWM is needed. Yet, this does not neces-sarily mean intensity stability and repeatability.

SXRD (Silicon X-tal Reflective Display) is Sony’s LCoS version D-ILA and HD-ILA are proprietary LCoS technologies from JVC.

High-temperature-polysilicon (HTPS) liquid-crystal-display panel technology isone of the most suitable microdisplay devices for projection displays because drivingcircuits that include high-mobility transistors can be formed on the TFT substrate.43

Epson version is also using microlens arrays on top of each of the HTPS pixels to in-crease the brightness of a projected image by minimizing the shadowing effect of inter-connection on top (Fig. 22.21).

Novel idea of image production was reported in Ref. 25. Flip-chip InGaN micro-pixelated LED arrays with high pixel density have been proposed as both illumina-tion and image-forming device. Resolution of the device presented is not high, just64 � 64 elements, with 50 mm pitch; each of which have a 20 mm emissionaperture and are matrix-addressable. This could be new idea of LED applicationin microdisplays.

22.2.3 Optics

Optics includes polarizing filters (in case of LCD or LCoS), dichroic mirrors or filters,mirrors, prisms and projection lenses. Projection system defines the performance of thewhole projector. Fortunately, recent developments in optics software allow predictingthe performance and carefully selecting the most optimal design. Wide presentation onprojection displays optics can be found in Ref. 1.

Illumination system is using a light integrator to homogenize the light sourceoutput. Rod integrator redistributes the light flow spatially using numerous reflections,providing uniform, flattop distribution.26 It has to be long because the number ofinternal reflections define the output uniformity but is directly proportional to thelength. For this reason integrator is inconvenient is small projectors. In order to geta compact homogenizer26 proposed to replace the integrator by a square-arrangeddouble-side microlens array with an extremely high fill-factor for high collection effi-ciency. Homogenizer was integrated with the LED light source and a projection

Polariser Polariser

PolariserPolariser

Black matrix(contrast)

Black matrix(contrast)

Incoming lightIncoming light

Output light Output light

Microlens

Figure 22.21 Light throughput improvement in Epson version of HTPS: top structure does notpass all of the incident light (left) while microlenses increase the amount of light passing (right).


engine, resulting in a pocket-sized projector with 35 lumens brightness small size(75 mm � 67 mm � 42 mm) total assembly.

LED luminous output is lower than that of other light sources. Then maximum LEDoutput utilization is important. Use of compound parabolic concentrators (CPC) is pre-sented in Ref. 23. CPC for multiple LED array package was designed. It was capable tocollect more than 90% of the LED array light output and transmit within the designedangle.

Proper design of the color combiner is important for total volume reduction. Novelcolor combiner was described in Ref. 44. It contains red, green, and blue laser diodesand a thermally expanded core fiber waveguide. Structure is simple resulting in the3.84 cm3 total volume.

Electrically tunable liquid crystal (LC) lens (Fig. 22.22) for pico-projectors wasdescribed in Ref. 45.

The focal length of this pico-projector is electrically tunable from 350 to 14 cm andthe tunable range is even wider than that of a manually focused pico-projector.Reported response times of turn-on and turn-off are 313 and 880 ms.

22.3 Applications

Largest projection displays use is in education: schools and universities widelyadopted this type of whiteboard alternative for its flexibility, cost and convenience.53

Projection displays are widely used in all types of venues: conferences, governmentmeetings and presentations, shows and concerts have found this type of display conve-nient for its large image size, ease of setup and maintenance and ever-increasingbrightness. Stage illumination and video rendering is unthinkable without theprojectors.49e52 Opening and closing ceremonies of the Rio 2016 Olympic Games

S’

Observation plane

LED

Relay lens

Pre-polarizer

d

Projectionlens

LC lensLCOSpanel

PBS

Figure 22.22 Electrically tunable liquid crystal lens application in pico-projector.Reprinted from Lin HC, Lin YH. An electrically tunable focusing pico-projector adopting aliquid crystal lens. Jpn J Appl Phys 2010;49(10):102502. Copyright 2010 The Japan Society ofApplied Physics.


in Brazil demonstrated how creative art can be created using large scale projection dis-plays. All architectural temporary art on buildings is using innovative ideas of projec-tions on surface. Non-contact measurements use the structured light to measure the 3Dgeometry of complex objects: patterned illumination48 is easy to project. Navigationsystems use small projectors to present maps directly of street surface or wall; HUDpresents the navigation instructions for driver directly on windshield of the car(Fig. 22.23); medical navigation systems project internal image on a patient skin;head-worn displays coupled with the human eye relay the information directly tothe eye (Fig. 22.24).5,6

All environment simulation, training and entertainment applications consider pro-jection displays as a must-have device17: only projection displays can project the im-age on complex surfaces (Fig. 22.25). Large areas, like ice-hockey field or Olympicstadium can be filled with live image. Entertainment and advertising get use of inter-active displays46 (Fig. 22.26).

Interactive Touch Projector allows users to manipulate content on displays at a dis-tance, even the objects in a room by manipulating objects by touch and drag in the livevideo. Result can also be overlayed on the remote display.46

Occupant

Foldmirror

Display

Asphericmirror

Windshield

Virtual imageat bumper distance

Figure 22.23 Head-up display is used in car for graphical information relay for driver.Left: projection on dedicated screen (Reprinted with permission from Exploride), right: directprojection on windshield (Reprinted from the March 2005 edition of Laser Focus WorldCopyright 2017 by PennWell).

Figure 22.24 Head-worn displays are coupled with the human eye. From left to right: 1) Pick-by-Vision solution xPick from Intel, developed by Ubimax (Reprinted with permission ReconInstruments, an Intel company); 2) Helmet Mounted Display (HMD) in Gen III helmet for F-35pilot (© Rockwell Collins); 3) Q-Warrior HMD. Reprinted with permission from BAE systems.


Projection finds its use in additive manufacturing as a structured light source, usedto selectively cure three-dimensional shapes with high resolution and speed.47 Pico,micro, and pocket projectors3,4,37 change the way how graphical information isexchanged and handled (Fig. 22.27). Projection technologies go into smaller dimen-sions, offering light processing capabilities at nanoscale.59

Figure 22.25 Projection displays are used in environment simulation (Flight simulator 180degree projection screen, picture courtesy of Pixelwix.Inc).

Figure 22.26 Interactive projection displays used in entertainment (left) and education (right).Reprinted with permission from TouchMagix Media Pvt. Ltd. 2017.

Figure 22.27 Miniature projectors can be used to project the image directly on skin for furtherinteraction. Note that projector is placed inside the CICRET bracelet.Reprinted with permission, courtesy of CN2P.


22.4 Summary

Projection display is established on the market as an easy to setup, portable, low-costdevice for graphical information presentation on large area. They are capable of anysize image, projectors brightness is increasing with the adoption of new laser lightsources. Use of light-emitting diodes (LEDs) and lasers in small-scale projectorsresulted in significant cost, size, and power consumption reduction. New solid-statelight sources increase the lifetime of projection displays, reduce the cost of ownership,relax the servicing burden, and inspire new applications.

References

1. Brennesholtz MS, Stupp EH. Projection displays. John Wiley & Sons; 2008.2. Pico projector market by type (embedded, non-embedded), technology (DLP, LCoS, LBS),

specification (brightness, resolution, power source), application (consumer electronics,business & education), components, and geography e Global forecast to 2022. Market-sandmarkets, SE 3125 2016.

3. Yan CJ, Liu X, Li HF, Xia XX, Lu HX, Zheng WT. Color three-dimensional display withomnidirectional view based on a light-emitting diode projector. Appl Optics 2009;48(22):4490e5.

4. Chen EG, Liu P, Yu FH. Optimization design of single freeform lens based illuminationsystem for CF-LCoS handheld pico-projectors. Optik Int J Light Electron Opt 2013;124(14):1912e6.

5. Holmgren DE, Robinett W. Scanned laser displays for head mounted displays. InstrumentDes 1990:1e13.

6. Cakmakci O, Rolland J. Head-worn displays: a review. J Display Technology 2006;2(3):199e216.

7. Derra G, Moench H, Fischer E, Giese H, Hechtfischer U, Heusler G, Koerber A,Niemann U, Noertemann FC, Pekarski P, Pollmann-Retsch J, Ritz A, Weichmann U. UHPlamp systems for projection applications. J Phys D 2005;38:2995e3010.

8. The brightest LCD digital projector. Christie Digital Systems, Inc.; 2017.9. White paper Laser-phosphor illumination in projectors. Barco NV; 2016.10. SKU: IRTNTR9248 Global light projector market 2016e2020. Infiniti Research Ltd.;

2016.11. Electro-optic modulators. RP photonics encyclopedia. RP Photonics Consulting GmbH;

2017.12. Ruan JA. Transmission and reflection dual operational mode MEMS display device.

Transmission and reflection dual operational mode MEMS display device. Displays 2015;37:11e8.

13. Beaulieu LY, Godin M, Laroche O, Tabard-Cossa V, Grutter P. A complete analysis of thelaser beam deflection systems used in cantilever-based systems.Ultramicroscopy 2007;107:422e30.


14. Maak P, Jakab L, Barocsi A, Richter PI. Recent developments and results in 2D acousto-optic light deflection. In: Proc. SPIE 3388, Advances in optical information processingVIII, vol. 48; 1998. p. 48e59.

15. Yeralan S, Doughty D, Blondia R, Hamburger R. Advantages of using high-pressure, shortarc xenon lamps for display systems. Projection displays XI. Proc SPIE 2005;5740:27e35.

16. Madamopoulos N, Papageorgiou F. Laser based projection system for displays, SignageIllumination. J Display Technol 2014;10(10):832e9.

17. Janssens P, Malfait K. Future prospects of high-end laser projectors. In: Emerging LiquidCrystal Technologies IV, 72320Y, vol. 7232; 2009. p. 1e12.

18. Recommendation BT. 2020e2: Parameter values for ultra-high definition televisionsystems for production and international programme exchange. International Telecommu-nication Union, ITU, (10/2015) 9 p.

19. Thomas JB, Bakke AM. A colorimetric study of spatial uniformity in projection displays.In: CCIW, vol. 5646. LNCS; 2009. p. 160e9.

20. Hong SJ, Kwon OK. An RGB to RGBY color conversion algorithm for liquid crystaldisplay using RGW pixel with two-field sequential driving method. J Optical Society Korea2014;18(6):777e82.

21. Darmon D, McNeil JR, Handschy MA. LED-illuminated pico projector architectures. SID2008;39(1e3):1070e3 [Micropatterned single lens];a Kim JJ, Chae S, Jeong KH. Micropatterned single lens for wide-angle light-emitting di-odes. Opt Lett 2010;35(6):823e5.

22. Sun WS, Tien CL, Ma CH, Pan JW. Compact LED projector design with high uniformityand efficiency. Appl Opt 2014;53(29):H227e32.

23. Peng CH, Li XN, Wang JW, Xiong LL, Zhang YX, Liu XS. A high power light emittingdiode module for projection display application. In: Proc. ICEPT-HDP; 2010. p. 1412e6.

24. Bierhuizen S, Weijers G. Improved color point stability of high brightness LED arrays forprojectors. In: Proc. International Display Workshop, vols. 1e3; 2007. p. 2357e60.

25. Gong Z, Gu E, Jin SR, Massoubre D, Guilhabert B, Zhang HX, Dawson MD, Poher V,Kennedy GT, French PMW, Neil MAA. Efficient flip-chip InGaN micro-pixellated light-emitting diode arrays: promising candidates for micro-displays and colour conversion.J Phys D-Appl Phys 2008;41(9):094002.

26. Pan JW, Wang CM, Lan HC, Sun WS, Chang JY. Homogenized LED-illumination usingmicrolens arrays for a pocket-sized projector. Opt Express 2007;15(17):10483e91.

27. Chellappan KV, Erden E, Urey H. Laser-based displays: a review. Appl Opt 2010;49(25):F79e98.

28. Steegm€uller U, K€uhnelt M, Singer F, Schwarz T, Albrecht T, Lutgen S, Reill W, Luft J,Brick P. High speed green frequency converted semiconductor laser for projection displays.Proc SID 2005;36(1):1608e9.

29. Tonga ZM, Chen X. Principle, design and fabrication of a passive binary micro-mirrorarray(BMMA) for speckle reduction in grating light valve (GLV) based laser projection displays.Sensors and Actuators a 2014;210:209e16.

30. Rohwer KC, Jorke H, Simon A. Stereoscopic laser display system using six primary colors.In: Proc. Projection Summit, vol. 60.5; 2013. p. 1e4.

31. Daniels M, Mehl O, Hartwig U. Laser-activated remote phosphor light engine for projectionapplications. In: Proc. SPIE XVI, vol. 9578; 2015. 95780N.

32. Dai JH, Hu TC, Lin TC, Lin TC. Phosphor wheel and projector having the phosphor wheel.US Patent US9503700. 2015.


33. Wang P, Zhang Y, Ruan C, Su L, Cui H, Yu WW. A few key technologies of quantum dotlight emitting diodes for display. IEEE J Selected Top Quan Electron 2017;23:1.

34. Akram MN, Chen X. Speckle reduction methods in laser-based picture projectors. Opt Rev2016;23(1):108e20.

35. Riechert F. Speckle reduction in projection systems. Dissertation. Universit€at Karlsruhe(TH); 2009.

36. Yao PH, Chen CH, Chen CH. Low speckle laser illuminated projection system with avibrating diffractive beam shaper. Opt Express 2012;20(15):16552e66.

37. Pan J-W, Shih C-H. Speckle reduction and maintaining contrast in a LASER pico-projectorusing a vibrating symmetric diffuser. Opt Express 2014;22(6):6464e77.

38. Wang L, Tschudi T, Boeddinghaus M, Elbert A, Halldorsson T, Petursson P. Specklereduction in laser projections with ultrasonic waves. Opt Eng 2000;39(6):1659e64.

39. Scholles M, Br€auer A, Frommhagen K, Gerwig C, Lakner H, Schenk H, Schwarzenberg M.Ultracompact laser projection systems based on two-dimensional resonant microscanningmirrors. J Micro/Nanolith MEMS MOEMS 2008;7(2). 021001-1-11.

40. Juska V, Svilainis L, Dumbrava V. Analysis of piezomotor driver for laser beam deflection.J Vibroengineering 2009;11(1):17e26.

41. Huang Z, Munro N, H€uhmer AF, Landers JP. Acousto-optical deflection-based laser beamscanning for fluorescence detection on multichannel electrophoretic microchips. Anal Chem1999;71(23):5309e14.

42. Ruckmongathan TN. Intensity modulation of light sources for gray scales in projectiondisplays. SID DIGEST 2012;43(1):591e4.

43. Tanaka T, Ito A, Yasukawa M. New technology for HTPS-LCD panels for projectionsystems. Soc Inf Display 2007;15(10):825e8.

44. Shin IH, Lee JJ, Kang HS. Novel color combiner composed of red, green, and blue laserdiodes and a thermally expanded core fiber waveguide. Opt Eng 2011;50(9):094005.

45. Lin HC, Lin YH. An electrically tunable focusing pico-projector adopting a liquid crystallens. Jpn J Appl Phys 2010;49(10):102502.

46. Boring S, Baur D, Butz A, Gustafson S, Baudisch P. Touch projector: mobile interactionthrough video. In: Proc. CHI2010 Human factors in computing systems, vols. 1e4; 2010.p. 2287e96.

47. Cheng YL, Kao HL. Study on visible-light-curable polycarprolactone and poly(ethyleneglycol) diacrylate for LCD-projected maskless additive manufacturing system. In: Proc.SPIE 9564, light manipulating organic Materials and devices II, vol. 95640K; 2015.p. 1e5.

48. Vairavan R, Retnasamy V, Shahimin MM, Sauli Z, Leng LS, Norhaimi WMW,Marimuthu R, Abdullah O, Kirtsaeng S. 3D mapping of breast surface using digital fringeprojection. SPIE 2017;10043(1004315):1e7.

49. Majumder A, He Z, Towles H, Welch G. Achieving color uniformity across multi-projectordisplays. In: Proc. Visualisation; 2000. p. 117e24.

50. Liu S, Ruan Q, Li X. The color calibration across multi-projector display. J Signal InfProcess 2011;2:53e8.

51. May BB, Cahill ND, Rosen MR. Calibration of a multi-projector system for display on acylindrical surface. In: Image processing workshop (WNYIPW); 2010. p. 6e9.

52. Sun Y, Dai S, Ren C, Chen N. Computer vision based geometric calibration in curved multi-projector displays. Proc ICCSIT 2010;6:342e8.


53. Brennesholtz MS. The evolution of projection displays. Part I: from mechanical scanners tomicrodisplays. In: Society for information display; 2015.

54. FengW, Zhang FM, Qu XH, Zheng SW. Per-pixel coded exposure for high-speed and high-resolution imaging using a digital micromirror device camera. Sensors 2016;16(3):331.

55. JepsenML. A technology rollercoaster: liquid crystal on silicon. Nat Photon 2007;1:276e7.56. Hornbeck LJ. From cathode rays to digital micromirrors: a history of electronics projection

display technology. Tex Instrum Tech J 1998;15:7e46.57. Bleha W, Sterling R. D-ILA technology for high-resolution projection displays. Proc. SPIE

5080, Cockpit Displays X 2003:239e49.58. The Digital Micromirror Device A Historic Mechanical Engineering Landmark. DLP0350.

Texas Instruments Incorporated; 2008.59. Kress BC, Meyrueis P. Applied digital optics: from micro-optics to nanophotonics. John

Wiley & Sons; 2009.


Index

‘Note: Page numbers followed by “f” indicate figures, “t” indicate tables.’

AABC model, 300ABC+f(n) model, 307Absolute band gaps, 339Absolute quantum yield, 125Absorbed photons, 125AC. See Alternated current (AC)Active cooling, 523ACU. See Angular color uniformity (ACU)Adaptive driving beam (ADB), 650Adaptive forward-lighting systems (AFS),

650ADB. See Adaptive driving beam (ADB)Adhesion and/or barrier layer, 211Adlayer-enhanced lateral diffusion

(AELD), 4AELD. See Adlayer-enhanced lateral

diffusion (AELD)AFM. See Atomic force microscopy (AFM)AFS. See Adaptive forward-lighting systems

(AFS)Agro-industrial application, 531ALD. See Atomic layer deposition (ALD)AlGaN. See Aluminum gallium nitride

(AlGaN)AlInGaNalloy system, 403e404quantum barriers, 307

AlInGaP, 418e419composition, 387LEDs, 720red LED, 416e417, 720TFLED, 225

AlInN, 597AlInN/(In)GaN system, 597All-optical switches, 601Alliances, 547e549AlN. See Aluminum nitride (AlN)Alternated current (AC), 591

Alternative crystallographic orientations,597e598, 598f

a-Alumina (a-Al2O3). See Syntheticsapphire (Al2O3)

Aluminates, 134e140Aluminum arsenide (AlAs), 86e87Aluminum gallium nitride (AlGaN), 387,

568e569AlGaN-based DUV-LEDs fabricated on

high-quality AlN, 568e574, 568fAlGaN/GaN buffer layer, 103e106alloy system, 402for DUV light sources, 554e555EL spectraof 225 nm-band AlGaN-QW withvarious quantum well thickness,574f

for various injection currents, 570e571,571f

NCs, 14UV-LEDs, 554, 554f

Aluminum nitride (AlN), 44, 402e403,568e569

AlGaN-based DUV-LEDs fabricated onhigh-quality, 568e574

“AlN-deposited” sapphire, 54atomic orientation on Si, 84fbuffer layer, 87e88epilayers, 8LT-buffer, 55e56NRs, 269

Ambient illuminance, 667Ammonia (NH3), 44, 253e254, 263e264,

286, 556, 558MBE, 3e4

Ammonothermal growth, 258Ammonothermal method, 286Angular color uniformity (ACU), 500,

534e535

Angular dependence. See Directionaldependence

Angular-resolved spectra, 233e234,236e238

Annular emission cone, 334fANSI contrast ratio, 667ANSI/INFOCOMM standard 3M-2011, 665Ant transmissive projector technologies,

738, 738fApatite structure phosphors, 174e176APT. See Atom probe tomography (APT)Arrhenius equation, 442e443Artefacts, 341Atom probe tomography (APT), 276Atomic force microscopy (AFM), 6e7, 7f,

90, 250, 560, 561fAtomic layer deposition (ALD), 90e93, 92fAuger coefficients, 307Auger mechanisms, 457e458Auger recombination, 275e276, 302e305,

305f, 381e382bandstructure of GaInN/GaN LED under

positive bias, 303fintegrated current of high energy Auger

peaks, 304fAutomotive headlamps, 647Automotive ICs, 703Automotive lighting, 543, 543f

applications, 647energy and environmental issues, 655forward lighting, 647e651human factor issues with LEDs, 652e654signal lighting, 651e652

Avago ASMT-QTC0e0AA02, 426Average index effect, 372e373Azimuthal anisotropy of far-field

distribution, 236(Ba,Sr)2Si5N8:Eu

2+ phosphor, 128e129(Ba,Sr)3SiO5:Eu

2+ phosphors, 129e131

BBa2SiO4 phosphor, 129e131Backlight unit (BLU), 619, 620t

advantages and history of LED BLUs,626e629

classification, 621e623direct-type structure, 621, 622fedge-type structure, 621

folded-mixing-LGP-type structure, 623hollow-type structure, 623

requirements for LCD BLUs, 625BAM. See Bit-angle modulation (BAM)BaMgAl10e2xO17:xMn4+, xMg2+ phosphor

(BMA), 164e165, 166fBaMgAl10O17 phosphor, 137e138Bare LED chip, 491Basal plane stacking faults (BPSFs), 285,

288, 315e317Base substrate material, 328BaSiF6:Mn4+ phosphor, 132e133“Bathtub” shape, 719Batwing lenses, 495Beam deflection, 748Beam-shaping, 363far-field, 370e371

Beam-steering effects, 361e363BEF. See Brightness enhancement film

(BEF)BenQ hybrid laser projectors, 745Beryllium oxide (BeO), 43Biaxial plane stress, 273Bimolecular transition state path, 29Binary micromirror array (BMMA), 747Binary pulse width modulation (BPWM),

696Biscyclopentadienyl magnesium (CP2Mg),

253e254, 263e264Bit-angle modulation (BAM), 696Bloch modes, 337e339BLU. See Backlight unit (BLU)Blue and green MQW, 32e35Blue GaInN/GaN LEDs, device structure

and performance, 315, 316tBlue GaN LED structure, 231Blue MQW, 32e35Blue-light-excited yellow-emitting Sr3SiO5:

Ce3+, Li+ phosphors, 129e131BMA. See BaMgAl10e2xO17:xMn4+,

xMg2+ phosphor (BMA)BMMA. See Binary micromirror array

(BMMA)BN. See Boron nitride (BN)Boltzmann transport equation (BTE), 456BTE-derived models, 461

Borates, 154e157Boron nitride (BN), 44

762 Index

Boron phosphide (BP), 87BoseeEinstein condensate, 405Bottom-up method, 244Bottom-up technique for GaN nanopillar

substrates, 263e269. See also Top-down technique for nanostructuredLED

air voids formation, 264fC-LEDs, NP-LEDs, and high-resolution

TEM image of region I, 266fEL peak wavelength and LeIeV

characteristics of two fabricated LEDs,267f

GaN NPs template, 263fRaman spectrum for GaN epilayer

overgrown, 266frelative IQE, 268fsurface morphology of overgrown GaN

NPs template, 265ftwo-dimensional FDTD and normalized

light output power, 268fBoundary condition, 353BP. See Boron phosphide (BP)BPSFs.SeeBasal plane stacking faults (BPSFs)BPWM. See Binary pulse width modulation

(BPWM)Bragg cell, 738, 748Bragg reflectors, 227Bragg’s diffraction law, 229Brightness enhancement film (BEF), 621BTE. See Boltzmann transport equation

(BTE)Buffer layer, 46, 54e55grown by PVD, 59strategies

AlAs, 86e87ALD of Al2O3, 90e93, 92fAlN buffer layer, 87e88other materials, 86e87SL structures, 88e90ZnO, 86e87

Bulk GaN substrates, 315

CC-DLTS. See Capacitance deep level

transient spectroscopy (C-DLTS)C-FCLED. See Conventional FCLED

(C-FCLED)

c-GaN NCs. See Cubic GaN NCs (c-GaNNCs)

C-LED. See Conventional-LED (C-LED)c-plane blue GaInN/GaN LEDs, 315C4. See Controlled collapse chip connection

(C4)(Ca,Sr)AlSiN3:Eu

2+ phosphor, 128e129Ca3Sc2Si3O12 phosphor, 129e131, 136Ca6BaP4O17 phosphor, 152Ca6BaP4O17:Ce3+ phosphor, 152e154Ca6BaP4O17:Eu

2+ phosphor, 152Ca6BaP4O17:Eu2+ phosphor, 152e154CaAl12O19:Mn4+ phosphor, 137e138CaAlSiN3:Eu

2+ phosphor, 128e129photoluminescence excitation, 128f

CAD tools. See Computer-aided designtools (CAD tools)

CAGR. See Compound annual growth rate(CAGR)

Capacitance deep level transientspectroscopy (C-DLTS), 476e478

Capture time, 462many-body formulation in InGaN/GaN

QWs, 462e467Carbonconcentration, 35e39, 38fin AlN, 35

incorporation, 31e32, 31fCarrierescape in QW structure, 306ffrequency amplitude, 678leakage, 305e308experimental EQE, 308frecombination processes, 306f

transport problems in multiple-quantum-well LEDs, 274e275

Casio laser-hybrid design, 745Catastrophic degradation, 450Catastrophic failure modes, 467e468Cathode ray tube (CRT), 619Cathodoluminescence (CL), 114e117, 177,

565e566, 565f, 590CaY2Al4SiO12 phosphor, 136CaY2Al4SiO12:Ce

3+ phosphor, 136CBL. See Current blocking layer (CBL)CCD. See Charge coupled device (CCD)CCT. See Correlated color temperature

(CCT)

Index 763

CDM. See Charged device model (CDM)Ce3+, ET models with, 169e183,

175fe176fCIE chromaticity coordinate diagram, 173f,

177fcolor variation with Tb3+ concentration

change, 174fenergy level model, 182fexcitation and emission spectra of

Sr3Al2O5Cl2, 178fphotoluminescent spectra of Ca4Y6(SiO4)

6O, 180fCENELEC. See European Committee for

Electrotechnical Standardization(CENELEC)

Certification of white LEDs, 549CFF. See Critical flicker fusion frequency

(CFF)Characterization techniques, 444e446

detectable signals and characterizationregions, 445t

Charge coupled device (CCD), 419e420Charge transfer band (CTB), 165e167Charged device model (CDM), 442Chemical bond imaging methods, 444e446Chemical doping, 377e379Chemical vapor deposition (CVD), 11e12,

396Chemical-mechanical polishing (CMP),

231, 329Chip bonding, 498. See also Wire bondingChip packaging

basic structure of LED packaging modules,495e498

high-power LED packaging, 497e498low-power LED packaging, 497, 497f

functions of LED chip packaging, 491e495,491f

design for X, 495electrical interconnection, 492encapsulation and protection, 491e492heat dissipation, 494light extraction, 492e494refraction, 493fsystem testing, 494e495

optical effectsof freeform lenses, 511e514of gold wire bonding, 502e505of phosphor coating, 505e511

processes in LED packaging, 498e502,499f

chip bonding, 498lens laying, 501e502phosphor coating, 500e501silicone injection and curing, 502wire bonding, 499

thermal design and processing of LEDpackaging, 515e524

Chip-on-board technology (CoBtechnology), 517e518

Christie DHD850-GS 6900 lm Full HD laserphosphor projector, 746

Christie HS Series, 746Christie laser phosphor projectors, 746, 746fChromaticity, 627, 671, 704coordinates, 534e535, 539, 540t, 671,

675e676spaces, 672standards, 672, 673t

CIE. See Commission Internationale deI’Eclairage (CIE)

CISPR. See International Special Committeeon Radio Interference (CISPR)

CL. See Cathodoluminescence (CL)Closed-loop LED micro-jet array cooling

system, 523e524, 523fCMP. See Chemical-mechanical polishing

(CMP)CoB technology. See Chip-on-board

technology (CoB technology)Codoped activators, LEDs phosphors by,

169e192Coefficients of thermal expansion (CTE),

515e517Coherence of laser photons, 747Collection plane placement, considerations

for, 353e355Colorcolor-conversion efficiency, 436COLORED rings, 500conversion elements, 430homogeneity of stacked device, 426of light, 536motivation for color tuning, 415e416multiplexing principles, 738, 739fpurity of LED, 675e676rendering, 704saturation of LED, 675e676

764 Index

sensing-related terms, 670e677CIE1931 XYZ chromaticity diagram,672f

color matching functions, 670fcoordinates of primaries of chromaticitystandards in CIE31 space, 673t

determination of dominant wavelengthand color purity, 676f

representative color spaces vs. NichiaLEDs gamut, 674f

spaces, 672, 673fspatial uniformity. See Angular color

uniformity (ACU)uniformity, 638e640

Color rendering index (CRI), 125, 507,534e535, 720

Colorimetric quantities, 534e535Colour tuneable LEDs. See also GaN-based

LEDsgroup-addressable pixelated micro-LED

arrays, 430e436motivation for color tuning, 415e416stacked LEDs, 416e429

Commercial driver, 709e712Commission Internationale de I’Eclairage

(CIE), 124e125, 169, 171f, 653,670e671

chromaticity, 139fCIE31 color space, 674CIE31 system, 671coordinates, 422e4241976 L*a*b* color space, 672

Communication interface, 688Complete band gaps, 339Complex LED video display, 690, 690fCompound annual growth rate (CAGR),

659e660, 737e738Compound parabolic concentrators (CPC),

754Computer-aided design tools (CAD tools),

456Conduction band, 284e285Conductive ODR, 210e211Conformal coating, 501Consortia, 547e549Constant current output driver, 701, 701fConstant reference current, 704e705Contact thermal resistance, 519e520, 520fContinuous freeform lens method, 514, 517f

Continuous-wave (CW), 3e4, 555e556Contrastcontrast-enhancing filters, 667daylight, 667ratio, 666e667

Contrast sensitivity function (CSF), 678, 678fControl electronics, 739Controllability of LED, 539Controlled collapse chip connection (C4),

211Controller, 688Conventional c-plane LEDs limitations,

273e277carrier transport problems in multiple-

quantum-well LEDs, 274e275efficiency droop, 275e276green gap, 274QCSE, 273

Conventional FCLED (C-FCLED), 219Conventional GaN-based LED, 211e212,

212fConventional polar c-plane GaInN/GaN

LEDs, 307Conventional SRH recombination

mechanisms, 457e458Conventional-LED (C-LED), 210e211,

244e247Cooling solutions for LED applications,

523e524closed-loop LED micro-jet array cooling

system, 523fCopper wire, 499Core-shell NCs (CSNCs), 16e18Correlated color temperature (CCT),

124e125, 429, 500, 531, 534e535,539, 540t, 652, 671

Corundum. See Synthetic sapphire (Al2O3)CP2Mg. See Biscyclopentadienyl

magnesium (CP2Mg)CPC. See Compound parabolic

concentrators (CPC)CPD. See Cycles per degree (CPD)“Crack-free” GaN epitaxial layer, 54e55CRI. See Color rendering index (CRI)Critical flicker fusion frequency (CFF), 682CRT. See Cathode ray tube (CRT)Crystalgrowth axis, 279IS, 556

Index 765

Crystal-site engineering, 126Crystallography, 82e83

alternative crystallographic orientations,597e598, 598f

facets, 219, 221fof wurtzite nitride, 277

CSF. See Contrast sensitivity function (CSF)CSL. See Current spreading layer (CSL)CSNCs. See Core-shell NCs (CSNCs)CTB. See Charge transfer band (CTB)CTE. See Coefficients of thermal expansion

(CTE)Cubic GaN NCs (c-GaN NCs), 13e14Cubic III-nitrides, 598e599Curing, 502Current blocking layer (CBL), 358e359Current droop. See Efficiency droopCurrent spreading layer (CSL), 349Current versus output power (IeL),

570e571, 572f, 576f, 579,579fe581f

Current-voltage relationship (IVrelationship), 456

CVD. See Chemical vapor deposition(CVD)

CW. See Continuous-wave (CW)Cycles per degree (CPD), 678e679Czochralski-grown crystals, 80

DD-ILA. See Direct Drive Image Light

Amplifier (D-ILA)Data delivery, 688, 689fData production and distribution, 686e692

approximation error, 692fgrayscale resolution impact on gamma

correction, 691fLED display structure, 690fLED drivers connection on tile, 688fmultiplexed drive configuration, 689frear view of assembled LED display, 686f

Daylight contrast, 667DBEF. See Dual brightness enhancement

film (DBEF)DBR. SeeDistributed Bragg reflector (DBR)Deep ultraviolet (DUV), 4, 553e554, 553f

lithography, 347e348Deep ultraviolet light-emitting diodes

(DUV-LEDs), 553e557, 553f

Defect-free intrinsic semiconductor, 382Deflectometry, 56e57Degradation. See also Laser diode

degradationcatastrophic, 450diffusion-related, 455gradual, 450LEDsunder electrical and thermal stress,467e481

evaluation, 444e447mechanisms, 447e452in LED ESD testing, 450e452in LED life testing, 448e450

rapid, 450Deionized water (DI water), 433e434Delta doping, 600Design for X, 495Device under test (DUT), 443e444Device-quality GaNepitaxial growth, 46materials, 54

DI water. See Deionized water (DI water)Dichroic mirrors for RGB primaries

production, 741, 741fDiffraction angle, 750Diffusiondiffusion-related degradation processes,

455in LEDs, 476e478mechanisms, 478e481dopant activation, 479, 479fdrift of impurities, 479e480, 480fkinetic energy transfer, 479, 480fnative defects, 480e481

over time, 468e469Digital micromirror device (DMD), 737, 751DMD 1 chip and color wheel system

structure, 752flight valve chips, 751, 751f

Dirac-d function, 390e393Direct Drive Image Light Amplifier (D-

ILA), 752Direct laser technology, 746Direct transition bandgap energy, 554e555Direct-gap AlGaInP alloy system, 386e387Direct-type BLUs, 638e640, 642t, 643fDirect-type structure, 621, 622fDirectional dependence, 712

766 Index

Directivity LEDs, 660e661Discontinuous freeform lens method,

513e514“E” light pattern, 515flight pattern for BK7 optical glass lens,

516fmicrographs of PMMA discontinuous

freeform lens, 516fDislocation density, 387, 398e399,

401e403Dislocation-free high-quality single crystals,

12e13Dispersion relations to solid-state physicist,

337e339Displayapplications, 754e756electronics structure, 683e686, 683f

LED tile, 683ftile structure and placement, 684f

parameters of LEDs, 665e682color sensing-related terms, 670e677radiant intensity, 665e670spatial distribution, 677e681temporal performance terms, 682

quality evaluation of LED, 665stream size, 691types of LEDs, 660e665

concept of modular display design,661f

large stationary outdoor LED display,662f

LED display installation on buildingfaçade, 664f

LED floor display with interactivefunction, 663f

rear view of outdoor RENTAL 768e3module, 662f

rotary LED display principle, 664fTWA Series LED video wall display,661f

Distributed Bragg reflector (DBR), 225DMD. See Digital micromirror device

(DMD)3DMD model PT-RS30K from Panasonic,

745, 745fDOE. See United States Department of

Energy (DOE)Donor concentration, 35e39, 39fDopant activation, 479, 479f

Dopingeffect, 599e601in nitride materials, 9e10, 10f

Double flip process, 103Drift of impurities, 479e480, 480fDrift-diffusion model, 456Drift-induced leakage, 304e305current, 307

Droop, 275. See also Efficiency droopcurrent, 302e304

Dry etching procedure, 263Dry-etch procedure, 244Dual brightness enhancement film (DBEF),

621DUT. See Device under test (DUT)DUV. See Deep ultraviolet (DUV)DUV-LEDs. See Deep ultraviolet light-

emitting diodes (DUV-LEDs)

EE-K diagram. See Energy vs. k-vector

diagram (E-K diagram)e-ph. See Phonon emission (e-ph)E2-high phonon modes of GaN substrates,

261EBL. See Electron-beam lithography (EBL)EBLs. See Electron blocking layers (EBLs)Economic Commission on Europe (ECE),

650ECR plasma. See Electron-cyclotron

resonance plasma (ECR plasma)ED transition. See Electric dipole transition

(ED transition)Edge ray principle, 513e514Edge-type BLUs, 635e638, 636t, 638f,

639te641tmeasured positions for 46}BLU, 642fmeasured positions for 7}BLU, 640foptical design considerations for

commercial example, 641trequirements specification for commercial

example 46}BLU, 641tEdge-type structure, 621, 622fEDS. See Energy dispersive spectroscopy

(EDS)ee scattering. See Electron-electron

scattering (ee scattering)EEFLs. See External electrode fluorescent

lamps (EEFLs)

Index 767

eeh process, 302EELS. See Electron energy loss

spectroscopy (EELS)Efficiency droop, 275e276, 299

GaInN/GaN LED efficiency, 299e300in GaInN/GaN LEDs, 300e302IQE, 301fcurve of LED, 301f

low-droop GaInN/GaN LEDs, 311e320physical mechanisms of current droop in

GaInN/GaN LEDs, 302e310Auger recombination, 302e305calculated IQE curves, 310fcarrier leakage, 305e308EQE, 311fother mechanisms, 309e310

thermal droop in GaInN/GaN LEDs,320e322

EIEs. See Electron injection efficiencies(EIEs)

EL. See Electroluminescence (EL)Elastic stiffness constants, 279Electric dipole transition (ED transition),

167e168Electrical imaging methods, 444e446Electrical interconnection, 492Electrical overstress (EOS), 467e468Electrical stress, LEDs degradation and,

467e481Electrically injected polariton LEDs, 405Electrically tunable liquid crystal lens

application, 754, 754fElectroluminescence (EL), 45e46, 93, 370,

569, 570fe571f, 574f, 576f, 590,619e620

intensities, 213e217, 222e224mapping, 444process, 124spectra for LEDs, 94f

Electromagnetic Compatibility Directive,722

Electromagnetic compatibility issues (EMCissues), 721e727

Electromagnetic environment (EMenvironment), 722

Electromagnetic interference (EMI),721e722

Electron blocking layers (EBLs), 274e275,305, 385e386, 568e569, 569t, 580f

Electron energy loss spectroscopy (EELS),444

Electron injection efficiencies (EIEs),556, 576e582

Electron-beam lithography (EBL),349

Electron-cyclotron resonance plasma(ECR plasma), 4

Electron-electron scattering (ee scattering),463

Electronic band gap, 336Electronic structure, 592e593Electrons, 336capture process, 463, 463f

Electrostatic discharges (ESD), 467e468methods for, 443e444equivalent circuit for, 444f

robustness test, 441testing, 443e444

Electrostatic Poisson equation, 462ELO. See Epitaxial lateral overgrowth

(ELOG)ELOG. See Epitaxial lateral overgrowth

(ELOG)EM environment. See Electromagnetic

environment (EM environment)EMC issues. See Electromagnetic

compatibility issues (EMC issues)EMI. See Electromagnetic interference

(EMI)Emission fluctuations, 565e566Emission microscopy (EMMI), 444Emission wavelength, 746Encapsulation, 491e492basic structure of LED packaging modules,

495e498encapsulated PC TFLED, 235e236functions of LED chip packaging,

491e495, 491fnitride LEDs and chip packagingoptical effects of freeform lenses,511e514


optical effects of gold wire bonding,502e505

processes used in LED packaging,498e502, 499f

End-user, 665

768 Index

Energyefficiency, 541and environmental issues, 655

estimated power and energy use offilament lamp, 655t

range, 335e336Star, 625

Energy dispersive spectroscopy (EDS), 444Energy protection agency (EPA), 541Energy transfer (ET), 124, 169models, 192e193

with Ce3+ as sensitizers, 169e183using Eu2+ as sensitizers, 183e192

new LEDs phosphors by codoped activatorsand, 169e192

Energy vs. k-vector diagram (E-K diagram),343, 344f

Enhanced non-radiative SRHrecombination, 467e468

Enhanced power extraction, 372e373Enhanced-spectrum PWM (ES-PWM), 702Environmental requirements, 625, 626tEnvironmentally friendly, 628EOS. See Electrical overstress (EOS)EPA. See Energy protection agency (EPA)Epifilm-transferred technology, 226e227Epitaxial GaN development history on

sapphire substrates, 43e46device-quality GaN epitaxial growth, 46early development, 44powder GaN, 44thin-film GaN, 45e46

Epitaxial lateral overgrowth (ELOG), 60,243e244, 557

Epitaxial materials, 327e328base substrate material, 328factors affecting internal quantum

efficiency, 327e328Epitaxial overgrowth of GaN on sapphire

substrates, 59e65growth of GaN on patterned sapphire

substrates, 63e65Pendeo epitaxy of GaN, 62e63selective area growth and epitaxial lateral

overgrowth of GaN, 60e62Epitaxial process, 3e4EQE. See External quantum efficiency

(EQE)Erbium (Er), 587

ES-PWM. See Enhanced-spectrum PWM(ES-PWM)

ESD. See Electrostatic discharges (ESD)ET. See Energy transfer (ET)Etch depth effect, 346e347, 361Etching process, 263ETSI. See European Telecommunications

Standards Institute (ETSI)Eu2+ ions, 123e126, 177ET models using, 183e192Eu2+eMn2+ system, 183e189Eu2+eTb3+ system, 189Eu2+eTb3+eEu3+ system, 192Eu2+eTb3+eMn2+ system, 190e192Eu2+eTb3+eSm3+ system, 192

Eu2+eMn2+ system, 183e189, 184fEu2+eTb3+ system, 189Eu2+eTb3+eEu3+ system, 192Eu2+eTb3+eMn2+ system, 190e192Eu2+eTb3+eSm3+ system, 192Eulytite-type M3Ln(PO4) phosphate

compounds, 151e152European Committee for Electrotechnical

Standardization (CENELEC), 722European Telecommunications Standards

Institute (ETSI), 722Ewald construction of Bragg’s diffraction

theorem, 229, 236Ex situ wafer curvature measurement,

56e57Exchangeability, 538, 538fExcitation, 124Extended defect-free high-quality single

crystals, 12e13External electrode fluorescent lamps

(EEFLs), 619e620External quantum efficiency (EQE), 64,

98e99, 209e210, 243e244, 289,299e300, 311f, 387, 555, 555f,573f, 576f, 580f

Extraction efficiency, 383e384

FFailure evaluationflow chart, 446e447LED failure evaluation flow chart, 446f

techniques, 444e446detectable signals and characterizationregions, 445t

Index 769

Far-fieldbeam profile, 337beam-shaping, 370e371emission patterns in zenith direction,

235e236pattern of LED, 341transform of the Fibonacci lattice, 341

FC techniques. See Flip-chip techniques(FC techniques)

FCC. See Federal CommunicationsCommission (FCC)

FCLEDs. See Flip-chip LEDs (FCLEDs)FDTD. See Finite difference time domain

(FDTD)FE-SEM. See Field emission scanning

electron microscope (FE-SEM)Federal Communications Commission

(FCC), 722Federal Motor Vehicle Safety Standard

(FMVSS), 647Fermi golden rule, 390e393Fermi-Dirac distribution, 309e310,

385e386, 459e461FFLs. See Flat fluorescent lamps (FFLs)FIB method. See Focused ion beam method

(FIB method)Fick’s second law solution in one

dimension, 468e469Field emission scanning electron

microscope (FE-SEM), 112e114,244e246

Filtering effect of dichroic mirrors, 741, 741fFinite difference time domain (FDTD),

243e244limitations for modelling PC/PQC LEDsboundary condition considerations, 353comparative normalisation of results,355

considerations for collection planeplacement, 353e355

light launch considerations, 352e353modeling tools, 355requirements for extent of cross-sectionalprofile, 355

methods, 350simulation, 257, 350e352, 354f

Fixed installation projectors, 737e738Flat fluorescent lamps (FFLs), 619e620Flat sapphire substrates (FSS), 63

Flat-Panel-Display Measurements (FPDM),665

Flat-plate vapor chamber, 522Flicker, 682Flip-chip InGaN micro-pixelated LED

arrays, 753Flip-chip LEDs (FCLEDs), 210e211Flip-chip techniques (FC techniques),

210e212Floating buck converter topology, 703e704Fluorescent lamps, 327FMVSS. See Federal Motor Vehicle Safety

Standard (FMVSS)Focused ion beam method (FIB method),

444e446, 471e472Folded-mixing-LGP-type structure,

623Forced convection, 523e524Foreign impurities, 480e481Foreign substrates based on oxides, sulfides,

and metals, 51, 53tForward bias, 457e458Forward lighting, 647e651, 648tlow-beam headlamp photometric

requirements, 649tphotograph of low-beam headlamp beam

pattern, 650fphotometric requirements for low-beam

headlamp patterns, 649fFourier transform (FT), 340e341FPDM. See Flat-Panel-Display

Measurements (FPDM)Free-standing GaN, 286“Free-standing” substrates, 52Freeform lenses, optical effects of, 511e514Freely dispersed coating method, 500Freestanding high quality GaN substrate,

258e262GaN nanorods, 259fGaN NR arrays with SiO2 passivated

sidewalls, 259finitial stage of HVPE regrowth, 259fNomarski illumination for GaN thick films,

260fresults of GaN thick films, 260f

Fresnel Equation, 421e422Fresnel loss, 331FSS. See Flat sapphire substrates (FSS)FT. See Fourier transform (FT)

770 Index

Full width at half maximum (FWHM), 3e4,54e55, 88e90, 127e128, 261,557e558, 558f, 560f, 675e676

Full-frame projectors, 750Full-width at half-maximum (FWHM),

142e144, 147e151FullWAVE program, 257FWHM. See Full width at half maximum

(FWHM); Full-width at half-maximum (FWHM)

GGaAlAs, 659e660GaAs. See Gallium arsenide (GaAs)GaAsP alloy, 386e387GaCl. See Gas-phase gallium chloride

(GaCl)GaInN/GaN LEDs, 299efficiency droop in, 300e302

Gallium (Ga), 44, 60vacancies, 480

Gallium arsenide (GaAs), 43, 386e387Gallium interstitials (Gai), 480Gallium nitride (GaN), 79buffer layer, 43

strategies, 86e93challenges for growth on silicon substrates,

83e86coalescence, 243e244device technologies

early device efforts, 93e95layer transfer, 101e103progress in large-area substrates, 95e101semipolar and nonpolar GaN LEDs onsilicon, 106e108

weak-beam dark-field TEM images, 99fepifilm-transferred technology, 226epilayers, 5e8epitaxial film, 49GaN-based LEDs, 79growth rate, 27e28

distribution, 28fhetero-epitaxy of, 25layers, 349LED, 387

on patterned silicon substrates, 103e106Mg doping of, 31e32, 31fnanowires and nanorods on silicon,

111e117

optical micrograph, 85fp-n junction LED, 387possible substrates for the epitaxial growth,

81tQDs, 399e401sapphire-based LED approach, 225on sapphire substrates for visible LEDsepitaxial GaN on sapphire substrates,43e46

epitaxial overgrowth of GaN on sapphiresubstrates, 59e65

GaN growth on nonpolar and semipolardirections, 65e67

LEDs on sapphire substrates, 67sapphire substrates, 46e52strained heteroepitaxial growth onsapphire substrates, 52e59

self-catalyst growth GaN NCs, 11e14silicon, 80e83on silicon substrates, 79e80strained heteroepitaxial growth growth

mechanism, 55e56, 56fwurtzite crystal structure, 388, 388f

Gallium phosphide (GaP), 44“Gamma”, 668, 668fcorrection, 670gamma-compressed RGB values, 669

GaN. See Gallium nitride (GaN)GaN FCLEDs with geometric sapphire

shaping structure, 218e224C-FCLED and SS-FCLEDchips, 223fcurrent-voltage characteristics, 223fdevices, 222fnormalized far-field patterns, 224f

crystallography facets, 221ffabrication steps for SS-FCLEDs, 218fGaN SS-FCLED, 220flight output power and wall-plug efficiency,

224fsapphire shaping structure, 220f

GaN FCLEDs with textured micro-pillararrays, 213e217

chip bonded to silicon sub-mount, 215fflat-surface FCLEDs and MPA-FCLEDscurrent-voltage characteristics, 216flight output power-current curves,216f

fabrication steps, 214f

Index 771

GaN FCLEDs with textured micro-pillararrays (Continued)

light extraction enhancement vs. MPAdepth, 217f

MPA surfaces of sapphire backside, 215fsilicon sub-mount before FC bonding, 214f

GaN nucleation layer (GaN NL), 263e264GaN thin-film photonic crystal LEDs (GaN

thin-film PC LEDs), 225e227epifilm-transferred technology, 226e227semiconductor wafer bonding, 225e226

GaN-based flip-chip LEDs and flip-chiptechnology, 210e212

background of flip-chip LEDs, 210e211flip-chip technology, 211e212structure of GaN-based FCLED, 211e212

GaN-based LEDs, 243e244. See alsoColour tuneable LEDs

conventional, 211e212, 212fdegradation of LEDs under electrical and

thermal stress, 467e481efficiency, 209e210HB-LEDs, 210e211modeling performance-limiting effects in,

456e467excess subthreshold forward currentinduction, 456e459

modeling GaN LEDs optical turn-on,459e467

GaN-based optoelectronic devices, 455GaN/AlGaN quantum cascade structure, 603GaN/AlN, 594

interdiffusion, 593e594QWs displaying ISB transitions in near-IR,

594superlattices, 592

GaP. See Gallium phosphide (GaP)Garnet-type green-emitting Ca3Sc2(SiO4)3:

Ce3+ phosphors, 129e131Garnets (A3B2C3O12), 142e144Gas-phase gallium chloride (GaCl), 286Gated pulse width modulation (GPWM),

696, 699fGaussian peak fitting, 154e155[(Gd1exLux)0.99]3Al5O12:0.03Ce phosphor,

135e136GEBL. See Graded electron blocking layers

(GEBL)Gemstone sapphires, 47

Gold wire, 499Gold wire bondingleads, 290e291optical effects, 502e505angular CCT distributions, 504fangular CCT measurement, 503fphosphor layers in LED module samples,502f

shape and height of gold wires, 502fGPWM. See Gated pulse width modulation

(GPWM)Graded electron blocking layers (GEBL),

385e386Gradual degradation, 450Green gap, 274Green MQW, 32e35, 36fGroup-addressable micro-LED array

implementation, 431e432adjacent microlight-emitting pixels, 432fgroup-addressable micro-LED, 432fe433f

Group-addressable pixelated micro-LEDarrays, 430e436

development status, 434e436architectures of regular micro-LED andthin-film micro-LED, 437f

micro-LED device with jet-printed redand green quantum dots, 438f

microphotograph of a completed micro-LED, 435f

RGB pixels set of pixels illumination,435f

group-addressable micro-LED arrayimplementation, 431e432

natural color mixing using miniature RGBpixels, 430

RGB pixels forming by jet-printing,432e434

Growthof low TDD AlN layers on sapphire,

557e561and structural properties, 593e594temperature, 18

Guided mode dispersion curves, 233e234

Hh-GaN NCs. See Hexagonal GaN NCs (h-

GaN NCs)Hamiltonian of crystal growth orientation,

284

772 Index

Hard failure, 450Hard nickel mask uniformity, 213HB GaN-based LEDs. See High-brightness

GaN-based LEDs (HB GaN-basedLEDs)

HBM. See Human body model (HBM)HCFLs. See Hot cathode fluorescent lamps

(HCFLs)HCL. See Human centric lighting (HCL)HD. See High-definition (HD)Head mounted display (HMD), 750Head-up display, 754e755, 755fHead-worn displays, 754e755, 755fHeat dissipation, 494problems, 290e291

Heated nitric acid solution (HNO3 solution),244e246, 253e254

Heaviside step function, 459e461Heteroepitaxy of nonpolar and semipolar

planes, 285Heterojunctions, 384e386Heterostructure field-effect transistors, 49Hexagonal GaN NCs (h-GaN NCs), 13e14HF solution. See Hydrogen fluoride solution

(HF solution)hhe process, 302HID. See High-intensity discharge (HID)High beam headlamp, 647e650, 648tHigh degree of polarized emission, 289High efficiency red LEDs, 397High extraction efficiency GaN-based LEDcurrent-voltage and intensity-current

characteristics, 251fon embedded SiO2 NR array and NPSS,

248e252GaN/sapphire interface, 252fLED with NPSS and a SiO2 PQC structure,

249fsapphire surface with NPSS and n-GaN

surface with SiO2 PQC, 250fHigh injection current, 276High IQE LED, 274High LED current, 700e701High light extraction packaging, 290e291High luminous efficiency, 626High rotational disc reactor, 25e26High temperature storage test, 441High-brightness GaN-based LEDs (HB

GaN-based LEDs), 209

High-definition (HD), 629e631High-efficiency green LEDs, 287High-electron mobility transistors.

See Heterostructure field-effecttransistors

High-indium-content alloys for IR emitters,587e589, 589f

High-intensity discharge (HID), 650e651High-powerblue LEDs, 329LED chip, 492, 492fLED packaging, 497e498, 498f

High-power Golden DRAGON ARGUSLEDs, 32}LCD television based on,640

High-qualitybuffer layers, 3bulk GaN, 285e286indium-rich InGaN alloys, 3

High-resolution X-ray diffraction(HRXRD), 56e57, 88e90

High-temperature (HT), 566High-temperature AlN (HT-AlN), 87e88High-temperature-GaN layer (HT-GaN

layer), 54e55High-temperature-polysilicon (HTPS), 753light throughput improvement in Epson

version, 753fliquid-crystal-display panel technology,

753Highly efficient and bright LEDs

overgrowth, 253e257air voids formation, 254fdiffusing reflectance spectra for both

samples, 255felectrical and optical properties of NAPSS

and conventional LED, 257fforward L-I-V characteristics for both

fabricated LEDs, 256fGaN NRs template, 253f

Highly reflective ODRs, 210e211Hillock morphologies, 287e288HMD. See Head mounted display (HMD)Hollow-type structure, 623Homoepitaxy and need for bulk GaN

substrates, 285e286Homogenizer, 753e754Homojunctions, 384e386Hooke’s law, 279

Index 773

Horizontal flow reactor, 4, 6f, 27e28Hot cathode fluorescent lamps (HCFLs),

619e620HRXRD. See High-resolution X-ray

diffraction (HRXRD)HT. See High-temperature (HT)HT-AlN. See High-temperature AlN (HT-

AlN)HT-AlN/LT-AlN method, 90HT-GaN layer. See High-temperature-GaN

layer (HT-GaN layer)HTPS. See High-temperature-polysilicon

(HTPS)Human body model (HBM), 442Human centric lighting (HCL), 531, 537Human factor issues with LEDs, 652e654

relative brightness of roadway pavementsurfaces, 654f

spectral distributionsof white LED and filament sources, 653fof yellow and red LED, 652f

Hydride vapor-phase epitaxy (HVPE), 258,286

Hydrogen, 472e473, 480Hydrogen fluoride solution (HF solution),

132e133

IICP. See Inductively coupled plasma (ICP)ICP-RIE. See Inductively coupled plasma

reactive ion etching (ICP-RIE)IEC. See International Electrotechnical

Commission (IEC)IEC 62776 standard, 538IES. See Illuminating Engineering Society

(IES)IeH model. See InokutieHirayama model

(IeH model)III-Arsenide system, 404III-N. See III-nitride (III-N)III-N heterostructures, 65III-nitride (III-N), 43. See also Nitride LED;

N-side up LEDs; Photonic crystalnitride LEDs

current status, 397e404polarization effects, 387e397methods for improving IQE, 393e396p-type doping problem and polarization-induced doping, 396e397

polarization-induced electric fields,389e393

spontaneous and piezoelectricpolarization, 388

properties of sapphire for substrates ofmaterials

chemical and thermal properties, 47e48materials properties, 48tstructural properties, 47

QD UV LEDs, 403e404semiconductors, 226, 299, 598e599UV LEDs, 402e404, 402fvisible LEDs, 397e401

III-nitride materials, 79for ISB optoelectronics, 591e601AlInN/(In)GaN system, 597alternative crystallographic orientations,597e598

cubic III-nitrides, 598e599effect of doping, 599e601electronic structure, 592e593growth and structural properties,593e594

optical characterization, 595e597quantum dots and nanowireheterostructures, 599

III-nitride visible LEDs, 387, 397e401external quantum efficiency variation with

peak emission wavelength, 399fhigh-resolution transmission electron

micrographs of defect distribution,400f

InGaN-AlGaN double heterostructureLEDs, 398f

IQE, 400fnormalized external quantum efficiency,

401fIeL. See Current versus output power (IeL)Illuminating Engineering Society (IES),

442Illumination system, 753e754Image generation, 747e753In situ nucleation/buffer layer, 59InAlGaNfor DUV light sources, 554e555InAlGaN-based DUV LED, 584fQDs, 403e404quaternary alloy, 565

Incandescent lamps, 327, 647

774 Index

Incoming video, 689raster size, 690stream, 686e687

Indium gallium nitride (InGaN), 401epilayers, 8InGaN-based green and yellow LEDs,

387LEDs, 416e417

epiwafer, 102e103nanowires, 405NCs, 14

Indium incorporation in nonpolar andsemipolar planes, 286e287

Indium nitride (InN), 8epilayers, 8NCs, 14

Indium phosphide (InP), 43Indium-tin-oxide (ITO), 213, 248e249,

290e291, 329CSL layers, 365

Individual RGB channels, 738, 739fInductively coupled plasma (ICP), 349etching, 213, 289

Inductively coupled plasma reactive ionetching (ICP-RIE), 249

Industry/factory/public lighting, 544, 544fInfrared (IR), 587emitters

high-indium-content alloys for IRemitters, 587e589

III-nitride materials for ISBoptoelectronics, 591e601

ISB devices, 601e604RE-doped GaN emitters, 590e591

optoelectronics, 591photodetectors, 601e602

InGaN. See Indium gallium nitride (InGaN)InGaN/GaNCSNC white light LED, 17fdots-in-NCs, 16e17InGaN/GaN QWs, many-body formulation

for capture time in, 462e467, 463fInjection efficiency, 299e300, 383e384Inkjet printers, 673e674InN. See Indium nitride (InN)InokutieHirayama model (IeH model), 189InP. See Indium phosphide (InP)Interactive projection displays, 755, 756fInteractive Touch Projector, 755

Interlaced black-and-white frames videotest, 682

Internal quantum efficiency (IQE), 51, 93,157e159, 209e210, 247e248, 273,299e300, 327e328, 347, 389, 456,555e556

factors affecting, 327e328marked increases, 561e567methods for improvingincreasing overlap by using thin QWsand QD, 393e394

polarization-matched LEDs, 395e396semipolar/nonpolar substrates, 394e395

International Committee for DisplayMetrology, 665

International Electrotechnical Commission(IEC), 534e535, 539, 540f, 722

International Special Committee on RadioInterference (CISPR), 722

International Telecommunication Union(ITU), 722

ITU Recommendation BT. 2020, 741e742Intersubband (ISB), 587devicesall-optical switches, 601IR photodetectors, 601e602light emitters, 602e604QCDs, 602

III-nitride materials for ISB optoelectronics,591e601

optoelectronics, III-nitride materials for,591e601

PL, 602e603transitions, 592

InxGa1exN alloy, 286e287IQE. See Internal quantum efficiency (IQE)ISB. See Intersubband (ISB)ITO. See Indium-tin-oxide (ITO)ITU. See International Telecommunication

Union (ITU)

JJet-printing, RGB pixels forming by,

432e434

KK2SiF6:Mn4+ phosphor, 132e133K2TiF6:Mn4+ phosphor, 132e133KBaPO4:Eu

2+ phosphor, 151

Index 775

Kinetic energy transfer, 479, 480fKrF, 101

excimer pulsed laser, 226e227KSrPO4:Eu

2+ phosphor, 151

LLaBaSiO3N phosphor, 140e142Lambertian lens, 495Lambertian radiation, 495, 623Lamps, 531

FFLs, 619e620fluorescent, 327incandescent, 327, 647tungsten-halogen, 647UHP, 737, 740Ushio UXW-15KD 15 kW water cooled

xenon lamp, 740, 740fXenon, 737, 740, 740f

Large-area substrates, progress in, 95e101Large-scale video displays, 665e666, 721

LED displays, 659e660Laser

extremely narrowband emission, 744flaser-assisted film debonding, 226e227micromachining, 419e420pico projector, 747

Laser beam diffraction, 748, 750fLaser diode degradation

diffusion-like process signatures in,469e474

experimental and calculated diffusioncoefficients

for hydrogen, 475ffor magnesium, 475f

extrapolated diffusion coefficientvs. bias current, 474fvs. thermal energy, 473f

increase in threshold current and decreasein subthreshold slope, 470f

variationin forward current and threshold current,471f

of threshold current in laser diodes stress,472fe473f

Laser diodes (LDs), 3, 10t, 50, 209,274e275, 299, 553e554, 553f

Laser lift-off techniques (LLO techniques),226, 226f, 436

LaSrSiO3N phosphor, 140e142

Lateral epitaxial growth (LEG), 60Lateral epitaxial overgrowth (LEO), 60Lateral overgrowth (LOG), 60Lattice symmetry, effect of, 340e341Layer transfer, 101e103LC lens. See Liquid crystal lens (LC lens)LCD. See Liquid crystal display (LCD)3LCD laser projector L1000 structure,

745e746, 746fLCoS. See Liquid crystal on silicon (LCoS)LDs. See Laser diodes (LDs)Lead frame, 492, 494Leaky mode coupling, 228, 228fLED ESD testing, degradation mechanisms

in, 450e452horizontal structure of LED device,

451fLED grown on MPSi (MPLED), 106,

108fe109fLED grown on NPSi (NPLED), 106,

108fe109fLED life testing, degradation mechanisms

in, 448e450catastrophic degradation, 450gradual degradation, 450rapid degradation in initial stage, 450reverse current, 449fsemi-log currentevoltage curve measuring

after long-term aging tests, 448fLED Si substrates (LED-Si substrates),

457e458LED SiC substrates (LED-SiC substrates),

457e458LEDs. See Light-emitting diodes (LEDs)LEE. See Light extraction efficiency (LEE)LEG. See Lateral epitaxial growth (LEG)Lens laying, 501e502LEO. See Lateral epitaxial overgrowth

(LEO)LGP. See Light guide plate (LGP)Li2ASiO4 phosphor, 129e131Life testing methods, 442e443Life time, 535acceleration of aging by means of

temperature, 536festimation, 442e443

LiGaO2, 51Lightenhancement, 236e237

776 Index

intensity distribution designs, 495, 496flaunch considerations, 352e353line, 345output, 694e695source, 740e747

technical considerations, 619e621Light emissioncharacteristics of GaN PC TFLEDs,

231e237cross-section of GaN PC TFLEDstructure, 233f

fabrication steps for GaN-basedTFLEDs with PC lattice structures,232f

far-field pattern normalized with peakintensity, 235f

GaN TFLED structure with PC,233f

light enhancement recorded at variousoutput collection angles, 236f

light output power-current curvecharacteristic, 234f

from GaN, 590Light emitters, 602e604basis on nitride MQWs, 10e11

Light extraction, 492e494from LEDs, 337methods of improving light extraction from

LEDs, 334e335N-side up over P-side up device

configuration advantages for,365e366

for nonpolar and semipolar LEDs,288e291

high light extraction packaging,290e291

increasing hextr via surface roughening,289

light extraction efficiency as limitingfactor, 288e289

TFFC LED, 289e290Light extraction efficiency (LEE), 243e244,

299e300, 422, 576e582, 582fas limiting factor, 288e289

Light guide plate (LGP), 621, 624, 624tLight output power (LOP), 299e300Light-current-voltage characteristics

(LeIeV characteristics), 213e217,231e232

Light-emitting diodes (LEDs), 43, 79e80,209, 243e244, 273, 327, 377e387,415, 491, 531, 553e554, 553f, 587,619e620, 647, 659e660, 742. Seealso Liquid crystal display (LCD)

advantages and history of LED BLUsadvantages of LEDs for LCD BLUs,626e628

development, 628e629chips, 219e222, 517, 522fcooling solutions for LED applications,

523e524degradation evaluation, 444e447failure evaluation flow chart, 446e447failure evaluation techniques, 444e446

degradation under electrical and thermalstress, 467e481

diffusion in LEDs, 476e478diffusion mechanisms, 478e481diffusion over time, 468e469signatures of diffusion-like processes inlaser diode degradation, 469e474

development, 386e387diffusion in, 476e478dimming technique, 725display parameters, 665e682display types, 660e665drivers, 659e660, 685, 688, 688fdynamics-related issues, 706e712driver topologies, 710fexperimentally measured nonlinearity,709f

nonlinearity, 708fPWM grayscale resolution limitation,707t

efficiency, 383e384extraction efficiency improving, 341e347effect of etch depth, 346e347

failure modes, 449forward current, 700e701front tip, 712GaN on silicon substrates, 79e80buffer-layer strategies, 86e93challenges for growth, 83e86device technologies, 93e117silicon, 80e83

homojunctions and heterojunctions,quantum wells and dots, 384e386

human factor issues with, 652e654

Index 777

Light-emitting diodes (LEDs) (Continued)LED LCD backlighting unit typesBLU classification, 621e623technical considerations for light source,619e621

lens application, 681light sourcesin connected world, 546e547light source-based projector, 737

light-trapping in, 330e334radiometry and discussion of solid angle,332e334

lumen maintenance test, 441luminance, 743market trends and technological

developments, 629e634methods of improving light extraction,

334e335optical design, 634e640p-n junction diodes, 377e380packaging, thermal design and processing

of, 515e524phosphors by codoped activators and

energy transfer, 169e192. See alsoPhosphor coating

CIE chromaticity coordinate diagram,171f, 173f

ET models using Eu2+ as sensitizers,183e192

ET models with Ce3+ as sensitizers,169e183

typical singly doped activators, 170tphosphors with different host systems,

134e168. See also Phosphor coatingaluminates, 134e140borates, 154e157Mn4+ doped red phosphors, 160e167narrow-band red nitride phosphors,157e160

phosphate, 151e154silicates, 140e146sulfides, 146e151tungstates/molybdates based redphosphors, 167e168

photoluminescence pulse rise and fall frontmeasurement results, 711f

power, 539projectors, 743e744PWM nonlinearity, 707f

recombination mechanisms, 381e383,383f

on sapphire substrates, 67technical considerations for optical films

and plates, 624e625technology in detail, 682e727data production and distribution,686e692

display electronics structure, 683e686,683f

LED dynamics-related issues, 706e712pixel control, 692e706

technology in detail, 682e727display directivity and LED directivitycomparison, 715f

electrical field strength, 724fEMC issues, 721e727LED axis definitions, 712fLED directivity influence, 712e719LED far-field pattern, 713fLED FFP parameters determination, 713fLED FFP variation, 714fradiation stability, 719e721in situ LED FFP measurement, 719f

types, 328e330N-side up LEDs, 329e330, 329fP-side up lateral current spreading LEDs,328e329, 328f

patterned substrate LEDs, 330Light-trapping in LEDs, 330e334radiometry and discussion of solid angle,

332e334Lighting applications, three-side-approach

for, 531e542Lighting equipment, 546Line-emitting rare earth ions, 123e124Liquid crystal display (LCD), 415,

619, 620f, 737e738. See alsoLight-emitting diodes (LEDs)

advantages and key technologies of LEDLCD televisions, 632e634, 633t

advantages of LEDs for LCD BLUs,626e628

environmentally friendly, 628high luminous efficiency, 626longer operating life, 627low operating DC voltage, 626package size and chroma selectionflexibility, 627

778 Index

quicker illumination to stable brightness,628

rapid switching speed, 627robustness, 628wide adjustment range for brightness,contrast, and chromaticity, 627

wide color gamut, 627wide operating temperature range, 626

requirements for LCD BLUs, 625environmental requirements, 625features, 625

technical considerations for optical filmsand plates, 624e625

Liquid crystal lens (LC lens), 754Liquid crystal on silicon (LCoS), 737light valve technology, 752fprojector, 752e753

LiSrPO4:Eu2+ phosphor, 151

Lithium niobate (LiNbO3), 748LeIeV characteristics. See Light-current-

voltage characteristics (LeIeVcharacteristics)

LLO techniques. See Laser lift-offtechniques (LLO techniques)

LM-80e08 standard, 442e446, 448LO. See Longitudinal optical (LO)Local power decoupling capacitors, 726LOG. See Lateral overgrowth (LOG)Longer operating life, 627Longitudinal optical (LO), 463, 592LOP. See Light output power (LOP)Louvers, 684, 684fe685fLow beam headlamp, 647e650, 649tLow operating DC voltage, 626Low power consumption, 626Low TDD AlN layers growth on sapphire,

557e561Low temperature storage test, 441Low Voltage Directive, 540e541Low-droop GaInN/GaN LEDs, 311e320.

See also Thermal droopdin GaInN/GaN LEDs

low-droop nonpolar/semipolar GaInN/GaNLEDs, 315e318

piezoelectric polarization and totalpolarization difference, 313f

polar c-plane, nonpolar/semipolar GaN,312e315, 312f

simulated band diagram, 314f

Low-droop nonpolar/semipolar GaInN/GaNLEDs, 315e318

(2021) and (3031) GaInN/GaN LEDs,317e318

comparison of EQE, 315fdevice structure and performance of blue

GaInN/GaN LEDs, 316tlight output power and EQE, 319fm-plane GaInN/GaN LEDs, 315e317modified ABC model for c-plane and

semipolarGaInN/GaNLEDs, 318e320simulated IQE curves, 320f

Low-power LED packaging, 497, 497fLow-pressure metal-organic chemical vapor

deposition (LP-MOCVD), 558Low-temperature AlN (LT-AlN), 58e59,

87e88Low-temperature buffer layer (LT buffer

layer), 54e56Low-temperature layer, 58e59Low-temperature-gallium nitride (LT-GaN),

58e59LP-MOCVD. See Low-pressure metal-

organic chemical vapor deposition(LP-MOCVD)

LT buffer layer. See Low-temperature bufferlayer (LT buffer layer)

LT-AlN. See Low-temperature AlN (LT-AlN)

LT-GaN. See Low-temperature-galliumnitride (LT-GaN)

Lu3exYxMgAl3SiO12:Ce3+ phosphors,

142e144Lumens (lm), 666Lumileds, 501Luminaire, 536, 543fLuminance, 666nonuniformity, 680e681, 681funiformity, 680

Luminescence from RE ions in solid hosts,590

Luminous efficacy, 125Luminous flux, 536, 539Luxeon, 497e498DCC as light source, 623, 623f

Mm-plane GaInN/GaN LEDs, 315e317M[Mg2Al2N4] phosphors, 159e160, 161f

Index 779

Machine model (MM), 442Macroblocks, 686e687Magnasonic LED Pocket Pico Video

Projector, 743Magnesium, 9e10, 478, 480Magnetic dipole transition (MD transition),

167e168Market needs, 531e539, 534fe535f

efficiency, 536exchangeability, 538life time, 535performance, 534e535safety, 534serving customer’s well-being, 536e537,

537fsmart and connected, 539

Market trends and technologicaldevelopments, 629e634, 633t

advantages and key technologies of LEDLCD televisions, 632e634, 633t

cost forecast for different BLU types, 630fforecast for number of LED packages per

television set, 631fmarket forecast for large BLUs, 630fnew display technologies using LEDs,

634technological development trends for

different products, 632tMarketsand-Markets Pico projector market

value, 737e738Master semiclassical model, 456MATLAB, 421e422Matsubara formalism, 465e466MattheweBlakeslee equilibrium model, 288MBE. See Molecular beam epitaxy (MBE)MCPCB. See Metal core printed circuit

board (MCPCB)MD transition. See Magnetic dipole

transition (MD transition)MDs. See Misfit dislocations (MDs)Mean free path (MFP), 508e511Mechanical axis, 712Mechanical shock test, 441Medium-gamut spaces, 672e673Meltback etching of substrate, 83MESA, 349e350Mesopic vision, 665e666Metal core printed circuit board (MCPCB),

621

Metal-organic chemical vapor deposition(MOCVD), 3, 25e26, 43, 79, 83,213, 243e244, 285e286, 398, 418

growth of GaN nanorods, 114e117growth of nitride semiconductorsblue MQW, 32e35carbon incorporation and Mg doping ofGaN, 31e32, 31f, 33f

GaN growth rate distribution, 28fgreen MQW, 32e35, 36fgrowth mechanism, 28e30horizontal flow production reactor, 27freaction diagram of bimoleculardecomposition path, 30f

transition energy of each combination ofmetals, 30t

two types of vertical flow reactors,25f

UV materials growth, 35e40Metal-organic vapor phase epitaxy

(MOVPE), 87, 588Metalorganics, 45MFP. See Mean free path (MFP)Mg dopingof GaN, 31e32, 31fMg-doped AlGaN, 40

Micro-jet cooling technology, 523e524Micro-LED arrays, group-addressable

pixelated, 430e436Micro-opto-electro-mechanical systems,

747Micro-patterned silicon (MPSi), 106Micro-pillar array (MPA), 210e211Microcantilever, 748, 749fMicrocavity effectin practice for unpatterned N-side up LED,

368e370utilization in N-side up LEDs, 366

Microelectromechanical systems mirror(MEMS mirror). See Bragg cell

Microlight-emitting diode, 430Micromesh filters, 667Microsoft high-performance computing

cluster environment (MS HPCenvironment), 355

Microvision, 737e738Miller indices, 277MinanoeBenitez design method, 512e513Miniature projectors, 756, 756f

780 Index

Miniature RGB pixels, natural color mixingusing, 430

Minimal strain effects, 287Minimum separable acuity, 678Misfit dislocations (MDs), 288ML. See Monolayer (ML); Multilayer (ML)MLCC capacitors, 726MM. See Machine model (MM)Mn2+ ions, 169e172Mn4+ doped red phosphors, 160e167Mn4+-activated fluoride phosphors,

132e133Mobile hole concentrations, 403Mobile LED screens, 665MOCVD. SeeMetal-organic chemical vapor

deposition (MOCVD)Modeling GaN LEDs optical turn-on,

459e467many-body formulation for capture time in

InGaN/GaN QWs, 462e467electron capture times, 467fsecond order Feynman diagrams, 464f

quantum capture model, 460fModelling PC-LEDs, 350e365beam-steering effects, 361e363effect

of etch depth on extracted power,361

of TCSL and passivation layers on PQCperformance, 365

electro-optical performance data, 364fFDTD simulation methods, 350e352limitations of FDTD for modelling PC/PQC

LEDs, 352e355P-side up

PC-LED performance, 358e361, 359fPQC LED electro-optical performance,363

unpatterned P-side up LED simulationexample, 356e358

Modern LED designs and enhancements,404e405

nanowire LEDs, 405polariton LEDs, 405RCLEDs, 404SLEDs, 404

Modified ABC model for c-plane andsemipolar GaInN/GaN LEDs,318e320

Modulation transfer function (MTF),715e716

Modules, 686, 686fe687fMoisture resistance test, 441Molecular beam epitaxy (MBE), 3, 44, 86,

386, 587growth techniques, 3e4MBE-based technique, 244nitride nanostructures based on NCs,

16e18nitride NC materials, 11e16PAMBE growth of nitride epilayers and

quantum structures, 4e11Molybdenum, 11e12Monochromatic light, 417e418Monolayer (ML), 593e594Monte Carlo optical simulation, 519Monte Carlo ray-tracing simulations, 507MOVPE. See Metal-organic vapor phase

epitaxy (MOVPE)MPA. See Micro-pillar array (MPA)MPLED. See LED grown on MPSi

(MPLED)MPSi. See Micro-patterned silicon (MPSi)MQB. See Multiquantum barrier (MQB)MQD. See Multiple quantum disk (MQD)MQWs. See Multiple quantum wells

(MQWs)MS HPC environment. See Microsoft high-

performance computing clusterenvironment (MS HPCenvironment)

MTF. See Modulation transfer function(MTF)

Multi-domain vertical alignment (MVA), 629Multi-phonon emission, 463Multilayer (ML), 556, 559f, 562fMultiple quantum disk (MQD), 112e114Multiple quantum wells (MQWs), 9, 65, 95,

213, 244e246, 327e330, 384e385,455, 561, 563fe564f, 566f

growth, 28LEDs, 274, 428carrier transport problems in, 274e275structures, 457e458

Multiquantum barrier (MQB), 556, 576,577fe578f

MVA. See Multi-domain vertical alignment(MVA)

Index 781

NN-doped GaP, 382N-doped layers, 328N-doping aluminum-rich AlGaN alloys, 3n-GaN layer, 213N-side up LEDs, 329e330, 329f. See also

III-nitride (III-N); Nitride LEDmicrocavity effect, 368e370in N-side up vertical LED, 367futilization, 366

PC-enhanced light extractionadvantages over P-side up deviceconfiguration, 365e366

N-side up PC-LED performance,367e370

N-side up PQC patterned LEDs,performance improvement for

enhanced power extraction, 372e373far-field beam-shaping, 370e371

NaAlSiO4:Ce3+, Mn2+ phosphor, 169e172

Nano-imprint lithography (NIL), 248e249,348

Nano-patterned silicon (NPSi), 106Nano-size LEDs, 243e244Nanocolumns (NCs), 11, 111

nitride nanostructures, 16e18CSNCs, 17e18InGaN/GaN CSNC white light LED,17f

quantum disks embedded, 16e17selective area growth, 18

Nanodisks (NDs), 16e17Nanopillars (NPs), 253, 263Nanorod-array patterned sapphire substrate

(NAPSS), 243e244Nanorods (NRs), 111

arrays, 244GaN nanowires and nanorods on silicon,

111e117Nanoscale epitaxial lateral overgrowth

(NELO), 243e244, 254e255electrical and optical properties of NAPSS-

and conventional-LED, 248fFESEMs, 245fof GaN-based light emitting diodes on SiO2

NAPSS, 244e248GaN/sapphire interface for GaN epilayer,

246fNanoscale masks, 253

Nanostructured LED, 243e244bottom-up technique for GaN nanopillar

substrates, 263e269top-down technique for nanostructured

LED, 244e262Nanowires, 111GaN nanowires and nanorods on silicon,

111e117heterostructures, 599, 600fLEDs, 405

NAPSS. See Nanorod-array patternedsapphire substrate (NAPSS)

Narrow-band red nitride phosphors, 157e160Narrowband spectral output of colored

LEDs, 652National Highway Traffic Safety

Administration (NHTSA), 651National Television System Committee

(NTSC), 627Native defects, 480e481Native substrates, 52, 53tNatural color mixing using miniature RGB

pixels, 430color mixing using micron-scale pixels,

431fmicropixelated red-green-blue LED, 430f

NCs. See Nanocolumns (NCs)NDs. See Nanodisks (NDs)Near ultraviolet (n-UV) LED chips,

123e124Near-IR ISB absorption, 597e598NELO. See Nanoscale epitaxial lateral

overgrowth (NELO)NELOG. See Nanoscale epitaxial lateral

overgrowth (NELO)“Neutral white” color, 531Newton’s law of cooling, 523NHTSA. See National Highway Traffic

Safety Administration (NHTSA)Nickel, 11e12NIL. See Nano-imprint lithography (NIL)Nitride epilayers, PAMBE growth of, 4e11Nitride LED, 377e387. See also III-nitride

(III-N); N-side up LEDschip packagingoptical effects of freeform lenses,511e514


782 Index

functions of LED chip packaging, 491e495,491f

GaN FCLEDswith geometric sapphire shapingstructure, 218e224

with textured micro-pillar arrays,213e217

GaN thin-film photonic crystal LEDs,225e227

GaN-based flip-chip LEDs and flip-chiptechnology, 210e212

LED packaging modulesbasic structure, 495e498processes, 498e502, 499f

light emission characteristics of GaN PCTFLEDs, 231e237

modern LED designs and enhancements,404e405

optical effects of gold wire bonding,502e505

p-type doping problem in, 396e397PC nanostructures and PC LEDs, 227e229reliability, 441

degradation mechanisms, 447e452evaluation of LED degradation,444e447

methods for electrostatic dischargetesting, 443e444

methods for life testing and lifetimeestimation, 442e443

testing, 441e444Nitride nanocolumn materials (Nitride NC

materials), 11e16AlGaN NCs, 14GaN NCs, 12fInN and InGaN NCs, 14InN growth, 15fovergrowth, 15e16, 16fself-catalyst growth of GaN NCs using

MBE, 11e14single non-tapered InN NC and InN NC

ensemble grown, 15fNitride nanostructures based on NCs, 16e18CSNCs, 17e18InGaN/GaN CSNC white light LED,

17fquantum disks embedded in NCs,

16e17selective area growth of NCs, 18

NitrideNCmaterials. SeeNitride nanocolumnmaterials (Nitride NC materials)

Nitride semiconductors, 4MOCVD growthblue MQW, 32e35carbon incorporation and Mg doping ofGaN, 31e32, 31f, 33f

GaN growth rate distribution, 28fgreen MQW, 32e35, 36fgrowth mechanism, 28e30horizontal flow production reactor, 27freaction diagram of bimoleculardecomposition path, 30f

transition energy of each combination ofmetals, 30t

UV materials growth, 35e40vertical flow reactor types, 25f

Nitride-based InGaN/GaN MQWs, 9Nitride-based white LEDs, 441Nitrogen interstitials (Ni), 480Nitrogen vacancies (VN), 480Non-contact measurements, 754e755Non-local TAT mechanisms, 456e459Non-radiative recombination, 299e300,

457e458Non-transparent metal contact stacks, 329Nondestructive characterizations, 446e447Nonpolar and semipolar LEDs, 277e285advantages, 276e277anisotropic strain influence, 283e285band profiles of InGaN MQWs for different

growth orientations, 282fcalculated polarization charge density, 281fchanges in piezoelectric polarization charge

with orientation, 278e281conventional c-plane LEDs limitations,

273e277crystallography of wurtzite nitride, 277epitaxial growth, challenges in, 285e288heteroepitaxy of nonpolar and semipolarplanes, 285

homoepitaxy and need for bulk GaNsubstrates, 285e286

Indium incorporation in nonpolar andsemipolar planes, 286e287

morphology of nonpolar and semipolarepitaxy, 287e288

strain-induced defect generation innonpolar and semipolar epitaxy, 288

Index 783

Nonpolar and semipolar LEDs (Continued)influence of polarization on band bending,

QCSE and carrier transport, 282e283light extraction for nonpolar and semipolar

LEDs, 288e291new and old coordinates, 279fpolar, nonpolar and semipolar planes in

wurtzite crystal structures, 278fwurtzite crystal planes, 277t

Nonpolar directions, GaN growth on, 65e67Nonpolar GaN LEDs on silicon, 106e108Nonpolar planes, 277Nonpolar substrates, 394e395

computed polarization charge density inInGaN/GaN QWs, 395f

nonpolar and semipolar planes in III-nitrides, 395f

Nonradiative process, 382Novel surface patterning design, 289NPLED. See LED grown on NPSi (NPLED)NPs. See Nanopillars (NPs)NPSi. See Nano-patterned silicon (NPSi)NRs. See Nanorods (NRs)NTSC. See National Television System

Committee (NTSC)Nucleation layer, 55e56

in strained heteroepitaxy, 59

OODRs. See Omnidirectional reflectors

(ODRs)Office lighting, 543OLEDs. See Organic light-emitting diodes

(OLEDs)OM. See Optical microscopy (OM)Omnidirectional reflectors (ODRs),

210e2111-Dimension (1D)

bulk drift-diffusion analysis, 462LED array, 664nanocolumns, 11PCs, 227Schr€odinger-Poisson equation, 313e315

Optical axis, 712Optical design

design factors, 634direct-type BLUs, 638e640edge-type BLUs, 635e638key design considerations, 635, 635f

Optical devices, 387Optical effectsof freeform lenses, 511e514continuous freeform lens method, 514discontinuous freeform lens method,513e514

SMS method, 512e513tailored freeform surface design method,511e512

of gold wire bonding, 502e505of phosphor coating, 505e511

Optical films and plates, 624e625Optical imaging methods, 444e446Optical microscopy (OM), 444“Optically flat” GaN epitaxial layer, 54e55Optics, 753e754Optoma Technology, 737e738Orange-yellow emitting Eu2+-activated

Li2SrSiO4 phosphors, 129e131Organic light-emitting diodes (OLEDs), 327Oscillator strength, 390e393decrease in, 390e393, 392f

OSRAM Opto Semiconductors, 637Outdoor lighting, 545, 545fOxidation barrier layer, 211(Oxy)nitride phosphors, 128e129

PP-doped layers, 328P-doping aluminum-rich AlGaN alloys, 3p-GaN layers, 213p-n junction (PNJ), 448e450diodes, 377e380, 659e660LED, 383e384

P-side up lateral current spreading LEDs,328e329, 328f

P-side up LEDs, PC-enhanced lightextraction in

fabrication of P-side up LEDs, 347e350CSL, 349EBL, 349fabricated PQC LED, 351fMESAs and contacts, 349e350NIL, 348PC patterning options, 347photonic crystal etching, 349thermal nano-imprint lithography, 348UV lithography, 347e348UV-NIL, 348

784 Index

PC-LED performance, 358e361, 359fPQC LED electro-optical performance,

363P-type AlGaN layer, 581p-type ohmic contacts, 102e103Package size and chroma selection

flexibility, 627Packagingsubstrate, 521e522thermal design, 517e519

PAMBE. See Plasma-assisted molecular-beam epitaxy (PAMBE)

Parabolic function, 716Partial directional polarisation-dependent

band gaps, 339Passivation layers effect on PQC

performance, 365Passive cooling, 523Passive driver, 709e711Patterned sapphire substrate (PSS), 59e60,

243e244, 288e289, 315, 335growth of GaN on, 63e65

Patterned substrate LEDs, 330PBG. See Photonic bandgap (PBG)PC. See Polycarbonate (PC)pc-wLEDs. See Phosphor converted wLEDs

(pc-wLEDs)PCBs. See Printed circuit boards (PCBs)PCs. See Photonic crystals (PCs)PE-CVD. See Plasma-enhanced chemical

vapor deposition (PE-CVD)Peak intensity axis, 712PEC etching techniques.

See Photoelectrochemical etchingtechniques (PEC etching techniques)

PECVD. See Plasma enhanced chemicalvapor deposition (PECVD)

Pendeo-epitaxy, 60of GaN, 62e63

Perfectly matched layer (PML), 352PFM. See Pulse frequency modulation

(PFM)PhCs. See Photonic crystals (PCs)Phonon emission (e-ph), 463Phosphate, 151e154Phosphor coating, 500e501conformal coating, 501freely dispersed coating method,

500

optical effects, 505e511arrangement of phosphor in white LEDs,506f

packaging methods in analysis, 508fphosphor geometry, 508e511phosphor location, 505phosphor thickness and concentration,507e508, 508t

phosphor-converted LEDs, 510fremote coating, 501self-heating, 519technologies, 500f

Phosphor converted wLEDs (pc-wLEDs),124e125, 127e128

Phosphors, 123e124embedded silicone matrix, 500geometry, 508e511layer, 500location, 505phosphor-free white light emission,

428e429thermal quenching, 515e517thickness and concentration, 507e508,

508tPhosphors for white LEDsfuture development, 192e193key parameters, 125e126new LEDs phosphorsby codoped activators and energytransfer, 169e192

with different host systems, 134e168principle on fabrication, 124e125state-of-the-art(oxy)nitride phosphors, 128e129Mn4+-activated fluoride phosphors,132e133

photoluminescence excitation, 127fsilicates phosphors, 129e132YAG:Ce phosphor and modification,127e128

Photoelectrochemical etching techniques(PEC etching techniques), 288e289

Photoemission intensity, 590Photolithography, 250e251Photoluminescence (PL), 6e7, 33, 86,

123e124, 275e276, 368, 555e556,563fe564f, 567fe568f, 587e588

intensity, 109fintensity of multi quantum well, 34f

Index 785

Photoluminescence (PL) (Continued)measurement, 261, 262fspectra of GaN on AlAs buffer on silicon,

86ftuning, 126, 172

Photoluminescence excitation (PLE), 127fof BMA, 166fof Ca2.985eyEu0.015MgySi2O7 phosphor,

145fCaAlSiN3:Eu

2+ and Sr(Mg3SiN4]:Eu2+

phosphors, 128fCe3+-doped Y3Al5O12 (YAG:Ce)

phosphor, 127fK2TiF6:Mn4+ phosphor, 133f

Photoluminescence quantum yields(PLQY), 131e132

Photometric quantities, 653Photometric system, 653Photon energy, 337Photonic bandgap (PBG), 227, 228f,

335e337Photonic crystal nitride LEDs. See also

III-nitride (III-N); Nitride LEDepitaxial materials, 327e328example ray paths inside LED, 332fimproving LED extraction efficiency,

341e347light extraction cone, 331flight-trapping, 330e334methods of improving light extraction,

334e335modelling PC-LEDs, 350e365PC-enhanced light extractionin N-side up LEDs, 365e373in P-side up LEDs, 347e350

PCs technology, 335e341types, 328e330

Photonic crystals (PCs), 225, 250effect, 18etching, 349nanostructures and PC LEDs, 227e229Brillouin zones for 2D square PC lattices,231f

photonic crystals with periodicity, 227f2D PC structure with Bragg diffractionphase-matching diagrams, 230f

various extraction methods using PCs,228f

PC-induced effective index, 233e234

technology, 335e341classes of PC device, 337e340regular PCs vs. PQCs, 340e341with triangular symmetry, 336fworkings, 336e337

Photonic quasi-crystals (PQCs), 248,340e341

Photonic quasi-crystals-light-emittingdiodes (PQC LEDs), 351f

fabricated, 351fFDTD limitations for modelling PC/PQC

LEDsboundary condition considerations, 353comparative normalisation of results, 355considerations for collection planeplacement, 353e355

FDTD simulation, 354flight launch considerations, 352e353modeling tools, 355requirements for extent of cross-sectionalprofile, 355

effect of TCSL and passivation layers, 365Photopic luminance function, 665e666Photopic vision, 665e666Photoreceptors, 665Physical vapor deposition (PVD), 59buffer layer grown by, 59

Pico projectors, 748electrically tunable liquid crystal lens

application, 754, 754fPiezoelectric polarization (Ppz), 278e280,

388GaN wurtzite crystal structure, 388f

Pixel control, 692e706attainable PWM grayscale resolution vs.

refresh frequency, 698t, 700tcolor gamut shift, 695fdriver output current vs. drop-out voltage,

702fexternal voltage source application, 705fluminous intensity variation, 694foutput current regulation, 704fprogrammable LED constant current driver,

703fPWM dimming time diagram, 695fseries resistor application for LED current

control, 700fswitch output driver topology, 701ftransferring image bitplanes, 696f

786 Index

voltage-current relationship, 693fwaveforms, 697f

Pixel(s), 430, 682e683intensity, 680e681pitch, 677spacing, 663e664

PL. See Photoluminescence (PL)Planck’s constant, 390e393Plasma enhanced chemical vapor deposition

(PECVD), 244e246Plasma-assisted molecular-beam epitaxy

(PAMBE), 3, 593e594growth of nitride epilayers and quantum

structures, 4e11AFM images of GaN grown, 7fAlN epilayers, 8developments and milestones duringMBE-grown LEDs and LDs, 10t

doping in nitride materials, 9e10, 10fGaN epilayers, 5e8InGaN epilayers, 8InN epilayers, 8light emitters based on nitride MQWs,10e11

nitride-based InGaN/GaN MQWs, 9plot of Ga flux vs. growth temperature, 6fRHEED patterns for GaN epilayers, 5f

Plasma-enhanced chemical vapor deposition(PE-CVD), 63, 347

Platforms, 547e549PLE. See Photoluminescence excitation

(PLE)PLQY. See Photoluminescence quantum

yields (PLQY)PML. See Perfectly matched layer (PML)PMMA. See Polymethyl-methacrylate

(PMMA)PNJ. See p-n junction (PNJ)Point spread function (PSF), 715e716effect, 309e310

Poisson equations, 377e379, 459Polar angle (Q), 334f, 713Polar c-plane, nonpolar/semipolar GaN,

312e315, 312fPolariton LEDs, 405Polarizationcontrol, 337direction, 282e283effects in III-nitride LEDs, 387e397

polarization-dependent band gaps, 339polarization-matched LEDs, 395e396,

396fpolarization-related internal electric field,

593Polarization-induced doping, 403p-type doping problem in, 396e3973D polarization-induced n-and p-type

doping, 397fPolarization-induced electric fields,

389e393decrease in oscillator strength, 390e393metal-polar nitride light-emitting diode,

391fred-shifted photon emission, 389e390reduction in overlap of electron and hole

envelope wave functions, 392fPolycarbonate (PC), 624Polymethyl-methacrylate (PMMA), 624Powder GaN, 44Power efficiency, 282e283Power extraction, 366Power function, 716PQC LEDs. See Photonic quasi-crystals-

light-emitting diodes (PQC LEDs)PQCs. See Photonic quasi-crystals (PQCs)Praseodymium (Pr), 587Printed circuit boards (PCBs), 497e498,

663e664Projectiondisplays, 754e755, 756fsystem, 753

Projector, 737, 751applications, 754e756screen, 739technologies, 738e754image generation, 747e753light source, 740e747optics, 753e754

“Pseudo-mask”, 62e63PSF. See Point spread function (PSF)PSS. See Patterned sapphire substrate

(PSS)Pulse frequency modulation (PFM), 696Pulse width modulation (PWM), 694e695dimming time diagram, 695f

Purcell effect, 228, 228fPVD. See Physical vapor deposition (PVD)PWM. See Pulse width modulation (PWM)

Index 787

QQCDs. See Quantum cascade detectors

(QCDs)QCL. See Quantum cascade laser (QCL)QCSE. See Quantum-confined Stark effect

(QCSE)QD. See Quantum dots (QD)QDIPs. See Quantum dot IR

photoconductors (QDIPs)QE. See Quantum efficiency (QE)“Quantum box”, 459Quantum cascade detectors (QCDs), 602,

603fQuantum cascade laser (QCL), 591, 604,

605fQuantum disks embedded in NCs, 16e17Quantum dot IR photoconductors (QDIPs),

601Quantum dots (QD), 384e386, 399e401,

588, 599, 600f, 746increasing overlap by using thin, 393e394

Quantum efficiency (QE), 125Quantum structures, PAMBE growth of,

4e11Quantum well IR photodetectors (QWIPs),

601Quantum wells (QWs), 9, 56e57, 93, 273,

302e304, 384e386, 554e555,563fe564f, 567f, 575f, 587, 593f,596f

capture process, 463increasing overlap by using thin, 393e394structures, 110f

Quantum yield, 125Quantum-confined Stark effect (QCSE),

65, 83, 106e108, 273, 300e302,389e390, 587e588

III-nitride properties, 390tQuasi-Fermi levels, 377e379Quaternary alloy InAlGaN, 565Quicker illumination to stable brightness,

628QWIPs. See Quantum well IR

photodetectors (QWIPs)QWs. See Quantum wells (QWs)

RRa. See Color rendering index (CRI)Radiant flux, 665e666

Radiant intensity, 665e670additional gamma correction effect on

image, 670fnonlinear human vision response

approximations, 669fpath of image compression using gamma

function, 668fRadiant power (W), 665e666Radiation stability, 719e721Radiative efficiency, 383e384Radiative mechanisms, 457e458Radiative recombination, 299e300, 307,

377Radio Equipment Directive Directives,

722Radio frequency (RF), 59, 111plasma, 4

Radiometry, 332e334Raman scattering, 261e262Raman spectroscopy, 56e57Random phase approximation (RPA),

464e465Random texturing, 335Rapid degradation in initial stage, 450Rapid switching speed, 627Rapid thermal annealing (RTA), 253e254Rare earth (RE)elements, 172elements, 587ion doping, 174e176

Raster principle, 682e683Ray-tracing simulation, 421e422RC. See Resonant cavity (RC)RCLEDs. See Resonant-cavity light emitting

diodes (RCLEDs)RE-doped GaN emitters, 590e591, 591fReactive ion etching (RIE), 249, 349Recombination mechanisms, 381e383Recording/transmitting signals, 546e547Rector configurations, 25e26Red-green-blue (RGB)devices, 416LEDs, 415e416, 418, 422e424, 426, 621phosphors, 554pixels forming by jet-printing, 432e434cross-sectional view of a single pixel,433f

jet-printing of fluorescent microspheres,434f

788 Index

Red-shifted photon emission, 389e390III-nitride properties, 390t

Reflective high-energy electron diffraction(RHEED), 5

patterns for GaN epilayers, 5fReflective projector technologies, 738, 738fRegulators’ views, 540e542, 541fRelaxation time, 327Reliability of nitride LEDs, 441degradation mechanisms, 447e452evaluation of LED degradation, 444e447testing, 441e444

Remote coating, 501phosphor coating, 501

Residential lighting, 546, 546fResidual stress, 261Residue theorems, 465e466Resonant cavity (RC), 404Resonant-cavity light emitting diodes

(RCLEDs), 404Response time of LED, 627RF. See Radio frequency (RF)RGGB LED clusters, 640, 644fRHEED. See Reflective high-energy

electron diffraction (RHEED)RIE. See Reactive ion etching (RIE)RMS. See Root mean square (RMS)Robustness, 628Rod integrator, 753e754Room temperature (RT), 555, 567fe568fCL spectrum, 114e117PL spectra, 112e114, 113f

Room temperature vulcanization (RTV), 684RoosbroeckeShockley model, 381Root mean square (RMS), 90Rough surface morphology, 287Rounding error, 691RPA. See Random phase approximation

(RPA)RT. See Room temperature (RT)RTA. See Rapid thermal annealing (RTA)RTV. See Room temperature vulcanization

(RTV)

SS-PWM. See Scrambled pulse width

modulation (S-PWM)SAE. See Society of Automotive Engineers

(SAE)

SAEG. See Selective-area epitaxial growth(SAEG)

SAG. See Selective-area growth (SAG)SAG-ELO. See Selective-area growth-

epitaxial lateral overgrowth(SAG-ELO)

SALEO. See Selective-area lateral epitaxialovergrowth (SALEO)

Salt atmosphere test, 441Sapphire, 211e212, 225, 258, 329, 403growth of low TDD AlN layers on,

557e561surface, 246

Sapphire shaping structure (SS structure),218

SS-FCLEDs, 218e219Sapphire substrates, 46e52, 243e244comparison to other substrates, 48e52native substrates, 52, 53tother foreign substrates based on oxides,sulfides, and metals, 51, 53t

SiC, 49e50Silicon, 50e51

LEDs on, 67properties of sapphire for substrates of III-N

materials, 47e48strained heteroepitaxial growth, 52e59buffer layer grown by PVD, 59development and demonstration, 54e55growth mechanism of GaN, 55e56, 56fprinciple of wafer curvaturemeasurement, 57f

two-step strained heteroepitaxy, 54e55wafer bowing during growth of GaN onsapphire, 56e59

Scanning electron microscopy (SEM), 12f,87, 213, 249, 422e424, 444

Scanning transmission electron microscope(STEM), 317e318

Schr€odinger equations, 459Scotopic curve, 665e666Scrambled pulse width modulation (S-

PWM), 696SDO. See Serial data output (SDO)SE. See Selective epitaxy (SE)Second generation LED stack with inclined

sidewalls, 419e424laser micromachining setup, 420fplot of light extraction efficiency, 422f

Index 789

Second generation LED stack with inclinedsidewalls (Continued)

scanning electron microscope of TP-LED,421f

SEM LED stack assembling, 423fsidewall light leakage still evident from

stacked LED, 423ftypical light rays propagating in TP-LED,

419fSecond harmonics generation (SHG), 561Second order Feynman diagrams, 464e465,

464fSecond-order non-linear MongeeAmp�ere

equations, 513Secondary ion mass spectroscopy (SIMS),

32, 32f, 86, 446e447, 468SEG. See Selective epitaxial growth (SEG)Selective epitaxial growth (SEG), 60Selective epitaxy (SE), 60Selective growth (SG), 60Selective lift-off (SLO), 101e102, 102fSelective-area epitaxial growth (SAEG),

60Selective-area growth (SAG), 60, 106e108Selective-area growth-epitaxial lateral

overgrowth (SAG-ELO), 106e108Selective-area lateral epitaxial overgrowth

(SALEO), 60Self-assembled GaN QDs, 403e404Self-catalyst growth of GaN NCs using

MBE, 11e14“Self-doping” effect, 65Self-heating of phosphor coating, 519SEM. See Scanning electron microscopy

(SEM)Semi-classical model, 459Semiconductor

device simulators, 456light sources, 553e554wafer bonding, 225e226

Semipolar directions, GaN growth on,65e67

Semipolar GaN LEDs on silicon, 106e108Semipolar planes, 277Semipolar substrates, 394e395

computed polarization charge density inInGaN/GaN QWs, 395f

nonpolar and semipolar planes inIII-nitrides, 395f

SensitizersET models using Eu2+ as, 183e192ET models with Ce3+ as, 169e183

Sensor electronic technology (SET), 556Separation by implantation of oxygen

process (SIMOX process), 95e98Serial data input (D), 688e689Serial data output (SDO), 688e689SET. See Sensor electronic technology

(SET)SG. See Selective growth (SG)Shallow etched PC structures, 346SHG. See Second harmonics generation

(SHG)Shockley, Read, and Hall theory (SRH

theory), 381e382coefficients, 299e300, 322recombination, 457e458

Shockley boundary condition, 379Shop lighting/accent lighting, 544, 544fShortperiod superlattice (SPS), 213Shower head reactor, 25e26SiC substrates. See Silicon carbide

substrates (SiC substrates)Side-emitting lenses, 495Signal lighting, 651e652Photometric and color requirements, 651t

Signatures of diffusion-like processes,469e474

Silane (SiH4) ), 253e254, 263e264SiLENSe software package, 313e315Silicates, 140e146phosphors, 129e132

Silicon (Si), 9e10, 50e51, 79advantages, 80e82AlN atomic orientation on, 84fcrystallography, 82e83

Silicon carbide substrates (SiC substrates),45, 49e50

Silicon nitride, 87Silicon substrateschallenges for GaN growth on, 83e86thermal expansion mismatch betweenGaN and silicon, 83e85

thermal management, 85e86GaN, 79e80

Silicon X-tal reflective display (SXRD),737, 753

Silicon-doped n-GaN layer, 231

790 Index

Silicon-on-insulator (SOI), 95e98Silicone gel, 502Silicone injection, 502SIMOX process. See Separation by

implantation of oxygen process(SIMOX process)

Simple LED display, 686e688, 687fSimplified Shockley diode equation,

458e459SIMS. See Secondary ion mass spectroscopy

(SIMS)Simultaneous multiple surfaces method

(SMS method), 512e513, 513fSingle plasmon pole dielectric function (SPP

dielectric function), 464e465Single quantum well (SQW), 313e315, 317,

384e385, 387InGaN LED structures, 397e398LED test devices, 457e458

Single-crystallineepitaxial GaN films, 45oxide materials, 51

Single-step laser-micromachining approach,419

SK growth mode. See StranskieKrastanowgrowth mode (SK growth mode)

Skew, 708SLEDs. See Superluminescent light-emitting

diodes (SLEDs)SLO. See Selective lift-off (SLO)SLS. See Strained layer superlattice (SLS)SLs. See Superlattices (SLs)Small-gamut spaces, 672Small-scale periodic pattering, 335“Smart and connected” lighting, 539SMDs. See Surface-mounted devices

(SMDs)SMPTE standard EG-18e1994, 679SMS method. See Simultaneous multiple

surfaces method (SMS method)Snell’s law, 421e422, 493e494, 513e514Society of Automotive Engineers (SAE),

647Soft failure, 450SOI. See Silicon-on-insulator (SOI)Solar cells, 375Solid angle, 332e334Solid-state lighting (SSL), 43, 511Spatial beam-steering, 337

Spatial distribution, 677e681comparison of high and low fill factor, 679fconventional LED construction, 680fMannos and Sakrison model of CSF, 678freal and virtual pixels explanation, 677f

Spatial modulation, 747e753Speckle, 747Spectra Physics, 419e420Spinel (MgAl2O4), 43Spontaneous polarization, 278, 388Sports lighting, 545, 545fSPP dielectric function. See Single plasmon

pole dielectric function (SPPdielectric function)

Spreading thermal resistance, 521e522SPS. See Shortperiod superlattice (SPS)SQW. See Single quantum well (SQW)Sr(Mg3SiN4]:Eu

2+ phosphor, 128fSr[LiAl3N4] phosphor, 159e160, 161fSr[LiAl3N4]:Eu

2+ phosphor, 129, 157Sr2.975exBaxCe0.025AlO4F (SBAF:Ce3+)

phosphor, 136e137Sr2MgAl22O36:Mn4+ phosphor, 137e138Sr2SiO4, 129e131Sr8MgLa(PO4)7 (SMLP) phosphor, 151Sr8MgY(PO4)7 (SMYP) phosphor, 151sRGB, 675SRH theory. See Shockley, Read, and Hall

theory (SRH theory)SrMgAl10O17:Mn4+ phosphor, 137e138SrxBa2-xSiO4:Eu

2+ solid solution phosphor,129e131

SS structure. See Sapphire shaping structure(SS structure)

SSL. See Solid-state lighting (SSL)Stacked LEDs, 422e424. See also White

LEDsinitial idea, 416e418initial version of LED stack, 418flight-emitting diode chip stacking, 417ftransmission of light rays, 417f

second generation LED stack with inclinedsidewalls, 419e424

third generation tightly integrated chip-stacking approach, 424e429

Stage illumination, 754e755Standard direct wafer bonding, 225e226Standard GaInN/GaN LED structure, 305Standing waves, 748

Index 791

Stationary installation, 661e662STEM. See Scanning transmission electron

microscope (STEM)Stoney’s equation, 58e59Strain-free high-quality single crystals,

12e13Strain-free homoepitaxy, 285e286Strain-induced

defect generation in nonpolar and semipolarepitaxy, 288

piezoelectric polarization, 280repulsive effects, 287

Strained heteroepitaxial growth on sapphiresubstrates, 52e59

buffer layer grown by PVD, 59development and demonstration, 54e55two-step strained heteroepitaxy, 54e55

growth mechanism of GaN, 55e56, 56fprinciple of wafer curvature measurement,

57fwafer bowing during GaN growth on

sapphire, 56e59Strained layer superlattice (SLS), 98e99StranskieKrastanow growth mode (SK

growth mode), 12, 399e401Street lighting, 542, 542fStress-free GaN substrate, 261Strip-type LED display modules, 663Structural design, 124, 133e135Subthreshold forward current, 456e459Sulfides, 146e151Superlattices (SLs), 88e90, 231, 557Superluminescent light-emitting diodes

(SLEDs), 404Suppression of vertical emission, 573Surface mount LEDs, 682e683Surface patterning, 335Surface reconstruction, 5e6Surface-mounted devices (SMDs), 495Surface-patterned LED, 346, 355Susceptor carrier, 56e57Sweeping turn signals, 651e652Switch driver, 700SXRD. See Silicon X-tal reflective display

(SXRD)Syndiant, 737e738Synthetic sapphire (Al2O3), 46e47

ALD of, 90e93, 92fSystem testing, 494e495

TTAB. See Tape-automated bonding (TAB)Tailored freeform surface design method,

511e512, 512fTape-automated bonding (TAB), 211TAT mechanisms. See Trap-assisted

tunneling mechanisms (TATmechanisms)

TCL. See Transparent conductive layer(TCL)

TCSL. See Transparent current spreadinglayer (TCSL)

TDD. See Threading dislocation density(TDD)

TDs. See Threading dislocations (TDs)TE modes. See Transverse electric modes

(TE modes)Technical/mathematical challenges,

549TEM. See Transmission electron

microscopy (TEM)Temperature cycling test, 441Temporal performance terms, 682Temporary installation, 661Texas Instruments, 737e738TFFC LED. See Thin-film flip-chip LED

(TFFC LED)TFLED. See Thin-film LEDs (TFLED)TFT-LCDs. See Thin-film-transistor liquid

crystal displays (TFT-LCDs)Thermal design and processing of LED

packaging, 515e524contact thermal resistance and TIM,

519e520cooling solutions for LED applications,

523e524self-heating of phosphor coatingspreading thermal resistance and packaging

substrate, 521e522thermal design of packaging, 517e519

Thermal droop, 299e302in GaInN/GaN LEDs, 320e322. See also

Low-droop GaInN/GaN LEDsSRH recombination and high-orderrecombination, 321f

thermal droop vs. temperature and EQE,322f

Thermal expansion mismatch between GaNand silicon, 83e85

792 Index

Thermal interface material (TIM), 519e520,520f

categories and general properties, 521tcontact face of an interface, 520f

Thermal management, 85e86, 494Thermal nano-imprint lithography, 348Thermal resistanceof LED, 720network, 517, 518f

Thermal shock test, 441Thermal stress, LEDs degradation and,

467e481Thermosonic bonding, 499Thin-film flip-chip LED (TFFC LED),

289e290Thin-film GaN, 45e46Thin-film LEDs (TFLED), 225Thin-film-transistor liquid crystal displays

(TFT-LCDs), 619Third generation tightly integrated chip-

stacking approach, 424e429Third-order non-radiative recombination,

307e308Thorium oxide (ThO2), 43Threading dislocation density (TDD),

243e244, 556Threading dislocations (TDs), 6e7, 99e100,

287e288, 565e566(3031) GaInN/GaN LEDs, 317e318Three-dimensional growth (3D growth), 5Three-side-approach for lighting

applicationsmanufacturers’ abilities, 539e540, 541tmarket needs, 531e539regulators’ views, 540e542

3M Company, 737e738Thulium (Tm), 587THz generation, 588e589TIM. See Thermal interface material (TIM)Time-and space-varying independent

speckle pattern, 747TIR. See Total internal reflection (TIR)Titanium oxide (TiO2), 554TM polarization. See Transverse magnetic

polarization (TM polarization)TMA. See Trimethyl-aluminum (TMA)TMAl. See Trimethyl-aluminum (TMA)TMG. See Trimethyl-gallium (TMG)TMGa. See Trimethyl-gallium (TMG)

TMI. See Trimethyl indium (TMI)TMIn. See Trimethylindium (TMIn)TO-18. See Transistor outline-18 (TO-18)Top surface metallurgy (TSM), 211Top-down method, 244Top-down technique for nanostructured

LED, 244e262. See also Bottom-uptechnique for GaN nanopillarsubstrates

freestanding high quality GaN substrate,258e262

high extraction efficiency GaN-based LED,248e252

highly efficient and bright LEDsovergrown, 253e257

nanoscale epitaxial lateral overgrowth ofGaN-based light emitting diodes,244e248

Total internal reflection (TIR), 210, 330effect, 210e211phenomenon, 492e493

TP-LED. See Truncated pyramidal geometry(TP-LED)

Traditional laser micromachining, 419e420“Traditional” IIIeV semiconductors, 52Transconductance amplifier, 709e711Transistor outline-18 (TO-18), 247Transmission electron microscopy (TEM),

6e7, 99e100, 99f, 231, 246, 252,444, 446e447, 560e561,562fe563f, 578f, 594, 595f

Transparent conductive layer (TCL), 213,329, 446e447

Transparent current spreading layer (TCSL),365

Transverse electric modes (TE modes),233e234, 595

Transverse magnetic polarization (TMpolarization), 594, 596f

Trap-assisted tunneling mechanisms (TATmechanisms), 456e458

Trimethyl indium (TMI), 34, 34fTrimethyl-aluminum (TMA), 29, 558Trimethyl-gallium (TMG), 28e29, 29f,

253e254, 263e264TMG-NH3, 28e29, 29f

Trimethylindium (TMIn), 253e254,263e264

Tristimulus values, 125

Index 793

“True bulk” GaN substrates, 52Truncated pyramidal geometry (TP-LED),

419e421, 419fTSM. See Top surface metallurgy (TSM)Tungstates/molybdates based red phosphors,

167e168Tungsten-halogen lamps, 647Tunneling, 463“Tunneling-into-traps” model, 457e458(2021) GaInN/GaN LEDs, 317e318Two-dimension (2D), 625

deflection axis gimbaled micromirror, 748,749f

growth, 5raster scan, 748, 748f

Two-step strained heteroepitaxy, 54e552D finite-difference time domain method

(2D-FDTD method), 182D square PC lattice, 229

UUBM. See Under bump metallurgy (UBM)Ultrahigh performance lamps (UHP lamps),

737, 740Ultraviolet (UV), 28, 209, 418

lithography, 347e348materials growth, 35e40

Ultraviolet LEDs (UV LEDs), 396. See alsoWhite LEDs

AlGaN-based DUV-LEDs fabricated onhigh-quality AlN, 568e574

growth of low TDD AlN layers on sapphire,557e561

increase in EIE and LEE, 576e582marked increases in IQE, 561e567maximum output power of AlGaN-and

InAlGaN-based DUV LED, 584fresearch background of DUV-LEDs,

553e557Under bump metallurgy (UBM), 211United States Department of Energy (DOE),

209, 541Unpatterned N-side up LED, microcavity

effect in practice for, 368e370Unpatterned P-side up LED simulation

example, 356e358, 356fFDTD simulation, 357fsimulation results, 357e358simulation set-up, 356e357

Unpatterned smooth-surface LED, 355Unselected LEDs, 638e640Ushio UXW-15KD 15 kW water cooled

xenon lamp, 740, 740fUV. See Ultraviolet (UV)UV LEDs. SeeUltraviolet LEDs (UV LEDs)UV Nano-imprint lithography (UV-NIL),

348

VV/III flux ratio, 11e12Valence band, 283e285Van Arkel-de Boer process, 44Vapor phase epitaxy (VPE), 386Vapor-liquid-solid (VLS), 11e12Vehicle headlamp systems, 650Vertical current spreading design, 329e330Vertical flow reactors types, 25e26, 25fVibration test, 441Video processor, 689, 691Vikuiti BEF II and BEF III, 624, 624tVirtual sampling planes, 352Visible LEDsblue, green, and white LEDs, 49GaN growth on nonpolar and semipolar

directions, 65e67sapphire substrates, 46e52epitaxial GaN on, 43e46epitaxial overgrowth of GaN, 59e65LEDs on, 67strained heteroepitaxial growth, 52e59

Visual acuity, 678VLS. See Vapor-liquid-solid (VLS)VolmereWeber growth mode (VW growth

mode), 12Voltage binning, 539VPE. See Vapor phase epitaxy (VPE)VPL-FHZ700L, 745VW growth mode. See VolmereWeber

growth mode (VW growth mode)

WWafer bowing during growth of GaN on

sapphire, 56e59Wafer carrier, 56e57Wafer-bonding technology, 225e226Wall-plug efficiency, 209e210“Warm white” color, 531Waveguide circle, 236

794 Index

Wavelength multiplexing technology(WMT), 744e745

Weber Contrast, 667Weber fraction, 667Weber’s law, 665e666Wentzel-Kramers-Brillouin formalism,

457e458Wetting layer, 211White LEDs, 123, 531. See also Stacked

LEDs; Ultraviolet LEDs (UV LEDs)certification, 549fabrication process, 123fields of application, 542e546

automotive lighting, 543industry/factory/public lighting, 544office lighting, 543outdoor lighting, 545residential lighting, 546shop lighting/accent lighting, 544sports lighting, 545street lighting, 542

LED light sources in connected world,546e547

phosphorsfuture development, 192e193new advances of future, 133e192

platforms, alliances, and consortia,547e549

technical/mathematical challenges, 549three-side-approach for lighting

applications, 531e542x-y coordinates determining white areas,

532tx-y diagram of color locations, 533f

White phosphor-converted LEDs, 652Wide adjustment range for brightness,

contrast, and chromaticity, 627Wide color gamut displays, 627, 675Wide operating temperature range, 626Wide-gamut spaces, 673e674Wire bonding, 492, 493f, 499copper wire, 499gold wire, 499

Wireless, 537WLAN Gateway, 537, 537fwLED solid-state lighting technology,

192e193WMT. See Wavelength multiplexing

technology (WMT)

Wurtzite (WZ), 554e555crystallography of wurtzite nitride, 277wurtzite III-nitride materials, 312

WZ. See Wurtzite (WZ)

XX ray diffraction full width at half maximum

(XRD FWHM), 35, 37fX-ray diffraction (XRD), 86X-ray diffraction and u-scan rocking curves

(XRCs), 557e558, 558f, 560fX-ray photoelectron spectroscopy (XPS), 87Xenon lamp, 737, 740arc lamp, 740, 740femission spectrum, 740, 740f

XPS. See X-ray photoelectron spectroscopy(XPS)

XRCs. See X-ray diffraction and u-scanrocking curves (XRCs)

XRD. See X-ray diffraction (XRD)XRD FWHM. SeeX ray diffraction full width

at half maximum (XRD FWHM)

Y(Y,Gd)3(Al,Ga)5O12:Ce phosphors,

127e128Y3Al5O12:Ce

3+ phosphor, 134Y3Mg2AlSi2O12 phosphor, 136YAG:Ce phosphor and modification,

127e128YAG:Ce yellow phosphor, 124Yellow-emitting Sr3Ce(PO4)3:Eu

2+

phosphors, 151e152Yellow-emitting Y3Al5O12:Ce

3+ (YAG:Ce)phosphor, 134

Yellowish-green Eu2+-activated CaLaGa3S7chalcogenide phosphor, 151

ZZ-Phosphor 3 LCD hybrid laser source

projectors, 745Zero-phonon line (ZPL), 163e164Zhaga consortium, 549Zinc oxide (ZnO), 51, 86e87Zinc selenide (ZnSe), 621ZPL. See Zero-phonon line (ZPL)

Index 795


Color Plate

Plate 1 (Chapter 4) (a) Optical microscope image of green LEDs grown on SOI (111) underelectrical probing and the I-V curve of an LED; (b) EL spectra of LEDs grown on Si (111) andSOI (111) under 100 mA current injection (Tripathy et al., 2007).

Plate 2 (Chapter 6) Photons from (a) a conventional flat-surface FCLED and (b) a MPA-FCLED for a DC injection current of 350 mA.

Plate 3 (Chapter 6) Normalized three-dimensional far-field patterns for (a) C-FCLEDs and (b)SS-FCLEDs.

798 Color Plate

Plate 4 (Chapter 6) Angular-resolved spectra for lattice constant (a) a ¼ 290 nm,(b) a ¼ 350 nm, and (c) a ¼ 400 nm, where GX direction points to the left and the GM directionpoints to the right. (d) Free-photon band structure calculated with n ¼ 2.42 for the transverseelectric modes. The red thick lines are the collinear coupled modes. The red dashed lines are thenoncollinear coupled bands. The boxes show the experimental regions for a ¼ 290, 350, and400 nm. The insets in the boxes are the calculated band structures.

Color Plate 799

Plate 5 (Chapter 6) Intensity map of the extracted light at a fixed wavelength l ¼ 470 nm for (a) a ¼ 290 nm, (b) a ¼ 350 nm, and (c) a ¼ 400 nm.Symbols show values calculated using Bragg’s diffraction theory fitted with effective refractive index neff ¼ 2.414 (cyan crosses for the GX direction,green circles for the GM direction, and blue squares for the GXGM direction). Top view of 3D far-field pattern for (d) a ¼ 290 nm, (e) a ¼ 350 nm, and(f) a ¼ 400 nm.

800Color

Plate

Documents

Nitride Semiconductor Light-Emitting Diodes (LEDs), Second Edition: Materials, Technologies, and Applications