Processing of Wide Band Gap Semiconductors

WIDE BANDGAP SEMICONDUCTORS

This page intentionally left blank

WIDE BANDGAP SEMICONDUCTORS Growth, Processing and Applications

Edited by

Stephen J. Pearton University of Florida Gainesville, Florida

NOYES PUBLICATIONS Park Ridge, New Jersey, U.S.A. WILLIAM ANDREW PUBLISHING, LLC Norwich, New York, U.S.A.

Copyright 9 2000 by Noyes Publications No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage and retrieval system, without permission in writing from the Publisher. Library of Congress Catalog Card Number: ISBN: 0-8155-1439-5 Printed in the United States Published in the United States of America by Noyes Publications / William Andrew Publishing, LLC Norwich, New York, U.S.A. 1098765432

1

Library of Congress Cataloging-in-Publication Data Pearton, S. J. Processing of wide bandgap semiconductors / by Stephen J. Pearton. p. cm. Includes bibliographical references. ISBN 0-8155-1439-5 1. Semiconductors--Design and construction. 2. Wide gap semiconductors. Compound semiconductors. I. Title. TK7871.85.P395 621.3815'2--dc21

2000 00-027325 CIP

MATERIALS SCIENCE AND PROCESS TECHNOLOGY SERIES

Editors Rointan F. Bunshah, University of California, Los Angeles (Series Editor) Gary E. McGuire, Microelectronics Center of North Carolina (Series Editor) Stephen M. Rossnagel, IBM Thomas J. Watson Research Center (Consulting Editor)

Electronic Materials and Process Technology CHARACTERIZATION OF SEMICONDUCTOR MATERIALS, Volume 1: edited by Gary E. McGuire CHEMICAL VAPOR DEPOSITION FOR MICROELECTRONICS: by Arthur Sherman CHEMICALVAPOR DEPOSITION OF TUNGSTEN ANDTUNGSTEN SlLICIDES: by John E. J. Schmitz CHEMISTRY OF SUPERCONDUCTOR MATERIALS: edited by Terrell A. Vanderah CONTACTS TO SEMICONDUCTORS: edited by Leonard J. Brillson DIAMOND CHEMICAL VAPOR DEPOSITION: by Huimin Liu and David S. Dandy DIAMOND FILMS AND COATINGS: edited by Robert F. Davis DIFFUSlON PHENOMENA IN THIN FILMSAND MICROELECTRONIC MATERIALS: edited by Devendra Gupta and Paul S. Ho ELECTROCHEMISTRY OF SEMICONDUCTORS AND ELECTRONICS: edited by John McHardy and Frank Ludwig ELECTRODEPOSlTION: by Jack W. Dini HANDBOOK OF CARBON, GRAPHITE, DIAMONDS AND FULLERENES: by Hugh O. Pierson HANDBOOK OF CHEMICAL VAPOR DEPOSITION, Second Edition: by Hugh O. Pierson HANDBOOK OF COMPOUND SEMICONDUCTORS: edited by Paul H. Holloway and Gary E. McGuire HANDBOOK OF CONTAMINATION CONTROL IN MICROELECTRONICS:edited by Donald L. Tolliver HANDBOOK OF DEPOSITION TECHNOLOGIES FOR FILMS AND COATINGS, Second Edition: edited by Rointan F. Bunshah HANDBOOK OF ION BEAM PROCESSING TECHNOLOGY: edited by Jerome J. Cuomo, Stephen M. Rossnagel, and Harold R. Kaufman HANDBOOK OF MAGNETO-OPTICAL DATA RECORDING: edited by Terry McDaniel and Randall H. Victora HANDBOOK OF MULTILEVEL METALLIZATION FOR INTEGRATED CIRCUITS: edited by Syd R. Wilson, Clarence J. Tracy, and John L. Freeman, Jr. HANDBOOK OF PLASMA PROCESSING TECHNOLOGY: edited by Stephen M. Rossnagel, Jerome J. Cuomo, and William D. Westwood HANDBOOK OF POLYMER COATINGS FOR ELECTRONICS, 2nd Edition: by James Licari and Laura A. Hughes HANDBOOK OF REFRACTORY CARBIDES AND NITRIDES: by Hugh O. Pierson HANDBOOK OF SEMICONDUCTOR SILICON TECHNOLOGY: edited by William C. O'Mara, Robert B. Herring, and Lee P. Hunt

vi

Series

HANDBOOKOF SEMICONDUCTORWAFER CLEANINGTECHNOLOGY:edited byWemer Kem HANDBOOK OF SPUTTER DEPOSITION TECHNOLOGY: by Kiyotaka Wasa and Shigeru Hayakawa HANDBOOK OF THIN FILM DEPOSITION PROCESSESAND TECHNIQUES: edited by Klaus K. Schuegraf HANDBOOK OF VACUUM ARC SCIENCE AND TECHNOLOGY: edited by Raymond L. Boxman, Philip J. Martin, and David M. Sanders HANDBOOK OF VLSl MICROLITHOGRAPHY: edited by William B. Glendinning and John N. Helbert HIGH DENSITY PLASMA SOURCES: edited by Oleg A. Popov HYBRID MICROCIRCUITTECHNOLOGY HANDBOOK, Second Edition: by James J. Licari and Leonard R. Enlow IONIZED-CLUSTER BEAM DEPOSITION AND EPITAXY: by Toshinori Takagi MOLECULAR BEAM EPITAXY: edited by Robin F. C. Farrow SEMICONDUCTOR MATERIALS AND PROCESS TECHNOLOGY HANDBOOK: edited by Gary E. McGuire ULTRA-FINE PARTICLES: edited by Chikara Hayashi, R. Ueda and A. Tasaki

Ceramic and Other Materials--Processing and Technology ADVANCED CERAMIC PROCESSINGAND TECHNOLOGY,Volume 1:edited by Jon G. P. Binner CEMENTED TUNGSTEN CARBIDES: by Gopal S. Upadhyaya CERAMIC CUTTING TOOLS: edited by E. Dow Whitney CERAMIC FILMS AND COATINGS: edited by John B. Wachtman and Richard A. Haber CORROSION OF GLASS, CERAMICS AND CERAMIC SUPERCONDUCTORS: edited by David E. Clark and Bruce K. Zoitos FIBER REINFORCED CERAMIC COMPOSITES: edited by K. S. Mazdiyasni FRICTION AND WEAR TRANSITIONS OF MATERIALS: by Peter J. Blau HANDBOOK OF CERAMIC GRINDING AND POLISHING: edited by loan D. Mavinescu, Hans K. Tonshoff, and Ichiro Inasaki HANDBOOK OF INDUSTRIAL REFRACTORIES TECHNOLOGY: by Stephen C. Carniglia and Gordon L. Barna SHOCK WAVES FOR INDUSTRIAL APPLICATIONS: edited by Lawrence E. Murr SOL-GEL TECHNOLOGY FOR THIN FILMS, FIBERS, PREFORMS, ELECTRONICS AND SPECIALTY SHAPES: edited by Lisa C. Klein SOL-GEL SILICA: by Larry L. Hench SPECIAL MELTING AND PROCESSING TECHNOLOGIES: edited by G. K. Bhat SUPERCRITICAL FLUID CLEANING: edited by John McHardy and Samuel P. Sawan

Other Related Titles HANDBOOK OF PHYSICAL VAPOR DEPOSITION (PVD) PROCESSING: by Donald M. Mattox

Preface

There has been a resurgence of interest in wide bandgap semiconductors in recent times, for two important classes of applications, namely blue/green light emitters and high power/high temperature electronics. For the first set of applications, ZnSe and GaN are leading the race, due to their direct bandgaps which give them a huge advantage over the indirect gap SiC. Nichia Chemical Industries of Japan, led by the efforts of Shuji Nakamura, announced its first blue GaN light-emitting diode product in 1993, followed by quantum well blue and green devices in 1995/1996. Currently, more than 10 million blue LEDs are sold per month. Attention has turned to work on high brightness (15 lumens per watt) white LEDs using a blue LED to excite a yttrium aluminum garnet phosphor. These may have applications in whitelighting situations, with the advantage ofmuch longer lifetimes than incandescent bulbs. Nichia has also demonstrated continuous-wave blue-violet laser diodes with room-temperature lifetimes acceptable for commercial applications. The active layer in these devices is InGaN with A1GaN cladding layers. The LEDs are useful for full-color displays, whereas the main commercial laser application is high density optical storage on CD-ROMs. A notable feature of current GaN light-emitter technology is the fact that all devices are currently grown heteroepitaxially on A1203, SiC, or magnesium aluminate substrates. The resulting high defect density (109-10 ~~cm 2) does not appear to affect the light output of LEDs, but can

io

VII

viii

Preface

cause problems in laser diodes through the migration of the p-contact metal along dislocations which may short-out the p-n junction. For this reason, there have appeared some novel lateral overgrowth techniques on SiO 2 patterned GaN templates that produce defect-flee regions above the masked areas. The active regions of the laser diodes are then processed in these areas, with the result that their lifetimes under high current operation are much longer than in devices grown in a blanket (non-patterned) fashion. While II-VI based laser diodes were the first to be demonstrated, their lifetimes are currently in the several hundred hours range and are limited by the ease of defect generation and migration in these soft materials. A typical laser diode structure consists of ZnSe/ZnMgSSe/ZnSSe/ZnCdSe/ ZnMgSSe/ZnSe layers grown on GaAs substrates by Molecular Beam Epitaxy. In this materials system, Metal Organic Chemical Vapor Deposition lags somewhat, due to poorer control over precursor purity and in-situ thickness control. This is in contrast to the GaN system, where MOCVD seems to have an advantage over MBE for photonic devices because of higher quality material due to the higher growth temperature. For the second major class of applications, high power electronics, SiC is by far the most mature, with diamond and GaN as other candidates. Diamond actually has the most appropriate material parameters, but problems with producing large single crystals and lack ofn-type dopability have retarded its progress. GaN has the advantage of the availability of heterostructures and excellent transport properties, but has relatively poor thermal conductivity. SiC has excellent thermal conductivity, demonstrated breakdown voltages of several kV and more well-developed substrates and device processing techniques. To date, the highest rfpower (850 W per mm at 850 MHz CW) was demonstrated by a 4H-SiC MESFET, and the highest total power (450 W pulsed at 600 MHz) produced by a SiC static induction transistor. A SiC power module containing four SITs has demonstrated a 1 kW capability at 600 MHz. The basic driving force is the requirement for electronics in adtomobiles, aircraft, and ships that can function directly on engines to lower the weight and cost of control functions. As an example, it is estimated that approximately 800 pounds could be eliminated on an F16 fighter jet if current mechanical, hydraulic, and pneumatic systems were replaced with advanced power electronics. Si-based electronics is limited to -- 100~ for reliability reasons, requiring active cooling systems. NASA and other agencies have needs for advanced electronics capable of operation at 600~ for temperatures <300~ it is likely that Silicon-on-Insulator technol-

Preface

ix

ogy will prevail. SiC has advantages over Si for power devices because of reduced parasitics, faster switching speed, and reduced thermal management. A key requirement is for SiC substrate micropipe and dislocation densities to be reduced from current levels. Several groups have demonstrated a positive temperature coefficient breakdown behavior in SiC, an important proof that SiC can have acceptable reliability. SiC has a thermal conductivity exceeding that of Cu, and is inert in virtually all chemicals. It is also radiation-hard with high saturation drift velocity. Initial technology insertions include HDTV receivers, power conditioning, and high power microwave applications operating at room temperature atmospheric ambients with high internal junction temperatures. SiC exists in many different polytypes, named according to the stacking sequence of Si and C atoms. For example, 6H material has a hexagonal lattice with an arrangement of six different Si and C layers before the pattern repeats itself. More than 200 different polytypes of SiC exist, and the exact physical properties depend on the crystal structure. The most common polytypes are 6H, 4H, and 3C. Initial work focused on 3C material because of its excellent transport properties. Emphasis is now being placed on the 4H and 6H polytypes. For these latter materials, single crystal substrates are commercially available (up to 3"~ currently). Additionally, the use of vanadium compensation produces semiinsulating substrates which are of interest for microwave power devices. 4H-SiC has a substantially higher carrier mobility than 6H-SiC (800 to 370 cmVV.s). In the latter polytype, there is an inherent mobility anisotropy that degrades conduction parallel to the crystallographic c-axis. Moreover, highly conductive 4H substrates are available for vertical power devices. There is also excellent progress on heteroepitaxy of 3C-SiC on both low-tilt-angle 6H substrates and on Si. In this volume, we have brought together experts in the fields of growth, processing, and characterization of wide bandgap semiconductors. The coverage includes growth and contacts oflI-VI compounds, processing techniques for SiC, GaN and diamond, and materials analysis of all wide gap semiconductors. We believe this is a very useful volume, as a valuable reference tool, both for people just entering this field, and for established researchers. Gamesville, Florida January, 2000

Stephen J. Pearton

NOTICE To the best of our knowledge the information in this publication is accurate; however the Publisher does not assume any responsibility or liability for the accuracy or completeness of, or consequences arising from, such information. This book is intended for informational purposes only. Mention of trade names or commercial products does not constitute endorsement or recommendation for use bythe Publisher. Final determination of the suitability of any information or product for use contemplated by any user, and the manner of that use, is the sole responsibility of the user. We recommend that anyone intending to rely on any recommendation of materials or procedures mentioned in this publication should satisfy himself as to such suitability, and that he can meet all applicable safety and health standards.

Contributors

Jeffrey B. Casady

Berthold Hahn

Mississippi State University Mississippi State, Mississippi

University of Regensburg Regensburg, Germany

Wolfgang Faschinger

Paul H. Holioway

University of Wurzburg Wurzburg, Germany

University of Florida Gainesville, Florida

Joseph R. Flemish

Tae-Jie Kim

Anadigics Warren, New Jersey

University of Florida Gamesville, Florida

Wolfgang Gebhardt

Jewor W. Lee

University of Regensburg Regensburg, Germany

University of Florida Gamesville,Florida

Donald R. Gilbert

Stephen J. Pearton

University of Florida Gainesville, Florida

University of Florida Gainesville, Florida

xi

xii

Contributors

Randy J. Shul Sandia National Laboratories Alburquerque, New Mexico Rajiv K. Singh University of Florida Gainesville, Florida Robert G. Wilson Hughes Research Laboratories Malibu, California

John M. Zavada US Army Research Office Research Triangle Park, North Carolina John C. Zolper Sandia National Laboratories Albuquerque, New Mexico

Contents

Doping Limits and Bandgap Engineering in Wide Gap IIVI Compounds . . . . . . . . . . . . . . . . . . . 1

WolfgangFaschinger 1.0

I N T R O D U C T I O N .............................................................................. 1

2.0

A B I N I T I O C A L C U L A T I O N S OF D O P I N G L I M I T A T I O N S ................................................................................... 6

3.0

THE 3.1 3.2 3.3 3.4 3.5

4.0

D O P I N G A N D B A N D S T R U C T U R E E N G I N E E R I N G .................. 23 4.1 Surface S e g r e g a t i o n ................................................................. 23 4.2 Superlattices ............................................................................. 25

5.0

O H M I C C O N T A C T T O p - Z n S e ...................................................... 29

6.0

C O N C L U S I O N S .............................................................................. 37

F E R M I L E V E L P I N N I N G M O D E L ......................................... 8 General C o n c e p t ........................................................................ 8 n - T y p e D o p i n g ......................................................................... 10 C o m p a r i s o n to G a A s and A I G a N ............................................ 15 p - T y p e D o p i n g ......................................................................... 17 C o m p e t i n g D o p i n g Limitation M e c h a n i s m s ............................ 22

R E F E R E N C E S .......................................................................................... 37 xm

xiv

Contents Epitaxial Growth of II-VI C o m p o u n d s by M O V P E ......... 42

Wolfgang Gebhardt and Berthold Hahn 1.0

I N T R O D U C T I O N ............................................................................ 42

2.0

B I N A R Y C O M P O U N D S ................................................................. 45 Z n T e ........................................................................................ 45 2.1 Z n S e ........................................................................................ 50 2.2 ZnS .......................................................................................... 57 2.3 2.4 Z n O ......................................................................................... 59 2.5 C d S e ........................................................................................ 60 2.6 CdS .......................................................................................... 61

3.0

T E R N A R Y A N D Q U A T E R N A R Y C O M P O U N D S ........................ 3.1 T h e r m o d y n a m i c s of Binary C o m p o u n d s ................................ 3.2 T h e r m o d y n a m i c s of Ternary C o m p o u n d s .............................. Z n T e S e .................................................................................... 3.3 3.4 Z n S S e ...................................................................................... 3.5 ZnxCd~.xSe ............................................................................... Z n C d S ..................................................................................... 3.6 Z n M g S e S ................................................................................ 3.7

4.0

61 62 63 65 66 68 71 71

C O N C L U D I N G R E M A R K S ............................................................ 72

R E F E R E N C E S .......................................................................................... 73

Ohmic Contacts to II-VI and III-V C o m p o u n d Semiconductors ..................................................................... 80

Tae-Jie Kim and Paul H. Hollowqy 1.0

I N T R O D U C T I O N ............................................................................ 80 1.1 E n e r g y Barrier and Current Transport M e c h a n i s m s ............... 82 1.2 Surface States and F e r m i L e v e l Pinning ................................. 89

2.0

O H M I C C O N T A C T S TO G a A s ....................................................... 98 2.1 In Situ Contact Scheme ........................................................... 98 2.2 Ex Situ Contact Schemes ...................................................... 102 2.3 Reliability o f and T h e r m a l l y Stable Contact Metallizations ........................................................................ 122 2.4 S u m m a r y ............................................................................... 125

3.0

O H M I C C O N T A C T S TO InP ........................................................ 3.1 Single Element/InP Metallizations ........................................ 3.2 Multi-element/InP Metallizations ......................................... 3.3 M e c h a n i s m s for F o r m a t i o n o f Ohmic Contacts ....................

126 126 128 129

Contents

xv

4.0

O H M I C C O N T A C T S TO G a N ...................................................... 130

5.0

O H M I C C O N T A C T S TO Z n S e ..................................................... 134

6.0

C O N C L U S I O N S ............................................................................ 137

A C K N O W L E D G E M E N T ....................................................................... 138 R E F E R E N C E S ........................................................................................ 138

Dry Etching of SiC .............................

151

Joseph R. Flemish 1.0

I N T R O D U C T I O N .......................................................................... 151

2.0

R E Q U I R E M E N T S OF D R Y E T C H I N G IN SiC D E V I C E F A B R I C A T I O N .............................................................. 152

3.0

C H E M I S T R Y OF SiC D R Y E T C H I N G ......................................... 154

4.0

METHODS FOR PLASMA-ASSISTED E T C H I N G OF SiC .......................................................................... 4.1 Plasma Etching ...................................................................... 4.2 Reactive Ion Etching ............................................................. 4.3 M a g n e t r o n Ion Etching o f SiC .............................................. 4.4 Electron Cyclotron R e s o n a n c e Plasmas ................................

156 156

158 163 165

5.0

PROFILE AND MORPHOLOGY CONTROL WITH E C R E T C H I N G .............................................................................. 168

6.0

S U M M A R Y .................................................................................... 174

A C K N O W L E D G E M E N T S ..................................................................... 175 R E F E R E N C E S ........................................................................................ 176

Processing of Silicon Carbide for Devices and Circuits .. 178

Jeffrey B. Casady 1.0

B A C K G R O U N D ............................................................................ 178 1.1 Historical D e v e l o p m e n t o f Silicon Carbide .......................... 180 1.2 Material Properties o f Silicon Carbide .................................. 182

xvi

Contents 2.0

S I L I C O N C A R B I D E D E V I C E P R O C E S S I N G .............................. 2.1 B u l k SiC G r o w t h ................................................................... 2.2 D o p i n g o f Silicon Carbide .................................................... Etching o f Silicon Carbide .................................................... 2.3 2.4 Oxidation and Passivation o f SiC ......................................... O h m i c Contacts to Silicon Carbide ....................................... 2.5 Schottky Contacts ................................................................. 2.6

189 190 194 201 206 212 215

3.0

S U R V E Y OF SiC D E V I C E S .......................................................... 3.1 L o w P o w e r SiC Devices ....................................................... 3.2 SiC M E S F E T s ....................................................................... 3.3 SiC P o w e r Devices ................................................................ 3.4 SiC Static Induction Transistor .............................................

218 220 221 222 227

4.0

SiC C I R C U I T S A N D S E N S O R S ................................................... 4.1 Operational Amplifiers ......................................................... 4.2 Digital Circuits ...................................................................... 4.3 Sensors ..................................................................................

229 229 229 229

5.0

C O N C L U S I O N S ............................................................................ 231

R E F E R E N C E S ........................................................................................ 233

Plasma Etching of III-V Nitrides

.......... 250

Randy J. Shul 1.0

I N T R O D U C T I O N .......................................................................... 250

2.0.

E T C H T E C H N I Q U E S .................................................................... 2.1 W e t Chemical Etching .......................................................... 2.2 RIE Etching ........................................................................... 2.3 H i g h - D e n s i t y Plasma Etching ............................................... 2.4 Additional Plasma Etch Platforms ........................................

3.0

P L A S M A C H E M I S T R Y ................................................................ 260 3.1 Gas Mixtures ......................................................................... 262 3.2 H y d r o g e n .............................................................................. 263 3.3 Sulfur Hexafluoride .............................................................. 265 3.4 N i t r o g e n ................................................................................ 267 3.5 A r g o n .................................................................................... 269 3.6 Additional Halogens ............................................................. 270 3.7 H y d r o c a r b o n s ........................................................................ 270

251 251 254 257 259

Contents xvii 4.0

P R E S S U R E .....................................................................................

273

5.0

I O N E N E R G Y A N D P L A S M A D E N S I T Y .................................... 273

6.0

T E M P E R A T U R E D E P E N D E N C E ................................................ 276

7.0

G R O W T H T E C H N I Q U E ............................................................... 279

8.0

E T C H P R O F I L E , M O R P H O L O G Y , A N D S T O I C H I O M E T R Y ... 280

9.0

P L A S M A I N D U C E D D A M A G E ................................................... 284

10.0 P L A S M A E T C H A P P L I C A T I O N S ................................................ 289 11.0 C O N C L U S I O N S ............................................................................ 292 ACKNOWLEDGMENTS

....................................................................... 294

R E F E R E N C E S ........................................................................................

294

Ion Implantation in Wide Bandgap Semiconductors ...... 300

John C. Zolper 1.0 2.0

I N T R O D U C T I O N .......................................................................... 300 I M P L A N T A T I O N I S O L A T I O N .................................................... 301 2.1

3.0

4.0

5.0

G r o u p III Nitrides ................................................................. 302

I M P L A N T A T I O N D O P I N G .......................................................... 309 3.1

Carrier I o n i z a t i o n in W i d e B a n d g a p S e m i c o n d u c t o r s .......... 309

3.2

I m p l a n t a t i o n A c t i v a t i o n T e m p e r a t u r e ................................... 312

3.3

G a N .......................................................................................

313

3.4

SiC ........................................................................................

319

3.5

D i a m o n d ............................................................................... 322

I M P U R I T Y R E D I S T R I B U T I O N ................................................... 325 4.1 4.2

G a N ....................................................................................... 325 SiC ........................................................................................ 331

4.3

D i a m o n d ............................................................................... 332

IMPLANTATION DAMAGE: CREATION AND R E M O V A L ..................................................................................... 5.1

333

G a N ....................................................................................... 333

xviii Contents

6.0

5.2

S i C ........................................................................................334

5.3

D i a m o n d ............................................................................... 338

DEVICE DEMONSTRATIONS 6.1

7.0

.................................................... 338

G a N ....................................................................................... 339

6.2

SiC ........................................................................................ 341

6.3

D i a m o n d ............................................................................... 343

F U T U R E W O R K A N D C O N C L U S I O N S ...................................... 346

ACKNOWLEDGMENT

.......................................................................... 347

R E F E R E N C E S ........................................................................................347

Rare Earth Impurities in Wide Gap Semiconductors ..... 354

John M. Zavada ABSTRACT

............................................................................................ 354

1.0

INTRODUCTION

2.0

B A S I C C O N C E P T S ....................................................................... 356

3.0

.......................................................................... 355

2.1

R a r e Ea r t h E l e m e n t s ............................................................. 356

2.2

W i d e G a p S e m i c o n d u c t o r s ................................................... 359

I N C O R P O R A T I O N OF RE A T O M S IN W I D E G A P SEMICONDUCTORS

.................................................................... 362

4.0

R E 3+ P H O T O L U M I N E S C E N C E

5.0

E L E C T R I C A L A C T I V A T I O N O F R E 3+ I O N S .............................. 383

6.0

S U M M A R Y .................................................................................... 388

ACKNOWLEDGMENT

.................................................... 367

.......................................................................... 389

R E F E R E N C E S ........................................................................................ 389

SIMS Analysis of Wide Bandgap Semiconductors .......... 393

Robert G. Wilson 1.0

INTRODUCTION

.......................................................................... 393

Contents

xix

2.0

W I D E B A N D G A P M A T E R I A L S D I S C U S S E D H E R E ................ 394

3.0

S E C O N D A R Y ION M A S S S P E C T R O M E T R Y ( S I M S ) ............... 394

4.0

S I M S I S S U E S ................................................................................. 395

5.0

Q U A N T I F I C A T I O N ...................................................................... 397

6.0

D I A M O N D ..................................................................................... 397

7.0

SiC .................................................................................................. 405

8.0

Z n S e ................................................................................................ 410

9.0

L i N b O 3 ( A N D LiTaO3) ................................................................... 412 9.1 H y d r o g e n in L i N b O 3 .......................................................................................................... 413 9.2 Rare Earths in L i N b O 3 and LiTaO 3 ....................................................................416

10.0 G R O U P I I I - N I T R I D E S .................................................................. 417 11.0 A C K N O W L E D G M E N T S

.............................................................. 424

R E F E R E N C E S ........................................................................................ 427

10 Hydrogen in Wide Bandgap Semiconductors

429

Stephen J. Pearton and Jewor W. Lee 1.0

I N T R O D U C T I O N ................................................................................

2.0

H Y D R O G E N I N C O R P O R A T I O N IN W I D E B A N D G A P S E M I C O N D U C T O R S .................................................................... 430 2.1 H y d r o g e n Plasma E x p o s u r e .................................................. 430 2.2 H y d r o g e n I m p l a n t a t i o n ......................................................... 434 2.3 Survey o f the Configurations o f H y d r o g e n in S e m i c o n d u c t o r s ..................................................................... 437

3.0

H Y D R O G E N IN G a N .................................................................... 445 M O C V D G r o w t h .................................................................. 447 3.1 Passivation o f Other A c c e p t o r Dopants ................................ 454 3.2 Diffusion o f H y d r o g e n in the Nitrides .................................. 459 3.3 Passivation in T e r n a r y Nitrides ............................................. 466 3.4 H y d r o g e n Incorporation During Processing ......................... 471 3.5 H in G a N / I n G a N Device Structures ...................................... 477 3.6 H in Cubic G a N .................................................................... 481 3.7

xx

Contents 4.0

H Y D R O G E N I N SiC ...................................................................... 482 4.1

T h e r m a l Stability o f D e u t e r i u m in 3 C - S i C ........................... 4 8 9

4.2

D i f f u s i o n o f H y d r o g e n in 6 H - S i C ......................................... 493

5.0

D I A M O N D ..................................................................................... 494

6.0

I I - V I C O M P O U N D S ...................................................................... 495

C O N C L U S I O N S ..................................................................................... 495 ACKNOWLEDGMENTS

....................................................................... 495

R E F E R E N C E S ........................................................................................ 496

11 D i a m o n d Deposition and Characterization

506

Donald R. Gilbert and Rajiv If. Singh 1.0

I N T R O D U C T I O N .......................................................................... 506

2.0

P R O P E R T I E S ................................................................................. 506

3.0

F A B R I C A T I O N ..............................................................................

4.0

5.0

6.0

509

3.1

H i g h T e m p e r a t u r e , H i g h Pressure ......................................... 509

3.2

C h e m i c a l V a p o r D e p o s i t i o n .................................................. 510

M O D I F I C A T I O N ........................................................................... 521 4.1

Structural M o d i f i c a t i o n ......................................................... 522

4.2

Electrical M o d i f i c a t i o n ......................................................... 525

C H A R A C T E R I Z A T I O N ................................................................ 525 5.1

M i c r o s c o p y ........................................................................... 527

5.2

S p e c t r o s c o p y ......................................................................... 528

A P P L I C A T I O N S ............................................................................ 532 6.1

T h e r m a l M a n a g e m e n t ........................................................... 532

6.2

Optics .................................................................................... 533

6.3

E l e c t r o n E m i s s i o n ................................................................. 534

6.4

M i c r o e l e c t r o n i c s ................................................................... 534

R E F E R E N C E S ........................................................................................

Index .............................................................................

536

543

1 D o p i n g L i m i t s and B a n d g a p E n g i n e e r i n g in Wide Gap ll-Vl Compounds WolfgangFaschinger

1.0

INTRODUCTION

Figure 1 shows the energy gap of wide gap II-VI compounds as a function of the cubic lattice constant. In addition to the "classical" II-VI compounds ZnS, ZnSe, ZnTe, CdS, CdSe, and CdTe, the range of available materials has been considerably enlarged in the last few years by the introduction of new compounds that form a metastable zincblende modification. The first metastable materials reported were manganese eompounds,[ 1] which were soon followed by the introduction of magnesium compounds with similar lattice constants and energy gaps.[ 2] All of these compounds crystallize naturally in a hexagonal or NaCl-structure and are insulators. Only when they are grown epitaxially on substrates with a zincblende structure do they form semieonducting zincblende modification. In addition, the wide gap II-VI spectrum has recently been further extended by the epitaxial realization of beryllium compounds.P] In contrast to the Mn- and Mg-compounds, these materials crystallized in the

Wide Bandgap Semiconductors zincblende structure even as bulk materials, but were only recently intensely investigated, partly due to their high toxicity. -I

"

-

'

"

4,5

I ....

''

I

'

'

4,0 >

'

I

'

'

I

"

"

I

'

~t, MgS Se MgTe

MnS II

3,5

aTe o. 3,0

I~r BeTe

~

2,5

=

2,0

ZnTe

CdSe

1,5 1 ~ 0

I

'

5,4

'

I

5,6

'

i

"

I

"

CdTe 'I

5,8 6,0 6,2 Lattice Parameter [/~ ]

~'

I'

6,4

'

6,6

Figure 1. Energy gap of wide gap II-VI compounds as a function of their lattice constant.

Altogether, wide gap II-VI materials cover a considerable part of the light speemma; from near infrared (CdTe) into the ultraviolet region (MgS, BeSe). One advantage is that nearly the entire range of compounds with different energy gaps exists at a given lattice constant, a necessary condition for the fabrication of heterojunction devices. These properties make II-VI materials very interesting for applications as light emitters and detectors in the blue and ultraviolet range. However, the application of these interesting properties has long been hindered by the limited dopability of these materials. Until 1990, CdTe (with an energy gap close to that of GaAs) was the only material that could be doped n- and p-type, while compounds with a larger gap could only be doped either n-type (e.g., ZnSe and ZnS) or p-type (e.g., ZnTe). All attempts to dope the other way led to strong compensation, so that no devices could be realized, although the material could be fabricated in outstanding crystalline quality. This picture changed drastically in 1990 when nitrogen plasma was introduced as a p-dopant which allowed p-doping levels around 1018cm -3

Doping Limits and Bandgap Engineering

3

for ZnSe. [4115]Suddenly, blue emitting devices came into reach and within a few months the first blue ZnSe based laser was reported.[ 6] This device operated only at liquid nitrogen temperature and with a very high threshold voltage. Soon the introduction of graded ZnSe/ZnTe contacts[7][ 8] and ZnMgSSe cladding layers for waveguiding[ 9] led to low threshold voltages and operation at room temperature. Room temperature cw operation for more than 100 hours has also been achieved.[ l~ A typical device structure is given in Fig. 2. The entire device, except the active zone, is lattice matched to the GaAs substrate. The active zone consists of CdZnSe, which is surrounded by ZnSSe waveguide- and ZnMgSSe cladding-layers. A typical emission spectrum of a diode fabricated in our lab is shown in Fig. 3, while the LI-characteristics are depicted in Fig. 4. As can be seen, threshold current densities as low as 340 A/cm 2 are obtainable even for simple gain guided structures without facet coating,[ 11]while the lowest reported value for index guided structures with facet coating is 240 A/cm2.[ 12] Threshold voltages well below 5 V have been obtained.[ TMThese devices are already nearly comparable to III-V lasers from an optical and electrical point of view. Therefore, the main issue for device optimization is, at the moment, the increase of the device lifetime, which seems to be limited primarily by stacking faults originating at the GaAs-ZnSe-interface.[ 14] Despite these impressive successes, one should note that the device structure given in Fig. 2 is a compromise governed by the obtainable doping. The use of CdZnSe as the active zone implies a number of disadvantages. In principle, it would be desirable to use ZnSe as the active zone. Its energy gap is really in the blue; it is a binary material with much narrower spontaneous emission linewidth;[ 15]and it is nearly lattice matched to GaAs, so that the width of the active zone could be optimized without limits imposed by the critical thickness. However, the maximum doping level in the ZnMgSSe cladding decreases strongly with an increasing energy gap, [16] limiting the energy gap of the cladding to values below about 2.9 eV at room temperature. For room temperature operation with low threshold currents, an energy difference of about 400 meV is needed between the cladding and the active zone to prevent carrier overflow and to allow reasonable light guiding.[ 17] As a consequence, the emission energy of the device is around 2.5 eV, which is more in the green than in the blue range of the spectrum. This energy corresponds to a Cd content in the active zone of about 25%. Since the lattice constant of CdSe is considerably larger than that of ZnSe, this high Cd content, in turn, seriously limits the thickness of the active zone to values below 5 nm.

Wide Bandgap Semiconductors

' ..... " "

i

.:

Gold I sulator ZnSe/ZnTe Graded Contact

'~

"

"

ZnMgSSe:N (p-Cladding)

~///~///////~//////~//~~ ~ / ~ Z ~ , . . ~ ..............

Zn S S e: N (p-Wave g u ide ) CdZnSe Quantum Well ZnSSe:CI (n-Waveguide) ZnMgSSe:C1 (n-Cladding) ~--

GaAs:Si Substrate

.J Figure 2. Layer sequence of a ZnSe based blue-green laser structure.

"

'

I

I

'"

'I

'

I

"

I

'

"

o..-I r~

.r I..--4

.... . . _ J I

525

,

~I

526

,

I

527

,

I

528

~.

I

529

Wavelength ( n m )

Figure 3. Stimulated emission spectrum of a blue green laser.

a

I

530

9

531

Doping Limits and Bandgap Engineering

I

'

'

'

I'

'

I

Pulsed Operation

~3

E 2

f~

9 "" f~

I

9

0

| I

0

9

I

,

200

i

,

_

9

9

J,h 400

9

|

600

Current Density ( A/cm 2 )

Figure 4.

Light output of a blue-green laser as a function of current density.

As a consequence, in spite of the use of nitrogen plasma, the device potential of II-VI compounds is still seriously restricted by compensation phenomena. An understanding of the processes leading to compensation is therefore of major importance. In the following, we will give a review on ab initio attempts to understand and describe compensation in widegap II-VI materials, and then describe, in more detail, a phenomenologieal model that is able to account for most of the experimentally observed doping limitations. This model will be further extended to give some insight into microscopic mechanisms leading to compensation during nitrogen doping, and will be used to explain important properties of contacts to p-ZnSe.

Wide Bandgap Semiconductors 2.0

AB I N I T I O C A L C U L A T I O N S O F D O P I N G LIMITATIONS

In a first approximation, a tendency to compensation is an intrinsic phenomenon of wide gap materials, since the energy gained by the formation of a compensating defect is on the order of the energy gap. As the energy gap increases, compensation becomes energetically more and more favorable. However, this simple analysis does not really explain the doping behaviour of II-VI materials, where materials with comparable energy gaps may differ dramatically in their dopability: CdS and ZnTe, for example, are comparable in their energy gaps (CdS can only be doped ntype, whereas ZnTe is only p-dopable). Therefore, a theoretical approach to the doping problem is more difficult. The major part of recent theoretical works in the field relies on ab initio calculations. Such calculations are usually based on the density functionals approach in the local density approximation, which allows the determination of the minimum energy configuration of a crystal containing dopants and other defects, without any experimental parameters. The starting point is an assumption of the charge distribution in the crystal (e.g., a superposition of atomic charge distributions) from which the potential distribution and the total energy of the crystal is calculated. This charge distribution is then varied in a self consistent way until a minimum of the total energy is reached. In this manner, Laks, Van de Walle, and coworkers calculated the total energy of all current intrinsic defects in ZnSe,[ 181 as well as different arrangements of important extrinsic pdopants in ZnSe,[ 19] and ZnTe.[ 2~ The resulting formation enthalpies for different defects were used to determine the concentrations of these defects as a function of temperature and chemical potentials. The chemical potentials are a measure for the stoichiometry of the host crystal as well as for the concentrations of extrinsic defects. The allowed range of chemical potentials is limited by the boundary condition that they must be smaller than the chemical potentials that would lead to the formation of competing compounds. From this boundary condition, the maximum solubility of the dopants is calculated. Some essential results of these works are: (i) the concentration of all intrinsic defects considered is too low to play a significant role in the compensation of extrinsic acceptors; (ii) Li as a small ion tends to be incorporated at interstitial rather than substitutional positions; and (iii) the decisive limiting factor for the pdoping with Na, Li, and N is not compensation but the solubility of these dopants.

Doping Limits and Bandgap Engineering

7

These statements are only partly in agreement with experimental results (SIMS data on ZnSe and ZnTe doped with radio frequency [RF] nitrogen plasma[ 21] support the last statement). The nitrogen concentration in ZnTe is one order of magnitude higher than in ZnSe, and in both cases, it is comparable to the free carrier concentration. On the other hand, this is not true for ZnMgSSe, where the free carrier concentration decreases several orders of magnitude with increasing energy gap, although the concentration of incorporated nitrogen stays approximately constant.[ 16] Additionally, the prediction of low concentrations for intrinsic defects is not supported by the evaluation of point defect concentration measurements, which conclude that high vacancy concentrations are present in these materials.[ 22] A similar formalism has been used by Chadi to calculate the stability of different defect configurations in ZnSe, ZnTe, MgSe, and MgTe.[ 23] This work focuses on the n-dopants A1, Ga, and C1 and concludes that the total energy of A1 and Ga in tellurides is considerably lowered by a relaxation in (111) direction (in analogy to DX centers in A1GaAs), so that these ions should not be suitable as n-dopants. In selenides, such DX-like centers are energetically unfavourable. A good n-dopability of MgSe is expected. C1 does not form DX-like centers, but is nevertheless a strongly localized defect in ZnTe. This fact is used to explain why the n-dopability of ZnTe is much lower than that of ZnSe; although for alloying of ZnSe with low Te mole fractions, even an improvement of the n-dopability is predicted. As in the case of p-doping, these conclusions only partly agree with available experimental data. Indeed, there is experimental evidence for the formation of DX-like centers in some tellurides,[24]-[ 26] and recent experiments support the existence of AX-like centers in nitrogen doped ZnMgSSe.[ 27] On the other hand, newer works demonstrate that n-doping of ZnTe can be achieved with CI[28] as well as with AI.[29] In contrast to predictions, the carrier concentrations in A1- doped ZnTe (1-4.1017cm 3) are even higher than in the case of C1 doping (3.1016cm-3). An alloying of Te to ZnSe does not lead to the predicted increase, but to a strong decrease in the free carrier concentration.[ 3~ Additionally, the free carrier concentration of ZnMgSe doped with C1 decreases strongly with increasing Mg content, and pure MgSe cannot be doped n-type.[ 31] Considering these facts, DX centers seem to limit the dopability of wide gap II-VI compounds only in some selected cases. Using the same approach, interesting results on nitrogen doping of ZnSe have been obtained by Garcia et al. who investigated the stability of

Wide Bandgap Semiconductors complexes involving charged native defects.[ 321 Their results reproduced the saturation in free hole concentration observed in nitrogen doped ZnSe for high total nitrogen contents (see Fig. 13), as long as the calculation is carried out for Zn-rich conditions. This is in partial agreement with experimental results showing that the maximum free carrier concentration in ZnSe:N is obtained for nearly stoichiometrir growth conditions and decreases only slowly with increasing the Zn to Se ratio.[ 33] However, the prediction that a more Se rich growth should lead to a higher doping level is in direct contradiction to experimental results which show a strong decrease in free carrier concentration even for a slight shift from stiochimetric towards Se rich growth conditions.[ 33]

3.0

THE FERMI LEVEL PINNING MODEL

3.1

General Concept

A more phenomenologieal treatment of doping limitations starts from the observation that for many semiconductors, the maximum dopability is in some way connected to the energetic position of the band edges. This is relatively well understood for III-V materials and has been described in detail by Walukiewiez.[ 341-[361He showed that the energetic position of deep levels induced by heavy irradiation with neutral ions is practically independent of the host lattice, but is fixed with respect to the vacuum level.[ 341One obvious reason for this behavior is that the dangling bonds at vacancies, which are strongly localized defects, do not "know" which atom is missing (e.g., for an As dangling bond in A1GaAs, an A1and a Ga-vaeaney are identical). This is also true for compensating defects which contain vacancies in the form of complexes. If one tries to introduce donors and to raise the Fermi energy above the energetic position of this compensating center, it becomes energetically favorable for the crystal to form the localized compensating center instead of adding the electron to the Fermi surface. This leads to a pinning of the Fermi level at the energetic position of the compensating center. In III-V materials, the nature of the compensating defects is known, and their energetic position can be calculated from their formation enthalpy. The maximum dopability of a compound can then be determined from the relative position of the band edges with

Doping Limits and Bandgap Engineering

9

respect to the Fermi energy. The calculated values are in agreement with experimental data. Several years ago, it was recognized that, at least qualitatively, a similar correlation exists also in II-VI compounds.[ 37][38]Attempts to dope the newly discovered cubic Mg compounds MgSe, MgTe, and BeTe made this still more obvious, as demonstrated in Fig. 5. The figure shows the relative band edge position of many II-VI materials plotted together with their dopability. Compounds with very high conduction bands (like MgSe and MgTe) cannot be doped n-type, while the sulphides with their very low valence band edges are not p-dopable.[ 39]

Vacuum Level i

2 1

["] not n-dopable I I not p-dopable n/p-dopable * extra ~olated = -

>

0

i

r I

1

I

!

I

EF, n

J

o/)

BeTe

Mg,S CdSelZnTe VlgTe

I-q

<

9

x

x

~

~x-x~A

-2 I

'

EF, p

-3 -4 Figure 5. Relative band edge positions of various II-VI compounds. The experimentally

obtained doping is indicated by different shadings.

However, a quantitative treatment analogous to III-V compounds is difficult, since the nature as well as the formation enthalpies of the compensating centers in II-VI's is a matter of controversy. There exist a number of models that all include the participation of vacancies. Marfaing proposed compensation reactions for both n-type[4~ and p-type material[ 41] and concluded that the free hole concentration in ZnSe should not increase

10

Wide Bandgap Semiconductors

above 10 TM cm -3. This study is in agreement with experimental data. [421 Hauksson et al. proposed a microscopic model of a deep donor complex acting as a compensating center for nitrogen in ZnSe, but made no conclusions on the maximum dopability.[ 43] The following shows that, in a semi-empirical approach, a quantitative description of the doping behaviour of most II-VI materials can be gained on the basis of the fermi level pinning model without an exact knowledge of the nature of compensating defects. Compared to Walukiewicz and Marfaing, this approach is reversed: instead of calculating absolute values for the maximum dopability from a microscopic model of the compensation process, the relative change in the dopability of mixed crystals is considered. The Fermi level pinning position is treated as a parameter (which influences the absolute values obtained and can therefore be used to adjust the calculation to experimental values), while the calculated relative change in dopability is uniquely determined by the change in the position of the mixed crystal band edges as a function of composition.

3.2 n-Type Doping To illustrate this approach, chlorine doping of ZnMgSe is described as an example. As a first step for a quantitative description, information on the variation of the band edge positions with composition must be obtained. For the given example, this determination has been done by the photolumineseenee experiment depicted in Fig. 6.[44] The ZnSe/ZnTe heterojunetion forms a staggered type II configuration. The prominent photoluminescenee signal in such a structure is given by a transition from the ZnSe conduction band to the ZnTe valence band[ 451 (upper insert in Fig. 6). An addition of Mg only to the ZnSe part of such a structure will induce changes in the position of the conduction and valence band of the ZnMgSe, but not of the ZnTe. Since the end point of the luminescence transition is unchanged, the change in the luminescence signal will only reflect the change of the ZnMgSe conduction band. Eventually, if the Mg concentration is high enough, the ZnMgSe conduction band will raise above the ZnTe conduction band, and the heterojunetion will become a type I structure with a ZnTe quantum well within ZnMgSe barriers (lower insert in Fig. 6). Provided the ZnTe quantum well is thick enough to prevent a significant confinement effect, the luminescence energy will then correspond to the ZnTe energy gap and will not be influenced by a further increase of the Mg content.

Doping Limits and Bandgap Engineering

11

w~w~'|lwwwwwww~l~ww~w~www|w~'wwwwwvw|www"~www"tw

ZnMgSe

3,4 3,2 3,0 > 2,8 .

~2,6

.

.

.

.

II

, -

~D

ZnMgSe/ZnTe MQW

CD

! '~

~ 2,4

"r" 'r

t

2,2-

1

~

~1~

~

2,0

~ ....

, ..........

0

N

......... 20 40 60 80 Mg Content ( % ) , . . . . . . . . .

, . . . . . . . . .

'

Figure 6. Photoluminescence signal of ZnMgSe/ZnTe double heterostructures as a function of the Mg content. Upper part: Near bandedge luminescence of the ZnMgSe. Lower part." Transition from the ZnMgSe conduction band to the ZnTe valence band. Figure 6 shows the energetic positions of the luminescence signals for a number of double heterostructure samples as a function of the Mg content. In the upper part of the figure, the luminescence from the band edge luminescence of the ZnMgSe is shown, while the energetic position of the transition from the ZnMgSe conduction band to the ZnTe valence band is drawn in the lower part. Since all structures were grown on GaAs substrates, and are therefore fully relaxed, the relatively poor crystalline quality of the samples leads to a broad luminescence signal and a relatively large error in the luminescence energy.

12

Wide Bandgap Semiconductors

Nevertheless, Fig. 6 allows some interesting conclusions. First, for high Mg content, the luminescence energy indeed saturates at 2.3 eV, which is the energy gap of ZnTe. This is in striking coincidence with the prediction given above and confirms the proposed interpretation. The saturation starts at a Mg content of 60%, which is the crossover point between the type II- and the type I-structure. This leads to the conclusion that the conduction band of Zn0.4Mg0.6Se and ZnTe are at the same energetic position. Second, the slope of the ZnMgSe band edge and the type II signal are identical within the experimental error. This can only be the ease if the increase of the energy gap is synchronized with an accompanying increase of the ZnMgSe conduction band edge position. Therefore, the conclusion from the figure is that Mg has a big influence on the position of the ZnMgSe conduction band, while the influence on valence band position is small (below 10%). Given this information on the conduction band position, the next step is a determination of the maximum doping level of ZnMgSe as a function of the Mg content. This has to be done carefully, since the obtained doping level is a strong function of the dopant flux. This behaviour is illustrated in Fig. 7, where the free electron concentration in ZnSe and Zn0.64Mg0.36Seis plotted as a function of the ZnC12 source temperature. As expected, the doping level increases exponentially with the source temperature for low temperatures, reflecting the nearly exponential temperature dependence of the ZnC12 flux. However, for a further increase of the source temperature, the free electron concentration does not saturate, but decreases rapidly for higher temperatures. This is a compensation phenomenon often encountered in II-VI doping which has nothing to do with a change of the band edge with respect to the position of the Fermi level as in the model described here. We will come back to this phenomenon in the case of nitrogen doping of ZnSe, where a similar behaviour is encountered. However, in our present context, the decisive point is that the maximum electron concentration is obtained at different ZnC12 source temperatures for different Mg contents. This fact must be taken into account, and the optimum source temperature has to be determined for each Mg content. The data obtained from such a procedure are shown in Fig. 8, where the maximum doping concentration is plotted as a function of the Mg content.J311 The doping level decreases nearly exponentially with increasing Mg content. To describe this behaviour in the framework of our model, we assume that the Fermi level is pinned at a constant energetic position,

Doping Limits and Bandgap Engineering

13

while the conduction band is shifted upwards in the way determined from Fig. 6. The relative change in the free electron concentration is then uniquely determined by the energy difference between the constant Fermi level position and the conduction band. From this difference, the corresponding free carrier concentration can be easily determined. Since the formation of compensating defects occurs at growth temperature, the calculation should be performed at growth temperature. We now treat the Fermi level position as a parameter to fit the experimental data. The result is shown as a solid line in Fig. 8 for a Fermi level position 130 meV above the ZnSe conduction band. The coincidence between experiment and calculation is striking. The firing parameter does not influence the slope, but only the absolute position of the calculated curve, with the exception of the saturation of the carrier density for low Mg contents, a consequence of degeneracy when the Fermi level comes close to the conduction band. Figure 9 shows that the limits for n-doping of other wide gap II-VI compounds can be explained by the same value of the pinning level.

r162

'~ 1El9 L) O 4.a I-(

1El8

o

L) 1El7 o ,I=,=)

II

ZnSe

9

ZnMg0.36Se

F==,-(

1El6 110

120

130

140 .150

160

170

180

Chlorine Source Temperature [ ~ ] Figure 7. Free electron concentration of ZnSe.CI and ZnMgSe.CIlayers as a function of the ZnCI2 source temperature. The solid lines are guides for the eye.

Wide Bandgap Semiconductors

.14

1E20

,

...

,

.

,

.

,

..

,

I

r162 !

o

1El9

~

1El8

-

.

,

ZnMgSe:CI

-

~

o

rj

1El7

o

1El6

' 0,0

'

' 0,1

.... ' 0,2

Mg

'

'

'

0,3

Content

in

'

'

0,4

'

0,5

ZnMgSe

Figure 8. Maximum free electron concentration in ZnMgSe as a function of the Mg content.

1E20

I

'

I

'

I

'

! ZnMgSe:CI 17 CdMgTe:I ZnTe:Cl ulation

!

0

'

1El9

o

o~==1

=

~D

1El8

O = o

1El7

1El6

I

0,0

O,2

0,4

0,6

ECB - ECB ZnSe ( eV ) Figure 9. Maximum free electron concentration in various II-VI compounds as a function of their conduction band edge position. Arbitrary zero of the energy scale is the ZnSe conduction band edge.

Doping Limits and Bandgap Engineering 3.3

15

Comparison to GaAs and AIGaN

Moreover, it is interesting to compare the n-doping behaviour of II-VI materials with that of III-V compounds, where similar limitations have been observed. GaAs, for example, has a relatively high conduction band edge compared to InP, and, correspondingly, a maximum reported n-doping of about 2.1019 cm -3 for GaAs[ 5~ is lower than the value of 1020 cm -3 obtained for InP.[5~ The limitation for GaAs corresponds to a position of the Fermi level of about 320 meV above the conduction band edge at a typical GaAs growth temperature of 600~ The band alignment between III-V and II-VI materials is much more uncertain than between different II-VI compounds, and it has been shown that the valence band offset of GaAs-ZnSe-heterostructures can be shifted by nearly 1 eV by varying the ZnSe growth parameters at the interface.[ 51]However, if one accepts older measurements that ZnSe and GaAs have typically a conduction band offset as low as 200 meV,[ 52] this would mean that the Fermi level pinning o c c u r s at a similar energy position even for III-V compounds. Limitations for n-type doping therefore seem to follow a kind of"universal" law in the sense that they are independent of host lattice, dopant, and even the class of material. The model can also be applied to A1GaN, although very few systematic data are available on n-type doping of this compound, and the uncertainty in the evaluation of the data is much larger. To illustrate this point, Fig. 10 shows the n-doping level for unintentionally doped MOVPE grown A1GaN.[ 53] Since highly resistive GaN layers have been obtained in the meantime by a better purification of the source gases,[ 54] it can be concluded that the unintentional doping is of an extrinsic nature. Although the doping is unintentional, the fact that very high doping levels are obtained for pure GaNjustifies the assumption that an excess of the dopant is offered for all compositions. The solid line represents results for the dopability based on the Fermi level pinning model, calculated for a growth temperature of 800~ and a conduction band offset of 75% of the energy gap difference between A1GaN and GaN.[ 55] As a consequence of the high growth temperature, which leads to a much higher carrier concentration for a given position of the Fermi level with respect to the conduction band, the slope of this curve is much smaller than in the case of II-VI materials (the calculated difference in the dopability due to the increased temperature is more pronounced at low doping levels and can be as high as four orders of magnitude for A10.4Ga0.6N). Although the scatter in the data is considerable, it is evident that the pinning model adequately

16

Wide Bandgap Semiconductors

describes the behaviour in the range of low A1 content, suggesting that the general concept of Fermi level pinning applies also to nitrides. The much stronger decrease of the carrier concentration at high A1 contents is understandable in analogy to the case of ZnMgSe described above. The maximum doping level is not simply obtained by offering as much of the dopant as possible, but by carefully reducing the dopant flux to an optimum value (compare with Fig. 7). If, in Fig. 8, the obtained doping level for a constant dopant flux is plotted instead of the maximum doping level, the decrease in the free electron concentration as a function of Mg content would be much stronger than predicted by the model. Since in the case of A1GaN the doping is unintentional, the dopant flux is assumed to be approximately constant, and a comparison to the calculation only makes sense in the range of relatively high carrier concentrations.

10 2! II 0~ 10 20 o o

=9

II 019

o o

II

0

~ 1018

at

0 o ~ 17 o10 o

III

1016 0

10 20 30 A1 Content in AIGaN ( % )

40

Figure 10. Free electron concentration in unintentionally doped MOCVD-grown AIGaN as a function of the AI content (after Ref. 53).

Doping Limits and Bandgap Engineering

17

3.4 p-Type Doping Since the Fermi level pinning model is very successful in the description of n-doping limitations, it is easy to apply the same model to pdoping data. The procedure is completely identical to that described above for n-doping. No assumption is made on the microscopic nature of the compensation mechanism, and the pinning position is determined by adjusting the calculation to available data for mixed crystals with varying composition. The results are summarized in Fig. 11, where experimentally obtained maximum p-doping levels for ZnSe,[421ZnMgSSe, [161and ZnSe/ ZnTe short period superlattices[ 3~ are plotted versus the energetic position of their valence band maximum. In contrast to the first three material systems, which have been doped with an RF plasma source, the ZnSe/ ZnTe samples were fabricated using a DC plasma, a difference which will be of importance for the following discussion. Arbitrary zero point on the energy scale is the ZnSe valence band. The valence band positions for ZnSSe and ZnMgSSe have been calculated under the assumption that the offset between ZnSe and ZnSSe is purely in the valence band[ 48] and the addition of Mg mainly influences the conduction band position (see See. 3.4 and Ref. 44). In the case of ZnSe/ZnTe, the valence band position has been determined by a Kronig-Penney calculation of the minibands. We will first concentrate on the data obtained with RF plasma. These data can be well described in the framework of the Fermi level pinning model under the assumption that the Fermi level gets pinned at an energy 120 meV above the ZnSe valence band maximum (left solid line). It is worth noting that this calculation has first been done purely on the basis of the ZnSe and ZnSSe data and predicted the doping behaviour of ZnMgSSe claddings that was later experimentally observed.[ 16] The technological implication of the Fermi level pinning interpretation is that the reason for the abrupt decrease of the doping level in ZnMgSSe claddings with increasing energy gap is the lowering of the valence band position of this compound that is caused by the increasing sulphur content. Therefore, a practical realization of deep blue lasers with ZnSe quantum wells as an active zone implies the development of cladding materials with a higher valence band position, but a comparable energy gap as ZnMgSSe. Proposals for the practical realization of such claddings will be given in Ch. 4.

18

Wide Bandgap Semiconductors

ZnSe/ZnTe----~

"~--ZnMgSSe

1020 '

10

QO

19

o

= O

1018

~, 10 17 o =

O

L) N O

101

6 Calculatioq Q ZnSe/ZnT~ ! ZnMgSSe|

1015 1014

-0,4

-0,2

0,0

0,2

0,4

0,6

0,8

1,0

EVB - EVB ZnSe ( e V ) Figure 11. Maximum free hole concentration in various II-VI compounds as a function of their valence band edge position. Arbitrary zero of the energy scale is the ZnSe valence band edge.

The free hole concentrations obtained with DC plasma doping are several orders of magnitude lower than those for RF doping, and doping levels obtained with this source for ZnSe are as low as 1014cm -3. Nevertheless, it is remarkable that the slope of the de plasma data is identical to that of the RF data. Consequently, these data can also be well described by a Fermi level pinning mechanism, with the only difference that the pinning position determined from the data is 460 meV higher, at 580 meV above the ZnSe valence band. According to the model, the pinning position depends on the nature of the compensation mechanism (see 3.1). The interpretation of this striking difference is that there must exist two different compensation paths for nitrogen in II-VI materials.

Doping Limits and Bandgap Engineering

19

It is obvious that such a difference in the compensation mechanism must be correlated to the nature of the plasma used. This becomes still more plausible if one takes a look at plasma emission spectra, which contain information on plasma composition. Emission spectra of RF plasma under two different operation conditions are shown in Fig. 12.[56] The upper part of the figure corresponds to the emission of the RF plasma in the so-called "high brightness mode," which is used to obtain high doping levels. The spectrum is composed of broad bands associated with excited nitrogen molecules, together with sharp features that can be ascribed to atomic nitrogen (marked with arrows in Fig. 12). The spectrum in the lower part of the figure is taken from a "low brightness mode" plasma, which is obtained if the plasma power is reduced. In this spectrum, the features of atomic nitrogen have nearly (but not completely) disappeared. In parallel, doping of ZnSe with this low brightness plasma leads to free hole concentrations in the 1015 to 1016 cm -3 range,[561 which is already close to what is achieved with the DC plasma. Obviously, effective p-doping of ZnSe is correlated to the presence of atomic nitrogen in the plasma, and a reduction of the atomic nitrogen content leads to a drastic decrease in the obtained doping.

I

'

I

'

I

~

I

'

'

I

w

t

I

9

I

2nd pos. series N2": C31-1u-> B311 g

atomic transitions

'

1st neg. series N2,: B2T_.u* -> X2T_.g

.

.

.

t

t

.

ff Plasma". Ha mode: ~4NI 110 W~ 7-10 -7 torr

~1

1st pos. series

300

400

500

600

I

700

800

900

IIl~

1000

1100

wavelength [nm]

Figure 12. Emission spectra of radio-frequency nitrogen plasma in the high brightness mode (upper part) and the low brightness mode (lowerpart) (from Ref. 56).

20

Wide Bandgap Semiconductors

Although no comparable emission spectrum of the DC plasma is available, the fact that this plasma source is operated at much lower power than an RF source (15 W compared to 150-300 W for an RF source) makes it reasonable to assume that the DC plasma does not contain any atomic nitrogen, and the much lower doping levels are a result of doping with excited nitrogen molecules. Two properties of RF plasma doped layers give additional insight in the processes occurring during compensation. First, p-ZnSe doped with RF plasma is rather unstable against annealing, and even anneals below growth temperature lead to a considerable increase in resistance,[ 57] approaching that of layers doped with DC plasma. Thus, it seems that the highly conductive "RF" state can be converted into the resistive "DC" state by a diffusion process. Second, the free carrier concentration in RF plasma doped layers decreases drastically if the atomic nitrogen flux is increased. This behaviour can be seen in Fig. 13, where the p-doping levels obtained by different groups[56][58] (solid symbols) are plotted as a function of the total nitrogen content incorporated in the layer. The variation of the atomic nitrogen flux was obtained by an increase of the RF plasma power used. For low nitrogen incorporation, the nitrogen was nearly completely electrically activated. When the total amount of nitrogen was increased, the free hole concentration first saturates and then drops abruptly for very high nitrogen content. In analogy to annealing, an excess of atomic nitrogen was another possibility to induce a transition from a highly conductive to a resistive state. All these experimental findings can be understood in a model that assumes the existence of two nitrogen complexes which contain one or two nitrogen atoms, respectively.[ 59] Many microscopic models of such centers have been proposed in the literature.[56116~-[62]We assume that the second one is energetically more stable and is formed automatically when nitrogen molecules are the doping species. However, its formation is kinetically limited as soon as atomic nitrogen is offered, since the nitrogen atoms first have to find each other by diffusion. It is now plausible that annealing has such a drastic effect on conductivity, and also helps to interpret the data shown in Fig. 13. As soon as too much atomic nitrogen is offered, the probability for the formation of nitrogen pairs increases. To quantify this latter point, the data in Fig. 13 have been modelled by a Monte Carlo simulation of the nitrogen pair formation process via diffusion (open symbols in Fig. 13, the solid line is a guide for the eye). This simulation assumes a nitrogen sublattice with fixed lattice points, and an average lattice constant determined by the absolute nitrogen content. One incoming nitrogen atom is then moved randomly for a fixed number of

Doping Limits and Bandgap Engineering

21

steps, and the probability of meeting another nitrogen atom is determined by averaging over 100 diffusing atoms at each given composition. This probability is then translated into a free carrier concentration by the assumption that for each nitrogen pair, two acceptors are lost, and a double donor is formed. The number of diffusion steps is calibrated by the condition that an absolute nitrogen concentration of 1018 cm "3 should correspond to a free hole concentration of 3.1017 cm "3 (an experimental finding of much of the research,[421156][58])and is then kept constant for all other concentrations. Under these assumptions, the experimental features, and especially the abrupt drop at high nitrogen content, can be exactly reproduced.

1018 . "J

}

-!

~1017. tO

i,

~

.

~1016O

r,.) ~1015

Q

Kurtzetal.

II

Ohkawa et al.

121

Monte Carlo

1014 .

t

1016

1017

1018

1019

Nitrogen Concentration ( cm -3 ) Figure 13. Free hole concentration in ZnSe doped with RF nitrogen plasma as a function of the total nitrogen concentration (solid symbols, Refs. 56 and 58) together with results of a Monte Carlo simulation (open symbols). The solid line is a guide for the eye.

22

Wide Bandgap Semiconductors

These ideas can be taken together with the Fermi level pinning model to explain the different behaviour for doping with RF and DC plasma, as shown in Fig. 11. As long as atomic nitrogen in a moderate concentration is offered, the compensating defect is the complex involving one nitrogen atom, which pins the Fermi level at a value of 120 meV above the ZnSe valence band edge. This pinning is responsible for the decreasing doping levels in ZnMgSSe claddings. If activated nitrogen molecules are offered, the energetically more favorable complex involving two nitrogen atoms becomes dominant and pins the Fermi level 580 meV above the ZnSe valence band edge. As in the case of n-doping, it is again of interest to compare these pinning positions with those occurring in III-V compounds. From the limited obtainable p-doping in InP with its relatively low valence band edge, one can estimate a pinning position about 500 meV below the GaAs valence band edge in these materials.[ 34] Under the assumption of 1.1 eV for GaAs/ZnSe[ 52] (as in the case of n-doping), this would correspond to a position about 600 meV above the ZnSe valence band edge. This is much higher than for RF plasma doping in II-VI compounds. From this, it may be concluded that the compensation mechanisms are very different in nature between III-V's and RF plasma doped II-VI's. This is not too astonishing if one considers the very special compensation situation which is encountered in nitrogen doping of II-VI compounds, as it has been described above. However, the value for the pinning level is close to what is observed for DC plasma doped II-VI compounds, suggesting that the compensation mechanism occurring in this case is closer to what is observed in III-V' s. As an important consequence, the compensation picture given here implies that the complex formed during RF plasma doping is metastable and can be converted into the stable configuration formed during DC doping by an annealing process. While this metastability does not seem to be a significant problem for the conductivity of the p-layers in a laser structure during operation, it has important consequences on the formation of Ohmic contacts to ZnSe, a point to which we will return in Ch. 5.

3.5

Competing Doping Limitation Mechanisms

Obviously, there exist upper doping limits which are determined exclusively by the positions of the band edges. However, this should not be misinterpreted in a way that Fermi level pinning is the only limiting

Doping Limits and Bandgap Engineering

23

mechanism. One obvious example of a competing mechanism is limited solubility of the dopant. This is the reason for the saturation observed in the n-doping level of CdMgTe:I visible in Fig. 8, as well as an analogous saturation in the p-doping level of ZnSeTe seen in Fig. 12. A similar mechanism is the formation of stable competing phases of the dopant with one of the constituents of the host. An example of this is the strong decrease in free hole concentration if Mg is added to ZnTe, which cannot be due to Fermi level pinning, since both ZnTe and MgTe have very high valence band edges. This behaviour has been explained with the formation energy of magnesium nitride, which is much higher than that of MgTe[ 63] and favors the formation of the nitride instead of the substitutional incorporation. Such a mechanism may also be present in selenides, where it has been shown that an addition of Mg to ZnSSe led to a slight decrease in the p-doping level[ 64] (although this may also be att~buted to an increase in the S sticking coefficient due to the presence of Mg and the resulting lowering of the valence bond edge). An analogous explanation has been given also for the relatively poor free hole concentrations in nitrogen doped CdTe.[ 63] The size of the dopant with respect to the host lattice may also be of importance in some cases. The high n-doping of CdMgTe plotted in Fig. 9, for example, is only obtained with the large halogen Iodine, which fits relatively well to the large CdMgTe lattice. If the halogens bromine or chlorine are used, the resulting doping levels are much lower, the lowest values resulting for the small chlorine ion.[47] Such effects may well be due to the formation of DX-like lattice deformations,[231which are more likely for small dopants. In a real crystal, the pinning mechanism competes with all these other limitations. In this context, the main value of the pinning model is that it allows a prediction of the maximum obtainable doping level if all competing mechanisms can be overcome.

4.0

D O P I N G AND BAND S T R U C T U R E E N G I N E E R I N G

4.1

Surface Segregation

It is evident from Fig. 11 that the key factor for the improvement of p-doping levels is to raise the position of the valence band edge. Since, in an MBE process, the incorporation of a dopant occurs at the surface of the growing crystal, the formation of compensating centers should rather be

24

Wide Bandgap Semiconductors

determined by the conditions at the surface than by those in the bulk. Consequently, it could be sufficient to increase the valence band edge at the surface in order to obtain a low degree of compensation. One possible way to do this is to take advantage of surface segregation. Segregation usually occurs if atoms of very different size are offered simultaneously, so that it becomes energetically favorable for the larger atom to occupy surface sites where more space is available than in the bulk. In the case of ZnSeTe, this effect is so strong that nearly no tellurium is incorporated during MBE growth. [65] In analogy, strong segregation can be expected if sulfur, as a very small atom, is offered together with the much larger tellurium. In the ease of ZnSTe (or ZnMgSTe) this combination offers, in addition, the possibility to be lattice matched to GaAs. In Fig. 14, the free hole concentration of MBE grown ZnSTe layers determined by CV profiling is plotted versus the bulk Te content of the layers (solid symbols).[ 66] The layers were grown using elementary Zn, S, and Te sources. The obtained carrier concentration is nearly independent of the bulk Te content, whereas the Fermi level pinning model would predict a very strong dependence on Te content (solid line, with a large error in Te content, since the dependence of the valence band on the Te content can only be estimated). This picture changes if one does not plot the bulk Te content, which is determined by x-ray diffraction, but the surface Te content, measured by x-ray photoemission spectroscopy for some selected layers (open symbols, the arrows connect identical samples). This surface Te content is always much higher than the bulk content and is above 20% even for layers with a bulk Te content as low as 5%. It must be noted that this strong segregation effect does not occur if a ZnS source is used instead of the elementary S source. As a consequence, the surface data points now lie close to the calculation, except for high Te content, where they are considerably lower. Since, in this range, the carrier concentration is already close to the 1019 em -3 range, this is probably an effect of limited solubility, in analogy to the ease of ZnSe/ZnTe superlattiees (see Fig. 11). Thus, surface segregation and the shift of the surface valence band edge associated to it allow an increase of the doping level. Unfortunately, there is only a limited potential for applications of this effect due to two negative by-products of the segregation. First, it causes a vertical gradient in composition which can be observed as a strong broadening of x-ray reflexes. Second, the large potential fluctuations associated with these inhomogeneities seem to have a tendency to trap carriers, so that transport finally occurs only via hopping.[ 66] Nevertheless, the conductivity of material with a carrier concentration close to 1019 em -3 is high enough to be useful for p-contacts, as will be shown in Ch. 5.

Doping Limits and Bandgap Engineering

25

10 20

10 = o

19

l

1018

l

l

i

.p,,~

t-d

10

17

o

~

Calculation I"! Volume Te Content (x-ray) I Surface Te Content (XPS)

10 z6

o

10~5 0

10

20 30 Te Content in ZnSTe

40

Figure 14. Free hole concentration of RF-plasma doped ZnSTe as a function of the bulk Te content (solid symbols) and the surface Te content (open symbols). The solid line is a calculation based on the Fermi level pinning model.

4.2

Superlattices

A second attractive possibility to influence the valence band position is the use of superlattices instead of mixed crystals. In that case, the positions of the first conduction and valence minibands that determine the superlattice energy gap can be shifted up and down by simply changing the superlattice period. Since the effective hole masses are much larger than those of electrons, confinement effects are much more pronounced in the conduction than in the valence band, a situation that is in favor of a higher valence band edge as compared to a mixed crystal with a comparable energy gap. That this really affects the doping properties is demonstrated in Fig. 15, where the free hole concentration of ZnSe/ZnTe superlattices doped with DC nitrogen plasma is plotted versus the superlattice period.[ 3~ The average Te content is 15% for all samples. An increase of the period from 2.4 to 5 nm leads to a strong increase of the hole concentration from 2.1016cm -3 to 3.1019cm"3. The solid line is the

Wide Bandgap Semiconductors

26

calculated carrier concentration according to our model for miniband positions determined from a Kronig Penney calculation of ideally abrupt superlattices, with heavy hole masses of 0.78 for ZnSe and 0.6 for ZnTe. In fact, this simple model reproduces the large increase in carrier concentration, except for very small periods, where, due to interdiffusion, the dopability approaches that of a mixed crystal of the same composition.

10 20

r

'~ 10

ZnTe Layer Thickness (ML) 1,5 2,0 2,5

1,0

19

1

3,0

Calculation Data

o 18 "'~ 10 I

o 017 =O 1 r,.)

l--

--

I

/

I

Mixed

O

~

10 ~6 2

3 4 5 ZnSe/ZnTe SL Period ( nm )

6

Figure 15. Free hole concentration ofZnSe/ZnTe superlattices with an average Te content of 15% as a function of the superlattice period. The solid line is the result of a model r the dashed line indicates the doping level obtained for a ZnSeTe mixed crystal with 15% Te.

Figure 16 gives the results of an analogous calculation for two examples of RF plasma doped superlattices, namely (MgS)n/(ZnSe)n and (Zn0.sMg0.sS)3J(ZnTe)n, two combinations that are reasonably well lattice matched to GaAs. The former has the advantage that it uses the same components as the quaternary ZnMgSSe, the latter contains ZnTe with its very high valence band edge (which in turn leads to a high heavy hole superlattice miniband). For comparison, the calculation and the data[ 16]for ZnMgSSe mixed crystals are also given in Fig. 16. The dotted part of the curve for (Zno.sMg0.sS)an/(ZnTe)n represents the region where the ZnTe

Doping Limits and Bandgap Engineering

27

thickness is lower than 2 ML, and the calculation is expected to be inaccurate due to interdiffusion.

I

g"" 10 Is

"

o

!

!

!

'

.... ZnMgSfZnTe . . . . . . MgS/ZnSe ZnMgSSe 9 ZnMgSSedata

\

o 9~

o 017 "~1

O

9

"

9 %~

% 9

"%

o

9 %

o 1016

9

9 %

9

9,

O O

% 9

1015

,

9 %

%

9149

.. 9

218

'

I

'

I

3.'0 3.2 3.4 Energy Gap ( eV )

'

I

3.6

Figure 16. Calculatedfreehole concentrationfor (MgS)n/(ZnSe)nand(Zno.sMgo.sS)3n/(ZnTe)n superlattices as a function of their energy gap. For comparison,calculated and experimental values for ZnMgSSe mixed crystals (Ref. 16) are also shown.

For MgS/ZnSe an increase of the hole concentration by a factor of five to ten compared to ZnMgSSe is expected. For cladding layers with an energy gap of 2.95 eV, as they are currently used in blue-green lasers, this would represent an increase of the hole concentration from 1 to 5.1017cm -3, which could help improve device performance. The predicted behavior of the Zn0.sMg0.sS/ZnTe superlattices is still more promising. Even combinations with an energy gap up to 3.2 eV are predicted to exhibit hole concentrations above 1017cm -3. For such a cladding material, the use of a ZnSe quantum well would no longer be unrealistic. Nevertheless, it has to be considered that the improvement in vertical transport, which is decisive for a laser device, will be somewhat smaller than the predicted increase in dopability. The carriers will have to tunnel through the barriers introduced by the superlattice structure. In addition, the conduction band edge of such a cladding layer would be

28

Wide Bandgap Semiconductors

already so high that the n-dopability is expected to be very bad. As a consequence, for a deep blue room temperature laser, we propose to use an asymmetric device design with a ZnMgSSe cladding on the n-side, and a Zn0.sMg0.sS/ZnTe cladding on the p-side. The schematic energy band diagram of such a proposed device is shown in Fig. 17. Since the dopability is directly connected to the band edge positions, the dopability scale given at the fight y-axis is directly related to the energy scale. As an additional feature, the use of ZnSTe, which should also allow high p-doping levels, is proposed as a lattice matched p-contact. Such asymmetric devices have already been realized, in principle, for the example ofZnMgSe/ZnSeTe LEDs. [67][68] n

...:. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

=

..........................

1014

........................................................................................................................... ..............................................................................................................................

10, r~

oo

oo

o

~2 >

c5

=1

0

N

c5

N

~

~._

N

i iiiii iiiiiiiiiiiiiii;iiiiiiiiiiiiiiiii'~ ".............................................................................................................................. 1 0 1 4

p

Figure 17. Schematic energy band diagram under fiat band conditions for a proposed asymmetric blue laser diode with a ZnMgS/ZnTe superlattice cladding and a ZnSe active zone.

Interesting new possibilities may arise from the use of the highly pdopable compound BeTe, which is nearly lattice matched to GaAs.[ 3] Superlattices between BeTe and ZnSe have already been produced in high quality. However, these superlattices are of the staggered type with an overlap of the two energy gaps of only about 1.9 eV.[ 3] They cannot be used as

Doping Limits and Bandgap Engineering

29

cladding layers in a laser. For this purpose, BeTe should be combined with a material with a much larger energy gap than ZnSe. The best candidate from a confinement and lattice match point of view is cubic MgS.[ 2] In fact, miniband calculations for such superlattiees reveal that they would be potential candidates for highly p-doped cladding layers with energy gaps in excess of 3.2 eV. However, the combination BeTe/MgS is somewhat exotic and may be difficult to realize. A more convenient altemative could be a ZnBeSe/ZnBeTe superlattice. The addition of Be to ZnSe reduces the lattice constant, while the addition of Zn to BeTe enlarges it, so that a combination of a selenide and a telluride with constant Zn to Be ratio, and equal thicknesses, is approximately stain balanced on GaAs. A Be content of about 50% is estimated to lead to approximately zero conduction band offset between the two materials, whereas the valence band of ZnBeSe is much lower than that of ZnBeTe. According to the miniband calculation, short period superlattices of this type can have energy gaps as large as 3.15 eV, if they can be fabricated without significant interdiffusion. At the same time, the valence band of these superlattices is more than 300 meV above that of ZnSe. The obtainable p-doping levels should consequently be limited only by solubility, and no longer by Fermi level pinning. Since it can expected that such a highly doped superlattice may be directly contacted with a metal, the valence band grading in this structure is not at the contact, but between the p-cladding and the p-waveguide.

5.0

OHMIC CONTACT TO

p-ZnSe

Soon after the introduction of nitrogen doping, it became evident that it was difficult to obtain good ohmic contacts to p-ZnSe. The reason for this was that the valence band of ZnSe is very low compared to the work function of even noble metals as Au or Pt. Consequently, a large Schottky barrier was formed between ZnSe and these metals. Additionally, the obtained doping levels in p-ZnSe remained limited to about 1018 cm "3, so that no effective tunneling through that barrier can lower the high contact resistance. Both problems are much less severe for ZnTe, which exhibits a higher valence band edge and much higher p-doping levels. As a result, several groups developed p-contacts to ZnSe that involve ZnTe. In order to reduce the problem of the large valence band offset between ZnTe and ZnSe, which acts as an additional barrier for holes, two

30

Wide Bandgap Semiconductors

strategies were followed. Fan et al. introduced a ZnSe/ZnTe superlattice with a quasi continuously varied Te content between the ZnSe and the ZnTe;[ 7] whereas Ishibashi et al. designed a series of ZnTe quantum wells in ZnSe which should allow sequential resonant tunneling of holes from the ZnTe through the quantum wells to the ZnSe.[ 81The decisive element for a good contact is believed to be an optimized grading in the first case and a precise design of the resonant tunneling structure in the second case. Finally, both approaches led to the development of ohmic contacts and allowed a reduction of the laser threshold voltage at room temperature to values well below 5 V[ 13] so that the p-contact problem for blue ZnSelasers seemed to be solved. New results, however, show that the contact properties are mainly governed by a diffusion process. This is demonstrated in Fig. 18, where IV characteristics of several ZnSe diodes are compared.[ 69] The contact used is a multi-quantum well structure similar to the one published in Ref. 8, followed by a ZnTe cap layer of 20 nm thickness. Based on the observation ofTaike et al., doping of the entire contact structure resulted in nonohmic contacts.[ 7~ The ZnTe quantum wells in the structure and the first 15 nm ofthe ZnTe cap were not doped. Only the topmost 5 nm of the ZnTe were heavily doped with nitrogen under the same plasma conditions as used for the p-ZnSe. This highly doped ZnTe layer was then contacted with gold which was evaporated, without breaking the vacuum, in a metallization chamber coupled to the MBE system. The only difference between the diodes was the substrate temperature at which the contact was grown; the temperature for the growth of the diode itself was 280~ for all samples. The solid curves in Fig. 18 represent diodes where the temperature for contact growth was set to 250, 230, and 200~ respectively. It was evident that a reduction of the growth temperature leads to a drastic improvement of the IV curve. This was a strong indication that the contact resistance was dominantly determined by a diffusion process and not, as commonly believed, by the exact design of the quantum wells in the grading or the resonant tunneling structure. A further reduction of the growth temperature to 160~ resulted in the deposition of polycrystalline ZnTe and a deterioration of the contact (see thin solid line in Fig. 18). The importance of diffusion is additionally underlined by the IV characteristics shown in Fig. 19. In this sample series, the growth temperature for all contacts was 200~ Only the thickness of the ZnTe cap was reduced from 10 to 5 nm until finally the cap was completely omitted. In

Doping Limits and Bandgap Engineering

31

order to guarantee a highly doped layer below the gold in these very thin contacts, in contrast to the samples shown in Fig. 19, the entire contact was doped with nitrogen. It can also be seen that the thickness of the ZnTe strongly affects the IV characteristics. For the contact where the ZnTe cap was completely omitted, the voltage which must be applied to obtain a current density of 600 A/cm 2 (a typical threshold current density for a ZnSe-based laser) was as low as 4 V, which was even lower than the best value reported for a similar contact structure.[ 13] This result can be also understood in terms of diffusion, if one assumes that the ZnTe cap acts either as a source or a sink for the diffusing species. A reduction of the ZnTe thickness limits the diffusion process by either reducing the amount of the species available for diffusion or the reservoir to which the diffusing species can migrate.

10,

,

'

" '

"'

'

'

~ZnTe

'!

'

I'

200 ~ ,

'

'

I'

160~

'

'

I'

230~

'

250~

8

"

n-ZnSe

4

Ii I ' : , 1

2

I! I ,I/j

0 I

-15

,

I

-10

.

I

-5

,

I

0

,

I

5

,

/I

i1

I

l

10

l

15

,,'1 i

I

20

i

I

25

.

l

30

Voltage ( V )

Figure 18. Current-density vs voltage for four ZnSe diodes with a ZnSe/ZnTe multiquantum well contact at the p-side. The structure of all diodes is identical, but the substrate temperature during the growth of the multi-quantum well contact was varied between 160 and 250~ The dotted lines represent results of a simulation.

32

Wide Bandgap Semiconductors

10 ,~

8

4

0 -2

0

2

4 6 Voltage ( V )

8

10

12

Figure 19. Current-density vs voltage for four ZnSe diodes with a ZnSe/ZnTe multiquantum well contact at the p-side. All contacts were grown at a substrate temperature of

200~ but the thickness of the overlyingZnTe cap layerwas varied between 0 and 10 nm. An additional experiment demonstrates that the diffusion induced deterioration of the contacts does not occur in the graded multi-quantum well itself, but rather in the underlying p-ZnSe. A diode was grown by MBE in which the p-ZnSe was replaced by p-ZnSTe, a mixed crystal which can by grown lattice matched on ZnSe for a Te concentration of 35% and doped p-type with nitrogen plasma.[661 This layer was followed by a ZnSTe stepgraded layer, consisting of five layers of 4 nm thickness for which the Te content was adjusted to 45%, 55%, 65%, 75%, and 85% by reducing the sulphur flux from the valved cracker sulphur source. This step-graded layer replaced the ZnSe/ZnTe multi-quantum wells used in the previous samples. The structure was terminated with a 20 nm ZnTe cap. The sample structure is shown as an insert at the left side of Fig. 20. Since high quality ZnSTe can only be grown at high substrate temperatures, the entire diode, as well as the contact, was grown at 280~ The IV characteristics of this sample are shown as a solid line in Fig. 20 Although it was grown at a high substrate temperature and with a relatively thick ZnTe cap, the IV curve is similar to the one of the best ZnSe diode. This diode was produced at a much lower substrate temperature and without a ZnTe cap to minimize

Doping Limits and Bandgap Engineering

33

diffusion. In the n-ZnSe/p-ZnSTe diode, the diffusion process observed in the pure ZnSe diodes was obviously absent or at least greatly reduced. To make sure that this improvement is not due to the replacement of the ZnSe/ ZnTe multi-quantum well by the ZnSTe grading layer, a second, nearly identical, sample was grown. As the only change, an additional p-ZnSe layer was introduced between the n-ZnSe and the p-ZnSTe. The sample structure is depicted as the right insert in Fig. 20. In contrast to the previous sample, the IV characteristics of this sample (dashed line in Fig. 20) was much worse (although still much better than for a pure ZnSe diode with a contact grown at such a high temperature), which means that interdiffusion is again important, in this case. From this we conclude that interdiffusion causes a deterioration of the p-ZnSe. On the other hand, if one succeeds in suppressing interdiffusion, the detailed design of the grading itself seems to be of minor importance, as can be seen from the fact that comparably good contacts can be achieved for gradings as different as a ZnSe/ZnTe multi-quantum well or a step-graded ZnSTe layer.

10 -

.

I

8

6 !

~

~[

I

n-ZnSe _ n-~~.~~~

/

I [ [

[.... p-ZnTe -1

/

iznsTe gradin~ i p,ZnS0.65Teo.351

; ,.

2 L)

0 ,

-2

I

0

,

I

2

,

I

,

I

4 6 Voltage ( V )

,

I

8

,

I

10

,

12

Figure 20. Current-density vs voltage for two diodes containing p-ZnSTe layers. The detailed structures are given in the inserts. The only difference between the two diodes is that the p-ZnSe layer was omitted in one case.

34

Wide Bandgap Semiconductors

The microscopic nature of the interdiffusion process that affects the p-ZnSe cannot be deduced directly from these experiments. However, additional insight can be gained by analyzing the IV characteristics in more detail. The low current for the contacts grown at 230 and 250~ can only be understood as a result of a tunneling process. In fact, a reasonable description of the IV curves can be obtained with a model for tunneling through a triangular barrier introduced by a highly resistive ZnSe layer of a given thickness. Although this model does not give a perfect fit of the observed IV curves, it describes at least qualitatively the observed shift of the experimental IV curves to higher voltages with increasing contact growth temperature (Fig. 18). Best fits are shown as dashed lines in Fig. 18. Although they are not unambiguous in the sense that both a variation of the barrier height and the free carrier density have a similar influence, one can nevertheless see clear trends. For a fixed barrier height of 0.5 eV (which is lower than that for gold on ZnSe,[ 71] but high enough to prevent significant thermionic emission), the thickness of the resistive layer used for the fit decreases from 145 nm for a substrate temperature of 250 ~ to 42 nm for a temperature of 200 ~ A change of the barrier height results in slight variations of these values, but the relative thickness change of the resistive layer remains constant. In Fig. 21, the thickness of the resistive layer obtained in this way is plotted as a function of temperature. These values are compared to a calculation giving the relative change of the diffusion length as a function of the temperature. For diffusion from a finite as well as for an infinite source, the diffusion length is proportional to the square root of the diffusion constant times the diffusion time [72] (which is approximately constant for all structures presented here). The diffusion constant D itself depends on temperature according to an Arrhenius law of the form: Eq. (1)

D = exp(-E a/kT)

where E a is the activation energy of the diffusion process and k the Boltzmann constant.[ 72] The relative change of the diffusion length obtained in this manner is plotted as solid lines in Fig. 21 with E a as a parameter. The comparison to the experimental values shows that the observed behavior can be understood for activation energies on the order Of 1 eV. This value is too low for diffusion via regular lattice sites, and corresponds rather to values observed for diffusion via interstitial sites or along dislocations.[ 72]

Dop&g Limits and Bandgap Engineering

35

160 140

1,o

~ 120

0,8

100 N

r~ ~,,J~

0,6

>

80

9 v,,,~

~

60

~

40

~

20

t~

~

0

*~,,I

o

0,4

o,2

0

190

200

210

220

230

240

250

0,0

Contact Growth Temperature ( ~ )

Figure 21. Thickness of a resistive layer, as obtained from a fit to the IV curves shown in Fig. 18, as a function of the temperature at which the contact was grown. The solid lines are calculated relative changes of the diffusion length with the activation energy E a as a parameter.

Based on these observations, our suggestion is that nitrogen diffuses along dislocations from the ZnTe cap and the grading into the p-ZnSe. The observation of such diffusion by secondary ion mass spectrometry (SIMS) has indeed been reported in the literature[ 7~ and could be driven by the different solubility of nitrogen in ZnSe and ZnTe.[ 2]] The free hole concentration in p-ZnSe would then be drastically reduced by the formation of nitrogen-nitrogen pairs (see See. 3.4 and Fig. 13) so that the diffusing nitrogen could lead to overcompensation in the p-ZnSe. These results can shed light on a few additional features of ZnSe/ ZnTe contacts. First, these contacts, and especially the resonant tunneling variant, seem not to be very reproducible. As an example, the device with a record lifetime of 101 hours reported by Ishibashi et al. needed a threshold voltage of 11 V[ 1~ (a value much larger than the lowest voltage of 4.7 V reported by the same group for a diode with a nominally identical

36

Wide Bandgap Semiconductors

contact).[ 13] This irreproducibility was still stronger when the results of different groups were compared. Several workers did not obtain ohmic contacts at all, although the same contact design and equivalent equipment was used.[ 7~ If the properties of the contact are determined by tunneling through a barrier created by nitrogen diffusion, even a slight variation of this barrier will have a strong influence on the IV characteristics, since the properties of the barrier appear exponentially in the tunneling probability. Thus, minor changes in growth conditions or composition of the plasma can lead to strong effects in IV characteristics. Second, aging experiments on laser diodes show that the threshold voltage tends to increase during operation.[ 73] This indicates that a contact degradation takes place during heavy duty operation. In the picture given here, this is an indication of nitrogen diffusion during device operation. Following the argument given above, even minor diffusion effects may be seen drastically in the IV curve. Although this diffusion during operation did not seem to be a factor that limited device lifetime, as long as the density of extended defects was high, it could become important for the structurally more perfect devices that are now available. Based on these observations, one can give some ideas how pcontacts to ZnSe can be further improved. A reduction of the dislocation density in the contact may help to suppress diffusion or at least to increase the activation energy of the diffusion process. In fact, the IV curve of the diode with a lattice mached ZnSTe spacer between the p-ZnSe and the graded contact (which is shown at the right side of Fig. 20) is better than that of a pure ZnSe diode grown at the same conditions. This indicates that the lattice matched spacer reduces interdiffusion. In that sense, the use of nearly lattice matched BeTe instead of ZnTe in contacts[ 3] may be an advantage. Completely avoiding nitrogen as a dopant in the contact region is still more promising. A first step towards this direction has been done by the use of p-Ge as a contact material.[ TM] An extension of this concept to graded ZnSe/Ge structures might even allow the realization of a lattice matched, nitrogen free contact, although significant difficulties for the epitaxy of ZnSe on Ge may be encountered in the practical realization of such a structure.

Doping Limits and Bandgap Engineering 6.0

37

CONCLUSIONS

Most experimental results on the maximum doping levels obtained in wide gap II-VI materials can be explained by a model that assumes a pinning of the Fermi level at a fixed position with respect to the vacuum level. In the case of n-type doping, this position is about 130 meV above the ZnSe conduction band edge for nearly all II-VI compounds. For pdoping, two different values (namely 120 meV and 580 meV above the ZnSe valence band edge) are obtained for doping with radio frequency and DC nitrogen plasma, respectively. This discrepancy can be understood by assuming the formation of complexes related to nitrogen-nitrogen pairs as the most stable compensating center, an assumption which also is able to describe the abrupt drop in the free hole concentration of RF plasma doped ZnSe if too much nitrogen is offered. In agreement with the model, an upward shift of the valence band edge through surface segregation, or the use of carefully designed superlattices, has the effect of increasing the obtained p-doping level. This fact can be used to gain more flexibility in the design of optoelectronic devices emitting in the blue spectral region. Finally, the formation of ohmic contacts to p-ZnSe is a process governed by diffusion. The results presented can be understood under the assumption that nitrogen diffuses along dislocations from the ZnTe cap into the underlying ZnSe. The excess of nitrogen leads to a highly resistive zone that must be overcome by a tunneling process and thus increases the operation voltages of ZnSe based diodes. This phenomenon may be reduced by the use of lattice matched contacts or contacts that do not involve nitrogen.

REFERENCES 1. Gunshor, R. L., Kobayashi, M., Kolodziejski, L. A., Otsuka, N., and Nurmikko, A. V., J. Crystal Growth, 99:390 (1990) 2. Okuyama, H., Nakano, K., Miyajima, T., and Akimoto, K., Jpn. J. Appl. Physics, 30:L1620 (1991) 3. Landwehr,G., and Waag, A., Proc. Int. Syrup. on Blue Lasers and LEDs, p. 17, Chiba (1996)

38

Wide Bandgap Semiconductors

4. Park, R. M., Troffer, M. B., Rouleau, C. M., DePuydt, J. M., and Haase, M. A., Appl. Phys. Lett. 57:2127 (1990) 5. Ohkawa, K., Karasawa, T., and Mitsuyu, T., J. Crystal Growth, 111:797 (1991) 6. Haase, M., Qui, J., DePuydt, J., and Cheng, H., Appl. Phys. Lett., 59" 1272 (1991) 7. Fan, Y., Han, J., He, L., Saraie, J., Gunshor, R. L., Hua, G. C., and Otsuka, N., Appl. Phys. Lett., 61:3160 (1992) 8. Ishibashi, A., and Mori, Y., J. Crystal Growth, 138:677 (1994) 9. Okuyama, H., Miyajima, T., Morinaga, Y., Hiei, F., Ozawa, M., and Akimoto, K., Electron. Lett., 28:1798 (1992) 10. Taniguchi, S., Hino, T., Itoh, S., Nakano, K., Nakayama, N., Ishibashi, A., and Ikeda, M., Electronics Letters, 32:552 (1996) 11. Albert, D., Olszowi, B., Spahn, W., Niimberger, J., Kom, M., Hock, V., Ehinger, M., Faschinger, W., Landwehr, G., J. Crystal Growth, 184/ 185:571(1998) 12. Kawasumi, T., Okuyama, H., Nakayama, N., Ishibashi, A., and Mori, Y., Electron. Lett., 31:1667 (1995) 13. Itoh, S., Nakayama, N., Matsumoto, S., Nagai, M., Nakano, K., Ozawa, M., Okuyama, H., Tomiya, S., Ohata, T., Ikeda, M., Ishibashi, A., and Mori, Y., Jpn. J. Appl. Phys, 33:L938 (1994) 14. Hua, G. C., Otsuka, N., Grillo, D. C., Fan, F., Han, J., Ringle, M. D., Gunshot, R. L., Hovinen, M., and Nurmikko, A. V., Appl. Phys. Lett., 65:1331 (1994) 15. Nurmikko, A. V., Jeon, H., Gunshor, R. L., and Han, J., J. Crystal Growth, 159:644 (1996) 16. Okuyama, H., Kishita, Y., Miyajima, T., and Ishibashi, A., Appl. Phys. Lett., 64:904 (1994) 17. Ishibashi, A., IEEE J. on Selected Topics in Quantum Electronics, 1:741 (1995) 18. Laks, D. B., Van de Walle, C. G., Neumark, G. F., and Pantelides, S. T., Phys. Rev. Lett., 66:648 (1991) 19. Van de Valle, C. G., Laks, D. B., Neumark, G. F., and Pantelides, S. T., Phys. Rev. B, 47:9425 (1993) 20. Van de Valle, C. G., Laks, D. B., Neumark, G. F., and Pantelides, S. T., Appl. Phys. Lett., 63:1375 (1993) 21. Fan. Y., Han. J., Gunshor. R. L., Brandt, M. S., Walker, J., Johnson, N. M., and Nurmikko, A. V., Appl. Phys. Lett., 65:1001 (1994) 22. Prior, K., Materials Science Forum, 182-184:11 (Trans Tech Publications), (1995)

Doping Limits and Bandgap Engineering

39

23. Chadi, D. J., Phys. Rev. Lett., 72:534 (1994) 24. Burkey, B. C., Khosla, R. P., Fischer, J. R., and Losee, D. L., J. Appl. Phys., 47:1095(1976) 25. Kachaturyan, K., Kaminska, M., Waber, E. R., Becla, P., and Street, R. A., Phys. Rev. B, 40:6304 (1989) 26. Terry, I., Penney, T., Molnar, S., Rigotty, J. M., and Becla, P., Solid State Commun., 84:235 (1992) 27. Han, J., Ringle, M. D., Fan, Y., Gunshor, R. L., and Nurmikko, A. V., AppL Phys. Lett., 65:3230 (1994) 28. Tao, I. W., Jurkowic, M., and Wang, W. I., Appl. Phys. Lett., 64:1848 (1994) 29. Ogawa, H., Irfan, G., Nakayama, H., Nishio, M., and Yoshida, A., Jpn. J. Appl. Phys., 33:L980 (1994) 30. Faschinger, W., Ferreira S., and Sitter, H., ,4ppl. Phys. Lett., 64:2682 (1994) 31. Ferreira, S., Sitter, H., and Faschinger, W., Appl. Phys. Lett., 66:1518 (1995) 32. Garcia, A., and Northrup, J., Phys. Rev. Lett., 74:1131 (1995) 33. Kurtz, E., PhD thesis, Univ. Wiirzburg, p. 32 (1996) 34. Walukiewicz, W., Materials Science Forum, 143-147:519 (1994) 35. Walukiewicz, W., Appl. Phys. Lett., 54:2094 (1989) 36. Walukiewicz, W.,J. Vac. Sci. Technol. B, 6:1257 (1988) 37. Ren, S. Y., Dow, J. D., and Shen, J., Phys. Rev. B, 38"10677 (1988) 38. Dow, J. D., Hong, R. D., Klemm, S., Ren, S. Y., Tsai, M. H., Sankey, O. F., and Kasowski, R. V., Phys. Rev. B., 43:4396 (1991) 39. Faschinger, W., Ferreira, S., Sitter, H., Krump, R., and Brunthaler, G., Materials Science Forum, 182-184:29, Trans Tech Publications, Switzerland (1995) 40. Marfaing, Y.,J. Vac. ScL Technol. B, 10:1444 (1992) 41. Marfaing, Y,J. Crystal Growth, 138, 305 (1994) 42. Qiu, J, DePuydt, J. M., Cheng, H., and Haase, M. A., ,4ppl. Phys. Lett., 59:2992 (1991) 43. Hauksson, S., Simpson, J., Wang, S. Y., Prior, K. A., and Cavenett, B. C., Appl. Phys. Lett., 61:2208 (1992) 44. Ferreira, S., Sitter, H., Faschinger, W., Krump, R., and Brunthaler, G., J. Crystal Growth, 146:418 (1994) 45. Rajakarunanayake, Y., Miles, R. H., Yu, G. Y., and McGill, T. C., Phys.Rev. B, 37:10212 (1988) 46. Langer, J. M., and Heinrich, H., Phys. Rev. Lett., 55:1414 (1985)

40

Wide Bandgap Semiconductors

47. Fischer, F., Waag, A., Bilger, G., Litz, T., SchoU, S., Schmitt, M., and Landwehr, G., J. Crystal Growth, 141:93 (1994) 48. Parbrook, P. J., and O'Donnel, K. P., Optical Properties of Wide Bandgap II-VI Superlattices, in: II-VI Semiconductor Compounds, (M. Jain, ed.), World Scientific, Singapore (1993) 49. Faschinger, W., Ferreira, S., Sitter, H., Krump, R., and Brunthaler, G., Materials Science Forum, 182-184:407, Trans Tech Publications, Switzerland (1995) 50. Tokumitsu, E., Jpn. J. Appl. Phys., 29:L698 (1990) 51. Nicolini, R., Vanzetti, L., Mula, G., Bratina, G., Sorba, L., Franciosi, A., Peressi, M., Baroni, S., Resta, R., Baldereschi, A., Angelo, J., and Geberich, W., Phys. Rev. Lett., 72:294 (1994) 52. Kowalczyc, S. P., Kraut, A., Waldrop, J., and Grant, R., J. Vac. Sci. Technol., 21:482 (1982) 53. Yoshida, S., Misawa, S., and Gonda, S., J. Appl. Phys., 53:6844, (1982) 54. Nakamura, S., Harada, Y., and Seno, M.,Appl. Phys. Lett., 58:2021 (1991) 55. Martin, G., Botchkarev, A., Rockett, A., and Morkoc, H., Appl. Phys. Lett., 68:2541 (1996) 56. Kurtz, E., Albert, D., Kraus, J., Hommel, D., and Landwehr, G., Proc. Int. Symp. on Blue Lasers and LEDs, Chiba, p. 429 (1996) 57. Einfeldt, S., Heinke, H., Behringer, M., Becker, C., Kurtz, E., Hommel, D., and Landwehr, G., J. Crystal Growth, 134:471 (1994) 58. Ohkawa, K., Tsujimura, A., Hayashi, S., Yoshii, Y., and Mitsuyu, T., Extended Abstracts of the Int. Conf. on Solid State Devices and Materials, Tsukuba, p. 330 (1992) 59. Faschinger, W., J. Crystal Growth, 197:557 (1999) 60. Kimura, K., Miwa, S., Yasuda, T., Kuo, L., Wang, T., Jin, C., Tanaka, K., and Yao, T., Proc. Int. Symp. on Blue Lasers and LEDs, Chiba, p. 167 (1996) 61. Hauksson, I, Simpson, J., Wang, S. Y., Prior, K. A., and Cavenett, B. C., Appl. Phys. Lett., 61:2208 (1991) 62. Zhu, Z., Brownlie, G., Horsburgh, G., Thompson, P. J., Wang, S. Y., Prior, K. A., and Cavenett, B. C.,J. Crystal Growth, 159:248 (1996) 63. Baron, T., Saminadayar, K., and Magnea, N., Appl. Phys. Lett., 67:2972 (1995) 64. Jobst, B., Strauf, S., B~iume, P., Kurtz, E., Schenk, H., Gutowski, J., Hommel, D., and Landwehr, G., Proc. Int. Symp. on Blue Lasers and LEDs, Chiba, p. 409 (1996)

Doping Limits and Bandgap Engineering

41

65. Turco-Sandroff, F. S., Nahory, R. E., Brasil, M. J. S. P., Martin, R. J., Besermann, R., Farrow, L. A., Worlock, J. M., and Weaver, A. L., jr. Crystal Growth, 111:762 (1991) 66. Kom, M., Niimberger, J., Faschinger, W., Spahn, W., Ehinger, M., and Landwehr, G., Proc.of the 23rd Int. Conf. on the Physics in Semiconductors, (M. Scheffler and R. Zimmermann, eds.), World Scientific, Singapore, p. 1023 (1996) 67. Faschinger, W., Krump, R., Brunthaler, G., Ferreira, S., and Sitter, H., Appl. Phys. Lett., 65:3215 (1994) 68. Krump, R., Brunthaler, G., Faschinger, W., Ferreira, S., and Sitter, H., Mater Sci. Forum, 182-184:349 (1995) 69. Niimberger, J., Faschinger, W., Schmitt, R., Kom, M., Ehinger, M., and Landwehr, G., AppL Phys. Lett., 70:1281 (1997) 70. Taike, A., Momose, M., Kawata, M., Gotoh, J., and Mochizuki, K., J. Crystal Growth, 159:714 (1996) 71. Suemune, I., Appl. Phys. Lett., 63:2612 (1993) 72. Schumicki, G., and Seegbrecht, P., Prozefltechnologie, Springer, p. 262ff (1991) 73. Ishibashi, A., IEEE J. Quantum Electronics, 1:741 (1995) 74. Ohki, A., Ohno, T., and Matsuoka, T., Proc. Int. Symp. on Blue Lasers and LEDs, Chiba, p. 252 (1996)

2 Epitaxial Growth of I I - V I C o m p o u n d s by MOVPE Wolfgang Gebhardt and Berthold Hahn

1.0

INTRODUCTION

This article reviews the present state of metal organic vapor phase epitaxy (MOVPE) of wide gap II-VI semiconductors, a class of materials, which is typically represented by ZnSe and related ternary and quaternary compounds. Epitaxial growth of large gap II-VI compounds has attracted a renewed interest, in the last years, since the successful operation of blue diodes and injection lasers based on ZnSe related material. Although MOVPE of II-VI compounds has not yet reached the same maturity as MBE growth, important progress has been made in the last years. Furthermore, we are convinced that there is no intrinsic problem connected with MOVPE which might hamper a successful production of II-VI devices. Both methods of epitaxy, MBE and MOVPE, still suffer from the lack of appropriate II-VI substrates, hence, (001)-GaAs wafers are widely used. The purity of MOVPE precursor compounds was, with exception of the hydrides, a severe problem in the past, but has now been solved for most of the precursor molecules given in Table 1.

42

Epitaxial Growth of ll-Vl Compounds by MOVPE

43

Table 1. Thermodynamic Data MOVPE Precursor Frequently Used for Growth and Doping of II-VI-Compounds[ 1]

Compound

t-(C4Hlo)OH

Shortnotation in MOVPE TBOH

82.2 (MP 25.5~

Vapor Pressure at 20~ (Pa) 5600(25~

-60.7

H2S t-(C4HIo)SH

Boiling Point (~ at Normal Pressure

TBSH

63

16 646

-41.5

H2Se

2926

(C2Hs)2Se

DESe

t-(C4Hio)2Se

DTBSe

122

580*

i-(C3Hs)ETe

DIPTe

49 at 18.12 Pa

282

(C2H5)2Zn

DEZn

118

1395

(CH3)2Zn-(C2H5)3N

DMZn-TEN

95

1733

(CH3)2Zn

DMZn

46

16 277 at 0~

(CH3-CsH4)2Mg

(M-CP)2Mg

C2H5I

El

72.3

13300

n-(C4Hlo)C1

n-BCI

78

11 000

(C2Ha)3Ga

TEGa

143

490

t-(C4HIo)PH2

TBP

54

216.9

t-(C4Hlo)AsH2

TBAs

66

19710

t-(C4HIo)NH2

TBN

46

33 701

21 at 40~

*This vapor pressure was derived from consumption measurements in the authors' group.

By far, most of the results presented in this article were obtained with a horizontal laminar flow reactor used in many commercial growth facilities. In our laboratory, an Aixtron MOVPE system AIX200 is used. These kind of reactor cells are operated between atmospheric and low pressure (1000 hPa-0.1 hPa). Their hydrodynamic properties, with

44

Wide Bandgap Semiconductors

respect to II-VI epitaxy, have been investigated[ 2] and reviewed by W. Kuhn. [3] There is a well known inhomogeneity of temperature and growth rate in cells along the reactor axis. The growth rate (GR) has a maximum at the gas inlet, which has to be taken into account when results of different groups are to be compared. The inhomogeneity in the used reactor is, at atmospheric pressure, about 10% over a 2" wafer and can be greatly reduced either by low pressure MOVPE or by installation of a rotating susceptor. In a typical MOVPE system three growth regimes may be distinguished (see Fig. 1): 9 A kinetic limited region at low temperatures where the rate of the surface reactions limits the GR 9A flat region which is defined by diffusion or transport

limited growth 9 A high temperature region in which surface desorption and thermal activation of further chemical reactions suppress growth increasingly Growth optimization should be performed (varying mass flow and temperature) within the diffusion limited regime where the GR is practically independent of temperature.

I

(9

I

I

transport

I I

(b) i

2 t--

kinetics (a)

'!' I

I I

0 o) o) o L_

..,.,.

I

I

I i

i

I i

ill.,

,,.,

1/T 0

Figure 1. Three regions of MOVPE growth.

i.

||

Epitaxial Growth o f l I - V l Compounds by M O V P E

2.0

BINARY COMPOUNDS

2.1

ZnTe

45

This substance grows easily on (001)-GaAs using DIPTe and DEZn, or DMZn-TEN as metal organics. The last mentioned adduct, first introduced by Jones, has been proved to be a very useful Zn-compound.[ 4] It is less volatile than DMZn, prevents prereactions in the gas phase, and leads to layers with a smooth surface and less deep center luminescence. The growth in our laboratory was usually performed with DIPTe and DEZn, or DIPTe and DMZn-TEN, at a substrate temperature of Ts = 340~ and a carrier gas flow of 5 stdl/min at atmospheric pressure. Typical input precursor flow rates are found in Fig. 2 which presents a typical plot of GR versus inverse temperature. The kinetic region in Fig. 2 fits a straight line which yields an activation energy (E~) and a prefactor of the surface reactions E r = 102 + 12 kJ/mol and A r = 6.108 pm/h for DIPTe + DEZn and E~ = 96 • 20 kJ/mol and Ar = 2" 108 ~tm/h for DIPTe + DMZn-TEN.

Growth Temperature [~ 400 '

I

380 '

360

I

'

:340

I

'

320

I

'

300

I

'

I

'

DIPTe+DEZn DIPTe+DMZn-TEN

s"A %

_

l.,-=..e C"

-1 ~IL%

s

%

s

E "--i

O---

~3'

t J

=__.J

s $

,,N=,

s

9

$ s s

r

~

J

J

o L_

0

[DEZn] =17.5 pmol/min [DMZn-TEN]=23.0 pmol/min [DI PTe]=34.1 pmol/min 0.1

0,1 ,

I

1.50

,

I

1.55

.,

l

1.60

,

i

1.65

~

i

1.70

,

.I

1.75

1000/TG[1/K] Figure 2. Growth rate of ZnTe versus 1/T for two different Zn precursors.P]

46

Wide Bandgap Semiconductors

Kuhn et al. investigated the co-pyrolysis DIPTe and DEZn in an isothermal quartz tube under H2-flow (see Fig. 3). They found the respective activation energies of 71 and 74 kJ/mol. The molar ratio of DIPTe and DEZn in this experiment ensured that the walls of the tube were homogeneously covered by ZnTe. As in pyrolysis of the separate metalorganics propene, propane, 2,3-dimethylbutane from DIPTe and ethene, ethane and n-butane from DEZn appear. Additionally, 2-methylbutane has been detected in significant concentrations which is diminished when He is used as carrier gas. Alkenes as reaction products give evidence for fl-elimination. Eq. (1)

(C2Hs)2Zn -~ C2H4 + C2H 6 + Zn

and Eq. (2)

(C3H7)2Te ~ C3H 6 + C3H s + Te

The production gain ofalkenes is reduced when He is the carrier gas, indicating that the carder H 2 also participates. Eq. (3)

(C2H5)2Zn + H 2 --~ C2H 6 + C2H5ZnH

and Eq. (4)

C2HsZnH ~ C2H 6 + Zn ..........................

r ...........................

I ......................

v ..........................

I .................

'

. . . . . . .

I ......................

100

' ......................

! .......................

..

:DPT:

80 a===

09 f-

.~ f-

60-

, . J

03

:~

40-

o

20-

e-

200

'

'

350

400

Temperature[~ Figure 3. Copyrolysis of DEZn and DIPTe.[S]

Epitaxial Growth of ll- VI Compounds by MO VPE

47

There is also a reaction of radicals which yields butane and substituted butanes. When DMZn is used, t-substitution is not possible. Pyrolysis experiments show a clear evidence for a contribution of H2-carrier gas.[ 6] Eq. (5)

(CH3)2Zn --~ CH 3. + - Z n CH 3

Eq. (6)

CH 3" + H 2 --~ CH 4 + H.

The additional effect of (C2H5)3N in the adduct compound DMZnTEN will be discussed with the growth of ZnSe. Before layer quality is discussed, we will consider the substrate preparation which was applied by us and by W. Kuhn in nearly all growth experiments of ZnTe. The polished surface of the GaAs wafers (not epiready) is degreased in propanol, then etched for 60 sec in HzSO4:HzO2:H20 in the proportion 4:1:1 at 40~ Immediately after this wet etching step, the substrate is rinsed in 18 Mr2 water. After drying by spinning, an in-situ deoxidation is followed by an annealing procedure of 15 min at 350~ in streaming Hz-gas. This process was later changed into a 15 min annealing step at 500~ ZnTe layers grown on GaAs suffer from a large lattice mismatch: Eq. (7)

f = as - at a t

which is -7.5% between 300~ and 400~ Here a s is the lattice constant of the substrate and a t that of the layer material under relaxed conditions. After the growth of a few monolayers, the strain is already largely relaxed by the formation of Lomer- and 60~ (see Fig. 4). This dislocation array yields a strain relaxation: Eq. (8)

6=

b.e(llO)

where the scalar product b'e(110) of the Burgers vector, and the unit vector along (110) is just the component ofb in this direction and d is the averaged distance between two misfit dislocations. The remaining unrelaxed strain at the interface ~1is given by the difference: Eq. (9)

~]1= 6 - f

48

Wide Bandgap Semiconductors

which is about -0.5%. When layers of various thicknesses are investigated (see Fig. 4), one finds a saturation of the relaxation at a thickness of about 0.8 ~tm. A very small residual compressive strain of about -0.05 % remains, even in thick layers, but is completely compensated for and even converted into tensile strain when cooled down to liquid Helium temperature for measurements. The dislocation density at the interface evaluated from high resolution transmission electron microscopy (HRTEM) images is about 6.10 ~2 cm -2, whereas, in a 3 ~tm thick layer, only about l0 s cm -2 can be determined by the broadening of x-ray rocking curve.

................... ' ............

I ...........

' ............

I ............

r .......

I . . . . .

' ........

I .........

'

.......

I

.......

r .............. [ ................

~ .............

e=,=~

c

0.0

o=,..

f.3

ffl &

II1 ---

9

/.D

9

-0.1

11.

-0.2 o

9 ion channeling DIPTe+DMZn-TEN 9 reflectance DIPTe+DEZn reflectance DIPTe+DMZn-TEN

. . . . . . . . . . . . . . . j . . . . . . . . . . . . . . . . 1 . . . . . . . . . . . . . . . . . . , . . . . . . . . . . . . . . . . . ! .................... ,. . . . . . . . . . . . . . . .

0.5

1.0

I .............

1.5

J .............

1 . . . . . . . . . . . . . . . . . , ................... !

2.0

.................. ~ .................... i . . . . . . . . . . . . .

2.5

~-. . . . . . . . . . . . . . .

3.0

Thickness [IJm] Figure 4. Parallel surface strain of ZnTe layersgrown on GaAs(100) as a function of layer thickness.[7H9]

Our laboratory has made some attempts to study the growth of cubic ZnTe on GaAs(111 A) 2 ~ off. The growth under standard conditions (DIPTe + DMZn-TEN, Ts = 340~ yielded layers which were nearly defect free. HRTEM images show a dark interface zone which extends about 10 nm into the ZnTe layer (see Fig. 5). The misfit dislocations have Burgers vectors parallel to the (111)-planes. Since these planes are glide planes of the zincblende structure, the dislocations are confined within these planes parallel to the interface. The FWHM of the x-ray (111)-reflection was 82 arc sec and, thus, close to the theoretical value of 72 arc sec for ZnTe-layers of 200 nm thickness.

Epitaxial Growth o f l l - V l Compounds by M O V P E

" .- -~,~~.,~- .

49

: ,~ ~ ,'.vr. . '~ : ~.

e. ~!

FWHM 24aw.sec GaAs( 14.1!

8Znre 2 (111) ~f cl~c

i O0 nm

4~

Omegaldegl

Figure 5. Cross section view of the ZnTe/GaAs (111) interface region together with the rocking curve of the (111 ) reflex.[ !~

Doping of ZnTe has been attempted with a variety of elements, however only p-doping has been achieved. The dopants used in our laboratory were: TBAs, TBP, TMBi [(CH3)3Bi ], EI, TEGa, EDMIn [(C2Hs)(CH3)2In ]. A successful doping up to p ~ 1018 cm -3 was obtained only with TBAs and TBP. Typical results from Hall measurements are given for three samples in Table 2 (see also Ref. 11).

Table 2. Acceptor Binding Energy AEA, Hole Concentration NA and the Compensation ND/NA in Three Samples Doped with TBP and TBAs

ZnTe:As- 12

Sample No.

ZnTe:P- 125

ZnTe:P- 124

AEA [meV]

46

35

58

NA [cm -s]

1.6.1017

7.8.1017

4.1.1016

N D [cm "3]

3.1.1016

1.6.1017

1.9.1016

No/NA

0.2

0.2

0.45

50

Wide Bandgap Semiconductors

The electron mobilities (~t) thus obtained are quite satisfactory. An example of ~t(T) is given together with a theoretical curve in Fig. 6.

10,8

1000

10"

10'e

750

E 10" ~'

~" "E

t

tO

'!ii I1'

[P]/Ere]: [P].~./Fe~

L

o

250

~ 10'2

10" 101o

5

10 15 20 1000/T [l/K]

25

00 1 ~50' 100....150' " " 200 ' ....250' ....300 Temperature [K]

Figure 6. Carrier concentrationand Hall mobilityof ZnTe:P.

2.2

ZnSe

The growth of ZnSe is usually performed with the same Zn sources as are commonly used for ZnTe. The usual Se sources are H2Se, DMSe, and DTBSe.[ 12] H2Se has the advantage of being readily available in relatively high purity. Disadvantages include its high toxicity and its tendency to prereact in the gas phase with the Zn-eompounds. Therefore, many laboratories have replaced H2Se by less hazardous compounds. Some have already been listed in Table 1. Heuken et al.[ 131have published a comparative study of ZnSe growth rates using various substituted selenium compounds (see Fig. 7), from which DTBSe allows by far the lowest growth temperatures (320~ < Ts). Further advantages of DTBSe are an exceptional broad region of diffusion limited growth (320~ < Ts < 360~ no detectable carbon incorporation, and no prereactions. High quality

Epitaxial Growth o f l l - V l Compounds by MOVPE

51

layers are thus grown at a relatively low selenium to zinc ratio (Se/Zn _=_ 2, compared to 18 necessary when H2Se is used).

550

500

450

400

350

300

o.1 f

......&

I t

--e-- DE_zn*DIPSe --"-- DEZn+DASe -+- DMZn+DESe

1.2

DEZn+DTBSe

1.3

_

1.4

1.5

1.6

1.7

1.8

1/TG [1 O00/K] Figure 7. ZnSe growth rates with DMZn and various Se-precursors.[ l 1][13]-[16]

Substrate Preparation. In the authors' laboratory, the standard procedure to grow ZnSe layers is the following: 9 The polished surface of the GaAs wafers (not epiready) is degreased in boiling propanol then rinsed in 18 MC~ water 9 A 60 see. etch at r.t. with H 2 S O 4 : H 2 0 in the proportion 4:1 9 Immediately after this wet etching step, the substrate is again rinsed in 18 M ~ water and dried by spinning 9 Lastly, an in-situ, 20 min desoxidation in flowing hydrogen at 500~ A slightly different procedure is applied to epi-ready GaAs substrates which, after wet etching, are only shortly in-situ annealed at 360~ G r o w t h Conditions with DTBSe. DMZn-TEN and DTBSe are used at Ts = 340~ with Hz-earrier gas flow of 5000 scem/min, Zn = 29 ~tmol/min, Se = 87 ~tmol/min, Se/Zn = 2-3. The pyrolysis of DMZn-TEN has been investigated in a clean growth reactor by recording the mass

52

Wide Bandgap Semiconductors

spectra between room temperature and 400~ The difference spectra reveal, as decomposition products, mainly ammonia, ethylene, and methane. The latter stems from DMZn in H 2, whereas NH 3 and C2H4 are products of the TEN pyrolysis. The amine is always adsorbed at the Se- or Sell-determined surface, a conclusion which was reached by in-situ experiments with a reflection difference spectrometer.[ 25] The complete decomposition of the amine on the surface leaves a Se-Zn-bond and releases NH 3, C2H4, C3H6, and C3Hs. Propene and iso-propane stem from the decomposition of DTBSe. Thus, the amine inhibits a quick reaction of Zn and Se on the surface. A model of the possible steps of this reaction has been suggested in Ref. 17. It is important to notice that H2-carrier gas does not participate in the thermal decomposition of DTBSe which disintegrates by ~-elimination. Therefore, H 2 may be safely replaced by an inert carrier gas like N 2 or He.[is] PL spectra may be used as a sensitive tool for the detection of point defects in the layers. As shown in Fig. 8, there is no deep luminescence in layers grown with DTBSe under the above mentioned conditions. Free and donor bound excitons dominate the near band gap luminescence. The donors are presumably chlorine atoms, impurities which are likely to originate in DTBSe. Although the purity of DTBSe has continuously improved, there is still much to be done to reduce background donor concentration. The density of line defects, such as dislocations and stacking faults, is another problem and is the principal source of degradation in II-VI LEDs and laser diodes. In order to reduce the number of ingrown dislocations, a 2-dimensional growth mode should be achieved as is commonly used in MBE. This is difficult to control in MOVPE, since all the powerful surface analytical methods, which need ultra high vacuum, cannot be used. In spite of this difficulty, MOVPE-growth was controlled by RHEED. For this purpose, growth was interrupted and the sample was cooled down to room temperature in flowing nitrogen. The sample was quickly transferred to a MBE-chamber. The RHEED pattern of samples grown under standard conditions (see above) showed a 3D-characteristic and [ 110]-oriented facets. When growth is continued in the MBE-chamber, a 2D-growth mode reconstitutes after a few monolayers. This procedure can be repeated several times with equal results. The results were independent of substrate preparation which was always the same. In Succeeding work, we were able to show that TEN in the adduct compound played a crucial role in the promotion of a 3D-growth mode during MOVPE. When DMZn was used, a 2D-growth mode could be achieved, although at relatively low temperatures.

Epitaxial Growth o f l I - V I Compounds by MOVPE

2.8

Energy [eV] 2.4

2.6

I

'

' I

'

""

2.2

I

"

'I

'

'

'

hh

.

,-.,

53

9

PL ZnSe 2K

"~

12

I,o

Z t"t

'

I

I

"

4380'

I

4 00

,

'

I

,

,

5000 Wavelength [A]

4400

.... 4 J l 2 0 -

i

"

'

.... 4~1.40 .... 4J160

.

.

.

.

.

5500

Figure 8. PL spectra of a 200 nm ZnSe layer at 2 K.

Growth Conditions with Other Se-Precursors. A comparative study of growth rates using DASe, DIPSe, DESe, and DTBSe together with DEZn at various VI/II ratios, has been carried out by Heuken et al.[ 13] In Fig. 7, the results are plotted versus 1/T. Data of another study using DEZn + DMSe and DEZN + MSeH are taken from Refs. 13, 16, and 15. In Refs. 14-16, there was broad experience with DMSe and DESe at growth temperatures above 400~ including photoassisted growth as low as 350~ This growth, however, needs H 2 as reactand (see below). In practice, it sometimes may be useful to choose an appropriate precursor in order to cover a convenient temperature range for growth. However, before planning the respective experiments, one should always ask if the chosen precursors arc available in the required purity. Photoassisted Growth. The lowest growth temperatures under standard conditions using alkyl selenides have been achieved, as shown above, with DMZn + DTBSe or TBSeH.[ 19l If necessary, the growth temperature Ts can be further reduced when the sample is irradiated with

54

Wide Bandgap Semiconductors

above bandgap light. In our laboratory, we have used a 150 W xenon lamp, which gives an intensity up to 45 mW/cm 2 at the sample surface. It was found that the diffusion limited region extended now to about 300~ when DMZn-TEN was used. In general, photoassisted growth facilitates doping but can also lead to an increase in the impurity background. Respective results have been published for DMSe and DESe by Fujita et al.[ 2~ and for DTBSe by Hahn et al.[211(See Fig. 9.) In general, the region of diffusion limited growth is shifted to lower temperatures and the reaction limited growth is governed by a lower activation energy.[ 171In the case of DTBSe + DMZn or DMZn-TEN, this yields, at 300~ a growth enhancement factor of 2.1 or 3.4 respectively. Fujita and Fujita summarized the known experimental facts and proposed the following model. The above band gap photons produce electron hole pairs with a quantum yield up to 10%. The holes accumulate at the ZnSe surface due to band bending.[ 221This effect changes the redox potential at the surface. Surprisingly, adsorption and decomposition of the Zn-precursor play important roles and determine the rate limiting step in photoassisted MOVPE (PA-MOVPE). The evidence was found by mass spectrometry [23] and surface sensitive experiments like RDS [24][25] and was also supported by ab-initio calculations.[ 26] When DMZn or DEZn molecules reach the positively charged surface, they bend and act as electron donors which react with the holes. Before a simple reaction scheme is formulated, the role of hydrogen should be emphasized. A Sedetermined surface grown by MBE shows reconstruction since the dangling bonds form dimers. However, if the surface is grown by MOVPE, all dangling bonds are occupied by hydrogen. Therefore, the reaction scheme may be written: Eq. (10)

(Se)surf-H + CH 3 Zn CH 3 ~ (Se)surf-Zn- CH 3 + CH 4 and

Eq. (11)

(Se)surf-Zn- CH 3 + CH 3 Se CH 3 --+ SeZn-(Se)surf CH 3 + C2H6

Here, (Se)sur f symbolizes a Se-atom at a normal lattice site but in surface position bound to two Zn-atoms in the lower lying layer. It is an essential aspect of this model, supported by all experimental results, that PA-MOVPE is a photocatalytic surface reaction. Furthermore, atomic hydrogen plays an important role, it covers the surface, is highly mobile, and reactive.

Epitaxial Growth of lI-VI Compounds by 510VPE

500

450

,

I

,-.,1

"

,,

I

,,

To [*C] 400

"

I

350

"

300

!

0 .......... 0 ~ ~ 0 - - - - - - - - - - - - - -

55

"

0"~

'

t.-

i

E "1 ,,._...=

L_ r

0.1"

Growth enhancement by illumination

--o-- DMZn+DESe ~ --..-- DMZn-TEN+DTBSe ~ I

1.3

'"

"

I

'

"

1.4

I

9

"

li~

1.5

1.6

"

"

'1

'

'

1.7

'"

'

!

1.8

1000/TG [l/K]

Figure 9. Growth enhancementunder irradiation with above band gap light; open symbols depict samples grown in photoassisted mode.[2~l[151

n-Doping. A variety of group-Ill-cations and group-VII-anions have been used, more or less successfully, as n-dopants of ZnSe.[ 27] The highest donor concentrations, well in the region of degeneracy, were obtained with C1 and I. Deep donor centers are not formed before very high concentrations (C1 and I > 1019 cm -3) are reached.[ 28] In Table 3, precursor molecules, doping results, and references are summarized.

Table 3. N-Type Doping of ZnSe with Group III and Group VII Elements

Dopant Atom

Precursor

I

C2H5I

Cl

n-C4HI0

Results

29, 30 n > 101Scm-3

27,31 27,32

AI Ga

Ref. #

(CH3)3Ga

n --__10Is cm"3, but forms deep centers

33, 34

56

Wide Bandgap Semiconductors

p-Doping. The p-doping of ZnSe and ZnS have a common problem believed to be due to the tendency in these compounds to have the Fermi level pinned in the upper region of the energy gap. This leads to donor states even with most of the group I and group V elements.[35][ 36] Only Li and N were found to be effective dopants yielding p > 1017 cm -3. However, lithium has been rejected because of its fast diffusion which occurs at moderate temperatures. This knowledge has led to numerous doping procedures in order to incorporate N-atoms into the lattice by MOVPE. They are summarized as follows" 9 Plasma doping with low pressure MOVPE[ 13] 9 MOVPE and PA-MOVPE with ammonia and substituted ammonia (organic amines)[37][38] 9 MOVPE with metal organic compounds with a direct nitrogen bond[ 39] 9 MOVPE and PA-MOVPE with azides or hydrazines[ 21][32][401141] The authors of this article have investigated a number of amines and found that, only those which have the structure R.NH 2, where R is an alkane or alkene radical, are effective p-dopants. The problem with most of these compounds, especially NH3, is the strength of the C-N bond which needs relatively high temperatures to dissociate. However, in order to achieve a sufficiently high doping concentration and to avoid interdiffusion in heterostructures, the growth temperature should be low. This is feasible when t-C4H10NH 2 (short notation TBN) is used. TBN has a weak C-N bond and already dissociates above 380~ The growth temperature can be further reduced by illumination. A standard procedure was described by Fujita et a1.[381[42] who used PA-MOVPE (1 > 300nm) with DEZn (6 ~mol/liter) + DMSe (12 ~mol/liter) at 200 Torr and 350~ The flow rate of TBN was varied between 6 and 12 I~mol/liter. A 30 minute post growth annealing in the reactor in flowing N 2 between 500 and 600~ was necessary to activate the nitrogen acceptors. Subsequent C-V-measurements revealed a hole concentration of up to 3.1017 cm -3. An extremely high purity of all precursors, especially with respect to halogen contamination, is an essential prerequisite of successful p-doping. The work cited above led to the suggestion that nitrogen is only incorporated if NH2-radicals reach the surface. On the other hand, hydrogen plays an important role and inactivates nitrogen acceptors. Therefore, it may be possible that at least one hydrogen atom stays either in the neighborhood of, or forms a bond with, the N-atom which occupies a Se-site.

Epitaxial Growth o f ll- VI Compounds by MO VPE

57

Therefore, compounds have been proposed which have a direct metal-nitrogen bond or which lack a NH-bond. Several compounds with direct metalnitrogen bonds have been synthesized and tested. However, no p-type conductivity has been found. [39][43] A member of the second group of compounds is ethylazide (C2HsN3) which easily decomposes into N 2 and N'. Although Nincorporation was achieved, [44]--[46] the layers showed high resistivity. The same is true for the application of substituted hydrazines, for which di-methyl- and phenyl-hydrazine (PhHz) have been used. The latter has also been applied by the present authors. Together with DTBSe + DMZn-TEN, very small molar concentrations of PhHz are sufficient to obtain samples in which nitrogen can be detected by SIMS. Under illumination with above band gap light, the optimum growth temperature is 300~ The N2-plasma treatment is a method borrowed from MBE where it has been quite successfully applied and leads to acceptor concentrations up to 4.1017 cm -3. In MOVPE growth and doping experiments, nitrogen incorporation has again been shown. However, p-conduction was not yet demonstrated.[ 13] For completeness, it should be mentioned, that p-ZnSe has been obtained by solid state diffusion using Li3N.[47]

2.3

ZnS

In early growth experiments, ZnS [48]-[50] (and ZnSSe) was grown with DMZn + H2S (+ H2Se) mostly between 3 50 and 400~ and often at a low pressure to avoid prereactions. Briot et al.[51] used DMZn-TEN + H2S and grew ZnS at 300~ with a reactor pressure of 40 Torr. The VI/II ratio was 5. The excellent optical properties of these layers were demonstrated by reflectivity and PL measurements.[ 52] Fujita et al.[ 53] replaced H2S by methylmercaptane (MSH) and grew ZnS between 400 and 500~ (see Fig. 10). In recent years, tertiary-butyl-mercaptane (TBSH) has become available.[ 54] Optimum growth conditions were found with DMZn-TEN +TBSH at a substrate temperature of Ts = 350~ and a reactor pressure ofP = 300 hPa.[ 55] Absorption measurements in the excitonic region of the material thus obtained have recently been published.[ 56] Low growth temperatures have also been reached with DMZn-TEN + DTBS. [57] (See Fig. 10.) Photoassisted growth allows lower growth temperatures as was explained for ZnSe. Fujita described the growth of ZnS from DMZn + DES under illumination with a xenon or Hg-lamp at temperatures between 100 and 150~ TMHowever, fragments of precursor molecules were incorporated leading to poor morphology of the ZnS-layers. The situation improved

58

Wide Ba n d g a p Semiconductors

when a ArF-Laser was additionally installed, leading to photodissociation of the MO-molecules and fairly good layers. In most practical applications, however, growth mastery should be attempted with as few external parameters as possible and the use ofDMZn-TEN +TBSH at Ts = 350~ may be preferred.

Ta[~ 700

600

500 450

400

350

300

250

t"

E ,,,n i.-...._a

0

-,--'

1

r

o 13 DMZn:TEN+TBSH

(.9

.

0.1

1.0

A

DMZn+DES

O

DMZn+MSH

[]

DEZn+DTBS

9

9 DMZn-TEN+DTBS .

I

1.1

.

,

J

.

1.2

,

I.

I

1.3

1.4

.

.

..

I

I

1.5

1.6

,

I

1.7

,

!

=

1.8

.I

1.9

=

2.0

1000/Ta[K] Figure 10. Growth rates of ZnS obtained with various precursors.[S3l[s4ltSTltSS]

Early doping experiments have been successfully performed with TMA1, HCI,[ 59] and EI.[ 6~ In the last mentioned reference, DMZn and HzS were used for growth which allowed growth temperatures as low as 260~ where the carrier concentration reached a maximum of about n = 2.1018 cm -3. In our laboratory, n-type ZnS was grown with DMZn-TEN + TBSH and n-BC1 as dopant at 330~ Degenerate carrier densities up to n < 5.10 TMcm -3 were easily reached.[ 611 Electroluminescence was another important field for industrial use of ZnS and especially of ZnS:Mn. For a description of the deposition of ZnS:Mn-layers by MOVPE, see Ref. 62.

Epitaxial Growth of ll- Vl Compounds by MO VPE 2.4

59

ZnO

ZnO has proven to be an excellent material for windows, antireflection coating, piezoelectric effects, and even for substrates on which to grow GaN. For this latter purpose, one needs single crystalline epitaxial layers grown on (0001)A1203 .[63] Conducting windows and antirefleetion coatings for solar cells [64]are deposited as polyerystalline layers on glass, CIS, [65] CdTe, and Si [66] from the gas phase where the crystallites are usually oriented with the e-axis normal to the substrate surface. Because of its large energy gap, the layers are transparent in visible light (~. > 380 nm). The transmission cutoff in the near IR is given by the plasma frequency and depends on donor concentration as well as on mobility (i.e., the carrier lifetime).[ 67] MOVPE of ZnO is usually performed with DEZn or DMZn and an oxidizing agent. Sometimes, the adduet DMZn:THF is used[ 68] where tetrahydrofurane (THF) may itself act as an oxidizer.[ 65] The substrate temperature ranges between 100 and 500~ Direct use ofO 2 or H20 in air, or a carrier gas such as N 2 or At', leads to prereaetions which may be partly avoided at reduced pressure. In the ease of CO 2, NO 2, and N20, however, the growth rate is rather small. Improvements were achieved by photoassisted MOVPE [67][69][70] and by using alcohols like t-BOH.[ 711 A comparative study of growth with H20 , CzH5OH , and t-BOH revealed superior properties oft-BOH, yielding defect free and highly oriented ZnO-films (see Fig. 11). [72] Kaufmann, et al. [71] investigated the precursor combinations tBOH, DMZn:THF, and DEZn:THF. They determined the activation energy in the kinetic region to be about 40 kJ/mol. There seems to be a diffusion limited region above 350~ for DMZn:THF where the growth rate reaches 0.45 ~tm/h. In all growth experiments, grain size grows with film thickness and is smaller, but of the same order of magnitude as the thickness. Films with good uniformity have been obtained. Analysis of the surface by REM shows erystallites with tetrapodlike morphology.[ TMSome solid Zn-preeursors have been considered, namely zinc-acetate, Zn(CH3COO)2174]; basic zinc-acetate, Zn4(CH3COO)6175]; and zinc-2-ethylhexanoate, Zn(C2HsCsHllCOO)2 .[76] These compounds, which are very cheap compared to DEZn, sublime at moderate temperatures and are used with a transport gas [74][76] or under vacuum conditions. [75] The oxygen is delivered by the acid radical. At 210~ a deposition rate of 5.4 ~tm/h was observed.[ TM]When this rate is plotted versus 1/T, an activation energy of 21 kJ/mole is obtained.

Wide Bandgap Semiconductors

60

T=[*C] 400 350

300

250

200

150

e-

E "-, 2 t_ .C:

o (.9

I

/"

_._ H,O

- - = - - t-C4H10OH

1

- - , - - CH3OH --v .... C=HsOH

[DEZn]=l.2sccm ,

9

I

1.6

,

I

1.8

,

I

2.0

,

I

,

2.2

I

I

1000/'1"=[1/K] Figure

11. The growth rates of ZnO films for varios oxidizing agents of DEZn.[723

In order to deposit stoichiometric material, it is advantageous to have independent control of the oxygen carrier. In some cases, a high ratio O/Zn 15 is used. Most as grown films are n-conducting (see Ref. 70). Oxygen vacancies probably act as donors. Intentional n-doping has been achieved with chlorine, when POC13 was used,[ 64] or with borane (B2H6) diluted with H2,[67] or even more efficiently with (CH3)3A1. Film resistivities as low as 0.008 W cm have been reported.[ 77] 2.5

CdSe

CdSe has been grown between 300 and 500~ on (111) GaAs in hexagonal, and on (001) GaAs in cubic, structure using the adduct DMCd.(C4HsS)2 and H2Se.[7s][791The VI/II ratio was about 3. The hexagonal layers showed better optical properties. The decomposition of DMCd starts at relatively low temperatures leading to prereactions especially with H2Se. This tendency is slightly inhibited when the thiophene adduct DMCd.(CiHsS)2 is used.[ 8~

Epitaxial Growth of ll-Vl Compounds by MOVPE

61

Parbrook et al. have grown cubic CdSe and Znl.xCdxSe (see also below) on (001) GaAs at 300~ under normal pressure using DMCd and DMSe. Limited diffusion growth is possible between 380 and 500~ with a VI/II ratio of 2.1 and a DMCd flow of 18 ~tmol/min. Although there is no evidence for hexagonal overgrowth, the layers have a high density of dislocations (Lomer and 60~ due to the 7% misfit between layer and substrate. The present authors used DTBSe + DMCd to obtain cubic CdSe layers on (001) GaAs at an optimum temperature near 360~ The VI/II ratio was 2. The layers were much less perfect than those of ZnTe on (001) GaAs which had about the same lattice misfit. This was probably due to the metastability of the cubic phase at growth temperatures. This conclusion was also drawn by Parbrook et al. since MBE samples showed the same width of x-ray rocking curves. No improvement of layer quality was observed when the adduct DMCd.(C4HsS)2 was used instead of DMCd. 2.6

CdS

Halsall et al.[81] found that CdS grows purely hexagonally only on the (111)A face of GaAs. A mixture of hexagonal and cubic phases was found when CdS was deposited on (001), (110), and (111)B faces. The precursor molecules were DMCd and HES, and the optimum growth temperature was 350~ Some growth improvement has been reported with DMCd-(C4H8S)2.[79] CdS on GaAs(111)A shows arrays of misfit dislocations confined near the interface. The relaxed epitaxial layers have mainly threading dislocations, but are of sufficiently high quality to allow a subsequent growth of wurtzite structure CdS/CdSe superlattiees.[821[831 MOVPE growth of purely cubic CdS has not yet been reported. However, mixed crystals of cubic ZnxCdl.xS were successfully grown on (001) GaAs at 400~ using DEZn, DMCd, and TBSH with x as low as 0.4 [84] (see also Sect. 3.6).

3.0

T E R N A R Y AND Q U A T E R N A R Y C O M P O U N D S

Ternary and quaternary solid solutions of II-VI compounds have been prepared and investigated in order to vary gap energies and lattice constants. The MOVPE technique makes the composition control especially easy, since the concentration of the constituents at the growing

62

Wide Bandgap Semiconductors

surface is regulated by the input partial pressure of the precursor compounds. Even a continuous grading of the energy bands can readily be achieved by MOVPE. If a ternary system mixed in the cation or anion sublattiee (A1.xBxC o r ACl.yDy) is t o be grown, it is necessary to allow an adequate overlap of the regions of optimal growth of the binaries AC and BC, or AC and AD. In general, the composition depends nonlinearly on the input partial pressures of the components. A quantitative description is given below.

3.1

Thermodynamics of Binary Compounds Under the conditions of thermodynamic equilibrium the binary reaction:

Eq. (12)

A + C 4-~ AC

is controlled by the reaction constant Kil_viwhieh is the ratio of the reaction rates defined by: Eq. (13)

Kii_w : = k ~ / k ~

The growth is completely determined by the chemical potential difference (A#) between the gas phase in the reactor and the solid interface. If we neglect the effect of a boundary layer, which may be correct in the ease of diffusion controlled growth, then we may write: Eq. (14)

A/t = R T ln(aK)

R is the universal gas constant and a the activity which replaces the concentration in non ideal solutions. Optimal growth conditions are usually found when the VI/II ratio R vI_II , given by the partial pressure in the gas phase, exceeds 2: Eq. (15)

R VI-lI P~176 =

>2

In this case, there is a depletion of cations at the semiconductor surface and the partial pressure ratio at the interface can be given by: Eq. (16)

P vl /P ll = (P~ v~)ZKll_ Vz/ a iz_vl

Note the quadratic dependence of the anionic partial pressure.

Epitaxial Growth of lI-Vl Compounds by MOVPE 3.2

63

Thermodynamics of Ternary Compounds

Given the ratio P v i / P m one would like to know the composition x of the growing layer. This is possible if the solid-vapor distribution function (SVDF) is known. Expressions for the SVDFs are given in Ref. 85. These authors use a model for regular solutions in which the interaction energy (f2) between neighboring atoms and the influence of the lattice distortion is taken into account.j86][87]The chemical activity is then given by: Eq. (17)

a i = x exp{(1-x)2f2/RT}

with the interaction constant if2 and

Tc = ~/R

given in Table 4.

Table 4. Interaction Constant (f2) and Critical Temperature (T~) for Three Ternary Systems

Compound

f2 (J/mol)

L (K)

Zn(SeTe)

14350

863

Zn(SSe)

6670

401

11300

680

(CdZn)Se

The equilibrium constant has to be determined from the well known relation: Eq. (18)

Ki=

exp{- AGi~

where the Gibb's free energy is given by: Eq. (19)

AGi 0 = Ani 0 + TAS i

using ASi extrapolated from 298K to the growth temperature T with:

64

Wide Bandgap Semiconductors T

Eq. (20)

AS = r! C--eTdT

with Cp(T)= a + bT + c7 2 The relevant values are given in Table 5. Table 5. Thermodynamic Data to Calculate the Free Energy of Formation

AS0298K J/mol

a

kJ/molk ZnS

-205

57.7

12.6

1.24

ZnSe

-160

70.3

11.99

1.38

ZnTe

-120

77.8

2.61

CdSe

-145

83.4

11.95

A n ~ 298K

-1.36

1.5

A solid-vapor distribution function (SVDF) can be derived which forms a unique relation between the ratio of the initial partial pressure F = p~176 +pOs) and the composition~. - -. [85]The following two equations for the SVDF are obtained when cationic mixed systems AxBl.xC are considered: Eq. (21)

F= x (1 - fu) +fn(1 + KAca.4c/KsAasA) "l

Eq. (22)

RvI_I](1-fvi) = 1 - ~ i

with the VI/II ratio: Eq. (23)

Rgl_ll= p~176

+ pOB)

The quantitiesfu andfv1are defined as the ratios of vapor pressure at the crystal surface Pk to input partial pressure pk~ of the component k:

Epitaxial Growth of ll-Vl Compounds by MOVPE

65

and ~

and describe the incorporated fraction of constituents. The SVDF for anionic mixed systems ACI_xDx obey similar equations: Eq. (24)

F = x (1 -fzI) +fvz( 1 + [K,4caAc/KBAaB,4]2) "1

Eq. (25)

Rv1-1il( 1 -~I) = 1-frz

Under the conditions of diffusion limited growth and choosing R vi_iz>2, the cationic component determines the growth rate and fn << 1. Then F can be obtained directly from: Eq. (26)

F

= X'Rvl_if

1

+ (1 - RzH[1)( 1 + [K,4caAc/Ks,4as,4] 2 )-1

For choosing Rvz_zz< 1 (surplus of the group VI component), the anionic component determines the growth and fvI << 1. In this ease, Eq. (24) simplifies to: Eq. (27)

F =x

An example is given in Fig. 12, where the SVDF of ZnSSe is plotted versus composition x for various values of R as the parameter. Near R = 1, the curves approach a straight line.

3.3

ZnTeSe

Layers of ZnTexSel. x have been grown at 340~ in an atmospheric pressure MOVPE process using DMZn-TEN, DIPTe, and DTBSe, and a ratio of VI/II = 3.[1~ as] Figure 13 shows a plot of the ratio of the input partial pressures versus the composition x. The full line was calculated from Eq. 26 with the respective parameters taken from Tables 6 and 7 with the exception of.(2 for which a much smaller value had to be chosen. The results of a former growth experiment with di-allylselenide (DASe) are also plotted and are seen to fit the calculated curve as well (i.e., the solid-vapor distribution function (SVDF) is the same for both Selenium compounds and depends only on the ratio VI/II as was expected from Sect. 3.1). The

66

Wide Bandgap Semiconductors

experiments with DASe were stopped when SIMS data revealed a considerable concentration of incorporated carbon. This effect might be reduced at higher growth temperatures. The layers have good optical quality. The PL spectra at x < 0.1 show a sharp excitonic structure. At higher Teconcentration (x > 0.1), the PL spectra broaden and shift to lower energies due to compositional disorder and exciton localization.[ 8s]

0.8 03

.(~4539 kJ/mol Kzns,/Kznse--9 T=340~

u.. 0.6 4-

=1

O)

u.. 0.4 v

1=2.1

O9 v LL

1=3

0.2

0.2

0.4 0.6 Composition x

0.8

Figure 12. Solid vapor distribution function (SVDF) of ZnSel.xS x using DMZn-TEN, DTBSe, and TBSH at Ts = 340~

3.4

ZnSSe

Good quality ZnSxSel. x layers (x < 0.15) grown by MOVPE were already reported in 1988.[89][9~ The atmospheric pressure growth was performed with DMZn, DMSe, and DMSH, a precursor combination which needs high growth temperatures (470-500~ Even successful ndoping was achieved with TEGa.[ 33] The same group was also able to reduce the growth temperature substantially (Ts > 270~ and to change x by PA-MOVPE.[ 91] Other high temperature growth was reported using DEZn, DESe, and DES at 515~ 92] In this reference, the high quality of the ZnSe/ZnSSe interface is demonstrated by measuring Brewster angle reflection spectra.

Epitaxial Growth o f lI- VI Compounds by MO VPE

67

1.0

$

Kz,,To/Kz,,~= 0.5 .0.=4490 kJ/mol

0.8

O

cO em

TDTBSe 0.6

o DASe

co r O CI.

0.4

E o

L)

0.2

0

0.2

0.4 0.6 [Se]/[Se+Te]

0.8

Figure 13. Composition x of ZnTexSel_x versus the ratio of input concentrations given by input partial pressures.

The p r e s e n t authors p e r f o r m e d the g r o w t h on G a A s ( 0 0 1 ) at

Ts = 340~ using DTBSe, TBSH, and DMZnoTEN at arbitrary compositions x. Figure 12[93] shows a plot of the partial pressure ratio versus composition. The full line was calculated with Eq. 26 and the parameters from Tables 6 and 7. The bowing is obviously due to the excess of group VI elements. Heuken et al. observed a linear dependence which is expected when VI/II = 1.[94] As shown in Fig. 14, the FWHM of the x-ray rocking curves has a minimum at about 7% when the lattice constants of ZnSSe and GaAs match.[ 93] These results suggest that the broadening at large x may be the result of two effects: 1) misfit dislocations which are generated when the thickness exceeds a critical value dc; and 2) alloy broadening due to concentration fluctuations. The growth of a thin ZnSe buffer layer seems to be always necessary in order to reduce the density of dislocations. This is especially true near x = 0.07, when the lattice parameters match (see Fig. 15 and Ref. 93). Obviously, sulfur reacts with the GaAs surface to form an interface of Ga2S3. n-Doping can be achieved in the same way as in ZnSe, e.g., with nBC1. Figure 16 (a) and (b) show Hall-data from three samples grown in our

68

Wide B a n d g a p S e m i c o n d u c t o r s

laboratory with different compositions X. [96] In these experiments, the partial pressure of C1 was always kept constant. However, the incorporation increases with composition x. Successful p-doping of ZnSxSel. x by MOVPE has not yet been reported.

1500

.

.

.

.

f

'

"

'

'"'

I

""

'

'

'

I

"

'"'"

"

I

.

.

.

.

X 1250

o 1000

r"

c) c)

[]

I~!

,,,,

"1

-DMZn-TEN -DTBSe -TBSH

a n,' 750 X n,' "r" 500

TG=340~

O

"!"

250

1.1_

0

0.2 0.4 0.6 0.8 C o m p o s i t i o n x: ZnS,Se~.,

Figure 14. FWHM of the x-ray rocking curve versus composition X. [931

3.5

ZnxCdl.xSe

ZnxCdl.xSe quantum wells are the active layers in ZnSe-based bluegreen lasers. Therefore some effort has been put into realization and optimization of the ZnxCdl.xSe growth. The problems which inhibited a quick improvement are: 1) low decomposition temperature of DMCd, which is the only appropriate precursor available; 2) the tendency of ZnxCdl.xSe to form microscopic inhomogeneities; and 3) quick diffusion of Cd even at moderate growth temperatures. The present authors used DMZn:TEN, DMCd, and DTBSe and found an optimum growth temperature of Ts = 340~ at a ratio VI/II = 3 and atmospheric pressure. The insert

Epitaxial Growth of lI-Vl Compounds by MOVPE

69

of Fig. 17 shows the full SVDP. The structural and optical quality of the layers with x < 0.1 is excellent, but at larger Cd-content, the FWHM of the x-ray diffraction increases dramatically, a tendency which is also observed in MBE-grown samples. The large FWHM is not only due to misfit dislocations but to microclustering, which limits the x-ray coherence and seems to be an intrinsic property of this system rather than an effect of the MOVPE-process. Some groups tried to improve the properties of Cdprecursor by using an adduct of DMCd with tetrahydrothiophen. They claimed some improvement which other researchers (including the present authors) denied.

,,.

(a)

O ~

~,~

..

,

"., ,~..

..~ ., .~

,~" . ,',,

~

'~

9

,

,, ,

... ~ "

"

" :.~ , ~

(b)

Figure 15. (a) and (b) show plan-view TEM images of ZnS0.15Se0.75 layers on GaAs(001). The thickness is 200 nm. The lattice misfit with GaAs is +0.3%. In (b) a 20nm ZnSe buffer has been grown between GaAs and ZnS0.1sSe0.75. Note the different scale of (a) and (b). The dislocation density in (a) is an order of magnitude larger than in (b).[95]

Alternative growth conditions have been published (DMZn, DMCd, and DMSe, normal pressure, Ts = 300-550~ 97] Since ZnxCdl.xSe is always used as well material for electron confinement, there was probably little motivation for doping experiments and therefore nothing about doping has been published. The problem of quick Cd diffusion in ZnSe/ ZnxCdl.xSe quantum wells has been investigated by HRTEM and optical methods. [98][99] The prefactor of the diffusion constant depends on the constraints of the experiment. For a free surface, this factor is more than two orders of magnitude larger than for a capped s u r f a c e , [99] whereas, the activation energy of diffusion is unaffected and about 2 eV.

70

Wide Bandgap Semiconductors

10"," " -

,

9 ,

-

,

-

,

-

l

-

,

-

-

9

10" lira. 1 9 ~'~'~8 ~ i l l ~ --i....1" i

9 "-----

&__&

--.~__&_

.

.

,

600

~

8oo

r

.

.

.

.

,

.

.

.

.

,

.

.

.

.

.

,,.

. . . .

,

.

.

.

.

ZnS.Se,. :[CI]/[VI]=O.06

~_

9 x=o.o6

-

9

0"

[c~l~q]=o.oe

' " ' - , - . ....,..

9

.~

=

10"

x=O.06 -'x=0.11 ...." - x=0.48 . l . ,

s

"11.

.... ,... -" 21111

.

A

,o

.

,5

i

,

~

Irr[1ooo~

i

,

i

.

=

2~

loo

1

35

. . . . -

5O

100

-...-L

=

. -

150 ~ m p e r a t u m [Ki

(a)

200

'.~ 250

(b)

Figure 16. (a) Carrier concentration n and (b) Hall mobility # of ZnSxSel.x :CI for various compositions x at a constant doping level.

2000 +

oO

.o

MBE 9 MOVPE

.-

1500

,

,

.!I-

(0 9

q-

LL

1000

~

0.8 -

---=~ A

DMCd DMCd-THT

0.6

A

r~ n, ,

oo 0 . 4 ~

_ ~ . / /

+

T0=340"C

q-

500

I

1

molar input flow ratio [Znj/[Zn+Cd]

o.,

,

I

0,2

,

o:,

I

0,4

,

-

L.

,

0,6

I

,.

I

0,8

c o m p o s i t i o n x ZnxCd,.~Se Figure 17. HRXRD FWHM of ZnxCdl_,,Se alloys grown by MOVPE and MBE. The insert depicts the composition of ZnxCdi.xSe versus ratio of input concentration.tZ~

Epitaxial Growth of l I - V l Compounds by MO VPE 3.6

71

ZnCdS

The growth of cubic Znl.xCdxS was first reported in 1989.[l~176 It covers a large range of gap energies and, at x = 0.57, the lattice matches with that of GaAs. The growth was performed with DEZn, DMCd, and DMSe. The growth temperature (Ts = 420~ had to be chosen high enough in order to obtain an acceptable growth rate and low enough to suppress the growth of a hexagonal phase. At Villi = 7, and a moderate x, the solid vapor distribution function (x versus Cd/Cd + Zn) is a straight line. The lattice matching shows up in a minimum of FWHM (~ 70 arc see) of the x-ray rocking curves.J1~ The same group successfully achieved the growth of the quaternary system Znl.xCdxSl.ySey[l~ with methylmercaptane (MSH) as sulphur precursor. In a subsequent paper, it was shown that photoirradiation (100 mW/cm 2 at ~, < 300 nm) not only increases the growth rate of most binaries, but can also change the composition of ternary compounds when the ratio of the initial partial pressures Cd/Cd + Zn is kept constant.[l~ TM] This method was recently used to grow the barriers of ZnxCdl.xS quantum wells by increasing the Znconcentration under illumination.[ 1~ An attempt to incorporate nitrogen using BNH 2 and photoirradiation was probably successful. However, low resistivity has not been achieved.[ 1~

3.7

ZnMgSeS

The system, Znl.xMgxSel.ySy, is used for cladding layers in bluegreen lasers and was first grown by MBE. However, the MOVPE-technique seems to be suitable to control the volatile elements Se and S very well, and is expected to give good reproducibility even for quaternary systems. Successful growth experiments have already been reported.[1~ 1~ et al. used an atmospheric pressure MOVPE system at a growth temperature of 480~ The precursors were DMZn, DMSe, DES, and bismethyl-cyclopentadienyl-magnesium, with the common short notation (MeCp)2Mg. The layers with compositions x,y = 0.1 have a specular reflecting surface, a small FWHM of the near-band-edge PL, and a FWHM of the x-ray rocking curve of about 170 arc see. Alternative precursors were used by the second group:[ 1~ DMZn:TEN, DTBSe, TBSH, and biscyclopentadienyl-magnesium (CP2Mg) at a growth

72

Wide Bandgap Semiconductors

temperature of T~ = 330~ and a reduced reactor pressure of 400 hPa. Prereaction between TBSH and CP2Mg led to white precipitates of MgS in the reactor.[ 1~ In a later paper,[ll~ the authors kept the low growth temperature of 330~ but replaced TBSH by (t-CaH10)2S (shortnotation DTBS), and CP2Mg, which is a solid, by (MeCP)2Mg, which is a liquid above 40~ The authors claim that this choice resulted in a reduction of inhomogeneities and a suppression of prereactions, which led to less structural defects in the layers. A systematic study of concentration fluctuations and inhomogeneities has not yet been published, however it was shown[ 1~ that those samples considered as "good" from their FWHM of PL had an estimated deviation from the average composition of about 1%. Recently, the successful operation of a MOVPE-grown blue-green laser diode with a ZnMgSeS cladding layer has been reported.[ ill]

4.0

CONCLUDING REMARKS

High quality material of most of the large gap II-VI-compounds have now been grown by MOVPE from commercially available precursors. The precursor purity has been continuously improved over the last years. Ternary and quaternary systems have been grown and their composition is easily controlled. A convenient method to grow p-ZnSe still remains a challenge. A few years ago, a breakthrough was expected from MO-MBE. As a combination of MBE and MOVPE, it should presumably have all the advantages of both epitaxial methods. However, the success of MO-MBE was, and is still, rather marginal. In our opinion, the reason is that its application severely underestimates the role of surface chemical reactions. As we have pointed out in this article, MOVPE needs specially designed molecules. Their architecture and their weak and strong bonds play an important role in the solid/gas-phase reactions in which, in some cases, even the carrier gas is involved. Furthermore, since MOVPE works at a finite pressure, the molecules have the chance to undergo many collisions; volatile products, especially hydrocarbons, are quickly released from the solid surface into the gas phase. However, in MO-MBE, the precursor molecule is, in most cases, at first destroyed in the cracker cell. Then the fragments, which contain C, CH, C2, and HE, reach the solid surface. Those parts which do not quickly leave by desorption have a good

Epitaxial Growth o f l l - V l Compounds by M O VPE

73

chance to be incorporated, and lead to unacceptable large carbon and hydrogen incorporation. Therefore, in most applications of MO-MBE, those advantages of MOVPE connected with solid/gas reactions and with the architecture of the molecules are lost, and only the easy flow control of volatile elements remains. Finally, we would like to emphasize the importance of new optical methods for in-situ growth control. A number of laboratories, including our own, already make use of ellipsometry or reflectance difference spectroscopy to monitor growth processes or to investigate layer by layer growth.[251[112]-[1161With these methods, very recently applied to surface analysis, most of the in-situ uncontrollability of MOVPE growth is eliminated. There is still pioneering work to be done in order to improve the somewhat difficult interpretation of the results. Another equally important development is the application of elaborate computer simulations of molecular dynamics in order to model chemical reaction pathways; an approach which is especially important for surface reactions which are not directly observable. This effort will greatly extend our knowledge of epitaxial processes on an atomic scale.

REFERENCES 1. CVD Metalorganics for Vapor Phase Epitaxy, Morton International, Advanced Materials, Danvers, MA (1992) 2. Sarkvan, W. G. J. H. M., Janssen, G., Croon de, M. H. J. M., and Gilling, L. J., Semicond. Sci. Technol., 5:16-35 (1990) 3. Kuhn, W. S., Qu'Hen, B., Gorochov, O., Triboulet, R., and Gebhardt, Prog. Crystal Growth and Charact., 31:45-117 (1995) 4. Jones, A. C., Wright, P. J., and Cockayne, B., dr. Crystal Growth, 107:297-308 (1991) 5. Kuhn, W. S., Qu'Hen, B., and Gorochov, O., Prog. Crystal Growth and Charact., 31:1-44 (1995) 6. Nhishio, M., Ogawa, H.,Jpn. J. Appl. Phys., 29:145 (1990) 7. Leiderer, H., Jahn, G., Silberbauer, M., Kuhn, W., Wagner, H. P., Limmer, W., and Gebhardt, W., J. Appl. Phys., 70:398 (1991) 8. Lovergine, N., Cingolani, R., Lee, G., Mancini, A. M., Vasanelli, L., Romanato, F., Drigo, A. V. and Matter, M., Appl. Phys. Lett., 63:3452 (1993)

74

Wide Bandgap Semiconductors

9. Kuhn, W. S., Lusson, A., Qu'Hen, B., Grattepain, C., Dumont, H., Gorochov, O., Bauer, S., Wolf, K., Wijrz, M., Reisinger, T., Rosenauer, A., Wagner, H. P., Stanzl, H., and Gebhardt, W., Prog. Crystal Growth and Charact., 31:119-177(1995) 10. Stanzl,H., Rosenauer, A., Wolf, K., Kastner, M., Hahn, B., and Gebhardt, W., Epitaxial Growth Processes, SPIE 2140:138 (1994) 11. Wolf, K., Stanzl, H., Naumov, A., Wagner, H. P., Kuhn, W., Hahn, B., Gebhardt, W., J. Crystal Growth, 138:412-417 (1994) 12. Kuhn, W., Naumov, A., Stanzl, H., Bauer, S., Wolf, K., Wagner, H. P., Gebhardt, W., Pohl, U., Krost, A., Richter, W., Diimichen, U., Thiele, K. H., d. Crystal Growth, 123:605-610 (1992) 13. Taudt, W., Wachtendorf, B., Beccard, R., Wahid, A., Heuken, M., Gurskii, A. L., Vakarelska, K., J. Crystal Growth, 145:582-588 (1994) 14. Fujita, S., Takeuchi, F. Y., and Fujita, S., Jpn. d. Appl. Phys., 27(11):L2019-L2021(1988) 15. Fujita, S., Tanabe, A., Sakamoto, T., Isemura, M., and Fujita, S., Jpn. J. Appl. Phys., 26(12):L2000-L2002 (1987) 16. Hirata, S., Fujita, S., Fujita, S., and Isemura M., J. Crystal Growth, 104:521-526 (1990) 17. Hahn, B., Preis, H., Schindler, M., Reisinger, T., and Gebhardt, W., J. Crystal Growth,179:415-422 (1997) 18. Taudt, W., Hardt, A., Lampe, S., Hamadeh, H, and Keuken, M., J. Crystal Growth, 170:491-496 (1997) 19. Nishimura, K., Nagao, Y., and Sakai, K., Jpn. J. Appl. Phys., 32" 1,428-1, 430(1993) 20. Fujita, S., and Fujita, S., J. Crystal Growth, 117:67-74 (1992) 21. Hahn, B., Deufel, M., Meier, M., Kastner, M. J., Blumberg, R., and Gebhardt, W., J. Crystal Growth, 170:472-475(1997) 22. Yoshikawa, A., and Okamoto, T., J. Crystal Growth, 117:107-110(1992) 23. Fujita, S., Hirata, S., and Fujita, S., J. Crystal Growth, 115:269-273 (1991) 24. Yoshikawa, A., Kobayashi, M., and Tokita, S., Phys. Stat. Sol., 187(b):315 (1995) 25. Kastner, M. J., Hahn, B., Blumberg, R., Sossna, E., Duschl, R., Gebhardt, W., J. Crystal Growth, 170:188-192 (1997) 26. Nakano, T., and Hirano, T.,J. Appl. Phys., 78(1):251 (1995) 27. Kamata, A., Uemoto, T., Okajima, M., Hirahara, K., Kawachi, M., and Beppu, T., J. Crystal Growth, 86:285-289 (1988) 28. Pereira, E., Vilao, Seitz R., Time-resolved luminescence of deep levels in ZnSe samples,,/. Crystal Growth, 189/190:546-550 (1998)

Epitaxial Growth o f l l - V l Compounds by M O V P E

75

29. Fujimoto, M., Suemune, I., Osaka, H., and Fujii, Y., Jpn. J. Appl. Phys., 32:L524--L527 (1993) 30. Yoshikawa, A., Nomura, H., Yamaga, S., and Kasai, H., Jpn. J. Appl. Phys., 27:L1948-51 (1988) 31. Stanzl, H., Wolf, K., Hahn, B., and Gebhardt, W., J. Crystal Growth,

145:918-923(1994) 32. Kukimoto, H.,J. Crystal Growth, 101:953-957 (1990) 33. Fujita, S., Terada, K., Sakamoto, T., and Fujita, S., J. Crystal Growth, 34. 35. 36. 37. 38. 39. 40. 41. 42. 43.

44. 45. 46. 47. 48. 49. 50.

94:102-108 (1989) Fujita, S., Asano, T., Maehara, K., and Fujita, S., Jpn. J. Appl. Phys., 32:L1153-L1156 (1993) Walukiewicz, W.,J. Crystal Growth, 159:244-247 (1996) Faschinger, W., J. Crystal Growth, 159:221-228 (1996) Taskar, N. R., Khan, B. A., Dorman, D. R., and Shahzad, K., Appl. Phys. Lett., 62(3):270 (1993) Ogata, K., Kawaguchi, D., Kera, T., Fujita, S., and Fujita, S., J. Crystal Growth, 159:312 (1996) Pohl, U. W., Freitag, S., Gottfriedsen, J., Richter, W., and Schumann, H., J. Crystal Growth, 170:144-148 (1997) Akram, S., and Bhat, I.,J. Crystal Growth, 138:105-109 (1994) Yoshikawa, A., Matsumoto, S., Yamaga, S., and Kasai, H., J. Crystal Growth, 101:305-3 10 (1990) Fujita, S., Tojyo, T., Yoshizawa, T., and Fujita, S.,J. Electr. Mat., 24(3):137 (1995) Rees, W. S., Jr., Green, D. M., Anderson, T. J., and Bretschneider, E., Wide Band Gap Semiconductors Symposium, (T. D. Moustakas, J. L. Pankove, Y. Hamakawa, eds.), pp. 281-286, Pittsburgh, PA, Mater. Res. Sec, (1992) Inoue, K., Yanashima, K., Takahashi, T. Hwang, J. S., Hara, K., Munekata, H. and Kukimoto, H.,J. Crystal Growth, 159:130-133 (1996) Hayashi, K., Appl. Phys. Lett., 65(16):2084 (1994) Kamata, A.,J. Crystal Growth, 145:557-561 (1994) Honda, T, Lim, S. W., Inoue, K., Hara, K., Munekata, H., Kukimoto, H., Koyama, F., and Iga, K., d. Crystal Growth, 170:503-506 (1997) Cockayne, B., and Wright, P. J.,J. Crystal Growth, 68:223-230 (1984) Fujita, S., Matsuda, Y., and Sasaki, A., d. Crystal Growth, 68:231-236 (1984) Yoshikawa, A., Yamaga, S., Tanaka, K., and Kasai, H., d. Crystal Growth, 72:13-16(1985)

76

Wide Bandgap Semiconductors

51. Briot, O., Briot, N., Abounadi, A., Gil, B., Cloitre, T., and Aulombard, R. L., Semicond. Sci. Technol., 9:207-209 (1994) 52. Abounadi, A., Di Blasio, M., Bouchara, D., Calas, J., Averous, M., Briot, O., Briot, N., Cloitre, T., Aulombard, R. L., and Gil, B., Phys. Rev. B, 50(16):11677 (1994) 53. Fujita., S., and Fujita, S., J. Crystal Growth, 117:67-74 (1992) 54. Armitage, D. N., Yates, H. M., Williams, J. O., Cole-Hamilton, D. J., and Patterson, L. P., J. Crystal Growth, 117:1082 (1992) 55. Lovergine, N., Longo, M., Gerardi, C., Manno, D., Mancini, A. M., and Vasanelli, L., J. Crystal Growth, 156:45-51 (1995) 56. Fem~indez,M., Prete, P., Lovergine, N., Mancini, A. M., and Cingolani, R., Phys. Rev. B, 55:7660-7666 (1997) 57. Obinata, T., Uesugi, K., Sate, G., and Suemune, I., Jpn. J." Appl. Phys., 34:4143-7 (1995) 58. Fujita,Y., Jpn. J. Appl. Phys. Lett., 35(7B):L919-22 (1996) 59. Yamaga, S., Yoshikawa, A., and Kasai, H., J. Crystal Growth, 86:252-256 (1988) 60. Yamaga, S., Yoshikawa, A., and Kasi, H., J. Crystal Growth, 106:683-689 (1990) 61. K. Ehrmann, Diploma Thesis, University Regensburg (1996) 62. Zhao, L., Zhong, G., Yang, B., and Fan, X., Inorganic and Organic Electroluminescence, (R. H. Mauch and H. E. Gumlich, eds.), p. 165, W&T, Berlin (1996) 63. Detchprohm, T., Hiramatsu, K., Sawaki, N., and Akasaki, I., J. Crystal Growth, 145:192-196 (1994) 64. Takato, H., Yamanaka, M., Hayashi, Y., Shimokawa, R., Hide, I., Gohda, S., Nagamine, F., and Tsuboi, H., Jpn. J. Appl. Phys., 31 :L1665-L 1667 (1992) 65. Olsen, L. C., Addis, F. W., Greer, D., Lei, W., and Abulfotah, F., Conference Record of the Twenty Fourth IEEE Photovoltaic Specialists Conference 1994, New York, NY, IEEE, 1:194-7 (1994) 66. Smith, F. T. J., Appl. Phys. Lett., 43(12):1108 (1983) 67. Yamada, A., Wenas, W. W., Yoshino, M., Konagai, M., and Takahashi, K., Jpn. J. Appl. Phys., 30:L1152-L1154 (1991) 68. Kaufmann, T., Fuchs, G., and Webert, M., Cryst. Res. Technol., 23:635-639 (1988) 69. Shimizu, M., Katayama, T., Shiosaki, T., and Kawabata, A., J. Crystal Growth, 99:399-402 (1990) 70. Shimizu, M., Katayama, T., Tanaka, Y., Shiosaki, T., and Kawabata, A., J. Crystal Growth, 101:171-175 (1990)

Epitaxial Growth o f l l - V l Compounds by M O V P E

77

71. Kaufmarm, T., Fuchs, Fuchs, G., Webert, M., Frieske, S., and G~ickle, M., Cryst. Res. Technol., 24:269-274 (1989) 72. Oda, S., Tokunaga, H., Kitajima, N., Hanna, J., Shimizu, I., and Kokado, H., Jpn. J. Appl. Phys., 24:1607-1610 (1985) 73. Yoshino, M., Wenas, W. W., Yamada, A., Konagai, M., and Takahashi, K., Jpn. J. Appl. Phys., 32:726-730 (1993) 74. Maruyama, T., and Shionoya, J.,J. Mat. Sc. Lett., 11:170-172 (1992) 75. Gyani, A. K., Khan, O. F. Z., O'Brien, P., and Urch, D. S., Thin Solid Films,

182:L1-L3(1989) 76. Kumar, N. D., Kamalasanan, M. N., and Chandra, S., ,4ppl. Phys. Lett., 65( 11): 1373(1994) 77. Sang, B., and Konagai, M., Jpn. J." ,4ppl. Phys., 35 :L602-L605 (1996) 78. Wright, P. J., and Cockayne, B., J. Crystal Growth, 94:97-101 (1989) 79. Jones, A. C., Rushworth, S. A., Wright, P. J., Cockayne, B., O'Brien, P., and Walsh, J. R., J. Crystal Growth, 97:537-541 (1989) 80. Jones, A. C., Semiconductor Science and Technology, 6:A36--40 (1991) 81. Halsall, M. P., Davies, J. J., Nicholls, J. E., Cockayne, B., Wright, P. J., and Russell, G. J., J. Crystal Growth, 91:135-140 (1988) 82. CuUis,A. G., Williams, G. M., Cockayne, B., Wright, P. J., and Smith, P. W., Microscopy of Semiconducting Materials, (A. G. Cullis, J. L. Hutchinson, eds.) pp. 217-22, Proceedings of the Royal Microscopical Society Conference, Bristol, UK, IOP (1989) 83. Bradley, L. V., Creasey, J. P., O'Donnell, K. P., Wright, P. J., and Cockayne, B., J. Crystal Growth, 159:551-554 (1996) 84. Dumont, H., Fujita, S., and Fujita, S., Appl. Surface Sci., 86:442-446 (1995) 85. Kisker, D. W., and Zawadzki, A. G., J. Crystal Growth,, 89:378-390 (1988) 86. Stringfellow, G. B., and Greene P. E., J. Phys. Chem. Solids, 30:1779-1791 (1969) 87. Stringfellow, G. B.,J. Crystal Growth,, 27:21-34 (1974) 88. Naumov, A., Stanzl, H., Wolf, K., Rosenauer, A., Lankes, S., and Gebhardt, W.,J. Crystal Growth, 138:595-600 (1994) 89. Tonami, Y., Nishino, T., Hamakawa, Y., Sakamoto, T., and Fujita, S., Jpn. J. Appl. Phys., 27(4):L506-L508 (1988) 90. Fujita, S., Isemura, M., Sakamoto, T., and Yoshimura, N., J. Crystal Growth, 86:263-267 (1988) 91. Fujita, S., Tanabe, A., Sakamoto, T., Isemura, M., and Fujita, S., J. Crystal Growth, 93:259-264 (1988)

78

Wide Bandgap Semiconductors

92. Inoue, K., Kuroda, T., Yoshida, K., and Suemune, I., Appl. Phys. Lett., 65(22):2830 (1994) 93. Gebhardt, W., Stanzl, H., Reisinger, T., Schiitz, G., and Kasmer, M., Materials Sci. Forum, 182-184:105-110 (1995) 94. SiSllner, J., Schmoranzer, J., Woitok, J., Hamadeh, Y., and Heuken, M., Materials Sci. Forum, 182-184:419-422 (1995) 95. Kastner, M. J., Leo, G., Brunhuber, D., Rosenauer, A., Preis, H., Hahn, B., Deufel, M., and Gebhardt, W., J. Crystal Growth, 172:64-74 (1997) 96. Gebhardt, W., Hahn, B., Stanzl, H., and Deufel, M., J. Crystal Growth,

159:238-243(1996) 97. Parbrook, P. J., Kamata, A., and Uemoto, T., J. Crystal Growth, 128:639-645 (1993) 98. Rosenauer, A., Reisinger, T., Steinkirchner, E., Zweck, J., Gebhardt, W., J. Crystal Growth, 152:42-50 (1995) 99. Chai, M. K., Wee, S. F., Homewood, K. P., Gillin, W. P., Cloitre, T., and Aulombard, R. L., Appl. Phys. Lett., 69(11): 1579 (1996) 100. Fujita, S., Funato, M., Hayashi, S., and Fujita, S., Jpn. J. Appl. Phys., 28(6):L898-L900 (1989) 101. Fujita, S., Hayashi, S., Funato, M., and Fujita, S., J. Crystal Growth, 99:437-440(1990) 102. Fujita, S., Hayashi, S., Funato, M., Yoshie, T., and Fujita, S., J. Crystal Growth, 107:674-678 (1991) 103. Fujita, S., Matsumoto, S., and Fujita, S., J. Electronic Mat., 22(5):521 (1993) 104. Dumont, H., Fujita, S., and Fujita, S., Mat. Res. Sec. Syrup. Proc., 340:509 (1994) 105. Dumont, H., Kawakami, Y., Fujita, S., and Fujita, S., Jpn. J. Appl. Phys., 34:L1336-L1339 (1995) 106. Dumont, H., Fujita, S., andFujita, S.,J. Crystal Growth, 145:570--575(1994) 107. Toda, A., Asano, T., Funato, K., Nakamura, F., and Mori, Y., J. Crystal Growth, 145:537-540 (1994) 108. Hamadeh, H., S611ner, J., Schmoranzer, J., and Heuken, M., J. Crystal Growth, 158:89-96 (1996) 109. Hamadeh, H., S/511ner, J., Hermans, J., Kiister, U., Woitok, J., Geurts, J., Bollig, B., and Heuken, M., J. Crystal Growth, 159:21-25 (1996) 110. Heuken, M., S611ner, J., Taudt, W., Lampe, S., J. Crystal Growth, 170:30-38 (1997) 111. Toda, A., Nakamura, F., Yanashima, K., and Ishibashi, A., J. Crystal Growth, 170:461-466 (1997)

Epitaxial Growth of lI-Vl Compounds by MO VPE

79

112. Razeghi, M., The MOCVD Challenge, Vol. 2, Institute of Physics Publishing, Bristol (1995) 113. Yoshikawa, A., Kobayashi, M., and Tokita, S., Phys. Stat. Sol., 187:315 (1995) 114. Stafford, A., Irvine, S., and Griffiths, C., J. Crystal Growth, 170:182-187 (1997) 115. Patrikarakos,D., Shukla, N., and Pemble, M.,J. Crystal Growth, 170:215-218 (1997) 116. Kobayashi, N., Kobayashi, Y., Uwai, K., J. Crystal Growth, 170:225-229 (1997)

3 O h m i c C o n t a c t s to I I - V I and III-V Compound Semiconductors Tae-Jie Kim Paul H. HoUoway

1.0

INTRODUCTION

New generations of high speed transistors, such as High Electron Mobility (HEMTs) and Heterojunction Bipolar Transistors (HBTs), have been developed from the superior transport properties of III-V semiconductors. Epitaxial growth techniques, such as MBE and CVD, on an atomic scale, enabled such development of heterojunction devices with different compositions of ternary or quaternary compounds (e.g., A1GaAs/GaAs and GalnP/InP) Therefore, bandgaps offer higher frequency and power characteristics unattainable in Si-based devices.[1]-[3] In addition to the superior transport properties, III-V and II-VI semiconductors have direct bandgaps, enabling quantum well heterostructures for optoelectronic applications such as light-emitting diodes (LEDs) and diode lasers (DLs).[ l] Monolithic integration of a photo diode and a Field Effect Transistor (FET) is possible because of the direct band gap of these compound semiconductors.[ 4] Voltage and current signals must be brought into, and taken out of, the semiconductors through electrical contacts, regardless of the applications. Ohmic contacts play a critical role by allowing current signals to 80

Ohmic Contacts to II- VI and III- V Compounds

81

exhibit a linear relation with respect to applied voltage, or to have a low enough contact resistance not to hinder performance of the devices, especially at high frequency and high power where parasitic contact resistance will limit both the frequency and power handling capabilities.[5][ 6] High specific contact resistance (>10 -3 f2-cm2) will limit the maximum frequency and cause excessive Joule heating and failure of the devices.J7][8] Ideally, an ohmic contact could be formed by selecting metal work functions depending on the type of semiconductors. Practically, formation of ohmic contacts is a complicated and challenging issue due to mismatch of levels at the interface, or due to a phenomenon called Fermi level pinning which involves surface states. A critical parameter in the formation of metal/semiconductors (M/S) contacts is the condition of the surface and particularly whether or not the Fermi energy level, E F, is pinned. Comprehensive surface science studies oflII-V surface and interfaces, for several decades, have shown that Fermi level pinning affects the formation of both Schottky and ohmic contacts, as will be reviewed below. It has been generally demonstrated that pinning of E F is consistent with the characteristic of the bonding in the semiconductor; with covalently bonded semiconductors most often exhibiting pinned surfaces. On the other hand, ionically bonded semiconductors normally exhibit unpinned surfaces. The type of bonding also correlates well with the electronegativity difference between the elements in a compound semiconductor.[ 9] For the semiconductors to be discussed in this chapter, the character of the bonding varies considerably. The III-V semiconductors tend to exhibit more covalent bonding with lower electronegativity differences (Ax) than do the II-VI semiconductors. Therefore, we would expect the former to exhibit pinned surfaces more than the latter. For example, GaAs bonding is only 32% ionic, A~ between Ga and As is only ~0.4, and E F is pinned in nearly all conditions. For InP, bonding is 44% ionic with A;~ ~0.6 and E F is partially pinned. In the AlxlnyGal_x_yN system, GaN bonding is 70% ionic, A)r is ~1.4 and E F does not exhibit pinning. In the II-VI system, ZnSe bonding is 63% ionic, A~ is ~0.8 and this semiconductor also does not exhibit E F pinning. The consequence of these systematic trends will become evident when we discuss the formation of ohmic contacts to these different semiconductors. Covalent bonding and low A)r for the III-V compound semiconductors leads to Fermi level pinning. Therefore, practical ohmic contacts to GaAs and InP devices are predominantly produced by controlling the barrier width rather than the barrier height. The barrier width is controlled

82

Wide Bandgap Semiconductors

by incorporating dopants into the interface region of semiconductors. Annealing of ex situ (i.e., post-growth) metallizations is the practical method to cause the incorporation of dopant through metallurgical reactions. When the interface regions of semiconductors are heavily doped by the metallurgical reactions, current transport across the metal/semiconductor contact is dominated by Field Emission (FE) or Thermionic Field Emission (TFE). In this respect, our primary interest is to understand the details of metallizations used for ohmic contacts. Generally, ohmic contacts to compound semiconductors are formed by a lift-off process using multiple (two or more) element metallization, such as AuGeNi (developed in 1967 for n-type GaAs).[ l~ This same metallization is often used for n-InP. While it has long been used and often studied, AuGeNi metallization exhibits problems which include irregular morphology,[11][ 12]poor thermal stability, [13][14] and interdiffusion.[ 15] Tremendous efforts have been made not only to solve the practical problems of the AuGeNi system, but also to understand the mechanism of AuGeNi contacts.[16]-[ TM]Non-alloyed and Au-free contact schemes have been demonstrated for n-GaAs and InP. [19]'[21] Not withstanding all of the studies, the technology of ohmic contacts is much more advanced than the basis for the understanding which is largely empirical. A more fundamental understanding is essential because electrical contacts are critical to both proper performance of a device as well as its lifetime. A great deal of effort has been focused recently on better understanding of both Schottky and ohmic contacts. Brillson has edited a general review of these subjects.[ 22] In the present review, the focus will be only on ohmic contacts to GaAs, InP, GaN, and ZnSe rather than their ternary or quaternary variants. The purpose of this review is to clarify the current understanding of the effects of interracial reactions occurring during formation of ohmic contacts, rather than to compile various recipes for their formation. As will be shown, new information on the metallurgical reactions taking place at the M/S interface will allow us to better understand when to expect ohmic behavior.

I.I

Energy Barrier and Current Transport Mechanisms

An enormous difference in the electrical properties between metal and semiconductor usually results in the formation of a rectifying barrier at the metal/semiconductor interface. In the case of GaAs and InP, rectifying barriers are formed during processing creating an ohmic contact because of Fermi level pinning.[ 231 Therefore, the interface between the metal and

Ohmic Contacts to II-VI and III-V Compounds

83

semiconductor always has a barrier to transport of charge. To make a good ohmic contact, this barrier must be negated. Ideally, either rectifying (Sehottky) or ohmic contacts may form as a result of charge transfer between metal and semiconductor to align the Fermi level across the interface when a metal is brought into electrical contact with a semiconductor. Figure 1 illustrates the idealized formation of a contact. Figure 1a shows the energy band diagram of a metal and a n-semiconductor before contact (unless otherwise specified, all discussion refers to n-type semiconductors), where ~m and ~ are the work functions for the metal and the semiconductor, EF~ and EF~ are the equilibrium Fermi energy levels of electrons in the metal and the n-type semiconductor, and ~ is electron affinity of the semiconductor. The values E c and E v are the energies of the bottom of the conduction and top of the valence bands in the semiconductor. No surface states exist within bandgap, thus E~ and E v are flat up to the surface before contact. Due to the potential difference (~m--f~Is) when the metal and semiconductor are joined, electrons in the conduction band of the semiconductor flow into the metal to occupy lower energy states. This results in a negative charge on the metal and a positive charge on the semiconductor. Since there are a large number of nearly-flee electrons in the metal, the charge distributes uniformly. However, the limited charge in the semiconductor causes the development of an electric field which bends the bands near the surface out to a distance over which majority carriers are depleted (by donation to the metal). The band alignment and bending shown in Fig. l b is observed. The length of the charge depleted region in the semiconductor (W) is called the depletion length (Fig. lb). In GaAs, W ranges from 20 nm to 1000 nm depending on the doping concentration at room temperature.[Z41 The resulting band bending due to the space charge region in the semiconductor is described by the spatial variation of electrostatic potential, V(x): Eq. (1)

V(x) = (1/q).[Evb-Ev(x) ] = (1/q).[Ec6-Ec(x)]

where the subscript b refers to bulk (i.e., x = 89 and subscripts V and C refer to the valence and conduction bands. With the assumptions of semiinfinite boundary conditions and no interface states, the potential as a function of x can be described by Poisson's equation: Eq. (2)

d2 V(x)/dx 2 = -p(x)/~es

where es is the dielectric constant of the semiconductor and p ( x ) is the spatial distribution of charge. Provided that there are no free carriers in the

84

Wide B a n d g a p S e m i c o n d u c t o r s

depletion region of n-type semiconductor and complete ionization, i.e., p(x) = N o, (No = donor concentration) solving Poisson's equation yields:

Eq. (3)

V(x) = VS + Em(x - x2/2 W)

where E m is the electric field at x = 0 given by (qNoW)/6 s, W is the depletion length given by (2SSVbi/qND) 1/2, and Vbi is the built-in potential defined by Ecb-Ecs(= Om-Os)" Here, Vs is the surface band bending at x = 0. Equation (3) shows that the band edge is a parabolic function of x from x = 0 to x = IT"(Fig. l b). The effect of image charge on the band bending was not considered.

Vacuum level

-

IZI i(i) ~,Vacuum s level

_ _

~ - _ ~_ _ Efs

Ec

Eft-Ev n-Semiconductor

Metal

(a) Before contact

I Efm

Os

b l _~~_ ' . .

_E_c- Ets

,.F x -0

~x

(b) At equilibrium Figure 1. Formationof energybarrierand band bendingin electricaland thermal equilibrium.

Ohmic Contacts to II-VI and III-V Compounds

85

The energy barrier, ~0b, is defined as the energy distance between the Fermi level and the edge of the majority-carrier band (Ec). The rPb of an ideal M/S junction is shown (Fig. 1 b) to be equal to" Eq. (4)

9

=

+

= a

,.-Zs

where Zs is the electron affinity of semiconductor as defined in Fig. 1a. Eq. (4) is the Schottky-Mott rule or the Schottky Barrier Height (SBH). The first term, (q),n-q)~), represents the built-in potential and the second term (q)~-Z), is the potential difference between the conduction band edge and the Fermi level in the semiconductor. There are three transport mechanisms for electrons (major carriers) to cross such an energy barrier.[251[261Thermionic Emission (TE) transport occurs by thermal activation of electrons over the energy barrier from the conduction band of the semiconductor into the metal (Fig. 2a). TE is normally dominant when the semiconductor is low doped up to 1016cm-3 at room temperature. The second current transport mechanism is Field Emission (FE), which is a quantum mechanical effect whereby electrons penetrate through the M/S energy barrier at the Fermi level under the influence of an electric field (Fig. 2b). For FE to be predominant, high doping above 1019 cm 3 at room temperature is necessary. Thermionic Field Emission (TFE) is a combination of the TE and FE mechanisms (Fig. 2c) in which the carriers are thermally activated above Fermi level to a region where they can tunnel efficiently through a narrower portion of the barrier. By assuming that only electrons traversing from the semiconductor to the metal with sufficient energies to overcome the energy barrier can contribute to transport, the current density from the semiconductor to the metal is given by: Eq. (5)

Js-~m =.,4*"T2"exp(-q cl)b/k T).exp( q Va/kT) = JrH.exp(-q cI)a/kT)'exp( q Va/kT)

where Va is the applied bias, Tis the absolute temperature, and gYbis the M/ S barrier height. A* is known as Richardson's constant and given by (4rutm*k2/h 3) where m* is the effective mass of electrons in the semiconductor, k is Boltzmann's constant, and h is Planck's constant. For free electrons, A ~ is 120 m/cm2/K 2. Jrtt (known as the thermionic current coefficient) and defined by: Eq. (6)

J'm = A*.T2

- 120(m*/m)T2[A/em 2] - 10.8(m*/m)(T/300) 2 [MA/cm 2]

86

Wide Bandgap Semiconductors

where m is the mass of a free electron. For free electrons at room temperature (T = 300K), Ja'rt is 10.8 Million Amperes (MA)/cm 2. By considering din__,,s, the net current across the M/S contact is given by: Eq. (7)

gaet = gm~s + Jm.s = JTa.exp(-q ~b/kT)'exp(q Va/kT)-JTH'exp(-q~b/kT) = JTH'exp(-q ~b/kT).[exp(q Va/kT ) - 1] - J

.[exp(q Z a / k r ) - 1]

where .Is is saturation eurrent and defined by: Eq. (8)

Js = A*'T2"exp(-q~b/kT) = 120(m*/m)'T2exp(-q~b/kT) [A/cm2]

Metal

n-Semiconductor

r ~x'~ e"

Ef

E-c- Ef

(a) Thermionic Emission

Ef . . . . .

EcEf

(b) Field Emission

Ef

C

E-c- Ef

(c) Thermionic-Field Emission Figure 2. Current transport mechanisms across metal/semiconductor interfaces.

Ohmic Contacts to II-VI and III-V Compounds

87

For a large forward bias V~ >> kT/q, Eq. (7) is reduced to "/net = Js.exp(q V~/kT). Equation 7 determines the current-voltage (I-V) characteristics of M/S junctions in low doped semiconductors, which is the most frequently used method to determine the barrier height. Saturation current (J~) is obtained by extrapolating./net at zero Va from a plot ofln(J,,et ) versus forward bias Va. Then the barrier height ~ob can be determined from Eq. (8). At a large reverse bias voltage, V~ << -kT/q, exp(-qVa/kT) is negligibly small compared to 1. Then J,,et ~ -J~, which indicates a constant saturation current independent of applied voltage. Tunneling of electrons was described to explain current transport under an electric field which was independent of temperature, first in metal[271 and in semiconductors.J28] Tunneling as a transport mechanism across the M/S junction was first suggested to describe the formation of ohmic contact in metal/CdS contacts.[ 29] Basically, tunneling currents are determined by the number of electrons impinging upon the junction and the probability of electron transmission through the potential barrier at the M/S junction: Eq. (9)

Jtunneling = q'n'v.P( ~b)

where q is the charge of an electron, n is the number of electrons impinging on the M/S junction, v is the thermal velocity of electrons, and P(qSb) is the tunneling probability. P(qSb) is obtained by solving Sehrodinger equations with appropriate boundary conditions typically in parabolic potential distribution such as Eq. (3). Since P(~b) is a very complicated function of the detailed band structure and doping concentration of the semiconductor,[3~ 31] approximations were used to calculate the tunneling probability.[8][3111321 Tunneling currents through a barrier of height q)b without applied bias is generally expressed as:[23][33] Eq. (10)

J oc exp (q~b/E00)

where Eoo = (h/4n').(ND/m *es)1/2. Typically, field strengths of 107-108 V/cm are needed to cause significant tunneling current.[ 3~ In semiconductors, such high field strengths are established in the heavily doped surface region near the M/S contacts. Once tunneling currents begin to dominate transport across the M/S junction, tunneling can provide higher current densities than thermionic emission. This is why n + or p+ doped surfaces frequently result in good ohmic contacts.

88

Wide Bandgap Semiconductors

Frequently, thermionic, field, and thermionic-field emissions are described in terms of specific contact resistance by differentiating Eqs. (7) and (10) with respects to Va. The specific contact resistance is defined by: Eq. (11)

pc = (dJ/dVa) l at Va = 0

where Va is the voltage applied across the contact. For thermionic emission over the Schottky barrier, the solution for Pc is given by: Eq. (12)

Pc = (k/qA*T).exp( ~b/kT)

From Eq. (12), the contact resistance for thermionic emission is independent of the doping concentration; the barrier height determines contact resistance. In the case of tunneling, the contact resistance is approximated by: Eq. (13)

pc oc exp( ~t,/Eoo )

where E00 = (h/4zc).(No/m *es) 1/2. The contact resistance is lower when N o (doping concentration in semiconductor) is large and thus the depletion distance (W) is reduced. For TFE, transport results from both tunneling and thermionic emission and contact resistance is given by: Eq. (14)

pc oc exp(~b/Eoo)/eoth(Eoo/kT )

where the temperature dependence is contained in the coth(E00/kT) term. Because coth(z) approaches unity for large z, Eq. (14) has been used to predict that TFE junctions behave: Eq. (15a)

Eoo/kT>>l

~

eoth(E00/kT) ~ 1

Ohmic by tunneling

Eq. (15b)

Eoo/kT = 1

~

coth(E00/kT) _> 1

Non-linear

Eq. (15c)

Eoo/kT<
~

coth(E0o/kT) >>1

Non-ohmic

For z = 2.7, the error in setting coth(z) = 1 is less than 1%. Therefore, it is possible to predict the doping concentration required to form a tunneling contact by setting Eoo/kT > 2.7. Eq. (16)

E00= 18.5 x lO'12.(ND/mroer)l/2> 2.7"kTeV

Ohmic Contacts to II-VI and III-V Compounds

89

where m r = m * / m , e~ = 6s/so and N D is the carrier concentration (cm-3). For n-type GaAs at 300K, m r --0.067, 6r -13, and Eq. (16) predicts ohmic behavior for N D > 1.2 x 1019 em-3.[s] The figure of merit for the formation of an ohmic contact is the specific contact resistance defined by Eq. (11). It is the contact resistance of an unit area for current flow perpendicular to the contact. Thus it has units of s z. Cox and Straek developed a bulk technique to measure contact resistance using different size circular dots defined on an epilayer.[ 341 This method is simple, but is only applicable to bulk structures (e.g., an epilayer on a conducting substrate), and its practical limit is ~ 10-5 ~-cm 2. The Transmission Line Method (TLM) is most frequently used to measure contact resistance in FET structures where current is confined in a highdoped conducting channel on a semi-insulating substrate.E3Sl Typically, the accuracy of TLM measurement is of the order of 10.6 ~-cm 2. For resistances of 10 -7 f~-cm 2, Kelvin Cross Bridge Resistor measurements are used, which require a relatively complicated geometry of the contact pads.j361 There are many other methods and variations to measure specific contact resistance.[371-[391 A particalar method should be selected depending on the type of application and required accuracy. In addition to the performance of ohmic contacts, the lifetime of ohmic contacts is often estimated by monitoring the variation of specific contact resistance over a period of time often versus temperature (to be discussed in Sec. 2.3 below). It should be noted that contact resistance measurements represent averaged electrical property of M/S contacts. This is because the dimensions of contact pads (dimension d in Fig. 3) typically range from-~ 10 (smallest) up to ~1000 gm, while the microstructure of the M/S contact is only homogeneous on the order of~l gm. In Fig. 3, current flow is not exactly perpendicular to the contact. The morphology of the contact is not ideal planar as depicted due to complicated metallurgical reactions, as discussed below.

1.2

Surface States and Fermi Level Pinning

In the development of Eq. (4), it is assumed that the M/S interface is free of structural defects and foreign atoms and is abrupt on the atomic scale (i.e., it is structurally ideal). In this case, the ideal barrier height (Oh) is determined by the work function of metal versus the electron affinity of the semiconductor. Experimentally measured barrier heights for many

90

Wide Bandgap Semiconductors

semiconductors, especially GaAs, show a very weak dependence on the metal work function.J4~ 42]

Metallized contact pad M/S~contact ~

~ / , ' /~

~

Semi-insulaing substrate conduction channel Figure 3. A schematicof Transmission Line Method (TLM)for measurementof specific contact resistance.

This deviation from ideal Schottky behavior was first attributed to the existence of surface states (intrinsic or extrinsic) by Bardeen in 1947.[4~ When surface states are present within the bandgap of semiconductors, they are occupied by electrons prior to contact formation with metals. Occupation is controlled by the Fermi level which is constant throughout the crystal. Consequently, these surface states pin the Fermi level resulting in band bending as shown in Fig. 4. As a result, the barrier height (Oh) is not given by Eq. (4) but must be modified by the details of Ess, the energy levels of the surface states. In Fig. 4, the barrier height can be described as Ob = Eg- Ess.[4~ As shown, the barrier height is nearly independent of the metal work function due to the Fermi level pinning. Fermi level pinning may be described by introducing extra charge (Qi), trapped in the surface states between E v to Ess, as shown in Fig. 4. Now charge neutrality is determined by: Eq. (17)

Qm + Qi + Qs = 0

Ohmic Contacts to II-VI and III-V Compounds

91

as contrasted to Qm + Qs = 0 in an ideal M/S interface. The potential variation (A), in Fig. 4 is due to Oi trapped at the surface states. The surface states may be a source of leakage current in M/S contacts and result in deviation from the ideal current-voltage characteristics predicted by Eq. (7), particularly in the low-voltage region.

Metal

n-Semi con ductor

Vacuum level A s

Om,T Erm _ __*_ _v_

Vacuum level ..

s Ec . . . . . Em

-I ss

~ I ' ,

Ev

Figure 4. Energy band diagram for a Schottky barrier formed by a metal and a semiconductor with surface states. Essis the surface state energy levels within the bandgap.
I n t r i n s i c State. Kurtin et al. proposed several intrinsic mechanisms as plausible causes of surface states, and suggested a correlation between a tendency to form surface states and the type of chemical bonding in the bulk. Based on measurements of the barrier heights of various metals on Si, GaSe, and SiO 2, versus electronegativities of the metals, it was suggested that the termination of periodic atomic arrangement at the surface of covalently bonded semiconductors is likely to give rise to surface states. [43] In Fig. 5, the S parameter (defined by S =/9~b//~, n where ~m is the electronegativity of the metal) is plotted as function of AZ, to show the general tendency of Fermi level pinning. As shown in Fig. 5, Si, Ge, GaAs, and InP belong to the group of semiconductors which exhibit Fermi level pinning (i.e., S - 0 . 1 ) . As a word of caution, Fig. 5 contains data which may

92

Wide Bandgap Semiconductors

be influenced by several effects in addition to Fermi level pinning. [44] To illustrate this, it has been shown experimentally[45]-[ 47] and theoretically[ as] that intrinsic states in the bandgap for the (110) surface of GaAs may play a minor or no role in Fermi-level pinning. Nevertheless, the use ofionicity of semiconductor bonding is still predictive for pinning of the Fermi level as mentioned in the introduction.[ 22]

AIN 9

eZnS

1.0

aS

SrTiO3 KTo03

AI zO3 9 9 9 SnOz ZnO SiOz Go20~

.8

/TO .,...

Lm_ 0 -1X bJ r-~ uJ z (~ o.

-6

,.'F

Si/eZnSe

_

(/) n,-.4 LIJ I--

eGoTeI

_z

cdSe

.2

Ge

GoAs/

~Si InSb~ ] I'n/ P 1

I

A 1

1

1

1

1

!

1

1

I

AX: ELECTRONEGATIVITYDIFFERENCE BETWEEN CONSTITUENTS OF SEMICONDUCTORS Figure 5. Generaltrendof Fermi level pinning in various semiconductors(from Ref. 43).

Unified Defect Model. To explain Fermi-level pinning, a large body of experimental data was analyzed and the result called the Unified Defect Model. Spicer et al. discovered that clean, cleaved (110) surfaces of III-V compounds, except for GaP, did not exhibit intrinsic states, but the surfaces were strongly perturbed by metal coverages ofa monolayer or less, resulting in Fermi-level pinning. [42][49] Figure 6 shows the pinned Fermi level of GaAs, GaSh, and InP versus several metals, and they are essentially independent of

Ohmic Contacts to II-VI and III-V Compounds

93

the metal. Spicer et al. proposed that pinning resulted from defects created by the interactions between the adatoms (metals and oxygen) and semiconductors. Unique to the UDM is that Fermi level pinning is a consequence of extrinsic effects, not intrinsic properties of the semiconductors.

FERMI LEVEL PINNING

evf l

/

/,/"

//

1 2~r-1 0.8~I

on ~p

/ / . / / /

ICB.M.

Go As (1101 .... o

~____a

........ o

o

0.4~-

Eg

o

i,-'A:S/:At

.... Go ",o " A:u. " 0 , /

vBM

i >LO

ev! / ,,I2: 0.6 ...... " 0.4

....

-

.-

."

_ "

,lCBM

LU Z ILl

0.2

o I," , - C s /

Go ....."

.Sb."Au,,0,y

.IVBM

,,.,"

,/

ev 1.2

In P (110 )

CBM

o A

O

0.8 0.4

--

A

Eg

l

VBM OVERLAYER PRODUCING PINNING ( S u b - M o n o l a y e r ) .

Figure 6. Fermi level pinningposition for n- or p-type (a) GaAs,(b) GaSb, or (c) lnP versus the overlayer adatom (from Ref. 49).

94

Wide Bandgap Semiconductors

Metal Induced Gap States. The Metal Induced Gap (MIG) model has been developed as an extension of the UDM.[5~ In the MIG model, surface states are postulated to be an intrinsic property of the semiconductor, and they exist over a continuum between the Valence Band Maximum (VBM) and the Fermi level. This is in direct contradiction to the discrete levels assumed in the UDM. The MIG states (surface states) are formed in the semiconductor at the initial interface, due to intimate contact with electrons from the metals. The tails of the metallic wave functions decay into the semiconductor and result in the interface states within the bandgap. Heine first pointed out that tails of the metal wavefunctions are derived from the Virtual Gap States (ViGS) of the complex band structure of the semiconduetor.[S~ In the MIG model, the character of the surface states changes across the bandgap from more acceptor-like, close to the Conduction Band Minimum (CBM), to more donor-like, close to the VBM. The energy at which the contributions from both bands are equal is called the branch point, and is located at mid-bandgap when the effective mass of electrons and holes are of equal value.[ 51] The Charge-Neutrality Level (CNL) is defined as the energy where the Fermi level coincides with the branch point. This determines the Schottky barrier heights. According to the MIG model combined with electronegativity concept, charge transfer occurs across the interface depending on the electronegativity difference between metals and semiconductors, and determines the final position of the Fermi level within the bandgap. For example, when the electronegativity of a metal is the same as the semiconductor, the final Fermi level pinning is located at the branch point of the semiconductor and there is no charge transfer. The final pinning position of the Fermi level should be above or below the branch point of the semiconductor when the metal exhibits a smaller or a larger eletronegativity than the semiconductor. With the Fermi level above or below the CNL, the net charge in the wavefunction tails of the metals has a negative and a positive sign, respectively, as a result of charge transfer between the metal and semiconductor. The Schottky barrier height of the M/S contact would be lowered with the Fermi level above the CNL, and increased with the Fermi level below the CNL, for n-type. The precise position of the Fermi level within the bandgap and corresponding barrier height depends on the occupation of the continuum of the MIG states, and the trend of occupation can be predicted from the electronegativites of metals and semiconductors:[ sll[s2] Eq. (18)

Ohmic Contacts to I I - V I and I I I - V Compounds

95

where tY~cnI is the CNL, Sx is a slope parameter defined as ~b/c m, and ~m and Zs are the eleetronegativities of metal and semiconductor, respectively. The slope parameter S x, (this is the same as S in Fig. 5) was related with the optical dielectric constant, Co, of the semiconductor, while (Y~cnl w a s theoretically predicted. Using Eq. (18), t~b -~ 0.7 eV above the VBM is predicted for Au/GaAs contact. In a recent refinement of the MIG model,[ 52] it was shown from experimental data[ 53] that the position of Fermi level pinning was a function of metal layer coverage. The Fermi level approached its final position as the surface was saturated from submonolayer coverage (isolated adatoms) to a continuous metallic film. The dependence of the Fermi level upon metal coverage (e.g., Fig. 7b), which can't be explained by the UDM, was explained in the MIG model by using the idea of metal-induced surface states at submonolayer coverage and by the continuum of metal-induced gap states at several monolayer coverage. At high coverage, only one Fermi level was determined, regardless of n- or p-type semiconductor. Chemical Reaction Models. Chemical reactivity between semiconductors and metals have been correlated to explain the deviation from the ideal Sehottky barrier; it was rationalized that chemical reactions would affect interfaeial electrical properties. The heat of phase formation was found to exhibit a linear relationship with the barrier heights (~b) between transition metals and Si.[ 54] Later, Freeouf proposed that ~b was determined by ~Siliciae (work function of silieides).[ 55J Even though this idea was developed to describe metal/Si contacts, it was later applied to various compound semiconductors and developed as the Effective Work Function (EWF) Model.E56] In the EWF model, ~b is given by: Eq. (19)

~b = ~eff-- Z instead of Ob =

(I)metal-

Z

where Z is the electron affinity of the semiconductor. r mainly due to the work function of the anion, ~Anion" The EWF model suggests that the Fermi level at the surface (or interface) is related to the work functions of microelusters of one or more interface phases resulting from either oxygen contamination (oxidation) or metal-semiconductor reactions with the metallization. Another chemical argument proposed is that chemical reactions at the interface on a microscopic scale modify the ideal Sehottky barrier via local charge transfer and creation of extrinsic interface states as a result of the interfaeial reaetion.[571[58] In this argument, discrete levels of defects, native or extrinsic, were used to account for the deviation from the ideal Sehottky behavior with chemical reactivity.

96

Wide Bandgap Semiconductors

Amphoteric Defect Model. Walukiewicz has attributed Fermi level pinning to amphoteric native defects in semiconductors.[59]-[ 61]This model relates the Fermi level pinning to the thermodynamic properties of the entire system of defects. A remarkable similarity was found between the semiconductor Fermi level (EFs) at the M/S interface and the Fermi level (EFI) in heavily irradiated III-V compound and column IV semiconductors (see Fig. 7). As shown in Fig. 7, the Fermi levels for n and p-type semiconductors merged into one level as the irradiating electron dose increased to greater than about 1017cm-2.[6~ Also note, there is a similarity between the convergence level in Fig. 7 and the pinned level shown in Fig. 6.

-/• 1.4-

>

I

C

t.u

%

1.0-

(a) _

9n - T y p e o p-Type ~ Semi-insul.

\eN

0.8~

_

" ~ - ' Z ~ . - - e._e_ ~ - e - & ,

0.6-

/

. . -

/

~_E 0 . 4 u_

I

GaAs

9 12 >, O}

1

1

N

0

0.2[ 1017

1 16

0

L 1018

1019

10 20

Electron Dose (cm -2) _A

>(1)

rw

1 29

0.8

~c-

~S

~9 0.4 LL

(b) -

,~

/.~

0 LAo V

~'-~-

..o- - z-o~-t== ~ , Ti/GaAs "'~

,,n-Type o p-Type , ,

1022

10 -1

1

10

102

Metal Coverage (~) Figure 7. Comparison of the Fermi level behavior (a) in electron irradiated GaAs and (b) for a submonolayer coverage of Ti (from Ref. 60).

Ohmic Contacts to 11-111and III-V Compounds

97

Table 1 lists the range of Fermi levei pinning positions deduced from the Schottky barrier heights for M/S contacts (Eps) and the Fermi level stabilization energy in heavily irradiated III-V compounds and column IV elemental semiconductors (EFt), plus the CNL predicted from the MIGS model (EcNL). Note that the Epi correlates better with the interface state pinning levels (E~s) than do the CNL values.

Table 1. Fermi Level Stabilization Energy in Irradiated Semiconductors (EFI) and at Metal-Semiconductor Interfaces (EFs)

Si InP GaAs

EEl (eV)

EFS (eV)

EClVL(eV)

0.4 1.0 0.5-0.7

0.3-0.4 0.8-1.1 0.5-0.7

0.36 0.76 0.5

Eclvl"represents Charge-NeutralityLevel from MIG model. All energies are with respect to the valence-band edges (from Ref. 60 and references therein).

According to this model, there is a Fermi-level stabilization energy (EFI) in covalent or weakly ionic semiconductors such as GaAs and InP, which is independent of the type of doping and the doping level. Therefore, this property is regarded as an intrinsic property of the semiconductors. As a consequence of this intrinsic property, native defects such as vacancies or substitutional dopants exhibit an amphoteric character depending on their energy level relative to the Fermi-level stabilization energy. For example, a Ga vacancy is a stable aceeptor in n-type GaAs, but it transforms to a donor complex (Asca + VAs) in p-type materials. This behavior results from a large electronic contribution to the total defect formation energy. Specifically, the formation energy of a Ga vacancy is lowered from --4eV to --0.2eV as the Fermi level varies from the VBM to the CBM under As-rich condition.[ 62] The defect formation energy varies until the Fermi level reaches the stabilized position (EF/), after which continued introduction of electrically active species does not affect the stabilized Fermi level. In the case of GaAs, the stabilized Fermi-level is located between E v +0.5 eV to E v +0.7 as can be seen in Table 1. In the amphoteric native defects model,

98

Wide Bandgap Semiconductors

the behavior of native defects is responsible for the Fermi level pinning, which appears to be similar to the UDM model. However, the behavior of defects is controlled by the stabilized Fermi level EFI, which is an intrinsic property of semiconductors similar to the MIGS model. Summary. For many years, all the models mentioned above were examined and compared. Nonetheless, the main idea of whether intrinsic or extrinsic effects play a primary role in Fermi level pinning still remains controversial.j63]-[ 66] It is now generally accepted that a single theory cannot explain Fermi level pinning. In the case of GaAs, it is readily and clearly seen that there is a range of-~0.3 eV in Fermi-level pinning (0.5-0.8 eV above the VBM) by reviewing the extensive experimental data collected to date. This variation implies that several mechanisms are simultaneously playing roles and even interacting with each other in the determination of SBH.

2.0

O H M I C C O N T A C T S T O GaAs

GaAs is the most widely used III-V semiconductor and as a result, ohmic contacts to it have been extensively studied. A few techniques have been developed to prepare an ohmic contact with no or little heat treatment after growth ofGaAs epitaxial layers. However, these in situ approaches to ohmic contact formation have not proven to be very practical. Therefore, the majority of this review will focus on contact schemes using deposition of metallization after the growth process has been completed and the sample removed from the MBE or MOCVD systems and stored in laboratory ambient for significant lengths of time. These techniques invariably required heating after deposition of the metallization, and they are called ex situ contact schemes.

2.1

In Situ Contact Scheme

In situ ohmic contact schemes do not utilize heat treatment to produce complicated alloying or metallurgical reactions for incorporation of doping elements into the surface of the GaAs substrate. Instead, very heavy doping is accomplished in situ during the growth of GaAs, or heterojunctions are formed to lower the barrier height between the contact metals and GaAs. In principle, heterojunctions may be either gradual or abrupt junctions, and both have been used for GaAs contacts. A gradual

Ohmic Contacts to II- VI and III- V Compounds

99

junction is one in which two bulk crystals are joined by a continuously varying composition (e.g., GaxInl_xAs/GaAs where x can vary from 0 to 1). An abrupt junction is a sharply defined interface between two homogeneous semiconductors (e.g., Ge/GaAs). For either doping or heterojunction formation, in situ contacts are nonalloyed because they are formed without post-deposition heat treatment. Nonalloyed contacts were developed to reduce some drawbacks of conventional metallization schemes, such as poor morphology, poor reproducibility, deep interdiffusion (deep junction), and poor thermal stability. They succeeded in minimizing some of these effects, but the penalty in product throughput was too great to be practical. Molecular beam epitaxy is frequently used to grow heavily doped n+-GaAs where n + can be as high as the mid-1019 cm -3 (i.e., above the critical carrier concentration calculated from Eq. 16 as being required to ensure TFE or FE transport across the M/S interface). The carrier concentrations produced by MBE are often as much as an order of magnitude above the values achieved in bulk crystal growth. MBE can produce such high carrier concentration because it is a non-equilibrium growth technique. Dopant incorporation may be controlled by surface kinetics rather than thermodynamic equilibrium conditions. Tin dopant has been studied because it is known to be less amphoteric as compared to Si or Ge. A free electron concentration as high as 6 • 1019 em -3 was achieved with Sn, resulting in specific contact resistance as low as 2 x 10 -6 ~_cm 2 simply by depositing metals without heat treatment.[ 67] Even though it tends to be amphoteric, Kirchner et al. reported--1 x 102~cm -3 Si doping which yielded contact resistances of ~1.3 x 10-6 ~_cm 2 with in situ metallization.[ 6s] As reported above, in situ formation of heterojunctions have been studied to form ohmic contacts. For example, Fig. 8 (a) and (b) show the band alignment between metal/GaAs and metal/n+-Ge/n+-GaAs, respectively. Because of Fermi level pinning, the Cb for metal/GaAs is 0.7-0.8 eV above the VBM as discussed already. The localized potential barriers of --0.45 eV at the metal/Ge interface and 0.06 eV at the Ge/GaAs heterojunction are much more favorable for ohmic behavior at room temperature .[691[7~ The dominant resistance will occur at the metal/Ge contact, but the Ge layer can easily be doped up to--102o cm -3, resulting in specific contact resistances of--10-6-10 -7 ~-cm 2 with smooth interfacial morphology.[69][ 71] The Ge/GaAs heterojunction is lattice matched (within 0.5%), and has compatible crystal structure. The thermal expansion coefficient of Ge (6.6 x 106/~ matches well with that of GaAs (6.0 x 10-6/~

100

Wide Bandgap Semiconductors

Metal

(a)

n+-GaAs

% Ec Ef

Ev

J Ec . . . .

Ef

(b)

I

Metal

!

,n§

,

n§

n-GaAs

Figure 8. Energy band diagram for metal contacts to GaAs showing the band alignment and bending for (a) a degenerately doped n-GaAs, and (b) the same GaAs with epitaxed Ge at the interface with a metal (from Ref. 7 l).

Another heterojunction scheme is to terminate the GaAs surface with Gal_xInxAS, in which the Fermi level is pinned in or near the conduction band as shown in Fig. 9.[72] Because there is a conduction band discontinuity between InAs and GaAs as shown in Fig. 9(e), a non-abrupt heterojunetion is necessary as shown in Fig. 9(d). In this contact scheme, tunneling is not required and low resistance contacts can be made for a wide range of doping without the need of alloying to form n + surface layers. InAs and Gal_xInxAs layers grown in situ by MBE produced contact resistances o f - 1 0 "7 f2-em2.[17][731However, this contact scheme is not generally practical because of MBE production throughput and

Ohmic Contacts to II-VI and III-V Compounds

101

inability to complete processing the device structure requires "dry" techniques in the MBE UHV environment.

Ec o)~////////////~

'

~,~M ! ETAL2//"

~

'

EF

n- GoAs

Ec b) ~

..

~..-~---~

. . .

EF

Ev ~,,~.__

n-InAs I.

c)-: MET.,;..,

.

.

.

.

.

'

. ,.~/,:.

Ec EF

s

ABRUPT NTERFAC :

:...,;;.::~;, I/,'.,

",9 , , . / / 3 " / . ,

NON ABRUPT INTERFACE

r

\

I

'

...........

.............; - ~ ~ ~ , - ~ Go i_x Inx As

Figure 9. Band diagram showing band b~nding and alignment for a metal conlact to (a) nGaAs, (b) n-InAs, (c) n-GaAs with a thin epitaxially abrupt layer ofn-InAs, and (d) n+-InAs layer on a graded n+-InxGal.xAS layer on n-GaAs (from Ref. 72).

102

Wide Bandgap Semiconductors

Fischer et al. [741175] have reported that sulfur passivation of the GaAs surface can at least partially remove surface pinning and result in ohmic contact formation. In their studies, solutions of P2Ss/Sx/NH4S were used to remove the native oxide on the GaAs surface and replace it with S bonding to the Ga and perhaps As. Samples of Si doped GaAs were then placed in a vacuum, S desorbed by heating to about 500~ and films of Au deposited in situ. Upon removal from the vacuum and without heat treatment, the contacts showed a linear I-V dependence demonstrating formation of an ohmic contact. Surface passivation with S, plus the H20/light treatment of GaAs surfaces reported by Woodall et al., [76]are the only known methods to at least partially unpin the GaAs surface for contact formation.

2.2

Ex Situ Contact Schemes Ex situ contact schemes are those using single or multilayer metal

thin films deposited in a separate chamber from that used in epilayer growth. Thermal annealing is generally required to form the ohmic contact. Single layer metallizations will be reviewed first, followed by a review of bilayer and multilayer metallizations. Single Metal/GaAs Contacts. While single metals on GaAs do not yield technologically feasible ohmic contacts, a review is worthwhile to learn about their interfacial reactions. Special attention has been given to several frequently used metals such as (Au, Ni, Pd, or Ge)/GaAs. It will become obvious that interfacial reactions between metals and GaAs play a critical role in the ex situ formation of ohmic contacts. It will also become obvious that equilibrium thermodynamics can be used to predict GaAs interfacial reactions, although non-equilibrium intermediate phases may also be encountered which may affect the contact properties. A u / G a A s Metallization. Au is frequently used as a contact metal for GaAs and other semiconductors because of its ease of deposition and etching, high conductivity which results in lower resistance interconnects, high ductility for bonding, and lack of oxide formation which results in good bonding and high reliability. Au begins to react with As depleted or near stoichiometric GaAs at ~250~ with rapid reaction at ---400~ resulting in large changes from the as deposited Schottky barrier heights.[771 Typically, Au/GaAs diodes exhibit a barrier height of--0.9 eV as deposited and low barrier Schottky or ohmic behavior after heat treatment.[ 781-[8~ The interfacial reaction between Au and GaAs upon annealing begins with the formation of a Au-Ga-As solid solution with very low As concentrations.[ 811

Ohmic Contacts to II-VI and III-V Compounds

103

The solubility of As in Au-Ga-As is so low that the reaction between Au and GaAs is commonly written: Eq. (20)

Au + GaAs ~ Au-Ga + As x (gas)

which represents the formation ofAu-rieh Au-Ga solid solutions with As x sublimation in the case of an open system (i.e., with a possible loss of Asx). In the ease of a closed system (i.e., no loss of As x possible), As forms precipitates because of the very low solubility of As in Au. [81]-[83]A sharp evaporation peak for As x has been observed by mass spectroscopy, suggesting rapid onset of a metallurgical reaction such as Au-Ga formation. Surface morphological changes have sometimes been interpreted to indicate formation of a liquid phase of Au-Ga alloys at the temperature ranges.[78][821184][85] However, in other cases, these morphology changes have been attributed to solid state transport without formation of a liquid phase.[ 81] While the bulk equilibrium phase diagram shows that a number of intermetallic phases are possible (Au7Ga 2, Au3Ga, Au2Ga, AuGa, or AuGa2), they have only been observed after cooling.j85] -[89] In contrast to the surface capillarity driven surface morphology mentioned above, solid state reactions between Au and GaAs often result in formation of elongated pyramidal pits (see Fig. 10a) bounded by {111 } planes and aligned in the [ 110] directions of GaAs after annealing above -3 50~ [90]-[92] Figure 10b is the corresponding schematic of Fig. 10a, and Fig. 10c is a bright field image of a {110} fractured cross section though a single pyramidal pit. Solid solutions of Au-Ga were found in the pyramidal pits. These solutions were sometimes separated from the GaAs substrate by an intermediate layer of Au-Ga compounds which probably formed upon cooling (i.e., they represent a divorced eutectic decomposition). The reactions pits form through solid-state dissolution of GaAs up to 450~ although a liquid reaction apparently increases their size and governs overall morphology above the melting temperature (--500~ of the Aurich solid solutions. The crystallographic orientation between the reaction products and the parent Au and GaAs has been attributed to the degree of misfit at the interface.j88][ 91] Others have suggested that the crystallographic dependence of GaAs dissociation kinetics may also influence their geometry.[ 92] In any case, formation of the pits on a GaAs surface during interaction with Au represents a very inhomogeneous reaction. The inhomogeneity of the interfacial reaction has also been attributed to the presence of native oxide which limits the ability of Au to react with the substrate.

104

Wide Bandgap Semiconductors

(a)

{111)As

[011"-] [o11 (b)

BaAs (100~

{ 111}Ga

(c)

Figure 10. Reaction "pits" from Au on GaAs after the Au has been chemically stripped. (a) Scanning electron micrograph of a pit. (b) Schematic representation of the crystallographic orientation of the pits. (c) Scanning electron micrograph of a fracture cross section through a pit (from Ref. 92).

Ohmic Contacts to II- VI and III- V Compounds

105

Upon reaction of Au with GaAs, the electrical characteristics of the interface have been observed to switch from Schottky to ohmic.[ 8~ To determine if this resulted from interfacial compound formation, Leung et al. [791and Lince et al.[941deposited Au-Ga phases (e.g., AuGa2) on GaAs. In general, these compounds did not result in ohmic contact formation because the surface of the GaAs remained pinned. In addition, the lack of or the limited interfacial reactions that were produced did not significantly modify the interfacial doping concentration, and Schottky contact behavior was maintained, although the quality of the Schottky contact (and barrier height) often was degraded. The chemical reactions between Au/GaAs were studied in detail by Mueller et al. to understand their effects upon the electrical properties of the Au/GaAs system.[8~ 811 It was shown that the Au/GaAs interfacial reactions were dependent upon the conditions of the initial interface. Surprisingly, an interface with a native oxide exhibited the most pronounced formation of reaction pits, consistent with the calculations of Mueller and Holloway showing that the phase stability of Au/GaAs is critically dependent upon the surface stoichiometry. Li and Holloway[771 showed that the Au/GaAs reactions could be modified extensively by supplying a flux of As to the surface during heat treatment. Liu and Holloway[ 921showed that the interfacial reactions pits underwent Ostwald ripening with time at temperature (i.e., some pits disappear while others grow during isothermal annealing). Appearance of the reactions pits results from dissolution of GaAs. Therefore, their disappearance indicates regrowth of GaAs, apparently from Ga and As in Au-Ga-As solid solutions. The reaction of Au with GaAs correlated with the electrical properties switch from rectifying to ohmic. The explanation for forming ohmic contacts was uncertain since the Fermi level was believed to remain pinned, and no dopant was added to the pure Au layer which could be incorporated during regrowth. The only possible dopant to be concentrated was the Si originally in the bulk GaAs. Segregation in the reactions pits of this Si dopant was detected by SIMS (see Fig. 11). Figure 11 is a SIMS image of m/e = 28 from a Si-doped GaAs after reaction with Au film at 450~ for 15 rain followed by chemical stripping of the Au film. Thus, the ohmic contact resulted from Si segregation in the pits causing a local n+region and TFE transport for ohmic behavior. Regrowth of the GaAs occurred as a result of the interfacial reaction, but did not play a role in the ohmic behavior. Ni/GoAs Metallization. Ni reacts uniformly with GaAs except where the native oxide-hydrocarbon interfacial contamination layer affects

106

Wide Bandgap Semiconductors

diffusion and compound formation.[95][96] Solid-state reactions between Ni and GaAs produced a hexagonal ternary phase after annealing at 100400~ for times of 5 min to 5 hours: Eq. (21)

xNi + GaAs = NixGaAs

g.

t Figure 11. SIMS imagefromm/e = 28 (i.e., Si§ showing segregation of Si in the reaction pits producedby Au on n-GaAs(fromRef. 93).

Diffraction analysis with TEM and XRD showed that NixGaAs could exhibit twinning and several epitaxial orientations with respect to the GaAs substrate. [97]-[106]The composition x was reported to vary between 2 and 4 without a change in crystal structure. Quantification of AES data yielded Ni2GaAs [96][97][99]while EDS and/or TEM measurement suggested NiaGaAs.[l~176176 1~ High spatial resolution EDS studies yielded NiE.4GaAs, as shown in Fig. 12.[1~ It is noted that compositions ranging from 2 to 4 were observed with the same NiAs-type hexagonal structure with varying Co/ao ratios.[a8][ 109] Studies of bulk material showed that the Ni-Ga-As ternary phase diagram exhibited five ternary phases with broad homogeneity ranges extending toward the binary phases.[ 1~ They all exhibited the hexagonal NiAs symmetry and were unstable in contact with GaAs.

Ohmic Contacts to II- VI and III- V Compounds

107

Figure 12. TEM micrograph of as-deposited GaAs/Ni2.4GaAs/Ge. Bright-field image of cross sectional view to show sharp interface between GaAs/NiE.4GaAs (from Ref. 108). While NixGaAs has often been observed in thin film reaction, it may not be an equilibrium phase. It was reported that NiaGaAs adopts a B8 structure with lattice parameters intermediate between those ofNi3.55Ga2.0 and NiAs B8 structures.[ 95][100]-[104] The hexagonal unit cell of NixGaAs has lattice parameters of ao~4A, co-~5A, similar to the Ni-As and Ni-Ga systems (NiAs:ao = 3.619/1,, %= 5.034A and Ni3Ga2:ao = 4.0A, %= 4.983A).E100]--[105] It has been speculated that NixGaAs is observed even though it may be metastable because of the epitaxial relationship with GaAs which would lower its nucleation barrier below that ofNiAs x and NiGay. [99]In addition, Ni is the mobile species at temperatures below-~300~ where Ga and As are relatively immobile, allowing formation of Ni~GaAs while avoiding separation of Ga and As.[1~176 1111Figure 12 shows an interfaeial morphology of a NiE.4GaAs/GaAs contact, which is rectifying with a barrier height of 0.9 eV.[ 1~ It should be noted that the interface is uniform in contrast to Au/GaAs as shown in Fig. 10e. While the NixGaAs phase is commonly observed in thin film systems at low temperatures, Ogawa showed that NiEGaAs decomposed into binary NiGa and NiAs upon annealing at 500~ for 5 min.[ 971Decomposition proceeded through NiAs precipitation in a matrix of Ni2GaAs at temperatures higher than 350~ After annealing at 500~ for 5 min, NiAs or As-rich phases were found near the GaAs and [3-NiGa near the surface. [97] However, Laval et al. and Guivareh et al. reported that Ni2GaAs epitaxed to (111) GaAs was more stable than on (100) and was still observed after annealing up to 600~ [99][106]On bulk GaAs at 600~ after 1 hour, NixGaAs decomposed into NiGa and NiAs.[ 1~ After the 600~

108

Wide Bandgap Semiconductors

anneal, [3-NiGa epitaxed to the GaAs substrate. Randomly oriented NiAs grains, about 0.7 mm in diameter, were observed.[ 99][1~ Sand et al. reported epitaxy for both NiAs and NiGa after annealing at 600~ for 1 hour.[ 11~ While the exact phases at the interface varied in all cases, Ni greatly improved the tmiformity of the reactions, presumably at least in part, because it forms both Ni-Ga and Ni-As compounds. With respect to electrical properties, Ni alone did not result in formation of ohmic contacts. The Schottky barrier height increased slightly (0.76 eV to 0.83 eV) upon formation of Ni2GaAs, then decreased upon decomposition into the binary phases.[ 99] Pd,/GaAs Metallization. Similar to Ni, Pd reacts with GaAs and rapidly and more uniformly penetrates the interracial native oxide, resulting in higher reactivity with GaAs. TEM micrograph images showed the formation of a PdxGaAs phase below the native oxide, where x is typically 2. [21][104][112]-[114] The exact stoichiometry of this reaction product varied, as did its crystallographic orientation with GaAs.[ 1121-[115]Below 300~ annealing partially decomposed the PdxGaAs into Pd-Ga and Pd-As binary phases, which caused the metal/GaAs interfacial morphology to become nonuniform.[ 116] Vacuum annealing above 450~ caused the loss of the Pd-As phases, presumably due to arsenic evaporation. Lin et al. reported that bulk diffusion couples of Pd (~0.6 mm thick)/GaAs showed three stable ternary phases at 600~ 116]They were characterized by broad homogeneity ranges, similar to Ni/GaAs. The thermodynamically stable phases with GaAs at 600~ were PdGa and PdAs 2. However, loss of As in an open system was expected to result in a PdGa/GaAs structure. Ge/GaAs Metallization. To demonstrate that regrowth of GaAs in the presence of Ge could lead to ohmic contacts, Li and Holloway[77][ 117] deposited layers of Ga, As, and Ge on GaAs and annealed the samples up to 500~ This resulted in the formation ofepitaxial GaAs doped with Ge and the formation of ohmic contacts, clearly demonstrating the feasibility of the solid phase epitaxial method of forming ohmic contacts to GaAs. In addition, they demonstrated that a flux of As onto the surface could stabilize GaAs against dissolution by Au, and could be used to regrow GaAs with dopant from a surface covered by Ge and Ga films. Refractory Metals/Go.As MetaUizations. High melting temperature metal elements such as Pt, Ti, and W are frequently used and studied for thermally stable contacts because the processing temperature can be as high as 800~ to anneal out damage by ion implantation.[ 1181Pt/GaAs produces Pt-Ga, Pt-As binaries upon annealing 400-500~ Similarly, a

Ohmic Contacts to II-VI and III-V Compounds

109

Ti/GaAs reaction produces Ti-As and Ti-Ga above 400~ [1191The melting temperatures of those reaction products are typically above 1000~ 12~ Electrically, these reaction products exhibit rectifying contacts. As a result, refractory metals (elements or compound) are more commonly used as constituents in multielement contact metallizations.[~2 ~1 More details will be presented below (in See. 2.3) in a review of the thermal stability and reliability of ohmic contacts. Multi-element/GaAs Contacts. Practical ohmic contact metallizations contain more than two elements. However, the prior knowledge obtained in single element/GaAs rnetallizations is useful in understanding complicated reactions between multielements and GaAs. Au/Ge/GaAs Metallization. The reactions and electrical properties of Au/Ge/GaAs contacts are quite different from the Au/GaAs system described above. An eutectic forms at 88-12 wt% Au-Ge with a melting temperature of 363~ However, even small concentrations of Ge (-~0.6 wt.%) drastically changed the morphology of Au-Ga intermetallic phases (e.g., Au7Ga2, Au3Ga ) and they form at progressively lower temperatures with increasing Ge concentration.[ ~22]In addition, a Au-Ge-As phase was observed to cover the contact after annealing at ~400~ with decreasing coverage at higher temperatures. Epitaxial regrowth of GaAs was reported after cooling to room temperature from higher temperatures. Longer time anneals caused the Au-Ge-As phase to disappear, perhaps from local melting, followed by solidification upon cooling.[ sS] For annealing above -~400~ a molten Au-Ge eutectic phase was detected followed by formation of Au-Ga compounds. These reactions resulted in an ohmic contact with a specific contact resistance of~l 0-5 ~-cm 2. Formation of ohmic contacts was attributed to incorporation of Ge during solidification of GaAs, leading to the formation of a heavily doped n + GaAs layer.liE3][TM]This is consistent with the observation ofepitaxially regrown GaAs after cooling to room temperature. The liquid phase could accelerate formation ofn + epitaxial GaAs, and Ge was responsible for both the low eutectic temperature as well as acting as the n-type dopant. While in-diffusion of Ge into GaAs was suggested, the n + layer was probably formed during cooling. Reduction in the barrier height from 0.77 eV to ~0.4 eV was also reported after annealing below the Au-Ge eutectic temperature, but this again probably resulted from surface doping by Ge. [1231[125] There were no reports of epitaxially regrown Ge films at the interface, therefore the reduced barriers did not result from formation of Ge at the interface (depicted in Fig. 8). Instead, regrowth of GaAs at the interface, doped with Ge, was responsible for the reduced barrier heights

110

Wide Bandgap Semiconductors

and ohmic properties. The reaction morphology was similar to Au/GaAs as shown in Fig. 10, even to the formation of pyramidal reaction pits, and even to lack of a uniform morphology. Ni/Ge, Pd/Ge, and Pd/Si/GaAs Metallizations. As shown by the above reviewed data, Au has only limited power to dissociate the lattice. Nickel and Pd are much more efficient at dissociating GaAs and bonding to both the Ga and the As to hold them in the thin film. As also pointed out, contacts containing Au show marginal thermal stability, irregular morphology, and degradation of contact resistance with time. The properties of Au-free metallizations were studied to determine if removal of Au would eliminate the problems. Bielement metallizations such as Pd/Si,[ 126] Pd/Ge, [127] and Ni/Ge[ TMwere tested. Without Ni or Pd, very little Ge was incorporated into GaAs after sintering at 450~ for 30 min.[128]-[13~ With a Ni overlayer on Ge to cause dissociation of GaAs, greatly enhanced incorporation of Ge resulted in ohmic contacts.[ 1281The Ge profile after annealing always followed the Ni distribution, suggesting a correlation between these elements during regrowth of GaAs.[ 13~ Marshall et al. showed that Ge/Pd/GaAs structures resulted in Pd dissociating the GaAs to form a ternary Pd2GaAs phase. This was followed by the formation ofa PdGe x intermetallic compound causing the release of Ga and As and regrowth of GaAs.[ 21][131][132]When regrowth occurs in the presence of a dopant (e.g., Ge), the dopant will be incorporated into the regrown layer and may result in ohmic contact formation. The concept of the regrowth mechanism was applied to explain ohmic behavior for Ni/Ge/GaAs metallizations.[ 21] In this model, Ni was postulated to react with GaAs to form NixGaAs, then decomposition of the ternary phase was driven by the lower free energy of formation of NiGe to result in: Eq. (22)

NixGaAs + xGe ~ xNiGe + GaAs(Ge)

where Ge incorporation is believed to result in n+-GaAs. Typically, the Ni and Pd germanides are formed by annealing above 400~ and they have a very smooth surface morphology and good thermal stability (up to 500~ due to their high melting temperature.[ 127]A minimum thickness ratio of Ni to Ge (29-3 8 at.%) was necessary to form ohmic contacts. Regrown GaAs was distinguishable from the single crystal wafer by a high density of stacking faults, microtwins, and precipitates.J21][ l~ The regrown GaAs was postulated to contain a high concentration of dopant (e.g., Ge with a concentration of ~ 1019 cm3). However, limited experimental data was presented to support the incorporation of Ge to form n+-GaAs.

Ohmic Contacts to II- VI and III- V Compounds

111

When the Pd/Ge layering sequence was reversed and samples of Pd/ Ge/GaAs were reacted below 400~ only a rectifying contact was observed. AES depth profiles showed that Pd-Ga and Pd-As phases were formed while Ge segregated near the surface to form PdGe and Pd2Ge.[ 129] The interfaeial barrier height was increased from -~0.7 eV to --0.8 eV by these phases. Reactions at 400-500~ were necessary to form ohmic contacts, and the Ge SIMS profile into GaAs correlated with the reacted layer.[ 13~ To explain these results, it was commonly postulated that Pd created Ga vacancies, which accelerated the diffusion of Ge into the nearinterface region to form n+-GaAs. The premise that diffusion of Ge caused formation of an n + layer was not directly supported by experimental data. Instead, incorporation of Ge during regrowth of GaAs was postulated by Marshall et al., without the need for solid state diffusion. As illustrated in Fig. 8, epitaxial Ge will potentially form nonalloyed contacts to n-GaAs. This was studied using Ge/Pd/GaAs metallization.[131 ][1321Upon annealing, Pd and Ge reacted to form PdGe while excess Ge was transported through the Pd layer to grow epitaxially on the GaAs substrate. To clarify the mechanism for ohmic transport, SIMS analysis was performed by removal of the vast majority of the GaAs substrate through etching of a parting A1GaAs layer. The concentration of Ge incorporated into the GaAs near-surface region was measured. [134] Concentrations ofGe of-~l x 1019cm-3 were correlated with the onset of ohmic behavior. Thus, epitaxial Ge was not the origin of the ohmic contacts; instead creation of an n § layer in regrown GaAs was concluded to have resulted in ohmic contact behavior. The characteristics of Si/Pd/GaAs were investigated and compared to those of Ge/Pd/GaAs. Again it was shown that an epitaxial Ge layer with a resultant low barrier heterojunction was not responsible for ohmic behavior.[ 126] Instead, a regrowth mechanism was suggested with excess Si reacting with PdaGaAs to produce the following: Eq. (23)

2Si + Pd4GaAs(Si ) ~ 2Pd2Si + GaAs(Si)

The critical product from this reaction is regrown GaAs doped with Si. This same mechanism was also tested in Ni/Si/GaAs which clearly showed that regrowth took place.[1351 Binary metallization of Au/Ni/GaAs is not discussed since this only results in formation of Schottky contacts. Without the presence of Ge, low resistance ohmic contacts cannot be formed.

112

Wide Bandgap Semiconductors

AuGeNi/GaAs-Diffusion Doping Model. AuGeNi contacts are the most heavily studied system for GaAs, partly because it is the oldest contact system. It was first introduced in 1967 by Braslau et al. to form ohmic contacts to GaAs-based microwave devices.[ 1~ The postulated mechanism of ohmic contact formation has undergone considerable evolution. The initial postulate was that a Au-Ge liquid eutectic formed and dissolved some GaAs which caused Ge to "diffuse" into the near surface region. Diffusion was possible because dissolution of Ga in Au would create a counter flux of Ga vacancies, which increased the transport of Ge by orders of magnitude.[ 1361The Ge created an n + layer and an ohmic contact. The role of Ni in this process was simply to prevent the AuGe melt from "bailing up" due to surface tension. Therefore, the three elements were evaporated to form a bilayer/substrate structure of Ni/AuGe/GaAs. Several investigators showed that with sufficiently high temperatures and long times, the sequence of the layers did not affect the final metallurgical or electrical properties.[ 137][138]A modified postulate was that Ni improved the surface morphology by improving the wetting of liquid Au-Ge to GaAs. It was observed that Ni and Ge accumulated at the interface with GaAs near the AuGe eutectic temperature, while Ga accumulated near the surface. [139]The evolution of the contact modeled continued when it was realized that the Au-Ge eutectic was not necessary and did not play a major role since Ge was gettered by Ni via a solid-state reaction.J14~ TM]The model evolved by recognition that Ni was correlated with the dissociation of GaAs, and the following reaction was suggested:[ 142] Eq. (24)

Au + Ni + GaAs --~ Au-Ga + Ni-As

Formation of nickel arsenides occurred at the GaAs interface and improved the surface morphology as illustrated in Fig. 13. [143]As shown in Fig. 13, without the first Ni layer on GaAs, Au/GaAs interfacial reaction dominated the interfacial morphology, which was not uniform as discussed above. However the reaction in Eq. 24 did not necessarily lead to formation of ohmic contacts; the barrier height, ~t~ increased from--,0.7 eV to ~0.9 eV as Ni accumulated at short times at the interface.j99][ 139]At longer annealing times, when ohmic behavior was observed, the surface morphology and contact resistance were both directly affected by the Ni to Ge ratio. Higher Ni concentrations led to better surface uniformity, while higher Ge concentrations led to lower contact resistance.[ 1441-[1471 The effect of Gewas still the creation of an n § surface layer by diffusion along a Ga vacancy flux.

Ohmic Contacts to II- VI and III- V Compounds

BE FORE ALLOYING

DURING HEATING ( BELOW 420*C)

i AutG,)'

Au

Ni

AF TER ALLOYING AT 440*C FOR 2~n

,

t

k'//

I

Au-Ge

SAMPLE A

GoAt

GoAt

"

~ A.(G',)

Au

N,

IJ-A. o , / / ' / A

Au (Ge,Go) ,,,m..

,,,.,Ira,.

(WITHOUT Ni FIRST LAYER)

113

4

Ni~Ge

i

Au- Ge

SAMPLE B

4' I AulGt,Go',

=Ni

/- ~" r .

(WITH 5nm Ni F I RST LAYER)

a,,,mm,

NixGoA$ NiAslGe) GoAs

GoAt

GoAs

Figure 13. Schematicillustration of the evolution of phases produced by the reaction of Au/ Ni/Ge layers on GaAs. Note that the first Ni layer on GaAs generally results in planar interface morphology. Generally,for a five elements system,the interface between reacted layers will not be planar (from Ref. 143).

An improved description of the evolution of the contact was suggested by the observation that Ni was always found to diffuse toward the interface and to react with GaAs. The Ge followed the Ni distribution. The Au was normally combined in Au-Ga compounds upon cooling. [85][138][139][141][142][148]-[150] Transmission electron microscopy (TEM) studies showed that annealing above 400~ for a few minutes followed by cooling to room temperature produced NiGe containing small concentrations of Ga and As, NiAs with Ge and Ga, and AuGa.[143][l 5~ The NiAs phase was in direct contact with GaAs resulting in a smooth interracial morphology (Fig. 13). A good ohmic contact with low contact resistance was attributed to a Ni2GeAs phase being in contact with n+-GaAs which resulted from diffusion of Ge from NiGe.[ 15~ In addition, the reaction was assumed to unpin E F, and the NiAs phases were suggested to be low barrier ohmic tunneling contacts where the contact resistance was a function of the NiAs(Ge) coverage at the interface. [23][143] At higher annealing temperatures, more NiAs phase with a highly oriented epitaxial

114

Wide Bandgap Semiconductors

relation was found at the interface.[ 85] The epitaxy was suggested to result from a close lattice matching between NiAs and GaAs. But even with NiAs at the interface, Ge diffusion to form n§ was necessary to explain ohmic behavior. Even though the Ga vacancy diffusion model for ohmic contacts was widely accepted, the activation energy for formation of contacts did not agree with the expected value for bulk diffusion. Therefore, Gupta and Kokle postulated that grain-boundary diffusion in the Au film was responsible for generation of Ga vacancies.[ 136] A Ga vacancy mechanism was also postulated to explain ohmic contact formation with Pd.[ 13~ Gupta and Kokle justified their grain boundary diffusion model based on the fact that contact resistance was high after either low or high temperature annealing, but low for intermediate temperatures.[ 136]This observation is generally true for any metallization leading to ohmic contacts on GaAs. Thus the "diffusion doping model" evolved to one in which it was presupposed that Ni dissociated the GaAs lattice and bonded to the As holding it at the interface. The Ga from this dissociation reacted with Au and formed Au-Ga solid solutions. Formation of AuGa x intermetallic phases was also believed to occur, especially during cooling after annealing. The formation of an n § surface region due to Ge diffusing along a counter flux of Ga vacancies was widely accepted as the mechanism of doping. Backside SIMS measurements in an alloyed AuGeNi system were used to demonstrate Ge doping of the near surface region,[ 152][153]but the profiles could not be analyzed in terms of a diffusion profile. In addition, the contact resistance correlated poorly with the detected Ge concentrations. Cross-sectional TEM showed that NiAs was in contact with the substrate. Based on these results, Bruce et al. suggested that n§ level was less important to contact resistance than the formation of the NiAs phase.[ 152]Further refinement of the model was clearly necessary. AuGeNi/GaAs-Solid Phase Regrowth Model. Because E F is pinned in GaAs (near mid bandgap at the surface), n § doping is critical to formation of an ohmic contact. The studies of interfacial reactions have been focused on the mechanism by which Ge occupies the Ga site to form n § GaAs. As pointed out above, while Au-Ga and Pd-Ga reactions have long been postulated to create Ga vacancies in GaAs, the data do not support this mechanism for Ge incorporation and doping. Therefore, the model of ohmic contacts formation in GaAs has evolved to the one of solid-phase regrowth, as discussed for Ni/Ge and Pd/ Ge metallization. Figure 14 is a simplified schematic to illustrate the evolution of phases in this model of ohmic contact formation with Au/Ge/

Ohmic Contacts to II- VI and III- V Compounds

115

Ni metallization. Between 100 and 300~ Ni reacts with GaAs to produce interfacial NixGaAs with low concentrations of Ge. Above this temperature range, the reaction normally results in "binary" phases consisting of NiGa x and NiAsy, both of which contain small concentrations of Ge. Simultaneous with Ni reacting with GaAs, Ni reacts with Ge to form NiGe, and Ge diffuses into Au. At longer times (Fig. 14b), the NixGaAs or NiGa and NiAs phases will be decomposed by formation of NiGe, whose free energy of formation is lower than that of either the ternary or the binary phases. The Ga and As released by this decomposition will lead to solid phase epitaxial regrowth of GaAs. Incorporation of Ge will cause the electrical properties to switch from rectifying to ohmic.

Ohmic contacts Au Ge Ni

GaAs

Au-Ge {de

Au-Ga

Ni-Ge

Ni-Ge

Ni-Ga-As-Ge

GaAs

(a) Interfacial reaction

GaAs

(b) Regrowth

Regrown GaAs (Ge) n+-GaAs

Figure 14. Idealized schematic of the evolution of phases during formation of an ohmic contact to n-GaAs using Au/Ge/Ni metallization. Even though planar interfaces are shown, non-planar reaction fronts are expected in five component systems.

Several key issues about the solid-phase regrowth model still need to be clarified to validate it. First, the evolution of phases at the interface needs to be clarified, and, since Ge is known to be an amphoteric dopant in GaAs, the mechanism(s) controlling selection of the Ga or As site in the GaAs matrix should be explained. A series of studies have been carried out

116

Wide Bandgap Semiconductors

in our laboratory to investigate the evolution of interfacial phases and their effect(s) on the formation of ohmic contacts. [74][75][77][93][108][118][154]-[157] Particular attention has been given to the reaction of Ni with GaAs and the evolution of the ternary hexagonal NixGaAs, since previous studies/21][135] showed that the contact properties were dependent on how this ternary phase evolved (see discussion of Ni/GaAs above). In previous experiments which demonstrated solid-phase regrowth, the entire layered structure was first deposited then annealed in one or two steps (e.g., first at 200-250~ prior to the final high temperature anneal).[ 135]As discussed above, multiple reactions occur simultaneously during this processing and it is unclear as to which reaction results in n§ doping. To better understand the reactions, Kim et al. deposited 650 A Ni films on GaAs and in situ annealed them at 300~ to form ~ 1300 A films of Ni2.4GaAs.[ 1~ This anneal was followed by in situ deposition of Ge and Ti films with various thicknesses, and evolution of the interfacial phases upon final annealing at 500~ was studied. Figure 15 is a schematic diagram of the experimental procedure. Ti rather than Au was used in the metallization to determine if the reaction sequence discussed above for the regrowth model (i.e., Ni dissociation of GaAs to form Ni-Ga-As-Ge phases which decompose to Ni-Ge and Ni-Ti plus epitaxial regrowth of Ge doped GaAs) was general. The general sequence of reactions was found for Ti/Ge/Ni metallization, supporting its general validity. In addition, the thickness of the Ge layer was varied from 300 A (thinner than the original 650 A Ni layer) to 750 A to control the degree of reduction of the Ni2.aGaAs phase to the NiGe phase, and thereby control the extent of GaAs regrowth. Ti was expected to assist the evolution of the interfacial phases at 500~ since Ni-Ti intermetallic compounds are formed with large, negative free energies of formation. By pre-reacting the Ni, the NiGe formed directly from Ni2.aGaAs + Ge, while the NiTi reaction occurred later after Ge had reacted with the Ni2.4GaAs layer. Separation of the reactions using this procedure was necessary since the overall metallurgical reactions at temperatures of -~500~ proceed rapidly over times of a few seconds.[23][148] In situ annealing to form Ni2.aGaAs allowed Kim to show that its evolution upon reactions with Ge and Ti proceeded along a route involving both transformation and decomposition.[ 1~ A flow chart showing the critical phases is shown in Fig. 16. In situ annealing at 300~ for 15 min produced Ni2.aGaAs, where the stoichiometry was determined using energy dispersive x-ray analysis (EDX) on a high resolution field emission STEM.[ lsS] After deposition of the Ge and Ti films on the Ni2.aGaAs and

Ohmic Contacts to II- VI and III- V Compounds

117

annealing at 500~ for 5 min, Ni2.4GaAs was decomposed to directly form NiTi and NiGe plus epitaxial doped GaAs as indicated by the dashed line, or decomposed into NiAs + Ni3Ga 2. This was followed by evolution into NiAs plus NiGeGa, and finally into TiNi + NiGe + GaAs(Ge), indicated by the solid line in Fig. 16.

Ge & Ti evaporation

Ni

I650A

Ni2.4GaAs

T-1300~

Ti Ge

Ni2.4GaAs

GaAs

Vacuum

I~ a t

nu = 2 910JScm4

GaAs

GaAs

annealing 500~

5rain

In situ annealing at 300~

Figure 15. Schematic of experimental procedure to investigate evolution of Ni2.4GaAs. See text for detailed explanation (from Ref. 108). GaAs + Ni

1

In situ annealed at 300"C, 15 minutes

Ni2.4GaAs /

/

/

Deposite Ge and Ti annealed at 500"C, 5minutes

~

!

I

NiAs + NisGa 2

I I ! ! I !, I

NiGeGa

! ! I

'

"

/

1

TixNi + NiGey + G a A s (Ge) Figure 16. Schematic of the evolution of phases on GaAs from an in situ anneal of Ni on GaAs at 300~ for 15 min followed by deposition of Ge and Ti and a subsequent anneal at 500~ for 5 min. See text for explanation (from Ref. 108).

118

Wide Bandgap Semiconductors

A representative set of data for the transformation and decomposition of Ni2.4GaAs phase are presented in Figs. 17-19. Figure 17 shows SIMS depth profiles of Ti/Ge/Ni2.4GaAs/GaAs. In Fig. 17a, the Ni, Ga and As signal are stable in the Ni2.4GaAs region, and the Ge and Ti layers are distinguishable at the surface. After annealing at 500~ for 5 min, the Ni and Ti signals are found together in the surface region, indicating formation ofNiTi x. The Ni2.4GaAs layer has been partially decomposed, but the Ge layer (250 A) was too thin to completely decompose this ternary film. The bilayer structure indicated by the SIMS data is shown by the cross section TEM mierograph in Fig. 18. The solid line in Fig. 18 indicates the location of the original interface between -~1300 A Ni2.4GaAs/GaAs. This interface obviously has moved towards the surface, indicating epitaxial regrowth of GaAs. In Fig. 18, a small arrow indicates trace of the decomposition of a Ni-As grain, resulting in the epitaxial regrowth of GaAs. Thus the area marked R in the figure is regrown GaAs. High spatial resolution EDX data shown in Figs. 19a and 19b were measured across an interface between regrown GaAs and a Ni-As (e.g., line scan 2 in Fig. 18) or Ni-Ga phase (e.g., line scan 1 in Fig. 18), respectively. These EDS results show that the Ni2.4GaAs transformed into Ni-As and Ni-Ga binaries. Since Ge is present in the Ni2.4GaAs and/or binary phases (4-12 at.% as shown in Fig. 19), it incorporated into the regrown GaAs to about 1020 cm-3.[ 158]This is sufficient to meet the requirements of TFE, as reported above, and therefore to achieve an ohmic contact. Electron diffraction analysis (e.g., from the structure shown in Fig. 18) showed several intermediate compositions for NiAs x and NiGay phases, most of which had crystal structures of the NiAs hexagonal type. They are believed to be intermediate phases along the reaction pathway followed to reach the equilibrium NiAs and Ni3Ga 2 phases.[l~ 159] Both NiAs• and NiGay phases must decompose to allow regrowth of Ge doped GaAs for ohmic contacts, and the decomposition products also depended critically upon the amount of starting elements. For example, in those samples where the amount of Ni was greater than the amount of Ge, the reaction to decompose Ni2.4GaAs or the binary phases did not go to completion. At times, this prevented the formation of an ohmic contact. When the ternary phase decomposed into the binary NiAs and Ni3Ga 2 phases, careful analysis with EDX and electron diffraction on the TEM Showed that Ge was incorporated predominantly at those regions where the binary phases decomposed and caused regrowth of GaAs. Figure 20 is a schematic summary highlighting the key information presented in

Ohmic Contacts to II-VI and III-V Compounds

119

Figs. 17-19 (i.e., the transformation of Ni2.4GaAs into NiAs x and NiGay binaries, and the decomposition of the binaries). The arrows in Fig. 20 indicate Ge indiffusion through the ternary/binary phases resulting in Gedoped regrown GaAs. Under some conditions, regrowth of GaAs occurred predominantly from decomposition adjacent to a NiGay binary phase at the interface, and ohmic characteristics were not observed for I-V data from these samples. In other cases, regrowth predominantly from a NiAs phase led to linear I-V data. Apparently, ohmic behavior was not uniform across the interface. The majority of the current was conducted across regions where the NiAs phase existed at the interface, consistent with the literature reporting a lower specific contact resistance as the percent of the interface covered by NiAs increased.[ 23][143]

(a) As-deposited le+5 ?

:' .....

I

GaAs

u~-a'--

"''"~'.,.'~""

"f

le+4

<_Xs-

G~

~1e+3 "

Ni2_AGaAs

Ti i Gel

le+2 le+1

!,,,

le+0

5

10 15 20 Sputter time (min)

25

30

Ga

|

-

1

{b) 5000C. 5 minutes

le+5 ..._1i

,.,.,..

le+4

~,~ le+3

FA.2 ~:t~x".N~ '1 C'-',. ,,,,/, ~.,..............,....,\

._zffl

~" le+2

...... ' ' "

t" ,==.

.......

-

....

I~lleII

le+1 le+0 0

5

10

15

20

25

30

Sputter time (min)

Figure 17. SIMS depth profile for GaAs/Ni2.4GaAs/250A Ge/300A Ti; (a) as deposited, (b) annealed at 500~ for 5 minutes (from Ref. 108). See text for explanation.

120

Wide Bandgap Semiconductors

Figure 18. TEM micrograph from GaAs/Ni2.4GaAs/250A Ge/300A Ti annealed at 500~ for 5 rain. This microstructure corresponds to the SIMS depth profile shown in Fig. 17 (b) (from Ref. 108). See text for explanation.

Not only was the regrown GaAs doped with about 1020 cm -3 Ge, but it contained-~2 at.% of Ni as well. Incorporation of this element should be expected although it had not previously been measured.[15s] Ni is a deep acceptor with a solubility limit of 1017-10 is cm "3 in n-type GaAs,[ 16~ and it would be expected to compensate some of the donor Ge. However, the Ge concentration of-~l x 102~ -3 should not be affected too dramatically. Li and Holloway also measured a free carrier concentration of about this magnitude in regrown GaAs.[ 77] The level of incorporation of Ge into regrown GaAs and into Ga sites are presumably controlled by the amphoteric doping model of Walukiewicz.[59]-[61][164] HyBrid Contact Metallizations. As already discussed, an in situ contact scheme such as In/GaAs can be grown by MBE to form ohmic contacts through InCraAs heterojunctions.[16s][166]Indium films of 50-200 A thickness were added to study the formation of graded GaxInl.xAS heterostructure in addition to the formation of n+-GaAs. [167] TEM images showed that the metal/GaAs interface was covered by regrown GaAs and GaxInl.• Compared to Ni/Ge bimetallization, the contact resistance was reduced from ~1.2 ~ m m to ~0.3 ~mm. These attempts at improvement have resulted in only marginal effects. Wang et al. suggested that for PdInGe/GaAs metallization, either heterojunctions or formation of n+-GaAs was dominant depending on the temperature.[ 16s] In this study, specific contact resistances of~2 x 10-7 ~-cm "3 was measured.

Ohmic Contacts to II-VI and III-V Compounds 121

(a) Regrown GaAs

,

60 It..

. . . . . . .

I

'

Ni

I

"~.

50

.....

-. . . . . . . . - I - _ _ t t ~

I ~' ........ s,

40 0

Nii.3As

|

I !

3o

^.

......... .,'x~

\,

ii/ /:, \ ". /I

10

/i

/

'iGa

-'-:,,-

I

. , j l ........... , i t . - - -9. d l -

0

§

....

-

-300

.... ~

-

-200

T .... ~

"

t

.

-100

.

.

0

.

...............

.................... .

, ........

.

100

200

300

Distance (~) (b) Regrown GaAs 60

,-

...........................

Ni:.lGa

~

.......

l

i

. . . . . . . .... . O" .... ....... "---~

50

9"

'

\\

// /

I

Ni

r o

..............................

..... 9

I

\\

........ Q

\ \\

'.

40

/

I

/

I

/

\\ i / - |~

-|p\ -

O 9 w,,,i

30

I

\

,,? -\ -t

Ga

/ r/

20

// /

10

/ /

..__-.---~ 0

I

:

ii l

-v- . .. .. .. . .

-300

" ~ v' "

. " " . "." .

-200

I/

_ ~- r " . . . .. . . ~(" . .

-100

'.,

Ge

'i

"

I ,

/~

~: ,i,- - - - . - - , ~

l./

............

9 ....

. . r! '

|I" ' , "

~ s ................................... ,"

9

0

,

i"'".

100

9

,

,"'

|

200

,

9 "'~ " , '

300

Distance (A,) Figure 19. High resolution EDS analysis across the interface (b) regrown GaAs/Ni2.1Ga (from Ref. 108).

(a) regrown GaAs/Nil.3As,

122

Wide Bandgap Semiconductors

(a) as-deposited

(b) SOOOC, 5 minutes

I

Ge

I

GaAs

I

Overall reaetioa:Ni,,GaAs +Ge +Ti = R-GaAs(Ge) +NiGa,+ NiAs, + NiTi,+ N i c e

Figure 20. Schematic of the evolution of Ni,,GaAs layer via formation of NiAs,, NiGa, and decomposition of the binaries after annealing at 500°C for 5 minutes. The arrows indicates direction of in-diffusion of Ge during the metallurgical reactions resulting in Gedoped regrown GaAs.

2.3 Reliability of and Thermally Stable Contact Metallizations As an aside, variants to AuGeNi systems have been studied in attempts to improve the thermal stability and surface morphology of nGaAs ohmic contacts. For high-temperature and high power devices, Interdifhreliable operations for long times at T > 300°C is sion and phase instabilities resulting in unnecessary reactions are of particular importance for shallow junctions. The reliability of contacts is a common and serious problem for any semiconductor. It is frequently found that the data of specific contact resistance versus annealing temperature exhibit a U-shaped curve (see Fig. 21), and the temperatures of the lowest resistance typically range from 450-550°C for GaAs and 300-400"C for InP. The specific contact resistance increases below and above this temperature range as shown in Fig. 2 1 for (a) Au-Ge films[80]and (b) for Ni/Ge films.[21]In a study of Au-containing

Ohmic Contacts to II-VI and III-V Compounds

123

metallizations by Kuan et al., contact resistance was reported to be -~1 x 10-6 f2-cm 2 after annealing Au~i/AuGe/GaAs at 410~ for 5 min. Excessive annealing at 450~ for 200 see resulted in an increased resistance to ---5 x 10-6 ~-cm 2 which was attributed to growth of a beta Au-Ga phase at the expense of the NiAs/GaAs coverage.[15~ Shih et al. correlated higher contact resistance to dilution of Ge in overgrown NiAs annealed at 600~ for 5 min. [143] Overgrowth of NiAs was correlated to non-ohmic behavior in W/Ni/Si/Ni/GaAs.[ 17~ Isothermal annealing at 400~ of Ni/Ge/Ni/ GaAs for 10 hours after contact formation caused a rise in contact resistance.[ 21] In the case of addition of Au to Ni/Ge/GaAs, Au indiffusion towards GaAs was suggested as the cause of increase of contact resistance after isothermal annealing at 400~ for 10 hours.[ TM] For In Au/Ge/Pd/InP contacts, annealing at 400-450~ resulted in Schottky contact, while 325350~ resulted in ohmic contacts. Protrusions of Au-In phases into InP with irregular interfacial morphology was suggested to be the cause of non-ohmic behavior.[ 172]Thus, Au diffusion and continued phase evolution in contact metallizations results in a long term thermal instability of low contact resistance. While this has been most often studied for Au in contacts to GaAs, it is a potential problem in any metallization where the phases formed by the metallurgical reactions are not allowed to go to completion for forming the equilibrium phases. This is normally the case since the GaAs wafer is so large compared to the metal thin films, that equilibrium is so far towards the Ga and As-rich end of the complex phase diagrams, that equilibrium is seldom achieved. In addition, the diffusion coefficient of Au is large and therefore the resistance instability is more dramatic than for other metals. Diffusion coefficients are large and interdiffusion rapid for metallization on InP devices. To improve thermal stability, W60N40has been added to the AuGeNi system to prevent the reaction between Au and GaAs while allowing the solid-phase reaction between the Ni and Ge layers to form the ohmic contact.[173] Both a W S i 2 and a Cr layer have also been inserted between the Au and Ge to prevent the formation of the Au-Ge eutectic.[ 13][174]These metallizations formed stable ohmic contacts with contact resistances of 105-10 -6 f2-cm 2, and exhibited smooth morphology above --600~ A Ta-Si-N barrier layer inserted between Au and AuGePt layers was reported to result in stable, smooth interface morphology.[ 122] For better stability in high temperature contacts, Zuleeg et al. replaced the Au in Au/Ge/Ni contacts with A1 (i.e., A1/Ge/Ni contacts). [175] They reported that a thermally stable specific contact resistance of ~ 10-6 f2-cm 2 could be achieved with very smooth surface morphology up to about 400~ Lampert et al. subsequently showed that the formation of ohmic contacts in this system depended critically upon the sequence of thin film deposition,

124

Wide Bandgap Semiconductors

with the A1/Ge/Ni/GaAs contacts becoming ohmic much more quickly than AI/Ni/Ge/GaAs contacts.[ 1~]-[156] They interpreted these observations in terms of Ni dissociation of GaAs and epitaxial regrowth of Ge doped GaAs due to formation of NiGe and AlxNiy intermetallic phases.[ 154]-[156][176][177]

1.0

!

!

? ,,_.,

> [-.. o~

10-2

< O u t.)

104

L) o~

104

1

I

I ..

6~

L

800

4O0

200

ANNEALING

TEMPERATURE

(~

(a) E E

I

I

I

SAMPLE-3

6 v u.I

o

l-. l.u

fls h0 < Z 0 0

0 500

.I 550

l

I

600

650

ANNEAL 1NG TEMPERATURE

700 ('C)

(b) Figure 21. (a) Specific contact resistance for conventional Au-Ge film (filled circle) or Gedoped isothermally regrown GaAs (from Ref. 80), (b) contact resistance from Ge/Ni/GaAs after annealing at various temperature for 5 minutes (from Ref. 21). Note U-shaped behavior of specific contact resistance with respect to annealing temperature.

Ohmic Contacts to II-VI and III-V Compounds 125

2.4

Summary

The formation of ohmic contacts to GaAs has been often studied and the model of their formation has undergone considerable evolution. Beginning with the empirical observation that Au-Ge/Ni formed a good ohmic contact to n-GaAs for Gunn diodes, the mechanism to explain this contact has evolved from one of surface tension and Au reaction, to one postulating doping due to a vacancy flux, to one of epitaxial regrowth after dissociation of the GaAs wafer. Each of the constituents of the contact, chosen empirically, contributes critically to the development of low resistance contacts. The role of Ni is to dissociate the GaAs lattice and to bond with both the Ga and As, often as a ternary phase, te keep them at the interface. Shortly after dissociation of the lattice, Ge reacts with the Ni bonded to Ga and/or As and forms NiGe, causing release of the Ga and As with Ge present to force regrowth of doped GaAs. Au helps drive the reaction towards completion by forming Au-Ga solid solutions. Based on this sequence of events, it is possible to develop rules for the formation of ohmic contacts to those semiconductors requiting heavy surface doping (e.g., when their Fermi level is pinned). The rules may be stated as follows: 1. The initial reaction between the semiconductor and metallization should dissociate the compound semiconductor. 2. The metallization should react with all elements in the compound semiconductor and hold them near the interface. 3. The metallization should have a subsequent reaction which leads to regrowth of the compound semiconductor. 4. A dopant or dopants should be present in the metallization. 5. The metallization should control the Fermi level during regrowth to ensure that dopant is incorporated on the proper lattice sites to result in a free carder density sufficient to yield low resistance ohmic contacts by tunneling or thermionie field emission transport. By following these rules, ohmic contacts should be possible for a large number of compound semiconductors.

126 3.0

Wide Bandgap Semiconductors O H M I C C O N T A C T S T O InP

InP has higher peak and saturation velocities of electrons than GaAs, rendering InP a potentially better material for high-speed and microwave devices.[ 178][179] However, InP is significantly more difficult to process since it decomposes at ~550~ versus GaAs at-650~ Due to the lower melting temperature of InP, metal/InP reactions are more rapid and morphology control can be more difficult. Since contact schemes for InP are similar to those being used for GaAs, it is worthwhile to review the interfacial and metallurgical characteristics and their effects on InP contacts. Katz [is0] reviewed the formation of ohmic contacts on InP based materials.

3.1

Single Element/InP Metallizations

Au/lnP Metallizations. Au is extensively used as a contact metal for InP, just as it is for GaAs devices. Upon annealing from just above room temperature to 400~ Au reacts with InP by In diffusing into Au to form Au-In solid solutions.[181]-[ 188] For example: Eq. (25)

Au + InP = Au-In + P

When the amount of indium in the Au films exceeds the solubility limits for solid solutions, a Au-In phase (Au3In) was formed.[ 182][1841[1851[187][189][19~ Eq. (26)

5Au + 3InP = Au3In + Au2P 3

It was suggested that phosphorus atoms either could leave the system or occupy nonlattice sites near the metal/InP interface.[ 185] It should be noted that Au did not react with As in Au/GaAs but that there was an extremely limited solubility of As in Au-Ga. The same may be true for P in Au-In. Annealing at 400~ for a few minutes changed the electrical property of the contact from non-ohmic to ohmic with-~l 0-4-10 "6~'~-cm2, which was attributed to formation of AuzP3 .[188][191] The surface morphology became rough when Au-In and AuzP 3 phases were formed.J182][ 185]Similar to Au/GaAs, the interface morphology was nonuniform due to formation of rectangular shaped reaction pits bounded by (111) planes.[185][Iss][ 1891The reaction pits were filled with Au-In.[ 1871Longer annealing times showed that Au-In solid solutions formed AuaIn upon cooling, and they were converted to Au9In4 as lateral spreading of the reaction zone took place.J188][19~

Ohmic Contacts to II-VI and III-V Compounds

127

At temperatures above -450~ Au2P 3 is dissociated and phosphorus evaporates as detected by a sharp evaporation peak at temperatures dependent upon the thickness of the Au film and independent of whether the substrate was n- or p-type.[ 184][187] Melting of the metallic layer was observed concurrent with the evaporation peak of phosphorus. For anneals above 450~ contact resistance increased by up to two order of magnitudes (to ~10 -4 ~-cm 2) upon dissociation of the Au2P3 phase.[191] Thus, the Au/InP and Au/GaAs interracial reactions are very similar to each other except that Au can form AUEP3 which decomposes above --450~ Ni/InP Metallizations. As deposited Ni/InP contacts exhibited rectifying behavior. The interfacial reaction in Ni/InP occurred upon deposition or after annealing below 250~ by Ni, producing amorphous NixInP, NiEP or NiaP.[192]-[196]Note that a rectifying ternary phase was formed near 250-300~ for Ni/GaAs, but it was crystalline. Annealing at higher temperatures above-300~ caused the amorphous phase to crystallize into a hexagonal NixInP (x-- 2.7), [192] monoclinic Ni2InP, and/or NiEP. [194][196]A low contact resistance (~10 -6 ~ - c m 2) was obtained with crystalline Ni2InP or Ni2P phases after annealing at 350~ and they were stable with time and temperature.[ 194][195][197][198]At 470 or 600~ for Ni-In-P, monoclinic Ni2InP was reported to be in equilibrium with InP at ~450~ after 3 month annealing.[ 1951This phase disappeared at 600~ At both temperatures, Ni2P, NisP 4, and NiP 2 were also in equilibrium with InP, but no Ni-In phases were found. [197] Ni2InP and InP had an epitaxial relationship with poor mismatch, and was suggested to be the cause of deviations from ideal I-V behavior after annealing at 5 0 0 ~ [199] It was suggested that the Ni/InP reaction proceeded from amorphization to segregation into Ni-P and Ni-In binary phases, and eventually recombination into the ternary phase.[ 1951With the Ni-P binary phases in contact with InP, contact resistances of 10-4-10 -6 f~-cm2 were measured. [~91 ][193][198] Pd/InP Metallizations. Pd/InP contacts were rectifying as deposited and an amorphous PdxlnP (x-- 3--4) ternary phase was reported. The PdxlnP remained amorphous but grew thicker at 175-225~176176176 Crys tallization into epitaxed cubic Pd2InP was observed upon annealing at 250~176176176 TM] Further annealing at 400~ caused decomposition of Pd2InP into Pdln and PdP 2. These two binary phases were stable after annealing at 500~ for 350 hours, but loss of P was detected when exposed to air or annealed at 650~ for 30 min. Thus Pdln and PdP 2 were reported to be the thermodynamic equilibrium phases with InP in a closed system at higher temperatures. The absence of a Pd-P phase was attributed to the

128

Wide Bandgap Semiconductors

sublimation of P due to its high vapor pressure.[2~1762~ Other bulk experiments to determine the Pd-In-P phase diagram identified several ternary phases, for example, Pd5InP, Pd3.sInP , and Pd5InEP2 (suggested to be PdEInP in the report by Ivey et al.). However, at 600~ Pd-P and Pd-In binary phases were found to be in equilibrium with InP. [203] It was noted that an important difference between Pd/InP reaction and Pd/GaAs reaction was that the initial PdxInP ternary phases produced by annealing at 300~ or below were amorphous, while PdxGaAs was hexagonal crystalline. Both systems have the same characteristics; the high temperature equilibrium phases in an open system are binary compounds, although the Pd-P and Pd-As phases are difficult to maintain because of evaporation. Ohmic behavior for Pd/InP was observed after annealing at ~300~ and was attributed to the presence of the PdIn phase.[ 2~ Particularly, PdIn/InP was ohmic as deposited with a contact resistance of 6 x 10.5 ~-cm2. [204]

3.2

Multi-element/lnP Metallizations

Pd/Ge/InP and Au/Pd/Ge/InP Metallizations. The use of Ge/Pd and Ge/Pd/Au for contact to InP followed the results for GaAs contacts. PdGe formed ohmic contacts to n-GaAs with an abrupt interracial morphology due to regrown GaAs.[ TM] The solid phase regrowth model has also been demonstrated for Ge/Pd/n-lnP,[ 2~ and was tested for the Ge/Pd/Z~ Pd/p-InP system.[ 2~ Regrown InP and PdGe and ohmic behavior were found after annealing above 400~ suggesting the following reaction:[ 2~ Eq. (27)

PdxInP(Ge, Zn) + Ge ~ InP(Ge, Zn) + PdGe

For Pd/Ge contacts, the relative amounts of Ge to Pd was a critical factor to lower contact resistance. When thin Ge layers were used (e.g., Pd/ Ge ratio > 2), the dominant metall~gical reaction products were PdP 2 and PdIn, and the contact resistance was -~10-5 ~-cm2.[ 2~ With Pd/Ge < 1, a contact resistance of--4 x 10-6 fl-cm 2 was obtained for annealing at 400450~ with only PdGe in contact with InP. Adding Au to Pd/Ge/InP alters the metallurgical reactions to form Au-In compounds at the metal/InP interface. Spatially, Au~0In3 was in contact with InP with an irregular interracial morphology. Aut0In 3, PdGe, and GeP were the reaction products after annealing at 325-450~ [207]Contact resistances of 2-4 x 10.6 ff)-cm2 was obtained at ~350~ The formation of GeP is different from GaAs since a GeAs phase was not reported.

Ohmic Contacts to 11-I/I and III-V Compounds

129

Au/Ni/InP and Au/Ge/Ni/lnP Metallizations. TEM images of Au/ Ni/InP after annealing at 250-400~ for 15 seconds revealed a layered structure of Aualn/NiP-NiP2/Au3In/InP.[ 2~ This indicates that Au diffused through the Ni layer and reacted preferentially with In, while Ni reacts with phosphorus at temperatures as low as 250~ (which is consistent with AES depth profiles). [2~ Gas mass spectrometry showed that the phosphorus loss was negligibly small, as compared to that from the Au/InP reaction metallurgical reactions near 400~ indicating that Ni captured P by forming Ni-P compounds.[ 21~ It was reported that these reactions resulted in ohmic contacts to n-type (doped to -- 1018 cm-3) without Ge[1991[21l] but non-ohmic contacts on p-InP.[ 2~ Presumably the ohmic contacts without Ge on n-InP results from dopant segregation to the interface during reactions, similar to Au/GaAs reactions.[ 93] Adding Ge to Au/Ni to form ohmic contacts to InP leads to the identical recipe of AuGeNi for n-GaAs contacts. With respect to the metallurgical reactions which lead to the ohmic contacts, below 250~ the formation of amorphous NixlnP and indiffusion of Au towards InP and outdiffusion of In toward the free surface are all observed.[ 212]Annealing at 250-350~ causes Ge to migrate into the Ni layer while Ni accumulates at the metal/InP interface, presumably dissociating the InP lattice to form amorphous NixlnP. Diffusion of Au and In through the intervening layers was also noticeable. [179][191][212][213]The contact resistance was reduced from 10-2 to 10-4-10-s f2-cm 2 during this reaction stage. [179][191][212] At 350~ 450~ where the lowest contact resistances were frequently measured, metallurgical reactions were more pronounced. Nickel accumulated at the InP interface as Ni2P, Ge reacted with Ni to form NiGe, and Au reacted to form Au-In compounds,[ 212][2141all of which are similar to the characteristics for AuGeNi/GaAs reactions.[179][2~ Spatially, the phases in contact with InP are Ni-P and Au-In, with Au-In and indium oxide phases at the metal/ambient interface. The Au-In phase at the InP interface causes irregular interfacial morphology. Elemental Ge coexisted with Au-In alloys at the metal/ambient interface rather than segregating at the metal/InP interface, again similar to GaAs with excess Ge.[ 212] 3.3

Mechanisms for Formation of Ohmic Contacts

As to the mechanism for ohmic contacts, the similarity to GaAs suggests strongly that the solid phase regrowth model should be applicable. In this model, again Ni dissociates the InP lattice and bonds to the P.

130

Wide Bandgap Semiconductors

A ternary phase may form at the interface, however, in the case of ternary decomposition to binary alloys, Au or a similar element may be necessary to hold the In at the interface. Formation of Au-Ni or NiGe then would cause regrowth of InP with incorporation of a dopant (Ge or Zn reported above). A significant difference between the two systems is that extremely low contact resistance (10 -7-10 -8~-em 2) was reported without using Ge,[ 215] particularly when Ni-P phases were formed at the interface with n-InP. The use of Ge is essential to formation of low resistance ohmic contacts to nGaAs. This difference with InP might result from stronger segregation of the dopant during metallurgical reactions, or since the Fermi level is only partially pinned for InP, formation of Ni-P may unpin the Fermi level and thereby significantly lower the barrier height. Thus, barrier lowering effects from unpinned Fermi levels may be an important factor in the formation of ohmic contact to InP.

4.0

O H M I C C O N T A C T S T O GaN

The situation for forming ohmic contacts to GaN is quite different from that for GaAs or InP in that Fermi level pinning does not occur. GaN devices are similar in this respect to the situation for ZnSe-based devices, as discussed below. Since the Fermi level is unpinned, formation of ohmic contacts can be predicted by the simple rule (Eq. 4) of the metal work function being less than the electron affinity of the semiconductor (~m < Xs) for an ohmic contact to n-GaN. [216][217] Since ~ G a N -- 4.14 eV, Foresi and Moustakis[ 218] showed that A1 (~gl - 4.08 eV) on n-GaN resulted in an ohmic contact, while Au contacts (~gu = 5.1 eV) were rectifying. Miller and Holloway also reported that Au formed rectifying contacts to nGaN.[216] Miller and Holloway studied Ag ( ( I ) g g = 4.3 eV) single component metallizations on n-GaN which resulted in weakly rectifying contacts in the as-deposited or heated to 500~ conditions. No interface reactions nor phase formation have been observed for either Au or Ag metallization on GaN epitaxial layers. However, Guo et al. showed that the Schottky barrier height for Pt and Pd varied with work function indicating a partially pinned surface level.[ 219] Wang et al. reported for the same system that the Schottky barrier height was the same for Pt and Pd.[ 220]These conflicting results clearly show that cleaning of the GaN surfaces is critical to successful and reproducible formation of ohmic contacts, as suggested by Murakami

Ohmic Contacts to 11-111and I I I - V Compounds

131

et al.[221]Thus, E r of GaN is regarded as unpinned, while partially pinned E F have been attributed to interfacial contamination. The characteristics ofmulticomponent contacts incorporating Ti are consistent with the conclusion of unpinned EF.[216][217][222]-[224]For pure Ti with a ~Ti - 4.3 eV, a weakly rectifying contact would be predicted. However, Ti will react with GaN to form TiN. With a theoretically predicted work function for TixN of 3.74 eV and reasonable electrical conductivities, this material should make a good ohmic contact to n-GaN. At the interface between a single or multicomponent metallization (e.g., Ti versus Ti/Pt/Au or Ti/A1/Ni/Au), TixN can be grown by interracial reactions during heating to temperatures ranging from 250~ (furnace anneal) to 900~ (rapid thermal anneal). The TiN interface phase is so thin that it is difficult to measure using the conventional thin film profiling techniques, such as AES, XPS, or SIMS. This interracial reaction product has only been suggested by limited depth profiling data and by ohmic electrical behavior after annealing. Very low specific contact resistances (~10-5-10 "7 f2-cm2) have been reported.j222][223]Ping et al. showed that A1 adjacent to n-GaN with a top Pd layer would also form an ohmic contact with a low specific contact resistance (10 -5 ~-cm 2) after heating to 650~ for 30 sec.[ 225] Even when the work function of the metallization is not appropriate, increased surface doping of n-GaN can be used to achieve ohmic contact. Miller and Holloway showed that incorporation of Si into a Au/Si/Ni metallization improved the performance.[ 2161Since the work function of both Au and Ni (5.2 eV) are greater than the electron affinity of GaN, neither should make an ohmic contact to n-type material. However, Si apparently helps create an n § surface layer which allows tunneling transport to result in an ohmic behavior. Another way to increase doping at the surface of n-GaN is to heat treat at high temperatures, resulting in the loss of nitrogen and formation of a shallow donor. As a result, Zolper et al. showed that high temperature heat treatment without a capping layer improved the properties of an A1/Ti ohmic contact, while it degraded the performance of a A u n t Schottky contact on n-GaN.[ 227] If the GaN surface was capped with an A1N layer before annealing, loss of N was prevented, the ohmic contact was worse, and the Schottky contact was constant. Lester et al.,[ 228] Cole et al.,[ 2291 Binari et al.,[ 23~ Lin et al.,[223] Revimov et al.,[ TM] and Smith et al.[232] all reported that formation of N vacancies during heat treatment or by interfacial phases increased the free electron concentration near the n-GaN interface and assisted in formation of ohmic contacts. Finally, Ingerly et al. heated Ptln 2

132

Wide Bandgap Semiconductors

surface layers on GaN and reported that decomposition of the surface layer resulted in formation of InGaN narrower bandgap material. A specific contact resistance of < 10-3 f2-cm 2 was reported.[ 233] In contact studies to n-GaN, thermal stability of the contacts has only been tested in a cursory fashion. Dt~bha et al. studied the contact resistance, topography, and composition profiles for Ti/Pt/Au, Au/Ge/Ni, A u ~ e and WSi x on GaN or In0.sGa0.sN thin films.[ 217] The WSi x contacts were found to exhibit excellent thermal stability (smooth surface morphology, no interdiffusion, and low contact resistance) even for heat treatments up to 800~ The minimum, specific contact resistance on In0.5Ga0.sN was 1.5 x 10-5 f~-cm 2. The Ti/Pt/Au, Au/Ge/Ni, and A u ~ e contacts showed much lower thermal stabilities, with Au/Be being unstable at temperatures as low as 400~ The instabilities were largely evident from changes in the surface morphologies resulting from voids or islands of metallization (from surface capillarity forces). There was no evidence for formation ofinterfacial phases, even though the interface width measured from AES depth profiles became broader, presumably due to roughening of the interface. In contrast to n-GaN, a metal or compound with a very large work function would be required to make an ohmic contact with p-GaN (electron affinity plus bandgap of 4.1 eV + 3.4 eV = 7.5 eV) as discussed above. Consistent with expectation, studies of the Schottky barrier height for Pt, Ni, Au, and Ti to p-GaN[ 214] showed that all contacts were rectifying and the barrier height decreased with increasing work function. In addition, Ishikawa et al. showed that the interfacial barrier height for contacts of Pt, Ni, Pd, Au, Cr, Ti, A1, and Ta also decreased as the work function increased. [221] Both studies are consistent with unpinned E F or only partially pinned E F. Since there are no metals with a work function >7.5 eV, ohmic contacts to p-GaN must be formed by either a compound with a very large work function/electron affinity, by doping the surface region to p++ with a lowered surface barrier, or by a graded surface barrier normally created during the last phases ofMBE, MOMBE, or MOCVD growth of the GaN epilayer. Nakamura et al. used Au or Au/Ni contacts to p-GaN for a demonstration of LEDs and electrically injected LDs based on GaN. [234][235] Khan et al. also reported using Au/Ni contacts on p-GaN. Other contacts metallizations to p-GaN include Ni by Sakai et al., [236] Au/Zn by Kuga et al.,[ 237] and Ti/Mo/Au by Goldenberg et al. [238] Trexler et al. studied the interracial reactions between Au, Au/Ni, Au/C/Ni, and Au/Cr metallizations on GaN doped with Mg (free hole density of 5 x 1016 to 2 x 1017cm 3)

Ohmic Contacts to II-VI and III-V Compounds 133 at temperatures up to 900~ [239] The Au contacts were rectifying and remained so even after heat treatment to temperatures up to 600~ There was very little interaction between the Au and GaN and no evidence for formation of an interfacial reaction phase or for diffusion of Au into the epilayer. For Au/Ni and Au/C/Ni metallizations, Ni dissociated the GaN lattice and led to a lower barrier height at the interface.[ 24~ After heat treatment at 600~ the Au and Ni had interdiffused and the width of the Ni/GaN interface had increased significantly, indicating dissociation of the GaN lattice. The reaction products of this dissociation are uncertain since there is low solubility of N in Ni, even though Ni will combine with Ga to form a solid solution and several binary intermetallic solid phases. While the Au/Ni contacts never achieved linear I-V ohmic behavior, the reverse bias breakdown voltage dropped from near 3 V to less than 0.5 V. Trexler et al. speculated that interracial carbon from contamination acted as a dopant and was incorporated into the lattice during this interfacial reaction.[ 24~ Abemathy et al. showed that C can act as an acceptor in GaN, thus it is reasonable to speculate that interfacial contamination detected by both AES and SIMS could lead to a higher free hole concentration at the interface.[ TM] Therefore, the contact structure of Au/C/Ni, where 10 nm of C was evaporated between deposition of Au and Ni, was tested to determine if increased doping and better ohmic contacts would be observed. As shown in Fig. 22, the current-voltage data for this contact scheme exhibited a higher resistance than did the AtffNi scheme. Obviously an evaporated film of C at the metal/semiconductor interface did not improve the ohmic behavior. This could result from a number of factors, including the fact that a monolayer interfacial C contamination layer should be sufficient to saturate the doping effects. In addition, interfacial contamination C is largely "adventitious" organic and hydrocarbon molecules which are similar to the C source used by Abemathy et al. to demonstrate p-doping of GaN. The C present from evaporation of a thin film is bound primarily as C sp 2 in a ring structure, which is notoriously poor as a doping source during thin film growth. Trexler et al. also studied the formation of Au/Cr contacts to p-GaN and found better linear I-V data than for Au/Ni or Au/C/Ni. Contacts were rectifying as deposited with a reverse bias breakdown voltage of ~1.5 V.[239] Upon annealing at 200~ or 400~ for 5 min, there was a slight decrease in conductivity across the contacts and samples. However after an RTA at 900~ for 15 sec., the I-V data were very linear with large current flow across the interfaces. Auger analysis showed very little reaction between

134

Wide Bandgap Semiconductors

the Cr and GaN at 200~ or 400~ but extensive dissociation of the GaN lattice and formation of a Cr-N-Ga ternary phase at the interface after the 900~ 15 see. RTA. Thus, Au/Cr appears to be a better contact metallization than does Au/Ni for p-GaN. While the specific contact resistance was not measured, it is expected to be very high compared to the values for ZnSe or GaAs (i.e., much higher than ~10 -5 f)-cm2). 2e-4

2e-4

le-4

<

. . . . . . . . . . . . . . . . .

4 . . . . . . . . . . . . . . . . . .

9. . . . . . . . . . . . . . . . .

5e-5

! i

C . . . . . . . . . . . . . . . . . .

!. . . . . . . . . . . . . . . . . .

9

- 5 e - 5 -~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : 9 Ni/C/~Au as-de;~:slted i : ! { 9 NilC/~Au 200*C,i 5 n~n -le-4

.~ .................

o ..................

~ ......

9

......

9 Ni/C/..Au 400"C, i 5 rain

O

Ni/A~

as-depos./ted

.: ....................

-2e-4

. . . . . .

i

:

O

Ni/A% 400~

5i re,in

-2e-4 -6

-4

-2

0

Voltage

2

4

6

(V)

Figure 22. Current versus voltage for A u ~ i and Au/C/Ni contacts deposited on p-GaN and heated for the temperatures and times shown. While the reverse bias breakdown voltages for the contacts are reduced by heat treatment, the contacts remain rectifying. See text for explanation (from Ref. 240).

5.0

OHMIC CONTACTS TO ZnSe

As discussed in the introduction, the largely ionic bonding and large AX for ZnSe based systems result in an unpinned E F .[242][243][244] Consequently, formation of ohmic contacts to ZnSe can be predicted by the simple rules discussed in the introduction, based on the value of the

Ohmic Contacts to II- VI and III- V Compounds

135

metallization work function versus the semiconductor electron affinity. Formation of an ohmic contact to n-ZnSe is relatively easy since the electron affinity is 4.09 eV, and n-ZnSe can be doped to free electron densities > 1019 em -3. Therefore it is easy to select a metal with a work function less than or nearly equal to 4.09 eV and achieve an ohmic contact, which can be improved by high surface doping (i.e., creation of an n + surface layer). For example, ohmic contacts to n-ZnSe with In (~In = 4.2 eV) and In-Sn alloys plus with Ti/Pt/Au alloys (r = 4.3 eV) have been demonstrated.[245][246][247] Even with an unpinned E F, heating of a deposited metal layer is normally required before ohmic behavior is observed.[ TM] Heating presumably assists an interfaeial reaction whose purpose is to eliminate or penetrate any interfaeial contamination layer which may be present. Liu et al. have discussed the surface preparation methods for ZnSe to minimize this eontamination.[ 248] Even with good cleaning, diffusion to eliminate or negate such an interfacial dielectric layer is commonly required because as-deposited contacts of In or Ti/Pt/Au on n-ZnSe were rectifying. After heat treatments at T > 200~ the contacts commonly switched from rectifying to ohmic. SIMS profiles of the In showed that the interface between the In and ZnSe was now very diffuse. This would be consistent with interdiffusion of In and/or roughening of the metallization thickness due to capillarity forces or displacement of the native ZnO or Zn(OH)2 present on the etched ZnSe wafer.[ 248] An upper limit to the specific contact resistance of In on n-ZnSe was reported by Wang and Holloway to be 10-3 g~-em2. They also reported that contacts heated to T > 350~ showed increased resistance.[ 247] In the ease of ohmic contacts to p-ZnSe, the situation is completely different, even though the Fermi level still appears to be unpinned. [245][246] In this ease, the sum of the electron affinity (4.09 eV) and the bandgap (2.67 eV) is so large (6.76 eV) that it is impossible to find a metal with a sufficiently large work function to create an ohmic contact, exactly analogous to p-GaN. As a result, all studies which have attempted to use simple metals or multiple layer metallization schemes have failed. For example, Fijol et al. showed that Au and Ag exhibited Schottky barrier contacts with minimum breakdown voltages of 3.0 and 2.3 eV, respeetively.[2491[25~ Oxygen helped lower the breakdown voltage, especially for Ag, and Akimoto et al. used an oxygen plasma to achieve a psuedo-ohmic contact using Au metallization.[ TM] Chen et al. only achieved a reduced breakdown voltage by placing Se monolayers at the Au/p-ZnSe interface.[ 243]

136

Wide Bandgap Semiconductors

Simple metallization schemes should, in principle, still be able to yield an ohmic contact, even with the mismatch between work function and electron affinity plus bandgap, if they could be used to create a p++ doped surface layer. However the maximum free hole concentrations achieved to date in ZnSe have been only at the mid 1017 em -3 level,[ 252] which is not sufficient to allow conversion of the interface transport from thermionie to thermionic field, or simple field emission. In addition, p-type conversion has only regularly and reliably been achieved during molecular beam epitaxy (MBE) growth of ZnSe. No dopant or regrowth scheme has been found which will force increased surface doping during contact formation due to interfacial reactions. Because of solubility limits and self compensation mechanisms for the common p-type dopants in ZnSe, it seems unlikely that such reaction/regrowth/doping systems can be found for p-ZnSe. The most successful ohmic contacts to p-ZnSe have been formations of multi-quantum well structures of p+-ZnTe/p-ZnSe.[253]-[255] The II-VI semiconductor p-ZnTe was used in these quantum well contacts because it can be doped to very high free hole densities during growth, and the band edges of ZnTe and ZnSe align properly for formation of resonant tunneling bandgap states. Therefore, formation of an ohmic contact to p-ZnTe is required to complete the ohmic contact to p-ZnSe. Trexler and Holloway showed that Au could be used to make good ohmic contacts to p-ZnTe with free hole concentrations of 3 x 1018 em-3.[256][257] As deposited, the Au formed a rectifying contact, but heat treatment at 200~ led to an ohmic contact due to disruption of the ZnTe interfacial contamination layer plus Au indiffusion without formation of an interface phase. For heat treatment above 250~ the contact resistance increased and an interface phase was detected by Auger and SIMS depth profiles. The quality of the ohmic contact was improved eonsiderabiy ifbilayer (Au~d), trilayer ( A u ~ t ~ d , Au~t/Ti, or Au/Mo/Pd), or quaternary layer (Au/Pt/Ti~i) metallization schemes were used on p-ZnTe, as shown by Kim et a1.,[258][259] Moehizuki et al.,[ 260] Ohtsuka et al.,[ 261] and by Ozawa et al. [262]Using two and three layer metallizations, specific contact resistances as low as 6 x 10-6 f2-em 2 were reported due to both increased surface doping ofp-ZnTe and reduced Sehottky barrier heights.[ 258] In an attempt to bypass the requirement of high surface doping on pZnSe, Lansari et al. used HgSe layers grown by MBE on p-ZnSexTe~_x layers to form low barrier contacts which were sometimes labeled "ohmic. ''[263][264]Films of HgSe form low barrier contacts to p-ZnSexTel, x because it is a semimetal with an electron affinity of about 6.1 eV. The

Ohmic Contacts to II-VI and III-V Compounds 137 predicted interfacial barrier for this film on p-ZnSe was 0.6 eV, and Fijol et al.[265]measured the barrier to be 0.5 5 eV. By addition of Te (ZnSel.xTex where x = 0.2) to the top-most epilayer, Lansari reduced the interfacial barrier to about 0.4 eV. This still leads to large contact resistances which result in high operating voltages and high junction temperatures for ZnSebased LEDs and diode lasers.[245][246][266] In addition, formation of the contact during MBE growth required deposition of rig in the system. Fijol et al. demonstrated an ex situ method for forming HgSe contacts to p-ZnSe using MBE grown p-ZnSe capped with a thick amorphous Se layer. [265] However the interfacial barrier is still too high to consider this an ohmic contact. Such an in situ capping procedure followed by the ex situ growth of the semiconductor contacting layers is a general scheme for ohmic contacts. For example, Ge-doped n§ for ohmic contacts was grown at the interface of As, or Ga capped GaAs, under a Ga or As flux, respectively, as reported by Li and Holloway.[ 77][118]In general, it is possible to use capped epitaxial layers and in situ or ex situ growth/regrowth to form either doped regions for ohmic contacts, or to form layers (both epitaxial and non-epitaxial) which do not exhibit interfacial contamination and therefore lead to ohmic contacts with low specific contact resistance.

6.0

CONCLUSIONS

Procedures to form ohmic contacts to GaAs, InP, GaN, and ZnSe were reviewed. For GaAs and InP, where Fermi level pinning has a significant effect on ohmic contacts, incorporation of dopants upon epitaxial regrowth of the semiconductor is the dominant method of forming ohmic contacts. The epitaxial regrowth of doped GaAs or InP is achieved by interfacial reactions which dissociate the semiconductor lattice. Subsequent reactions decomposed the phases bonded to the semiconductor elements and allow them to epitaxially regrow in the solid phase. General guidelines for the use of this technique for solid phase regrowth were given and are expected to be applicable to many semiconductor systems. These guidelines are as follows: 1. The initial reaction between the semiconductor and metallization should dissociate the compound semiconductor.

138

Wide Bandgap Semiconductors 2. The metallization should react with all elements in the compound semiconductor and hold them near the interface. 3. The metallization should have a subsequent reaction which leads to regrowth of the compound semiconductor. 4. A dopant or dopants should be present in the metallization. 5. The metallization should control the Fermi level during regrowth to ensure that the dopant is incorporated on the proper lattice sites to result in a free carrier density sufficient to yield low resistance ohmic contacts by tunneling or thermionic field emission transport.

For GaN and ZnSe, the Fermi level is not completely pinned and therefore it should be easier to form ohmic contacts. This is true for n-type materials where metals with work functions less than the electron affinities result in ohmic contacts, often after heat treatment to penetrate interracial contamination layers. However, formation of ohmic contacts to unpinned p-type surfaces is difficult, since the sum of the semiconductor work function and electron affinity is larger than the work function of any known metal. In this case, graded bandgaps or multi-quantum well contacts have proven best for ZnSe. Contacts to p-GaN are still being developed.

ACKNOWLEDGEMENT This work was supported by ONR Grant N00014-92-J-1895 and AFOSR Grant F49620-96-1-0026.

REFERENCES 1. Haigh, D. and Everad, J., GaAs Technology and Its impact on Circuits and Systems, Peter Peregrinus Ltd., London, UK (1989) 2. Ali, F., and Gupta, A., HEMTs and HBTs: Devices, Fabrication, and Circuits, Artec House, Massachusetts (1990) 3. Feng, M. S., Hsin, Y. M., and Wu, C. H., Mat. Res. Soc. Symp.Proc., 300:61 (1993) 4. Tell, B., Andrew S. H., Liao, K. F., and Bridges T. J., IEEE Trans on Elec. Dev., ED-32:2319-2321 (1985)

Ohmic Contacts to II- VI and III- V Compounds

139

5. Heime, K., InGaAs Field-Effect Transistors, Research Studies Press LTD, England (1989) 6. Christianson, K. A., Moglestue, C., and Anderson, W. T., Solid-State Elec., 38:1623 (1995) 7. Anderson, W. T., and Christou, A., Semiconductor Device Reliablity, p. 423, Kluwer, Boston (1990) 8. Barnes, A., SPIE Journal, 1632:98 (1992) 9. Wolfe, C. N., Holonyak, N., and Stillman, G. E., Physical Properties of Semiconductors, p. 59, Prentise-Hall, NY (1989) 10. Braslau, N, Gunn, J. B., and Staples, J. L., Solid-State Elec., 10:381(1967) 11. Gupta, R., Khokle, W. S, Wuerfl, J., and Hartnagel, H. L, J. Electrochem. Soc., 137:631 (1990) 12. Ball, R. K., Thin Solid Films, 176:55 (1989) 13. Wilier, J., Ristow, D., Kellner, W., and Oppolzer, H., d. Electrochem. Soc., 135:179 (1988) 14. Hugon, M. C, Agius, B., and Vamiere, F., J. Appl. Phys., 72:3570 (1992) 15. Kolqwa, E., Nieh, C. W., Flick, W., Molarius, J., and Nicolet, M.-A., Mat. Res. Soc. Proc., 126:289 (1988) 16. Rideout, V. L., Solid-State Elec., 18:541 (1975) 17. Piotrowska, A., Guivarc'h, A., and Pelous, G., Solid-State Elec., 26:179 (1983) 18. Shen, T. C., Gao, G. B., and Morkoc, H., J. Vac. Sci. Technol. B, 10:2113 (1992) 19. Mehdi, I., Reddy, U. K., Oh, J., East, J. R., and Haddad, G. I., d. Appl. Phys., 65:868 (1989) 20. Murakami, M., Price, W. H., Greiner, J. H., and Feder, J. D., J. Appl. Phys., 65:3546 (1989) 21. Tanahashi, K., Takata, J. J, Otuki, A., and Murakami, M., J. Appl. Phys., 71:4183 (1992) 22. Brillson, L. J., Contacts to Semiconductors:Fundamentals and Technology, Noyes Publ., Park Ridge, NJ (1993) 23. Marshall, E. D., and Murakami, M., Contacts to Semiconductor: Fundamentals and Technology, Ch. 1, (L. J. Brillson, ed), Noyes Publ., Park Ridge, NJ (1993) 24. Sze, S. M., Physics of Semiconductor Devices, p. 78, Jon Wiley & Sons Inc., New York (1981) 25. Sze, S. M., Physics of Semiconductor Devices, p. 254, Jon Wiley & Sons Inc., New York (1981) 26. Sah, C.-T., Fundamentals of Solid State Electronics, p. 488, World Scientific Publishing Co., River Edge, NJ (1991)

140

Wide Bandgap Semiconductors 27. Flower, R. H., and Nordheim, L., Pro. Roy. Soc., A, 119:173 (1928) 28. 29. 30. 31.

Stratton, R., Proc. Roy. Sot., B, 68:746 (1955) Kroger, F. A., Dimer, G., and Klasens, H. A, Phy. Rev., 103:279 (1956) Stratton, R., Phys. Rev., 125:67 (1962) Conley, J. W., Duke, C. B., Mahan, G. D., and Tiemann, J. J., Phys. Rev., 150:466 (1966) 32. Chang, C. Y., Fang, Y. K., and Szc, S. M., Solid-State Elec., 14:541 (1971) 33. 34. 35. 36.

Padovani, F. A., and Stratton, R., Solid-State Elec., 9:695 (1966) Cox, R. H., and Strack, H., Solid-State Elec., 10:1213-1218 (1967) Bergcr, H. H., Solid-State Elec., 15:145-158 (1972) Proctor, S. J., and Linholm, C. W., IEEE Elec. Dev. Lett., EDL-3:67 (1982) 37. Henry, H. G., IEEE Trans. Elec.Dev., 36:1390-1393 (1989) 38. Harrison, H. B., and Reeves, G. K., Mat. Res. Soc. Symp. Proc., 260:31-41 (1992) 39. Kovacs, B. and Mojzes, I., IEEE Trans. Elec. Dev., ED-33:1401-1403 (1996) 40. Bardeen, J., Phy. Rev., 71:717 (1947) 41. Cowley, A. M., and Sze, S. M.,J. Appl. Phys., 36:3212 (1965) 42. Spicer, W. E., Chye, P. W., Skeath, P. R., Su, C. Y., and Lindau, I., J. Vac. Sci. Technol., 16:1422 (1979) 43. Kurtin, S., McGill, T. C., and Mead, C. A., Phys. Rev. Lett., 22:1433 (1969) 44. Schluter, M., Phy. Rev. B, 17:5044-5047 (1978) 45. Gudat ,W., and Eastmen, D. E.,J. Vac. Sci. Technol., 13:769 (1976) 46. Spicer, W. E., Lindau, I., Gregory, P. E., Garner, C. M., Pianetta, P., and Chye, P., J. Vac. Sci. Technol., 13:780 (1976) 47. Tang, S. Y., Mei, W. N., and Xu, G.,J. Vac. Sci. Technol. B, 2:393 (1984) 48. Schilfgaarde, M. V., and Newman, N., Phys. Rev. Lett., 65:2728 (1990) 49. Spiccr, W. E., Lindau, I., Skeath, P., and Su, C. Y., J. Vac. Sci. Technol., 17:1019(1980) 50. Heine, V., Phys. Rev. A, 138:1689 (1965) 51. Monch, W., Semiconductor Surface and Interfaces, Ch. 6, SpringerVerlag, Berlin (1993) 52. Monch, W., J. Vac. Sci. Technol. B, 6:1270 (1988) 53. Cao, R., Miyano, K., Kendelewicz, T., Lindau, I., and Spicer, W. E., Phy. Rev. B, 39:12655 (1989)

Ohmic Contacts to II- VI and III- V Compounds 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75.

76. 77.

141

Andrews, J. M. and Phillips, J. C., Phys. Rev. Lett., 56:35 (1975) Freeouf, J. L., Solid-State Comm., 33:1059 (1980) Freeouf, J. L., and Woodall, J. M.,Appl. Phys. Lett., 39:727 (1981) Brillson, L. J.,J. Vac. Sci. Technol., 15:1378 (1978) Brillson, L. J., Chang, S., Shaw, J., and Viturro, R. E., Vacuum, 41:1016 (1990) Walukiewicz, W., J. Vac. Sci. Technol. B, 5:1062 (1987) Walukiewicz, W., Phy. Rev. B., 37:4760 (1988) Walukiewicz, W., Appl. Phys. Lett., 54:2094 (1989) Zhang, S. B., and Northrup, J. E., Phy. Rev. Lett., 67:2339 (1991) Sankey, F., Allen, R. E., Ren, S.-F., and Dow, J. D., J. Vac. Sci. Technol. B, 3:1162 (1985) Spicer, W. E., Newman, N., Kendelewicz, T., Petro, W. G, Williams, M. D., McCants, C. E., and Lindau, I., J. Vac. Sci. Technol. B, 3:1178 (1985) Tersoff, J., Metallization and Metal-Semiconductor Interfaces, (I. P. Batra, ed.), Plenum Press, New York (1988) Spicer, W. E., and Green, A. M.,J. Vac. Sci. Technol. B, 11:1347 (1993) Barnes, P. A. and Cho, A. Y.,Appl. Phys. Lett., 33:651 (1978) Kirchner, P. D., Jackson, T. N., Dpettit, G., and Woodall, J. M., Appl. Phys. Lett., 47:26 (1985) Stall, R. A., Wood, C. E. C, Board, K., Dandekar, N., Eastman, L. F., and Devlin, J.,J. Appl. Phy., 52:4062 (1981) Ballmgall, J. M., Wood, C. E. C., and Eastman, L. F., J. Vac. Sci. Technol. B, 1:675-681 (1983) Devlin, W. J., Wood, C. E. C., Stall, R., and Eastman, L. F., Solid-State Elec., 23:823 (1980) Woodall, J. M., Freeouf, J. L., Pettit, G. D., Jackson, T., and Kirchner, P., J. Vac. Sci. Technol., 19:626 (1981) Kumar, N. S., Chyi, J.-I., Peng, C. K., and Morkoc, H., Appl. Phys. Lett., 55:775 (1989) Fisher, V., Holloway, P. H., Ristolainen, E., and Schoenfeld, D., J. Vac. Sci. Technol. B, 12:1419 (1994) Fisher, V., Viljoen, P. E., Ristolainen, E., Hollway, P. H., Lampert, W. V., Haas, T. W., and Woodall, J. M., Advanced Metallizations for Devices and Circuits-Science, Technology and Manufacturability, (S. P. Muraka, A. Katz, K. N. Tu, and K. Maex, eds.), Mat. Res. Soc. Symp. Proc. 337:413, Pittsburgh, PA (1994) Woodal, J. M., Metal Contacts to III-VSemiconductors, p. 223 (1986) Li, B., and Holloway, P. H.,J. Appl. Phys., 71:4385 (1992)

142

Wide B a n d g a p S e m i c o n d u c t o r s

78. Gyulai, J., Mayer, J. W., Rodriguez, V., Yu, Y. C., and Gopen, H. J., J. Appl. Phys., 42:3578 (1971). 79. Leung, S., Yoshiie, T., Bauer, C. L., and Milnes, A. G., J. Electrochem. Soc., 132:898 (1985) 80. Holloway, P. H., and Mueller, C. H., Thin Solid Films, 221:254 (1992) 81. Mueller, C. H., Hollway, P. H., and Connell, R. G., Advanced Metallizations and Processing for Semiconductor Devices and Circuits II, (A. Katz, S. P. Muraka, Y. I. Nissim, and J. M. Harper, ed.) p. 481, MRS Pittsburgh, PA (1993) 82. Kinsbron, E., Gallagher, P. K., and English, A. T., Solid-State Elec., 22:517-524 (1979) 83. Weizer, V. G., and Fatemi, N. S., J. Appl. Phys., 64:4618-4623 (1988) 84. Mojes, I, and Veresegyhazy, R., Thin Solid Films, 144:29 (1986) 85. Kim, T., and Chung, D. D. L., Phil Mag. A, 62:283 (1990) 86. Kumar, K., Jpn. J. AppL Phys., 18:713 (1976) 87. Vandenberg, J. M., and Kinsbron, E., Thin Solid Films, 65:259 (1980) 88. Yoshiie, T., and Bauer, C. L.,J. Vac. Sci. Technol. A, 1:554 (1983) 89. Kumar, K., Jpn. J. ,4ppl. Phys., 18:713-716 (1976) 90. Pecz, B., Jaroli, E., Radnoczi, G., Veresegyhazy, R., and Mojzers, I., Phys. Stat. Sol., 94:507 (1986) 91. Bauer, C. L., Surface Sci., 168:395 (1986) 92. Liu, L.-M., and Holloway, P. H., Mat. Res. Soc. Proc., 142:607 (1992) 93. Holloway, P. H., Yeh, L. L-M., Powell, D. H., and Brown, A., Appl. Phys. Lett., 59:947 (1991) 94. Lince, J. R., Thomas, C., and Willimas, R. S., J. Mater. Res., 1:537 (1986) 95. Sand, T., Keramidas, V. G., Yu, A. J., Yu, K-M., Gronsky, R., and Washburn, J., J. Mater. Res., 2:262 (1987) 96. Solomon, J. S., and Smith, S. R., J. Vac. Sci. Technol. A, 5:1809 (1987) 97. Ogawa, M., Thin Solid Films, 70:181 (1980) 98. Chen, L. J., and Hsieh, Y. F., Proceedings of the 41st Meeting of the Electron Microscopy Society of America, p. 234, San Francisco (1983) 99. Lahav, A., Eizenberg, M., and Komem, Y., J. Appl. Phys., 60:991 (1986) 100. Sands, T., Keramidas, V. G., Washburn, J., and Gronsky, R., Appl. Phys. Lett., 48:402 (1986) 101. Sands, T., Chang, C. C., Kaplan, A. S., Keramidas, V. G., Krishnan, K. M., and Washburn, J., Appl. Phys. Lett., 50:1346-1348 (1986) 102. Sands, T., and Keramida, V. G., Thin Solid Films, 136:105-122 (1986)

Ohmic Contacts to II-I/I and III-V Compounds

143

103. Sands, T., Chang, C. C., Kaplan, A. S., Keramidas, V. G., Krishnan, K. M., and Washburn, J., Appl. Phys. Lett., 50:1346-1348 (1986) 104. Sands, T., Keramidas, V. G., Washbum, J., and Gronsky, R., Appl. Phys. Lett., 48:402 (1986) 105. Chen, S. H., Carter, C. B., and Palmstrom, C. J., J. Mater. Res., 3:1385 (1988) 106. Guivarc'h, A., Guerin, R., Caulet, J., Poudoulec, A., and Fontenille, J., J. Appl. Phys., 66:2129 (1989) 107. Lin, J.-C, Zheng, X.-Y., Hsieh, K.-C., and Chang, Y. A., Mat. Res. Symp. Proc., 102:233 (1988) 108. Kim, T.-J., Evolution of lnterfacial Phases and Their Effects on Ohmic Contacts to n-GaAs in Ni-Ge-Ti Metallizations, Ph.D. dissertation, University of Florida (1996) 109. Guerin, R., and Guivarc'h, A., J. Appl. Phys., 66:2122 (1989) 110. Sand, T., Keramidas, V. G., Yu, K. M., Washburn, J., and Krishnan, K., J. Appl. Phys., 62:2070(1987) 111. Lin, J.-C., Zheng, X.-Y., Hsieh, K.-C., and Chang, Y. A., Mat. Res. Soc, Symp. Proc., 102:233-236 (1988) 112. Zhao, Y. C., Wu, Z. Q., and Zhang, S. Y., Philosophical Magazine A., 60:63-81 (1989) 113. Kuan, T. S., Freeouf, J. L., Batson, P. E., and Wilkie, E. L., Jr. Appl.Phys., 58:1519-1526 (1985) 114. Cole, M. W., Han, W. Y., Casas, L. M., and Jones, K .A., J. Appl. Phys., 77:5225-5230 (1995) 115. Sands, T., and Keramida, V. G., Thin SoBd Films, 136:105-122 (1986) 116. Lin, J.-C., Hsieh, K.-C., Schulz, K. J., and Chang, Y. A., J. Mater. Res., 3:148-163 (1988) 117. Li, B., and Holloway, P. H., J. Vac. Sci. Technol. A, 9:944-948 (1991) 118. Zheng, X.-Y., Schulz, K. J., Lin, J.-C., and Chang, Y. A., J. the LessCommon Metals, 146:233-239 (1989) 119. Kim, K.-B., and Sinclair, R., Mat. Res. Soc. Symp. Proc., 148:21-27 (1989) 120. Ko, D.-H., and Sinclair, R., J. Appl.Phys., 72:2036-2042 (1992) 121. Che, J. S., Kolawa, E., Nicolet, M.-A., and Ruiz, R. P., J. Appl. Phys., 75:7373-7381 (1994) 122. Kim, T., and Chung, D. D. L., dr. Vac. Sci. Technol. B, 4:762 (1986) 123. Iliadis, A., and Singer, K. E., Solid-State Elec., 26:7 (1983) 124. Kulkarni, A. K. and Lukowski, J. T., J. Appl. Phys., 59:2901 (1986) 125. Ilieadis, A.,J. Var Sci. Technol. B, 5:1340 (1987) 126. Wang, L. C., Zhang, B., Gang, F., Marshall, E. D., Lau, S. S., Sands, T., and Kuech, T. F.,J. Mater. Res., 3:922 (1988)

144

Wide Bandgap Semiconductors

127. Tsuchimoto, J., Shikata, S., and Hayashi, H., J. Appl. Phys., 69:6556 (1991) 128. Anderson, W. T., Jr., Christou, A,. and Davey, J. E., J. Appl. Phys., 49:2998 (1978) 129. Grinolds, H. R., and Robinson, G. Y., Solid-State Elec., 23:973-985 (1980) 130. Chen, C. L., HoUis, M. A., Mahoney, L. J., Goodhue, W. D., Manfra, M. J., and Murphy, R. A., J. Vac. Sci. Technol. B, 5:902-907 (1987) 131. Marshall, E. D., Chen, W. X., Wu, C. S., Lau, S. S., and Kuech, T. F., Appl. Phys. Lett., 47:298-300 (1985) 132. Marshall, E. D., Zhang, B., Wang, L. C., Jiao, P. F., Chen, W. X., Sawada, T., Lau, S. S., Kavanagh, K. L., and Kuech, T. F., J. Appl. Phys., 62:942 (1987) 133. Kim, T.-J., and Holloway, P. H., unpublished. 134. Palmstrom, C. J., Schwarz, S. A., Yablonovitch, E., Harbison, J. P., Schwartz, C. L., Florez, L. T., Gmitter, T. J., Marshall, E. D., and Lau, S. S., J. Appl. Phys., 67:334 (1990) 135. Sands, T., Marshall, E. D., and Wang, L. C.,J. Mater. Res., 3:914 (1988) 136. Gupta, R. P., and Khokle, S., Solid-State Elec., 28:823 (1985) 137. Christou, A., Solid-State Elec., 22:141 (1979) 138. Marlow, G. S., Das, M. B., and Tongson, L., Solid-State Elec., 26:259 (1983) 139. Robinson, G. Y., Solid-State Elec., 18:331 (1975) 140. Iliadis, A., and Singer, K. E., Solid-State Comm., 49:99 (1984) 141. Relling, E., and Botha, A. P.,Appl. Surf. Sci., 35:380 (1989) 142. Ogawa, M., J. Appl. Phys., 51:406 (1980) 143. Shih, Y.-C., Murakami, M., Wilkie, E. L., and CaUegari, A. C., J. Appl. Phys., 62:582 (1987) 144. Patrick, W., Mackie, W. S., Beaumont, S. P., and Wilkinson, C. D. W., Appl. Phys. Lett., 48:986 (1986) 145. Kovacs, B., Mojes, I., Veresegyhazy, R., Nemeth-Sally, M., and Pecz, B., Vacuum, 40:129 (1990) 146. Buhlmann, H. J., and Llegems, M., J. Electrochem. Soc., 138:2795 (1991) 147. Chua, S.-J., and Lee, S.-H.,Jpn. J. Appl. Phys., 33:66 (1994) 148. Wittmer, M., Pretorius, R., Mayer, J. W., and Nicolet, M.-A., Solid-State Elec., 20:433 (1977) 149. Heiblum, M., Nathan, M. I., and Chang, C. A., Solid-State Elec. 25:185 (1982)

Ohmic Contacts to II- VI and III- V Compounds

145

150. Kuan, T. S., Batson, P. E., Jackson, T. N., Rupprecht, H., and Wilkie, E. L., J. Appl. Phys., 54:6952 (1983) 151. Han, W. Y., Lu, Y., Lee, H. S., Cole, M. W., Casas, L. M., DeAnni, A., Jones, K. A., and Yang, L. W., J.Appl.Phys., 74:754-756 (1993) 152. Bruce, R. A., Moore, W. T., Lester, T., Clark, D. A., and Springthorpe, A. J., Paper presented at Microscopy of Semiconducting Materials Conf., p. 671, Oxford (1989) 153. Schilfgaarde, M. V., and Newman, N., Phys. Rev. Lett., 65:2728 (1990) 154. Lampert, W. V, Haas, T. W., and Holloway, P. H., AdvancedMetallization and ProcessingofSemiconductor Devices, p. 941, (A. Katz, Y. I. Nissim, S. P. Murarka, and J. M. E. Harper, eds.) MRS, Pittsburgh, PA (1992) 155. Lampert, W. V., A Study of Aluminum-Germanium-Nickel Ohmic Contact Metallurgical Effects at the Gallium Arsenide Interface, Ph.D dissertation, University of Florida (1992) 156. Lampert, W. V., Haas, T. W., Lambers, E. S., and Holloway, P. H., Semiconductor Heterostructuresfor Photonic and Electronic Applications, pp. 695-700, (D. C. Houghton, C. W. Tu, and R. T. Tung, eds.), MRS Proc., Pittsburgh, PA (1993) 157. Fischer, S., Wetzel, C., Hailer, E. E., and Meyer, B. K., Appl. Phys.Lett., 67:1298-1300 (1995) 158. Kim, T.-J., Holloway, P. H., and Kenik, E. A.,AppL Phys.Lett., submitted. 159. Kim, T-J., and Holloway, P. H., unpublished. 160. Matard, A., Defect electronics in Semiconductor, Ch. 3, Wiley, New York (1971) 161. Ennen, H, Kaufmann, U., and Schneider, J., Appl. Phys. Lett., 38:355 (1981) 162. Hizem, N., Bremond, G., Mayet, L., Gavand, M., Gregoire, J., Guilot, G., and Ulrici, W.,Appl. Surf. Sci., 50:490 (1991) 163. Chang, J. C. P., Otsuka, N., Harmon, E. S., Melloch, M. R., and Woodall, J. M., Appl. Phys. Lett., 65:2801 (1994) 164. Walukiewicz, W., Mat. Res. Soc. Symp. Proc., 300:421, San Francisco (1993) 165. Otsuki, T., Aoki, H., and Kano, G., J.,4ppl.Phys., 63:2011-2014 (1988) 166. Jan, C.-H., Swenson, D., and Chang, Y. A., J.Appl.Phys., 68:6458-6462 (1990) 167. Oku, T., Wakimot, H., Otsuki, A., and Murakami, M., J. Appl. Phys., 75:2522(1994) 168. Wang, L. C., Wang, X. Z., Hsu, S. N., Lau, S. S., Lin, P. S. D., Sand, T., Schwarz, A. Z., Plumton, D. L., and Kuech, T. F., dr. Appl. Phys., 69:4364-4372 (1991) 169. Wilson, C., O'Neil, A., Baier, S., and Nohava, J., Mat. Sci. & Eng. B, 29:54 (1995)

146

Wide Bandgap Semiconductors

170. Takata, H. J., Tanahashi, K., Otuski, A., Inui, H., and Murakami, M., J. Appl. Phys., 72:4191-4196 (1992) 171. Kawata, H. R., Oku, T., Otsuki, A., and Murakami, M., ,/. Appl. Phys., 75:2530-2537 (1994) 172. Jian, P., Ivey, D. G., Bruce, R., and Knight, G., J. Electro. Mad., 23:953-962 (1994) 173. Kolqwa, E., Nieh, C.-W., Flick, W., Molarius, J., and Nicolet, M-A., Mat. Res. Soc. Proc., 126:289 (1988) 174. Gupta, R., Khokle, W. S., Wuerfl, J., and Hartnagel, H. L., J. Electrochem. Soc., 137:631 (1990) 175. Zuleeg, R., Friebertshauser, P. E., Stephens, J. M., and Watenabe, S. H., IEEE Electron Device Lett., EDL-7:603 (1986) 176. Lin, X. W., Lampert, W. V., Swider, W., Haas, T. W., Holloway, P. H., Washbum, J., and Liliental-Weber, Z., Thin Solid Films, 253:490 (1994) 177. Lin, X., Lampert, W. V., Haas, T. W., Holloway, P. H., Liliental-Weber, Z., Swider, W., and Washburn, J., J. Vac. Sci. Technol. B, 13:2081 (1995) 178. Itho, H., and Ohata, K., IEEE Trans. Elec. Dev., ED-30;811-815 (1983) 179. Del Alamo, J. A., and Mizutani, T., Solid-State Elec., 31:1635-1639 (1988) 180. Katz, A., Handbook of Compound Semiconductors." Growth Processing Characterization and Devices, (P. H. Holloway, and G. E. McGuire, eds.), Noyes Publ., Park Ridge, NJ (1995) 181. Hiraki, A., Shuto, K., Kim, S., Kammura, W., and Iwami, M., Appl. Phys. Lett., 31:611 (1977) 182. Piotrowska, A., Auvray, P., Guivarc'h, A., Pelous, G., and Henoc, P., J. Appl. Phys., 52:5112 (1981) 183. Camlibel, I., Chin, A. K., Ermanis, F., DiGiuseppe, M. A., Lourence, J. A., and Bonner, W. A., J. Electrochem. Soc., 129:2585-2590 (1982) 184. Mojzes, I., and Veresegyhazy, R., Thin Solid Films, 144:29 (1986) 185. Fatemi, N. S., and Weizer, V. G., J. Appl. Phys., 65:2111-2115(1989) 186. Fatemi, N. S., and Weizer, V. G., J. Appl. Phys., 67:1934 (1990) 187. Veresegyhazy, R., Pecz, B., Mojzes, I., and Gombos, G., Vacuum, 40:189 (1990) 188. Weizer, V. G. and Fatemi, N. S., J. Appl. Phys., 68:2275-2284 (1990) 189. Wada, O.,J. Appl. Phys., 57:1901 (1985) 190. Barnard, W. O., Malherbe, J. B., Auret, F. D., and Myburg, G., Thin Solid Films, 215:42-49 (1992) 191. Erickson, L. P., Waseem, A., and Robison, G. Y., Thin Solid Films, 64:421-426 (1979)

Ohmic Contacts to II- VI and III- V Compounds

14 7

192. Sands, T., Chang, C. C., Kaplan, A. S., and Keramidas, V. G., Appl. Phys. Lett., 50:1346-1348 (1987) 193. Appelbaum, A, Robbins, M., and Schrey, F., IEEE Trans. Elect. Dev., ED-34:1926-1031 (1987) 194. Stremsdoerfer, G., Wang, Y., Nguyen, D., Clechet, P., and Martin, J. R., J. Electrochem. Soc., 140:2022-2028 (1993) 195. Mohney, S. E., and Chang, Y. A.,J. Appl. Phys., 78:1343 (1995) 196. Persson, L, Bouanani, M. E., Hult, M., Whitlow, H. J, Andersson, M., Bubb, I. F, Johnston, P. N., Walker, S. R., Cohen, D. D., Dytlewski, N., Zaring, C., and Ostling, M., J. Appl. Phys., 80:3346-3354 (1996) 197. Mohney, S. E., and Chang, Y. A.,J. Mater. Res., 7:955-960 (1992) 198. Fatemi, N. S., and Weizer, V. G., J. Appl. Phys., 73:289-295 (1993) 199. Clausen, T., and Leistiko, O., Physica Scripta, T54:68-70 (1994) 200. Caron-Popowich, R., Washburn, J., Sands, T., and Kaplan, A. S., J. Appl. Phys., 64:4909 (1988) 201. Ivey, D. G, Jian, P., and Bruce, R., J. Electro. Matl., 21:831-839 (1992) 202. Stremsdoerfer, G., Calais, C., Martin, J. R., and Clechet, P., J. Electrochem. Sot., 137:835-838 (1990) 203. Suznne, E. M., and Chang, Y. A., Mat. Res. Syrup. Proc., 260:519-524 (1992) 204. Kuphal, E., Solid-State Elec., 24:69-78 (1981) 205. Schwarz, S. A., Palmstrom, C. J., Schwartz, C. L., Sands, T., Shantharama, L. G., Harbison, J. P., Florez, L. T., Marshall, E. D., Han, C. C., Lau, S. S, Allen, L .H., and Mayer, J. W., J. Var Sr Technol. A, 8:2079 (1990) 206. Park, M.-H., Wang, L. C., and Hwang, D. M., J. Electro. Matl., 25:721-725 (1996) 207. Jian, P., Ivey, D. G., Bruce, R., and Knight, G., J. Electro. Matl., 23:953-962 (1994) 208. Ivey, D. G., Bruce, R., and Piercy, G. R., J. Electro. Matl., 17:373-380 (1988) 209. Clausen, T., Leistiko, O., Chorkendorff, I., and Larsen, J. Appl. Surf Sr 74:287-295 (1994) 210. Mojes, I., Veresegyhazy, R., Kovacs, B., and Pecz, B., Thin Solid Films, 164:1-4(1988) 211. Fatemi, N. S., and Weizer, V. G.,Mat. Res. Sor Symp. Proc., 318:171-176 (1994) 212. Ivey, D. G., Wang, D, Yang, D., Bruce, R., and Knight, G., J. Electro. Matl., 23:441-446 (1994)

148

Wide Bandgap Semiconductors

213. Bahir, G., Merz, J. L., Abelson, J. R., and Sigmon, T. W., J. Electro. Mad., 16:257-262 (1987) 214. Morais, J., Fazan, T. A., Landers, R., and Sato, E. A. S., J. Appl. Phys., 79:7058-7061 (1996) 215. Fatemi, N. S., and Weizer, V. G., Mat. Res. Soc. Syrup. Proc., 260:537-542 (1992) 216. Miller, S., and Holloway, P. H., J. Elect. Mad., in press (1996) 217. Durbha, A., Pearton, S. J., Abemathy, C. R., Lee, J. W., and Holloway, P. H., J. Vac. Sci. Technol. B, 14:2582 (1996) 218. Foresi, J. S., and Moustakas, T. D., Appl. Phys. Lett., 62:2859-2861 (1993) 219. Guo, J. D., Feng, M. S., Guo, R. J., Pan, F. M., and Chang, C. Y., Appl. Phys. Lett., 67:2657 (1995) 220. Wang, L., Nathan, M. I., Lam, T. H., Khan, M. A., and Chen, Q., Appl. Phys. Lett., 68:1267 (1996) 221. Ishikawa, H., Kobayashi, S., Koide, Y., Yamasaki, S., Nagai, S., Umezaki, J., Koike, M., and Murakami, M., J. Appl. Phys., 81" 1315-1322 (1997) 222. Fan, Z., Mohammad, S. N., Kim, W., Aktas, O., Botchkarev, A. E., and Morkoc, H., Appl. Phys. Lett., 68" 1672-1674 (1996) 223. Lin, M. E., Ma, Z., Huang, F. Y., Fan, Z. F., Allen, L. H., and Morkoc, H., Appl. Phys. Lett., 64:1003-1005 (1994) 224. Luther, B. P., Mohney, S. E., Jackson, T. N., Khan, M. A., Chen, Q., and Yang, J. W., Appl. Phys. Lett., 15"57 (1997) 225. Ping, A. T., Khan, M. A., and Adesida, I., J. Electro. Mad., 25:819 (1996) 226. Jenkins, D. W. and Dow, J. D., Phys. Rev. B, 39:3317 (1989) 227. Zopler, J. C., Rieger, D. J., Baca, A. G., Pearton, S. J., Lee, J. W., and Stall, R. A., Appl. Phys. Lett., 69:538 (1996) 228. Lester, L. F., Brown, J. M., Ramer, J. C., Zhang, L., Hersee, S. D., and Zolper, J. C., Appl. Phys. Lett., 68:539 (1996) 229. Cole, M. W., Eckart, D. W., Han, W. Y., Pfeffer, R. L., Monahan, T., Ren, F., Yuan, C., Stall, R. A., Pearton, S. J., Li, Y., and Lu, Y., J. Appl. Phys., 80:278 (1996) 230. Rinari, S. C., Rowland, L. B., Kruppa, W., Kelner, G., Doverspike, K., and Gaskill, D. K., Electron. Lett., 30:1248 (1994) 231. Ruvimov, S., Liliental-Weber, Z., Washburn, J., Duxstad, K. J., Haller, E. E., Fan, Z.-F., Moharmned, S. N., Kim, W., Botchkarev, A. E., and Morkoc, H., Appl. Phys. Lett., 69" 1556 (1996)

Ohmic Contacts to II- I/I and III- V C o m p o u n d s

149

232. Smith, L. L., Bremser, M. D., Carlson, E. P., Weeks, T. W., Huang, Y., Kim, M. J., Carpenter, R. W., and Davis, R. W., Mat. Res. Soc. Symp. Proc., 395:861 (1996) 233. Ingerly, D. B., Chang, Y. A., Perkins, N. R., and Kuech, T. F., Appl. Phys. Lett., 70:108 (1997) 234. Nakamura, S., Mukai, T., and Seno, M., Jpn. 3. Appl. Phys., 30:L1998 (1991) 235. Nakamura, S., Mokia, T., and Seno, M.,Appl. Phys. Lett., 62:1786 (1996) 236. Sakai, H., Koide, T., Suzaki, H., Yamaguchi, M., Yamasaki, S., Koike, M., Amano, H., and Aksaki, I., Jpn. J. Appl. Phys., 34:L1429 (1995) 237. Kuga, Y., Shirai, T., Haruyama, M., Kawanishi, H., and Suematsu, Y., Jpn. J. Appl. Phys., 34:4085 (1995) 238. Golderberg, B., Zook, J. D., and Ulmer, R. J., Appl. Phys. Lett., 62:381 (1993) 239. Trexler, J. T., Electrical Contacts to p-Type Zinc Telluride and Gallium Nitride, Ph.D. dissertation, University of Florida (1997) 240. Trexler, J. T., Miller, S. J., Holloway, P. H., and Khan, M. A., GaN and Related Materials, (R. D. Dupuis, J. A. Edmond, F. A. Ponce, and S. Nakamura, eds.) 395:819, MRS, Pittsburgh (1996) 241. Abernathy, C. R., MacKenzie, J. D., Pearton, S. J., and Hobson, W. S., Appl. Phys. Lett., 66:1969 (1995) 242. Xu, F., Vos, M., Waver, H., and Cheng, H., Phys. Rev. B, 38:13418 (1988) 243. Chen, W., Khan, A., Soukiassian, P., Mangat, P. S., Gaines, J., Ponzon, C., and Olego, D.,J. Vac. Sci. Technol. B, 12:2639 (1994) 244. Holloway, P. H., Kim, T.-J., Trexler, J. T., Miller, S., Fijol, J. J., Lampert, W. V., and Haas, T. W.,Appl. Surf. Sci., 117/118:362 (1997) 245. Fijol, J. J., and Holloway, P. H., Crit. Rev. Sol. St. Matl. Sci., 21:77 (1996) 246. Holloway, P. H., Fijol, J. J., Park, R. M., Calhoun, L. C., Jones, K. S., Simmons, J. H., Zory, P., and Anderson, T. J., Proc. Twenty First Stateof-the-Art Prog. Compound Semiconductors (SOTAPOCS XXI), 94-34:2 (S. N. G. Chu, F. Ren, V. Malhotra, and D. P. Van, eds.), The Electrochemical Society, Pennington, NJ (1995) 247. Wang, Y.-X., and Holloway, P. H., Vacuum, 43:1149 (1992) 248. Liu, L.-M., Lindauer, G., Alexander, W. B., and Holloway, P. H., J. Vac. Sci. Technol. B, 13:2238 (1995) 249. Fijol, J. J., Calhoun, L. C., Park, R. M., and Holloway, P. H., Jr. Electron. Matl., 24:143 (1995)

150

Wide Bandgap Semiconductors

250. Fijol, J. J., Trexler, J. T., Calhoun, L., Park, R. M., and Holloway, P. H., J. Vac. Sci. Technol., 14:159 (1996) 251. Akimoto, K., Miyajima, T., and Mori, u J. Crystal Growth, 115:683 (1991) 252. Park, R. M., and Calhoun, L., Phys. Today, (1996) 253. Fan, Y., Han, J., He, L., Saraie, J., Gunshor, R., Hagerott, M., Jeon, J., Nurmikko, A., Hua, G., and Otsuka, N., Appl. Phys. Lett., 61:360 (1992) 254. Fan, T., Han, J., He, L., Saraie, J., Gunshor, R., Hagerott, M., Jeon, H., Nurmikko, A., Hua, G., and Otsuka, N., J. Vac. Sci. Technol. B, 11:1748 (1993) 255. Hiei, F., Ikeda, M., Ozawa, M., Miyajima, T., Ishibashi, A., and Akimoto, K., Electro. Lett., 29:878 (1993) 256. Trexler, J. T., Fijol, J. J, Calhoun, L. C., Park, R. M., and Holloway, P. H., J. Cryst. Growth, 159:723 (1996) 257. Trexler, J. T., Fijol, J. J, Calhoun, L. C., Park, R. M., and Holloway, P. H., J. Electron. Matl., 25:1474 (1996) 258. Kim, D.-W., Kwak, J. S., Park, H.-S., Kim, H. N., Lee, S.-M., Sim, C.-S., Noh, S.-K., and Baik, H. K., J. Electron. Matl., 26:83 (1997) 259. Kim, D.-W., Kim, H.-S., Kwak, J. S., and Baik, H. K., unpublished (1997) 260. Mochizuki, K., Jerano, A., Momose, M., Taike, A., Kawata, M., Gotoh, J., and Nakatsuka, S., J. Appl. Phys., 78:3216 (1995) 261. Ohtsuka, T., Yoshimura, M., Morita, K., Koyama, M., and Yao, T., Appl. Phys. Lett., 67:1277 (1995) 262. Ozawa, M., Hiei, F., Takasu, M., Isibashi, A., and Akimoto, K., Appl. Phys. Lett., 64:1120 (1994) 263. Lansari, Y., Ren, J., Sneed, B., Bowers, K., Cook, J., and Schetzina, J., Appl. Phys. Lett., 61:2554 (1992) 264. Lansari, Y., Cook, J., and Schetzina, J., J. Elect. Matl., 22:809 (1993) 265. Fijol, J. J., Trexler, J. T., Calhoun, L., Park, R. M., and Holloway, P. H., J. Vac. Sci. Technol., 14:159 (1996) 266. Kim, J. R., and Jones, K. S., Crit. Rev. Sol. St. Matl. Sci., 21:1 (1996)

4 D r y E t c h i n g of SiC Joseph R. Flemish

1.0

INTRODUCTION

As the quality of semiconducting silicon carbide improves, there have been continued efforts toward developing commercially viable electronic components from this wide-bandgap semiconductor with a particular focus on high-temperature power-switching applications. The two SiC polytypes which have received the greatest attention for power devices are 6H and 4H. Much of the work in SiC devices in the late 1980's and early 1990' s employed the 6H polytype owing to the development and commercial availability of single-crystal wafers of sizes greater than one inch in diameter. More recently, the 4H polytype has displaced 6H for electronic device fabrication due to its superior electron mobility combined with advances in 4H-SiC boule and epilayer growth. Realization of advanced device structures fabricated from SiC require the ability to etch this material smoothly, controllably, and with a minimal amount of damage to the underlying material. Methods for dry etching SiC are necessitated by the lack of wet etching techniques which are compatible with SiC device fabrication schemes. The majority of reports on SiC etching have utilized traditional reactive ion etch (RIE) reactors in conjunction with fluorinated gas mixtures. Compared to other plasma etching methods, RIE exploits the dc self-bias which 151

152

Wide Bandgap Semiconductors

develops on the substrate to assist in the sputter removal of volatile etch products. However, high levels of bias lead to a greater degree of etch damage caused by energetic ion bombardment. For a given plasma density, the ion bombardment energy can be reduced in several ways. For example, plasmas in RIE reactors can be magnetically enhanced so that the process can be run at a lower pressure with high etch rates and with lower self-bias developing on the substrate, resulting in less ion-damage. Alternative reactor configurations which de-couple the sample bias from the plasma power have attracted attention in recent years for both etching and deposition processes. Examples of these methods include the use of electron cyclotron resonance (ECR) and inductively-coupled plasmas. This chapter discusses the applications of etching in SiC device fabrication, the chemistry of plasma-assisted SiC etching, and presents most of the published results to date on dry etching of SiC.

2.0

R E Q U I R E M E N T S O F D R Y E T C H I N G IN SiC DEVICE FABRICATION

The difficulty associated with selective-area doping of SiC has forced many device fabrication schemes to rely on the blanket epitaxial growth of doped layers and their selective removal by dry etching methods. As a result, some common needs for dry etching during SiC device fabrication include: mesa isolation, gate recess etching for planar FETs, channel etching for vertical FETs, and making ohmic contacts to buried layers. To illustrate these applications, Fig. 1 shows cross sectional schematics of several types of SiC power switches and amplifiers which have either been fabricated or proposed. For high voltage applications (>1 kV), bipolar devices such as thyristors have much lower on resistances than unipolar devices. Thyristors with structures as shown in Fig. 1a have been fabricated from both 6H and 4H SiC.Ill[2] In order to achieve low onresistances, the most successful devices have utilized heavily doped n-type substrates with a stack of n-p-n-p+ epitaxial layers. In the o f f state, the voltage blocking capability of the device depends largely on the quality, thickness, and doping density of the middle epilayers. In order to switch the device to its on state, current must be injected through an ohmic contact into one of the buried layers. Therefore, fabrication of a SiC thyristor requires two etching steps which must be carefully controlled: a deep (> 5-10 micron)

Dry Etching of SiC 153 etch for electrically isolating the device; and a shallow etch for recessing down to contact the buried n-type layer. The demands placed on the two etching steps are different. The first requires a rapid process with high etch selectivity for SiC relative to the masking material. The second requires that the etch rate be controllable so that the resulting surface is smooth and free of residues.

OHMIC GATE

OHMIC GATE

t ~

,,c,=r "c" I. METSA f

ANODE ~

/ SOURCE

l

n

lyl:~

SCHOTTKY GATE

DRAIN

I

n" channel p " buffer

.

.

.

.

.

.

.

.

.

n - SiC substrate

CATHODE

(a) Thyristor

(b) MESFET

SOURCE i

SCHOTrKY

l:.:.-.-:.----:.:--:.:-.:--' ........... ~ ......... ~ SGHOTTKY --:-:-:-:-:.:.:.:-,

....

o.

/GATE

n" drift ration ff

SOURCE

~

SiC subslrato

SOURCE

i~ r

Jl n"

n+

MOS GATE

+-++

J-Pi-- i._. '

drift region

n+ SiC substrato

I:-:_-.:_--'._=."..=:: '_::::-_-" . _ : : - ~ : ~ ~..'_;:(:_=:- ~."_-:;_ :-~.._:.'-:_::!

DRAIN

(c) Static Induction Transistor

DRAIN (d) UMOSFET

Figure 1. Schematics of SiC power switching devices (a) thyristor, (b) MESFET, (c) static induction transistor, and (d) UMOSFET.

154

Wide Bandgap Semiconductors

High frequency unipolar devices which have been fabricated from SiC include planar MESFETs and vertical static induction transistors.[ 3]In the fabrication of both of these devices, a Schottky contact is formed on a buried n-type layer which is exposed by dry etching through an upper n § source/drain layer. In this case, the ideal etch process must impart minimal damage to the underlying material so that a high quality Schottky barrier can be formed. Another advanced structure which places perhaps the highest demands on the dry etching process is vertical MOSFET devices (UMOSFETs) which have been demonstrated.[ 11 In these devices, the active channel of the MOSFET is formed on the sidewall of an etched trench so that the requirement of smooth, damage-free, residue-free etching is again particularly critical. Additionally, sharp comers at the base of sidewall features, which can result from anisotropic etching, must be minimized in order to avoid premature gate breakdown. This breakdown can occur due to the electric field enhancement which arises at sharp edges. Therefore, for trench-based MOS devices, an ideal etch would also yield a tapered transition between the etch sidewall and trench bottom. Future SiC power devices which are envisioned include MOS-controlled thyristors and insulated gate bipolar transistors (IGBTs), which are analogous to their Si-based counterparts.[ 4] Most of the viable options for these devices will require the integration of vertical UMOSFET switching elements with thyristor-type structures. In order to realize these next-generation devices, very high demands will be placed on the dry etching processes used in their fabrication.

3.0

CHEMISTRY OF SiC DRY ETCHING

Glow discharges or plasma have been used extensively over the last several decades to promote chemical reactions between gas-phase species and solid-phase materials. Most dry etch processes employ halogenated gas mixtures in order to form volatile products upon reaction with the material being etched. Fluorinated gas mixtures have been found to be quite effective for etching SiC, as they also have for etching Si, SiO 2, SiN x, and silicides of Ta, Ti, and W. Common commercially available gaseous or liquefied sources for fluorine include CHF3, CF4, C2F6, SF6, and NF 3. Most of these sources are nonflammable and relatively harmless; with the exception of NF3, which although highly toxic is often employed due to its efficiency in generating highly reactive species. The chemistry of etching

Dry Etching of SiC 155 Si, SiO2, and SiN x in fluorinated plasmas has been extensively studied and modeled with perhaps the greatest emphasis on CF4/O 2 mixtures. Consequently, much insight can be gained into the chemistry of SiC etching from these results.J5]-[ 14]However, the study of plasma chemistry can be complicated by the presence of many short-lived radical and other excited species which can rapidly react or recombine. For example, when silicon is etched in a fluorine plasma, the only downstream product which can be detected is SiF 4. However, spectroscopic results and thermodynamic arguments suggest that F atoms are the primary reactant and that SiF2 is the initial reaction product. Studies using optical emission spectroscopy have shown that the radicals CF3, CF2, CF, and F all exist in CF 4 plasmas. Various gas additives can have profound effects on etch behavior. The addition of oxygen to a C F 4 plasma reduces the concentration of CF and CF 2 (due to their reactions with oxygen to form CO and CO2).[15][161 Oxidation ofCF x radicals prevents their recombination with F radicals and, consequently, the F-atom concentration increases along with the rate of etching for silicon compounds. As more O 2 is subsequently added, the etch rate eventually reaches a maximum and then decreases, due to the eventual dilution of F by 0 2 and also to strong competition for adsorption sites on the surface from O-atoms. As a consequence of such site-competition among adsorbates, the maximum in the etch rates for Si and SiO 2 films does not always correspond to a maximum in the F-atom concentration. In contrast, the addition of H 2 to the gas mixture can have an effect opposite to that of 0 2. In this case, H radicals can combine with atomic F to form HF. This addition will lower the etch rate of Si, and has been used successfully to increase the selectivity of etching SiO 2 o v e r Si. [17] The etching characteristics of silicon carbide follow trends similar to those of Si, as both CF x and SiF x products can form by reaction ofF with SiC. This reaction can proceed without bias applied to the substrate indicating that SiC is etched chemically by F radicals and does not require the assistance of energetic ions. However, application of a negative bias to the substrate dramatically increases the etching rate as ion bombardment enhances the reaction and subsequent removal of volatile products. Gas additives can also have important effects. For example, the addition of O 2 can enhance the etch rate of SiC just as it does for other Si-containing materials. In this case, 0 2 not only enhances the atomic fluorine concentration but can also participate directly in the etching by reacting with carbon to form volatile CO and CO 2. In some cases, the addition of O 2 can have deleterious effects for SiC etching. It has been known most notoriously to

156

Wide Bandgap Semiconductors

enhance micromasking effects.[ is] These effects arise because most commercial plasma chambers use electrodes and other chamber components made of aluminum which can sputter and redeposit on the SiC surface. Oxidation of these A1 deposits decreases their volatility and results in dispersed microscopic regions of contaminants which unevenly mask the underlying SiC. Conversely, the addition of H 2 to the gas mixture can suppress micromasking effects as will be discussed in See. 4.2. Chlorine-containing gas mixtures can also be used for effectively etching many silicon compounds. However, their use for SiC compounds has not been extensively reported. Recently, McDaniel et al.[191 compared etching of SiC using C12, IBr, SF6, and NF 3 with additions ofAr, H2, o r 0 2. In contrast to employing F-containing gas mixtures, the use of C12 or IBr exhibits a threshold level of de self-bias on the substrate, below which no etching occurs. In this case, the reaction and removal of etch products needs to be induced by ion-bombardment because of the lower volatility of the heavier halides of silicon and carbon. Nevertheless, high etch rates with smooth resulting surfaces have been obtained from C12 and IBr chemistries.

4.0

METHODS FOR PLASMA-ASSISTED E T C H I N G O F SiC

4.1

Plasma Etching

The term "plasma etching" most often refers to plasma-assisted etching which occurs at relatively high pressures (e.g., > 10-1 torr) with the sample sitting on the grounded electrode of a parallel plate reactor of the type depicted in Fig. 2. [16] This type of reactor is most common among commercially available systems owing to its success in silicon device processing. Alternatively, plasma etching may also be performed in a barrel reactor or downstream from a plasma source where the plasma is generated inductively using rfpower or in a microwave cavity. Typically, in the case of plasma etching, only a small sheath potential (10-20 eV) develops on the sample and thus the energy of bombarding ions is relatively low. Therefore, plasma etching processes of this type can be relatively gentle to the underlying semiconductor and tend to be largely isotropic in nature.

Dry Etching of SiC 157

(

o.

TUN.NONE WO.K I

GROUND

........

JL ?:"......... "!"::..................... :::":........... :~:i i r

_~

PLASMA

....... [i,

7

!i . . . ~

ISUBSTRATE I

lil !

TO

Ill........... PUMPS

uB r.AT

I li] //

RF1

oR " - ~ , , ~

GROUND

~

I

TUNINGNETWORK

..

I

I

Figure 2. Schematic of rf powered parallel plate reactor for use in plasma etching or RIE.

Early studies showed that sputter-deposited amorphous SiC films could be etched at reasonable rates in plasmas using fluorine-containing gas mixtures; and that A1 or Cr films could be used as masking materials with a high degree of selectivity.[ TMIn 1985, Dohmae et al.[21] reported plasma etching of ~SiC using C F 4 and 0 2 at pressures of 0.2 to 2.0 torr. The samples were placed on the grounded electrode in a parallel plate reactor powered at 13.56 MHz. These studies, performed at a pressure of 0.6 torr and rfpower density of 0.50 W/cm 2, showed that a gas composition of 70% 0 2 yielded the highest etching rate (17 nm/min). Moreover, the etch rate was found to be proportional to the power density and to decrease sharply with increasing pressure. The morphologies of the etched surfaces were not reported by these authors. In 1987, Kelner et al.,[ 12]using a similar configuration, reported etch rates of 24 nm/min for ~SiC using SF 6 at 0.2 torr and a power density of0.18 W/cm 2 (indicating that SF 6 was more effective than CF 4 alone for etching SIC). In this study, the etching rate also increased linearly with power density. However, extremely rough surface morphologies resulted from these conditions due to micro-masking effects. In a similar study in 1986, Palmour et al.[ 231 used C F 4 and 0 2 at pressures ranging from 0.5 to 2.0 torr with an rf power density at 30 kHz ranging from 0.081 to 0.326 W/cm 2. They reported that etch rates were in the range of 1-55 nm/min, but that the rates were inconsistent and dark surface layers

158

Wide Bandgap Semiconductors

of 20-150 nm thick formed on the SiC surface. Little other work on traditional plasma etching has been reported for SiC. In a less traditional manner, plasma etching of 6H-SiC was reported by Luther et al.[ 24] using remote microwave (2.54 GHz) plasma. In their remote-plasma arrangement, the SiC was placed in a resistance heated furnace, downstream from the microwave cavity. The plasma was generated at a power of 400 W and pressure of 1 torr from a mixture containing 95% Ar with various ratios of NF 3 and 0 2. The etching was described as being smooth and nearly isotropic with sloping sidewalls extending deep underneath the A1 masking layer. The etch rate was found to have a strong dependence on both temperature and the ratio of NF 3 to 0 2, being a maximum of 220 nm/min in a 82% O2/NF3 mixture and a temperature of 325-330~ This rate was approximately ten times greater than those that had been previously reported for rf-plasmas.

4.2

Reactive Ion Etching

The vast majority of reports on SiC etching have utilized the conventional RIE method using rf (13.56 MHz) power in a parallel-plate reactor of the type shown previously in Fig. 2. In contrast to plasma etching, in an RIE process, the electrode on which the substrate is placed is not grounded, but instead is used to supply the rfpower to the plasma. In this case, a large negative dc self-bias develops on the sample. This de bias attracts ions from the plasma which assist the etching process through enhanced formation and sputter-removal of volatile products. This process is sometimes descriptively referred to as "reactive sputter etching" in some of the older literature. Self-bias potentials of hundreds of volts, potentially damaging to the underlying material, are typical for RIE processes. RIE of [3-SIC using CF4/O 2 mixtures was reported by Palmour et. al.[ TMThe etch rate was found to be a maximum of approximately 20-24 nm/min in the pressure range of 20-50 mtorr and at a power density of 0.55 W/cm 2. At pressures above this range, the rate decreased sharply due in part to a reduction in the dc self-bias. At pressures below this range, the rate decreased sharply due to the decrease in the density of reactive species in the plasma. In contrast to previous studies, the etch rate was not significantly affected by the addition of up to 50% 02 to the gas mixture. However, similarly to previous studies, extremely rough surfaces resulted with features described as "sharp columns" and "spikes."

Dry Etching of SiC 159 In 1990, Pan and Steckl[ 2s] compared the etching behavior of SiO2, Si, and polycrystalline I3-SiC in SF6/O2, CHF3/O2, and CBrF3/O2 plasmas at a power density of 0.46 W/cm 2, pressure of 20 mtorr, and gas-flow of 20 seem. Using optical emission spectroscopy, they measured the relative concentrations o f F and O radicals as a function of oxygen addition and detected both CF x and CO x products as shown in Fig. 3. It was found that the de selfbias increased from-285V to -400V as 0 2 was added, and that the highest etch rate of 53.3 nm/min occurred using SF 6 with 35% 0 2. The maximum rate corresponded to the F radical concentration reaching a maximum value. The etched profiles were found to be more tapered using CBrF3/O 2 and to be more anisotropic using either SF6/O 2 or CHF3/O 2. In this study, rough etched surfaces were also observed. Under most of their conditions, etching of SiO 2 was slightly faster than SiC, although a slight selectivity of 3:2 for SiC o v e r SiO 2 could be obtained using the CBrF3/O 2 mixture.

50

500

.....a,_,

_

[F]

.._

:~ m

4o

>~" 400

n0

300

-~m" un I n " "

m

,-O 200

(3' o"O" I

20

--"O

.. ,,..~~ -

[o]

30

%" m~

20

go

10

I

I

I

40

60

80

0

m

z

Co -4

-<

PERCENTAGE OF 0 2 Figure 3. Change in dc self-bias and relative concentrations of F and O with oxygen addition.t2s] A comparison of the RIE etch rates of amorphous, polycrystalline 13SiC, and single-crystal (100) 13-SICin CF4/O 2 has been reported by Padiyath et al.[ 261and is shown in Fig. 4. The etch rate ofpolycrystalline material was found to be much higher than for single-crystal films. This effect was attributed to the large number of high-energy sites associated with dangling

160

Wide Bandgap Semiconductors

bonds at grain boundaries. Moreover, amorphous films which are known to contain significant amounts of hydrogen etched far more rapidly than the crystalline films. This study illustrated the dependence of the etch rate on the degree of crystallinity, but left unanswered the question of whether the etching behavior is different for different polytypes of SiC or for different crystallographic faces of a given polytype. Although no micrographs of the etched materials were presented, the investigators reported that the microroughness of the etched films decreased with increasing crystallinity.

"E" 120 / "~~~ a-SiC'H E E (-. 80 _ poly 13-SiC U.I I-CE -r- 40 g..-o -si -o-o..o\ O I--uJ i

. ,,...

0

\

I

I

I

I

20

40

60

80

"o

PERCENTAGE OF 0 2 Figure 4. Etch rate comparison of amorphous SiC, poly 13-SICand single-crystal ~-SiC. [26]

Significantly, in 1992 Steckl and Yih [18] reported that the micromasking effect which led to rough etched surfaces could be eliminated by adding H 2 to the gas mixture (for RIE processes using CHF3/O2 chemistry). This effect (produced using 18% 0 2 in CF4) is illustrated in Fig. 5 which shows the resulting surfaces without H 2 and with the addition of 17% H2.[27] Using Auger electron spectroscopy, A1 was detected on the rough surfaces which had been etched without the H 2 additive, and was not detected on the smooth surfaces etched in the presence of H 2. This study ascertained that A1 can be sputtered from the rf electrode and redeposited on the SiC surfaces. These A1 residues then coalesce and form micro-clusters

Dry Etching of SiC 161 which mask the subsequent etching of the underlying SiC. The mieromasking effects of A1 are made worse when 0 2 is added to the gas mixture, owing to the strong tendency of A1 for oxidation and the extremely non-volatile nature of aluminum oxide. Unfortunately, many commercial RIE systems utilize electrodes which are constructed of A1 and the addition of H 2 dramatically reduces the etch rate. Moreover, it has been shown that the atomic H generated in H 2 plasmas can significantly passivate the aceeptors in p-type SIC.[ 19]

[ J

Figure 5. Etching of 6H-SiC by RIE with CF4/I 8% 0 2 (a) without H2, and (b) with H2.[27|

162

Wide Bandgap Semiconductors

In subsequent papers, Yih and Steckl evaluated a variety of gas mixtures for etching (001) [3-SIC and compared the etching behavior and the tendency of residue formation.[28]-[ 29] They later extended this study to include single-crystal (0001) 6H-SiC.[ 3~ Upon the addition of H 2, A1Hx species were detected in the plasma by optical emission spectroscopy, leading to the speculation that the redeposition of sputtered A1 was thus prevented by the formation of volatile A1 compounds. The data of Yih and Steckl on 6H-SiC are shown in Fig. 6 for cases of no H 2 added and with the minimum amount needed to suppress residue formation. Here, the relative reactivity of these gases and the degree to which 0 2 and H 2 addition affects the etch rates are illustrated. Without either additive, the reactivity of the gases varied as NF 3 > SF 6 > CF 4 > CHF 3. When the minimal amount ofH 2 was added, the reactivity of the two fluorocarbon gases became comparable, as did the reactivities of the SF 6 and NF 3 plasmas. Moreover, when H 2 was added, the maxima in the etch rates corresponded to mixtures containing approximately 10% 0 2 in each case. It was also reported that the micro-masking effect can be eliminated with the use of a graphite sheet covering the powered electrode.[ 3~ However, side effects which have been reported include polymeric deposition and flaking within the reactor, contamination of the pumping equipment, and mass-loading effects. Casady et al. [31l showed that RIE using N F 3 at the relatively higher pressure of 150-350 mtorr can result in high etch rates and residue-free surfaces without the use ofH 2 or O 2 additives. They reported etch rates of approximately 150 nm/min at an rfpower density of 1.7 W/cm 2 and a pressure of 225 mtorr. At these higher pressures, the corresponding de biases were reported to be in the range of only -25 t o - 5 0 V, suggesting that these conditions may be advantageous for recessing down to active regions of SiC devices. When comparing studies reported in the literature, it appears that, under most conditions, [3-SIC films etched approximately 20-50% faster than 6H-SiC, although accurate experimental comparisons have not been made. It could be speculated that the differences in the etch rates are due more to differences in the dangling bond densities and the corresponding reactivities of the crystal faces, than to the different crystal structures. For example, each atom on a cubic (001) face has two dangling bonds; whereas only one dangling bond exists on a (111) face or similarly on the (0001) face of hexagonal SiC. RIE of (111) I3-SiC films has been reported by Wu et al.,[ 32] albeit under conditions where a direct comparison with reports on (001) oriented films cannot be readily made.

Dry Etching of SiC 163

60

'

_'

'

I

'

'

'

I

'

'

'

I

'

'

'

I

'

'

0

'

.

.

.

I

'

'

'

I

'

'

50~ E 40,

"'.~

3o

""

'E

..'.<

...........

i '"

cF,,

I

'

'

'

I

'

'

'

H2 ADDED

_

.....

-

40 ..............

......

....

30 20

-r 20 O !-. m 10

0' 0

'

- :

....

10 "'~CHE ............A..... """ ~,~ 100 20 40 60 80 PERCENT OXYGEN ADDED

0

,

0

(a)

.

,

I

20

=

.

,

I

,

,

40

.

I

60

,

=

,

I

,

.

80

_

.

100

PERCENT OXYGEN ADDED

(b)

Figure 6. 6H-SiC etch rates using (a) various fluorine-containing mixtures and (b) mixtures with the minimum amount of H E needed to suppress micromasking.[ 3~

4.3

Magnetron Ion Etching of SiC

In a RIE reactor, the plasma density can be enhanced and the corresponding dc self-bias reduced with the proper application of magnetic fields (as illustrated in Fig. 7). This method is oiten termed magnetron ion etching (MIE) or magnetic field-enhanced RIE (MERIE). In this configuration, the permanent magnets confine the electrons in the plasma, thereby increasing the ionization cross section.[331The MIE system can then sustain relatively denser discharges with lower ion energies (dc self-bias), and do so at pressures down to approximately 1 mtorr. Importantly, lower ion energy leads to a correspondingly lower degree of damage induced by ion bombardment. The MIE technique has been used to etch single-crystal (0001) 4H-SiC, u s i n g S F 6 both with and without the addition of 0 2 and H2 .[34] In these experiments, when the samples were mounted on an A1203-coated, rfpowered (13.56 MHz) electrode, problematic A1 micro-masking effects occurred (despite the addition of H 2 to the gas mixture). This effect was eliminated by covering the electrode with a Si wafer coated with indiumtin oxide (ITO). Figure 8a shows the MIE etch rate and dc self-bias as a function of rf power density at a pressure of 2 mtorr and flow of 5 sccm SF 6. Significantly, for a given power density, the etching rate was more

164

Wide Bandgap Semiconductors

than an order of magnitude greater than for the corresponding RIE process, and the dc self-bias was nearly four times lower. At the moderate power density of 0.5 W/cm 2 and self-bias of- 100 V, the observed etch rate of 450 nm/ min was among the highest reported to date. The rate was found to be relatively independent of the addition of oxygen up to 20% 0 2, but decreased sharply with further additions (as shown in Fig. 8b). The rate was fastest at pressures in the range of 2-4 mtorr (Fig 8c). In this case, the dc self-bias was not significantly affected by variations in pressure, and ranged from 75-80 V. The etch rate increased slightly with a gas flow rate of up to 10 sccm, indicating that the supply of SF 6 can be rate-limiting. The anisotropic nature of this process is illustrated in Fig. 8d which shows deep (2 I~m) trenches etched using a sputtered ITO mask.

QUADRUPOLE MAGNET CONFIGURATION

9

: . . .

.

.

. .

.

.

.

.

.

..

(,( (,Elk.,, CHAMBER ToP

i')))

,,._

~

i ~

~ ~

: ,i.l.l,l,.

iq.l.l.i

ii.1.1,,

i,.i.i~

.l.l.m-

..1=1.~

,.11.1,,~ , l . l . l . l l

,ll, l,l.

,i.1.1,i. ,i=1.1~

~.1,,~

~.ii.i

1,,i.i~

:1 MAGNETIC FIELD LINES

~.I.I=t-

INTERNAL MAGNETS , .,~.,~

,,~.w.~

~.m.,.

,,,.,,,~...i.~.

p~.~

,.,'I"M

~,,'~

~'.'~"P

RF TO CATHODE

"

,,-~.~

.,-~"~

,~.!,.~

POLE PIECES

Figure 7. Arrangement of magnets and cathode in MIE reactor.

Dry Etching of SiC 165

500 ._r

9' '

"I"

'"'I"

"'

"

I

'

/

'

''

'

120

350

tO0

400

E

8O

"~ 3oo C LU ,< 200 n-r" O 100 I--iii

40

m

300 E ~250

~,,

iii I-< rr 200

03

"1-

C3 0

20-.~ i &, ii,

0

i i . I ....

0

I,,,,l,,,,l

0.2

....

0.4

LU

'' ''

I''''1'

I,1,,I

....

....

0

10

P O W E R D E N S I T Y ( W / c m 2)

(a) 450

r.... i ....

,'•,, ,, . . . .

"''1

'''

"i''''1"'

"

150 100

0.6

I''''

I.,,11.,,,1,,,,1,,1,

20 30 40 50 60 PERCENT 0 2 ADDED

70

(b) ,"" ' ', ' "-"

E:

rE

400

E

ua :350 F-

< n-

-r 300

O I-u.l

250

J i l . l / i | * l ,

0

1

,1

, l , l , | i l l . ' l l '

2 3 PRESSURE

4 5 (mTORR)

'|

6

(c) Figure 8. M I E etch rate as a function of (a) power density and (d) micrograph of trenched etched in 4 H - S i C .

4.4

(d) (b) oxygen addition (c) pressure

Electron Cyclotron Resonance Plasmas

An excellent review of the principles of ECR and various reactor configurations has been presented by Asmussen.[ 351 Important advantages of ECR relative to RIE are: the sample can be biased independent of the plasma power by applying rf power to the substrate holder; and a high density discharge can be generated at lower pressures and done so in an electrode-less manner. Plasma etching of SiC in an ECR reactor has been shown to result in smoothly etched surfaces without the formation of micro-masking residues.[361 These studies employed a reactor which used a multipolar ECR source consisting of six permanent magnets surrounding a six inch diameter microwave (2.45 GHz) cavity constructed of an aluminum

166

Wide Bandgap Semiconductors

alloy (shown schematically in Fig. 9). The sample sat approximately 8 cm below the bottom of the ECR source on an electrode which could be rfpowered (13.56 Mz) to provide a desired value of dc bias. A 50 sccm 18% O2/CF 4 mixture at 1-2 mtorr was used to investigate the etching behavior of (0001) Si-face, 6H-and 4H-SiC. It was found that for ECR etching, ITO masks endured far better, offered greater selectivity, and adhered better to the SiC than A1, Ni, or Cr masks. As shown in Fig. 10a, the etch rate increased linearly with microwave power and nearly so with applied substrate bias (Fig. 10b). Increasing pressure had a strong effect on decreasing the rate, (Fig. 10c) due to the reduced efficiency of the ECR coupling and the reduced gas-phase mean-free-path. The rate of etching was also affected by the proximity of the sample to the plasma source, increasing by a factor of three as the distance decreased from 8 to 1 cm (Fig. 10d). No significant differences in ECR etch rates had been found for 6H versus 4H polytypes or for different doping types.

TUNERS.---.~

p-wave 2.54 GHz GAS IN

:

QUARTZ WINDOW

CAVITY SAMPLE LOAD LOCK TO TURBO PUMP

" RF POWERED ELECTRODE (13.56 MHz)

Figure 9. Schematic of microwave ECR plasma reactor with rf biasing capability.

ECR etching was also investigated by Ren et al. [37] who found that using NF 3 led to higher etching rates than SF 6, although SF 6 resulted in

much smoother etched surfaces. Reports by McDaniel et al.[ 19]showed that NF 3 resulted in higher etch rates than SF 6 only at lower pressures. When

using NF 3, they found a decreasing etch rate as pressure increased, analogous

Dry Etching of SiC 167 to previous reports which employed CF4/O 2. However, when using SF 6 they found that the etch rate increased with pressure, in the range of 1-10 mtorr.

9

80 e~

/

,

,i,

80 e~

i

-IOOV~

6o

-

o,'

60

.,4r

/ e,-

/

m

I--

40

0 / /

f

20

I-"

/

-20V

/ /

~ 2o

kill

500 W

/ /

<1: I:1=

/

-r O

/

/

40

/ /

I--

w

0

0

' 260 '4()0 '600 " 800 I~-WAVE POWER (W)

's6 '75 BIAS

(a) 100 -

"i00

(-V)

(b) 4O

t'~

r

75

30

o t,-

C

..

u.I 5 0 I,,==: nt"r9 2 5 0I-I,M O - -

-lOOV "o. 5 0 0 W

~.. 2 0
~, - 2 0 V " "800 W

--.. ,.4;

600 W zero bias

.

I--.

Ill

"

3 ' 6 PRESSURE

(c)

' 9 " (mTorr)

12

0 DIST.

" 2 FROM

' z~ ' Ei " 8 ECR CAVITY (cm)

(d)

F i g u r e 10. Etch rate variation o f 6 H - S i C in CF4/18%O 2 plasmas with (a) m i c r o w a v e power, (b) applied bias, (c) pressure and (d) sample proximity to E C R source.

A comparison of ECR etching and RIE was performed to assess the degree of etch-induced damage and the tendency for residue formation characteristic with each process.[ 27][38]In these experiments, samples of ntype 6H-SiC having epilayers with n = 5 x 1015/cm 3 were first subjected to a standard peroxide-based Si cleaning procedure. ECR etching was performed at 600W and 2 mtorr with an imposed de bias o f - 2 0 V . The RIE was performed at an rfpower density of 0.59 W/cm 2 at 25 mtorr. The same gas flow of 50 seem of a mixture of 18% 02/CF 4 was used for each case, but, in the case ofRIE, 10 seem ofH 2 was added to prevent the occurrence of micro-masking. Auger electron spectra showed that the ECR etched

168

Wide Bandgap Semiconductors

surfaces and sidewalls were free of any detectable residues containing A1, F, or O. In contrast, the RIE surface composition showed approximately 9% A1 and 6% O. Moreover, larger amounts of A1 and O were detected on the RIE sidewall along with trace amounts of F. The reverse-bias I-V characteristics of Pd Schottky diodes fabricated on the etched surfaces showed that leakage currents were 1000 times greater on RIE samples. Furthermore, C-V measurements showed a 40% reduction in the donor concentration up to a depth of 0.6 ~tm in the RIE samples indicative of ioninduced damage. No reduction in donor concentration was measured for the ECR samples, although the reverse bias leakage currents were 100 times higher than for unetched control samples. The thermal stability of dry etch damage created in n-type SiC was investigated by Pearton et al., [39] who identified threshold energies for ion-damage in both RIE and ECR Ar plasmas. They showed that annealing of ion-damage does not occur until temperatures of greater than 700~ and that some remnant disorder remains even after annealing at 1050~ Due to the relative stability of damage in SiC, it can be concluded that any etching process for SiC device fabrication should be designed to operate below the damage threshold.

5.0

PROFILE AND M O R P H O L O G Y C O N T R O L W I T H ECR E T C H I N G

In the fabrication of advanced devices from SiC, the profile of the etched sidewall, the morphologies of the etched surfaces, and the degree of damage to the underlying SiC are all issues of high concern. For example, fabricating a vertical power-MOSFET, where the FET channel is fabricated on the sidewall of an etched trench, requires not only smooth surfaces with low damage, but also a tapered transition from the sidewall to the trench bottom. This minimizes electric field enhancement that would otherwise occur at the sharp edges of the gate contact. In SiC etching experiments using ECR plasmas, the level of imposed bias was found to have a strong effect on both the profile and surface morphology.[ 4~ Figure 1 l a shows the comer of mesa features etched at a high bias of-100 V, and Fig. 1 lb shows a low bias of approximately-8 V. The ITO masking films were retained on these samples so that the degree of undercutting could be observed. The surface features that resulted under high bias conditions were found to be relatively smooth both on etched sidewalls and on horizontal surfaces. On the other hand, the high bias

Dry Etching of SiC 169 consequently led to undesirably enhanced etching at the base of the sidewalls. This effect, referred to as trenching, results from the deflection of energetic ions at glancing angles off the sidewalls, thus enhancing the sputtering component of the etching mechanism in the affected regions.[ 16] Therefore, the degree of trenching becomes increasingly more pronounced as the etching proceeds deeper. This effect is also prevalent in RIE and can be observed in the micrographs of etched SiC presented in much of the literature. From an applications perspective, a profile of this nature is illsuited for device fabrication (if an active element of the device, such as the gate of a FET, is to be fabricated in or on the etched feature). In contrast, etching, with low bias applied to the substrate, avoided the trenching effect and led to a more tapered transition from the sidewall to the bottom surface. However, the resulting surface morphology tended to be textured and the sidewalls possessed many jagged edges originating from rough features initially present on the mask edges. The consistent appearance of rough features on samples etched at low bias and smooth features on samples etched at high bias suggests that the ITO mask edges can be effectively smoothed by ion-bombardment during the etching process. In light of this speculation, experiments were performed where the samples were subjected initially to a brief high-bias step, prior to prolonged etching at lower bias. Figure 1 lc shows smoother features which resulted when the sample was etched for 2 min at-100 V prior to prolonged etching at-8 V. Therefore, a two-step bias process can result in a morphology which is far more suited to the fabrication of a trench device element than are the morphologies which result when a constant level of bias is applied throughout the process. The proximity of the sample to the ECR source has been found to strongly affect both the anisotropy and rate of etching. Figure 12 shows the etched profiles as viewed from the comers of adjacent square mesa features with the masking ITO layers still remaining on the structures. In both cases, no bias was applied to the substrate. By measuring the vertical etch depth (d) and the lateral etch depth (/), the anisotropy can be quantified as A = (d- l)/d. The anisotropy of etching was approximately 0.8 when the sample was 8 cm from the bottom of the ECR cavity (Fig. 12a). The anisotropy value decreased to 0.5 when the substrate was elevated to within 1.5 cm of the ECR source (Fig. 12b). The rate of etching also increased with the proximity of the sample to the plasma source. Three important consequences of increased proximity are: the sample is bombarded by a higher density of reactive species; there is likely a significant degree of sample heating from the plasma and from microwave absorption; and the reactive species bombard the sample with a wider angular distribution.

170

Wide Bandgap Semiconductors

Figure 11. (a) Morphology of ECR etched samples with (a)-100 V shows smooth surfaces and sidewalls with trenching effect; (b) -8 V shows rougher surfaces without trenching; (c) two-step bias process to obtain smooth features without trenching.

Dry Etching of SiC 1 71

Figure 11.

(Cont 'd.)

Figure 12. Profiles of 6H-SiC ECR etched with no applied bias have anisotropy of A=0.8 with sample 8 cm from ECR source. (b) A=0.5 with sample 1 cm from source.

(a)

172

Wide B a n d g a p S e m i c o n d u c t o r s

Figure 12. (Cont'd.)

The profile of the etched SiC sidewall can be alternatively controlled by using a thick masking layer of a material which can be eroded gradually at the mask edges.J41][42] One approach is illustrated in Fig. 13 where the sloping edges of a mask can be translated to the etched SiC profile. The resulting sidewall angle is determined by both the initial slope of the mask edge (0m) and the selectivity of etching. Assuming that the etching process is anisotropie, the angle resulting after etching to depths d s and d m in the SiC and mask, respectively, will be given by: Eq. (1)

tan (0~)= d~ tan (Ore)

dm

where (d/dm) is the selectivity of etching SiC relative to the mask and ~ is the sidewall angle of the etched SiC relative to the horizontal surface. It is known that SiO 2 films chemically etched in buffered HF solutions will develop an edge taper of approximately 30 ~. Therefore, by controlling the selectivity of the dry etching process, this method can be applied to SiC using SiO 2 masks. In this case, a variation in selectivity from 0.5 to 3.0 will result in a variation in the SiC sidewall from 16~ to 60 ~

Dry Etching of SiC 1 73

TAPERED MASK

SUBSTRATE

ANISOTROPIC ETCH

SUBSTRATE

(a)

(b)

Figure 13. Diagram of how angled sidewalls are obtained by using tapered mask and varying the etch selectivity.

In a demonstration of this method, a 0.7 ~m layer of SiO 2 was deposited by plasma enhanced chemical vapor deposition onto a 4H-SiC substrate to serve as the masking layer.[ 41] The PECVD film was then densified by annealing at 1100~ in O 2 to improve the film durability. Figure 14 shows the etching rates of SiC and SiO 2 as a function of the amount of N20 added to 10 seem of a gas mixture of 18% O 2 in CF 4 for conditions of 650 W power, 2 mtorr, and applied biases of zero and-50 V. The gas composition and sample bias were both found to affect the etching selectivity. Without N20 the etch rate ofSiO 2 was nearly twice that of SiC. However, as N20 was added, the etching became selective in favor of SiC, with the maxima for different bias values occurring at greatly different gas ratios. Interestingly, at low ratios of O/F, an increase in bias favors etching of SiO2, whereas, at high ratios of O/F, an increase in bias appears to favor etching of SiC. To exploit the etch selectivity for the purpose of controlling the profile angle, test patterns were defined photolithographically on the SiO2/SiC samples. Using the photoresist as a mask, the SiO 2 was removed in the exposed regions with a buffered HF solution. The chemical etch resulted in a 30 ~ slope to the SiO 2 mask edge as shown in Fig. 15a. After removing the resist, the sample was etched under conditions of no bias and an O/F ratio of 0.26 (predicted to yield a selectivity of approximately 1.5 in favor of SIC). The resulting SiC profile corresponding to a 0.6 ~m deep etch is shown in Fig. 15b after removal of the remaining SiO 2 film. The measured sidewall angle was 48 ~ corresponding to an actual selectivity of 1.7. Lanois et al. [42]demonstrated considerably better control using SF6/O 2 mixtures with ECR plasmas where the selectivity of SiC/SiO 2 could be

1 74

Wide Bandgap Semiconductors

varied from 1 to 6 by varying the amount of 0 2 in the gas mixture. In this ease, the sidewall angles could be tailored from 35 ~ to 80 ~ This technique may eventually prove valuable for fabrication oftreneh-MOS devices and for edge termination in high voltage devices.

O/F INPUT RATIO 700 0.10 I ' I'

0.22 I

I

I

'

'

I

0.46 ' ' '

c-

0.58 '

''

I

0.70 '

'

I

o SiC

E 500

A SiO 2

u.I 400 I-,< n" 300 -r O 200 LU

I

10sccm CF4/18%0 2

600 ~

0.34 '

Vdc

= -50V

100 i

0

5

10

I

I

15

I

t

2O

ADDED N20 FLOW (sccm) Figure 14. ECR etch rates of SiC and SiO2 in CF4/18%02 plasma with addition of N20.

6.0

SUMMARY

Realization of advanced device structures fabricated from semiconducting SiC requires the ability to etch this material smoothly, controllably, and with a minimal amount of damage to the underlying material. In this chapter many of the published results on dry etching of silicon carbide using various techniques have been reviewed and compared. The vast majority of reports on SiC etching have utilized traditional reactive ion etch (RIE) reactors in conjunction with fluorinated gas mixtures. These processes have typically been plagued by the formation of mieromasking residues. These residues can be minimized or eliminated by the addition of H 2 to the gas mixture at the expense of lower etch rates and the potential passivation of aceeptors. While RIE etch rates are enhanced by the dc

Dry Etching of SiC 175 self-bias which develops on the substrate, these high levels of bias lead to a greater degree of etch damage. Recently, alternative methods ofmagnetron enhanced RIE (MIE) and electron cyclotron resonance (ECR) plasma etching have been demonstrated. MIE has resulted in extremely high etch rates and ECR etching has resulted in smooth, residue-free surfaces with an ability to control the etched profiles.

9

)~.~

,":..'~.~

..~-:.:!

,

.~..

Figure 15. Profiles of (a) SiO 2 chemically etched in BHF using photoresist mask and (b) SiC with angled sidewall etched using tapered SiO 2 mask.

ACKNOWLEDGEMENTS I gratefully acknowledge the help and support of my former colleagues at Army Research Laboratory, Ft. Monmouth with whom I have collaborated in the research and development of SiC devices. The contributions of Dr. Kezhou Xie and Dr. Walter Buchwald are especially appreciated.

176

Wide B a n d g a p S e m i c o n d u c t o r s

REFERENCES 1. Weitzel, C. E., Palmour, J. W., Carter, C. H., Jr., Moore, K., Nordquist, K. J., Allen, S., Thero, C., and Bhatnagar, M., IEEE Trans. Electron Dev., 43(10):1732 (1996) 2. Xie, K. Z., Zhao, J. H., Flemish, J. R., Burke, T., Buchwald, W. R., Lorenzo, G., and Singh, H., IEEE Electron Dev. Lett., 17(3):142 (1996) 3. Siergiej, R. R., Clarke, R. C., Agarwal, A. K., Brandt, C. D., Burk, A. A., Jr., Smith, T. J., and Orphanos, P. A., IEDM Tech. Digest, p. 353 (1995) 4. Baliga, B. J., IEEE Trans. Electron Dev., 43(10):1717 (1996) 5. Flanun, D. L., and Donnelly, V. M., Plasma Chem. & Plasma Process., 1:317 (1981) 6. Dalvie, M., and Jensen, K. F., J. Electrochem. Soc., 137(4): 1062 (1990) 7. Field, D., Klemperer, D. F., and Wade, I. T., J. Vac. Sci. Tech. B, 6(2):551 (1988) 8. Dalvie, M., and Jensen, K. F.,J. Vac. Sci. Technol. A, 8(3):1648 (1990) 9. Blom, H.-O., Berg, S., Nender, C., and Norstrom, H., J. Vac. Sci. Technol. B, 7(6):1321 (1989) 10. Field, D., Klemperer, D. F., and Wade, I. T.,J. Vac. Sci. Technol. B, 6(2):551 (1988) 11. Clarke, P. E., Field, D., Hydes, A. J., Klemperer, D. F., and Seakins, M. J., J. Vac Sci. Technol. B, 3(6): 1614 (1985) 12. Kuo, Y.,J. Vac. Sci. Technol. ,4, 8(3):1702(1990) 13. Venkatesan, S. P., Trachtenberg, I., and Edgar, T. F., J. Electrochem. Soc., 137(7):2280 (1990) 14. Schoenbom, P. R., Patrick, R., and Baltes, H. P., J. Electrochem. Soc., 136(1):199(1989) 15. Mogab, C. J., Adams, A. C., and Flamm, D. L., J. Appl. Phys., 49:3769 (1978) 16. Mogab, C. J., VLSI Technology, (S.M. Sze, ed.), pp. 303-345, McGrawHill, New York (1983) 17. Ephrath, L. M.,J. Electrochem Soc., 126:1419 (1979) 18. Steckl, A. J., and Yih, P. H., Appl. Phys. Lett., 60:1966 (1992) 19. McDaniel, G., Lee, J. W., Lambers, E. S., Pearton, S. J., Holloway, P. H., Ren, F., Grow, J. M., Bhaskaran, M., and Wilson, R. G., J. Vac Sci. Technol. A, 14:826 (1997) 20. Chang, C. Y., Fang, Y. K., Huang, C. F., and Wu, B. S., J. Electrochem. Soc., 132(2):418 (1985)

Dry E t c h i n g

of SiC

177

21. Dohmae, S., Shibahara, K., Nishino, S., and Matsunami, H., Jpn. J. Appl. Phys., 24 :L873 (1985) 22. Kelner, G., Binari, S. C., and Klein, P. H., J. Electrochem. Sot., 134(1):253 (1987) 23. Palmour, J. W., Davis, R. F., Wallett, T. M., and Bhasin, K. B., J. Vac. Sci. Technol. A, 4:590 (1986) 24. Luther, B. P., Ruzyllo, J., and Miller, D. L.,Appl. Phys. Lett., 63" 171 (1993) 25. Pan, W. S., and Steckl, A. J., J. Electrochem. Soc., 137(1):212 (1990) 26. Padiyath, R., Wright, R. L., Chaudhry, M. I., and Babu, S. V., Appl. Phys. Lett., 58:1053 (1991) 27. Flemish, J. R., Xie, K., Buchwald, W., Casas, L., Zhao, J. H., McLane, G., and Dubey, M., Mat. Res. Soc. Symp. Proc., 339:145 (1994) 28. Yih, P. H., and Steckl, A. J.,J. Electrochem. Soc., 140:1813 (1993) 29. Yih, P. H., and Steckl, A. J.,J. Electrochem. Soc., 142:2853 (1995) 30. Yih, P. H., and Steckl, A. J.,J. Electrochem. Soc., 142:312 (1995) 31. Casady, J. B., Luckowski, E. D., Bozack, M., Sheridan, D., Johnson, R. W., and Williams, J. R., J. Electrochem. Soc., 143(5):1750 (1996) 32. Wu, J., Parsons, J. D., and Evans, D. R., J. Electrochem. Soc., 142(2):669 (1995) 33. Rossnagel, S. M., Thin Film Processes II, (J. L. Vossen, and W. Kern, eds.), Academic, San Diego, pp. 11-77 (1991) 34. McLane, G. F., and Flemish, J. R., Appl. Phys. Lett, 66(26):3755 (1996) 35. Asmussen, J.,J. Vac. Sci. Technol. A, 7(3):883 (1989) 36. Flemish, J. R., Xie, K., and Zhao, J. H., Appl. Phys. Lett., 64:2315 (1994) 37. Ren, F., Grow, J. M., Bhaskaran, M., Lee, J. W., Vatuli, C. B., Lothian, J. R., and Flemish, J. R., Mat. Res. Soc. Symp. Proc., 421:251 (1996) 38. Xie, K., Flemish, J. R., Zhao, J. H., Buchwald, W. R., and Casas, L., Appl. Phys. Lett., 67:386 (1995) 39. Pearton, S. J., Lee, J. W., Grow, J. M., Bhaskaran, M., and Ren, F., Appl. Phys. Lett., 68(21):2987 (1996) 40. Flemish, J. R., and Xie, K., J. Electrochem. Soc., 143(8):2620 (1996) 41. Flemish, J. R., Xie, K., and McLane, G. F., Mat Res. Soc. Proc., p. 421 (1996) 42. Lanois, F., Lassagne, P., Locatelli, M. L., Appl. Phys. Lett., 69:36 (1996)

5 Processing of Silicon Carbide for Devices and Circuits Jeffrey B. Cas ady

1.0

BACKGROUND

Most traditional integrated circuit technologies using silicon devices are not able to operate at temperatures above 250~ especially when high operating temperatures are combined with high-power, high-frequency, and high-radiation environments. Much attention has been given to SiC, currently the most mature of the wide-bandgap (2.0 eV < Eg< 7.0 eV) semiconductors, as a material well-suited for high-temperature and efficient high-power operation. High-temperature circuit operation from 350~ to 500~ is desired for use in aerospace applications (turbine engines and the more electric aircraft initiative), nuclear power instrumentation, satellites, space exploration, and geothermal wells.[ 11-[61 As an example of one such application, the more electric aircraft concept offers substantial system-level benefits which are similar to the more electric tank, more electric ship, and more electric automobile concepts. By replacing bulky hydraulic-driven flight control actuators and engine-gearbox driven fuel pumps with power electronics capable of high efficiency and high operating temperatures, tremendous weight savings can be induced. Currently, hydraulic-based systems also provide cooling to aircraft electronics, so the removal of the hydraulic systems will increase 178

Processing of Silicon Carbide for Devices and Circuits

179

the operating temperature for on-board electronics. Thus, a distributed flight control system which would allow the elimination of cooling systems is envisioned using wide-bandgap devices.[ 2]Over 90% of closed loop environmental control system cooling requirements on modem fighter aircraft are utilized for the cooling of electronics (including radar electronics). Reducing or eliminating this need for cooling would reduce aircraft weight, maintenance support, down time, and ground supplies, while increasing the aircraft' s performance, flying range, efficiency, and reliability. Other advantages of distributed flight control include elimination of long and heavy wiring/shielding runs to actuators and sensors, reduction of the number of unreliable connector pins between control electronics and sensors/actuators, and a decreased chance of catastrophic failure since the control electronics would be located in multiple positions as compared to one centralized location.[2]A pictorial summary of the more electric aircratt concept is shown in Fig. 1.

Integrated Power Unit

Internal Engine Electric Starter/Generator

Solid-State Remote Terminals

Electric-Driven Flight Actuators

Fault-tolerant, solid-state electrical distribution Electric Anti-lcing Electric-Actuated Brakes

Figure 1. Conceptual summary of the more electric aircraft concepts applied to a currentgeneration fighter aircraft.

In addition to these applications, SiC has potential for use in numerous other high-power, high-frequency, and radiation-resistant applications.[ 7]--[9] Silicon carbide (SIC), aluminum nitride (A1N), gallium nitride (GaN), boron nitride (BN), diamond, and zinc selenide (ZnSe) are the primary widebandgap semiconductors now being developed for use in the aforementioned

180

Wide Bandgap Semiconductors

applications. Of these, SiC presently has several advantages including commercial availability of substrates since -~1991 (see Refs. 10 and 11), known device processing techniques, and the ability to grow a thermal oxide for use as masks in processing, device passivation layers, and gate dielectrics. In addition, SiC's high thermal conductivity (about 3.3 times that of Si at 300 K for 6HSIC), high electric field breakdown strength (about 10 times that of Si for 6HSIC), and wide bandgap (about 3 times that of Si for 4H-SiC and 6H-SiC) make it a material ideally suited for high-temperature, high-power, highfrequency, and high-radiation environments. The first polytypes of SiC to appear in bulk wafer form[l~ 11]were 6HSiC and 4H-SiC, which has helped SiC to emerge as one of the relatively mature wide-bandgap semiconductor technologies. A complete list of known SiC vendors in 1997 is shown in Table 1 for reference, although the list is certain to expand. Lattice mismatches of only 1% for A1N,[121 and 3% for GaN,[ 131114]exist when these materials are grown on 6H-SiC substrates. Thus, SiC processing is otten intimately linked with A1N and GaN electronic and optical device fabrication as well. For example, commercially available GaN blue (peak wavelength of 430 nm) LED's manufactured on 6H-SiC substrates were released in 1995.[1~ By 1998, this GaN on SiC blue LED represented the highest volume blue LED sold. The small lattice mismatches with AIN and GaN, as well as the abundance of polytypes in SiC, combine to make SiC a material with an immense potential for use in heterostructure electronic devices which take advantage of the differing bandgaps, carrier mobility's, etc.

1.1

Historical Development of Silicon Carbide

With all the recent attention given to SiC, an often surprising fact is that growth of SiC crystals by the Acheson technique dates back to 1893. Discovery and identification of a SiC Light Emitting Diode (LED) occurred in 1907 when H. J. Round published a short article entitled "A Note on Carborundum. ''[15] Semiconductor pioneer W. Shockley, inventor of the bipolar junction transistor in 1947, recognized the potential of SiC in his now prophetic, introductory remarks at the First International Conference on Silicon Carbide in April 1959, as illustrated by the following quote: "Today, in the electronics .field there are probably two areas of special interest. One of these is miniaturization, the process of maldng devices small complicated and fast; the other has to do with problems of new environment, such as higher temperatures and radiation resistance ....

Processing of Silicon Carbide for Devices and Circuits

181

Now, the big question is this: How is theproblem of high temperature going to be solved? What are the horses to put one's money on?... One approach is the logical sequence we see here: Ge, Si, SiC, C in that sequence .... The SiC situation suffers from the very same thing that makes it so good. The bond is very strong and so all processes go on at very high temperature .... Another aspect of the silicon carbide situation is similar to past situations in the semiconductor field. The lesson is that one should not give up too soon and one should not always look for gold at the ends of new rainbows .... The situation may be similar with silicon carbide. The material problem will have to be extensively worked on. Perhaps one day.., large single crystals o f silicon carbide will be grown easily .... These are difficult questions."[ 16]

Table 1. List of Known SiC Vendors (1997)

Supplier Epitronics, an ATMI company Cree Research, Inc.

Location 7 Commerce Drive Danbury, CT 06810 USA

30 mm 4H and 6H-SiC bulk

2810 Meridian Parkway

50 mm 4H and 6H-SiC bulk/epi

Durham, NC 27713 USA FTTIKS Enterprise

Comments

3-4-13 Nihonbashi Chuo-ku, Tokyo, Japan

6H-SiC bulk and 4H-SiC epi on 6H-SiC

HOYA Corporation

3-3-1 Musashino, Akishima Tokyo 196, Japan

3C-SiCon Si

Sterling Semiconductor

Sterling, VA

50mm 4H-SiC bulk/epi

Nippon Steel Advanced Technology Research Laboratories

5-10-1 Fuchinobe Sagamihara Kanagawa 220 Japan

6H-SiC and 4H-SiC 25 mm wafers available (1995)

182 1.2

Wide Bandgap Semiconductors Material Properties of Silicon Carbide

SiC is part of a family of materials which exhibit a one-dimensional polymorphism called polytypism. An almost infinite number of SiC polytypes are possible, and approximately 200 polytypes have already been discovered.[ 17] A listing of some of the more common polytypes include 3C, 2H, 4H, 6H, 8H, 9R, 10H, 14H, 15R, 19R, 20H, 21H, 24R, and others (see Table 2). SiC polytypes are differentiated by the stacking sequence of each tetrahedrally bonded Si-C bi-layer. With the exception of 2H and 3C, all of the polytypes form 1-D superlattice structures.[ 181Even though individual bond lengths are nearly identical, as shown in Table 2, the crystal symmetry is determined by the stacking periodicity. The polytypes are divided into three basic crystallographic categories: cubic (C), hexagonal (H), and rhombohedral (R). Table 2. Lattice Structure Properties of Selected SiC Polytypes

Polytype

Lattice Structure

Lattice constant a (A)

Lattice constant c (A)

, , .

3C-SiC

face-centered cubic

4.3590

2H-SiC

hexagonal

3.0817

4H-SiC

hexagonal

3.079

6H-SiC

hexagonal

3.0817

15.1183

8H-SiC

hexagonal

3.079

20.146

15R-SiC

rhombohedral

3.079

37.78

21R-SiC

rhombohedral

3.079

52.88 67.995

5.0394 10.254

27R-SiC

rhombohedral

3.079

33R-SiC

rhombohedral

3.079

83.10

51R-SiC

rhombohedral

3.079

128.434

75R-SiC

rhombohedral

3.079

188.867

84R-SiC

rhombohedral

3.079

211.539

87R-SiC

rhombohedral

3.079

219.094

Cubic Silicon Carbide. Cubic SiC has only one possible polytype, and is referred to as 3C-SiC or J3-SiC. Each SiC bi-layer can be oriented into only three possible positions with respect to the lattice, while the tetrahedral bonding is maintained. If these three layers are arbitrarily

Processing of Silicon Carbide for Devices and Circuits

183

denoted A, B, and C, and the stacking sequence is ABCABC..., then the crystallographic structure is cubic zincblende. This arrangement is known as 3C-SiC or 13-SIC. The number 3 refers to the number of layers needed for periodicity. The smallest bandgap (---2.4 eV)[ 19] and one of the largest electron mobilities (~800 cm2/V.s in low-doped material)[2~ of all the known SiC polytypes is possessed by 3C-SiC. It is not currently available in bulk form, despite bulk growth of 3C-SiC having been demonstrated in a research environment.[ all Cubic SiC has been grown on Si with limited success and incorporated into heterostructure devices, despite the nearly 20% lattice mismatch between 13-SIC and Si. Recent advances in heteroepitaxial growth of I3-SiC include low temperature (750~ CVD of [3-SIC on silicon achieved by researchers at AlliedSignal,[ a2] and the growth of 3C-SiC on 6H-SiC substrates.[23][TM Hexagonal and Rhombohedral Silicon Carbide. If the stacking of the bi-layers is ABAB..., then the symmetry is hexagonal and referred to as 2H-SiC. All of the other SiC polytypes excluding 2H and 3C are a mixture ofzincblende (cubic) and wurtzite (hexagonal) bonding. An equal number of cubic and hexagonal bonds make up 4H-SiC, while 6H-SiC is composed of two-thirds cubic bonds and one-third hexagonal bonds. The overall symmetry is hexagonal for both polytypes, despite the cubic bonds which are present in each. The hexagonal structures are collectively grouped as o~-SiC.The only SiC polytypes currently available in bulk wafer form are 6H-SiC and 4H-SiC. Polytypes with rhombohedral symmetry have also been found and are denoted by the letter R. Advanced Technology Materials, Inc. (ATMI) have also grown 15R-SiC substrates.[ 251 Electrical Properties and Impurity Centers in 3C, 6H, and 4H Polytypes. Basic electrical properties obtained from the literature for the 3C, 6H, and 4H polytypes of SiC are shown in Table 3. The data cited are dependent upon the temperature, test methods used, and quality of the material, as well as dopant concentrations and species. Accuracy of these material properties is extremely important for device modeling, yet material properties are often cited erroneously or improperly in the literature. To fully analyze the data provided in Table 3, one should carefully examine the given references. The properties for Si[TM are also shown for reference. Three key categories where SiC enjoys inherent advantages over Si and GaAs for high-temperature, high-power, and high-frequency operation are thermal conductivity, electric field breakdown strength, and bandgap. SiC also has a much higher thermal conductivity (| than GaAs at 300 K, (8 to 10 times), [a6] a bandgap of approximately twice the bandgap of GaAs, while still possessing a peak saturation velocity (v~at)of 2 x 107 cm/s, just as in

184

Wide Bandgap Semiconductors

GaAs. As a partial result of the saturation velocity overshoot phenomenon in GaAs, SiC's Vsat at high electric fields is superior to that of GaAs. The electron velocity of silicon, silicon carbide, and GaAs are compared in Fig. 2a where SiC's superiority at high electric field (>300 kV/cm) can be observed.

Table 3. Silicon Carbide Material Properties QUANTITY

3C-SiC

4H-SiC

6H-SiC

Sil281

Eg (eV) at T < 5 K

2.40 [19]

3.26 [27]

3.02 [27]

1.12

Ecdt(MWcm)

2.12 [29]

2.2 [29]

2.5 [29]

0.25

Eerit ('ND) (kV/cm)

10.64.ND 0.142

O K (W/cm'K) at 300 K*

3.2 [32]

3.71321

4.9[ 33]

1.5

n i (cm "3) at 300 K ~

1.5 x 10 "1

5 x 10 -9

1.6 x 10 -6

1.0 • 1050

v~t (cm/s)

2.0 x 107 [lo]

2.0 x 107 [30]

1.0 x 107

[parallel to c-axis]

10001351

4001351

1400

0 . 7 - 0 . 8 3 [ 34]

6[35]

1 15 [40]

101 [40]

800 [20]

/'/e (cm2/Vs)

m_L/mjj at 300

K

~h (cm2/Vs)

40 [39]

2" (eV) at 300 K refractive index (n)

2.7

es

9.72[ 42]

471

2.70 [4]

2.95[ 4]

3.15

2.712

2.7

3.5

9.661421

11.7

*doped at -~1017 cm "3 **Nc, Nv ~ 1019 cm -3

All of the quantities shown in Table 3 are temperature dependent to differing extents. Electron mobility (ge) and hole mobility (Ph) are carrier velocity per unit drift field, and are critically important device parameters, affecting the microwave performance, transconductance ( g ~ , output gain of FET' s, on-resistance of power FET's (Ro~), and other parameters. The low, anisotropic electron mobility in 6H-SiC is one of the primary reasons for the popularity of 4H-SiC for electronic applications, which has a higher and much less anisotropic electron mobility. In fact, (P-t)/(P II) is about 0.70.83 at 300K in 4H-SIC,[34][35] while the same ratio is about 6 in 6H-SiC.[ 35] Of course, mobility is also strongly dependent upon temperature and doping density. In Fig. 2b, the room temperature mobilities of Si, 4H-SiC, and 6HSiC are compared as a function of total doping density. The mobility values

Processing of Silicon Carbide for Devices and Circuits

185

for 4H and 6H-SiC were least squared fits (of experimental data taken in the basal plane) to an empirical relation for mobility in silicon.P 5]For reference, the mobility values for n and p-type silicon at room temperature are also shown.[ 28] The superiority of 4H-SiC over 6H-SiC with respect to mobility values can clearly be seen from this figure, especially when considering the severe anisotropic nature of mobility in 6H-SiC as discussed earlier.

10 7

108

103

104

105

106

107

Electric Field Intensity, vlcm (a)

14007= .............................

'~

1

4H-SiC (N)

................................. =........................~ - - i

10001

,., 4 H - S i C

................................................................ : --.

>

~

" ................................................................. ................

800

-

~

-

t

i

1200 .-.

y. . . .

(AI)

6H-SiC

(N)

-..

6H-SiC

(^i)

+

Si (Phos)

+

Si (Boron)

400- ~ "" i

1E+15

'

"'-,~i

i

i

1E+16

1E+17

1E+18

1E+19

1E+20

Total Impurity Concentration (crn -3)

Co) Figure 2. (a) Electron velocity vs. electric field in SiC. Higher saturated electron velocity and breakdown field of SiC result in 4 times higher power density than silicon or GaAs. Taken from Ref. 4.(b)Electron and hole mobility's in silicon and silicon carbide at room temperature as functions of total dopant concentration. Mobility's for SiC are for the basal plane. Silicon data from Ref. 28 and silicon carbide data from Ref. 35.

,186

Wide Bandgap

Semiconductors

The intrinsic carrier concentration (ni), defined by Eq. (1), is directly proportional to N c and N v, which are the conduction band and valence band density of states, respectively. N c and N v have an empirical temperature (T3/2) dependence. However, as a result of the fundamental change in energy between the electron states, n i has an exponential dependence upon temperature, as well as Eg. [37] In Fig. 3a, n i as a function of reciprocal temperature is shown for 4H-SiC, illustrating that device operation in excess of 1500~ is theoretically possible. Intrinsic carrier concentration (ni) is important in high-temperature device applications (pn junction leakage currents in devices are normally proportional to n i or ni 2) when diffusion currents dominate the total leakage current. It should be noted that in wide bandgap semiconductors, the generation current resulting from electron hole pairs thermally generated in the depletion region can actually dominate the total leakage current. Electron effective masses (m• = 0.42 m o and mll - 0.29 m o in 4H-SiC)[ 34] which are also a function of the band curvature have not been analyzed as a function of temperature, and work remains in that area. Eq. (1)

n~ =x[Nc.Nvexp (-e~/~-~r) n,., = n~ exp (~/~r) T (~ -" 73

227

1727 liaJa

~!!!~i ! nl

l x l O + 1 6 c m "3

I

I

I

,

i

ii!i

i !

i

ili

i

~

4.-s,c 11 (a)

71i:Oi

1x10 +lOcm -3 : ,,;:~_.,.~ .~ , /

/

._iiii! -

l x 1 0 +02 c m "3

l~gliil i'i

i

i ,

'

[

i

!~

I ~ ! i : i i i In I~ ~ ~ ~ Ii

9 ~-~i'! EG,,3,24oV

|

~ili

~ i ~

l x 1 0 " 1 0 c m "3

!/

i/

~ ! :

n p~i.

I""" 10

::ilUl

! ! ::! U

JRI

L~i'i''

N C ,, 1.25x10 TMx (T/3IX)) I"s cm "3 !

!

~;~i~ . . .!. i ~:~ p

J

/

ill

!

I I

~iliJl i i

n !

'

~

!ii!li

o

, I

'

iili~i

1

i

i

0.1

1000/T (K "1) Figure 3. (a) Intrinsic carrier concentration as a function of reciprocal temperature ( 1 0 0 0 / T ) in units of inverse Kelvin for 4H-SiC. (b) Temperature-dependent bandgap values for different polytypes of SiC. Taken from Ref. 38.

Processing of Silicon Carbide for Devices and Circuits

(4.2K) I~X

(RT) ~,

187

(RT)

, ....-' 1 '. ~

r

~-'

"---.... S i C

,o.

&.l~& 300

(4H) s.au 2H 9 ~',.

"~r "

-,x2a5

~,

3.,127

a.a~

3.20

IIIH) . 3.00-2.995 ,2.972 ,,z957

115fl1,,--

3.022 2.999 2.964

2.9o

121 R) z.au, ~1" t~

($H) z.m i

2.80"2.02 ~.77

(24R1 t ~

~1~1%~'~

2.70,2.71

" 2.4,R,'-,, "~ (3C) ~ . ~

2.827 2.80

\

2.7, ;LeO 2.40

F-......

,~1.360

2.387

3s _l

I[_ l__l Ti~l~lUm

Ibwgy ( ; a ~ . Exawn S ~

F i g u r e 3.

- 2.30 %% % X, 9 o 2=O I l I l {Ki - ~

EG(ov) = EGx(eV). ( B E ) x EneflW, ( B E ) x = 0.027 ev

(Cont 'd.)

Typical bandgap values are obtained from photoluminescence studies performed at liquid He temperatures (~4.2 K) under very low pressures (--~10-II T). [17][19][27] Energy bandgap as a function of temperature for polytypes 2H, 4H, 6H, 15R, 21R, 8H, and 3C of SiC have been reported, with the temperature dependencies shown by Fig. 3b.[TMHigh doping levels lead to band gap narrowing (BGN) effects in semiconductors, but have not been extensively studied in SiC. Therefore, the effective intrinsic carrier concentration (hie) relationship with doping has not been firmly established. Arrangement of next neighbors in the lattice is the same for all SiC polytypes, but crystallographically inequivalent lattice sites exist in different

188

Wide Bandgap Semiconductors

polytypes. Thus, electronic properties, such as effective mass, carrier mobility, and bandgap, vary widely between different polytypes of SiC, as shown in Table 3. Measurement of all material properties shown in Table 3 should be investigated carefully since varying material quality (mixed polytype material can easily be obtained) can drastically affect measured parameters. An example of variability in reported material constants is the range in thermal conductivity reported for SiC as shown in Fig. 4.

~'~12

g,07 "

E

'

.>

-

"r0

-

96H-SiC . 4H-SiC

~ 6-3

,L 3C-SiC

uo 4 -

= 2j,

m ~

~

.

,, p-

A 9

14

19

24

29

34

39

44

Temperature (K)

(a) I~ Low

I~ Median

4"2

"

~---

"~2-

8 t-I--

D High

6H

4H SiC Pol~ype

3C

(b) Figure 4. Reported values from literature of thermal conductivity for SiC at (a) low temperature and (b) room temperature. Range of room temperature data is reported as high, median, or low. Taken from compiled data in Ref. 95.

Processing of Silicon Carbide for Devices and Circuits

189

Electrically active impurities in semiconductors are normally substitutional dopants occupying vacant lattice sites. Dopants for SiC include N, P and As (n-type), and A1, B, Be, Ga, O, and Sc (p-type), with A1 being the most common p-type dopant because it has the shallowest acceptor level.[ 17] Undoped SiC is typically n-type from residual nitrogen, and has a slight green tint in color for 6H-SiC. Color of the material depends upon the specific polytype, however. Donor activation energies are often found to vary over a wide range, depending upon measurement technique, material quality, polytype and dopant concentrations. Activation energies also vary depending upon the substitutional site occupied in the lattice (cubic or hexagonal). For n-type 3C, Hall measurements have yielded nitrogen activation energies from 18-48 meV. In 6H-SiC, two donor levels have been found depending upon the occupancy site. Site 1 (hexagonal site) is from 84-100 meV, and site 2 (cubic site) is from 125-150 meV.[ 17]In 4H-SiC, donor levels are 45 meV, and 100 meV for site 1 and site 2, respectively.[ 17] The fact that most dopant levels are deeper than those found comparably in silicon explains the partial carrier freeze-out found in SiC at room temperature, since the thermal energy (kT/q) is only---25.9 meV at 300 K. Since carriers are ionized by electric fields in many devices (most notably field effect transistors), lower temperature operation is also feasible. For example, SiC Junction Field Effect Transistors (JFET's) have been operated to temperatures as low as 77 K, because of field ionization of dopants.[ 41] For p-type A1 doped SiC, an average acceptor energy level of approximately 200 meV is found for all pol3~,pes. [17]Other p-type dopants such as boron have deeper acceptor levels (ranging from approximately 320 meV to 735 meV), and are not as commonly used because of increased concerns regarding carrier freeze-out. Boron's most common use has been in work involving deeper (>0.8 ~tm) p-type implants where boron is preferred over aluminum because of its lighter mass.

2.0

SILICON CARBIDE DEVICE PROCESSING

Now that a brief description of SiC material properties has been given, examining processing issues of the semiconductor for combined high-temperature, high-power, and high-frequency applications is now appropriate. SiC device processing has rapidly evolved since the commercial availability of SiC substrates in 1991. In this section, the major aspects of SiC device processing are discussed, beginning with bulk material growth.

190 2.1

Wide Bandgap Semiconductors Bulk SiC Growth

Historically, bulk growth of SiC has been perhaps the most significant problem limiting the usefulness of SiC in electronic applications. [43}-[59]Singlecrystal wafers of 6H-SiC have been available commercially only since 1991 (from Cree Research, Inc. )[101 and 4H-SiC wafers have only been available since 1994 (from Cree in 1994 and ATMI in 1995).[1~ 11] Research environments, but not commercial ones, have produced 3C-SiC wafers[ 211and 15R-SiC[ 1l] wafers. An excellent review of commercial SiC boule growth by seeded sublimation is given by V. F. Tsvetkov et al.[431 Single crystal boules grown from either a melt or solution would require excessive temperatures (>3200~ and very high pressures (> 100,000 atm) which precludes this growth method from being utilized. Thus, in the absence of other viable melt-growth techniques, physical vapor deposition via seeded sublimation is the most commonly used approach. The vapor phase of SiC (typically Si, Si2C, SIC2) is deposited upon a SiC seed crystal at high temperatures (>2000~ Examples of a 4H-SiC boule grown via the sublimation process is shown in Fig. 5 and example wafers shown after slicing and polishing are shown in Fig. 6. Typical wafer diameters are 3 5 mm, although Cree Research is currently offering the sale of 2 inch (~50 mm) diameter wafers beginning in 1997. One major problem with the sublimation growth technique has been the formation of mieropipe defects. Although defects such as mieropipes are not found in Lely platelets of SiC, this type of growth results in irregular shaped SiC substrates which are unsuitable for commercial SiC device produetion.[5~ 551Mieropipes are bulk defects (voids) which propagate the length of the boule from the seed crystal, and are also found to propagate through subsequent epitaxially grown SiC layers. Mieropipes have hexagonal cross-sections with diameters from about 0.1 ~tm to 5 ~tm. [43][46] Mechanisms causing the mieropipes have not been clearly identified in the literature, but 13 possible thermodynamic, kinetic, and technological mechanisms have been identified.[ 43] A good discussion of various defects (hexagonal pits, mieropipes, screw dislocations, hillocks, etc.) and possible causes is found in Ref. 56. In physical vapor transport grown 6H-SiC substrates, all mieropipe defects were positioned along the lines of super screw dislocations with Burgers vectors of at least four times that of the eaxis lattice constant.[ 461Mieropipe defect densities (MDD), found in densities of 1000/em 2 in the early 1990's, were reported as reduced to 3.5/em 2 at the research level on a 30 mm (1.18 inch) 4H-SiC wafer in 1995.[43]Typical commercial wafers in 1997 possess mieropipe defects densities ranging

Processing of Silicon Carbide for Devices and Circuits

191

from 50-200/cm 2. One fact often overlooked by device engineers is that the mieropipe defect density is non-uniform and often locally clustered on the wafer. The non-uniform density is a real benefit to device engineers seeking large-area regions for power devices or complex circuits since 1 cm 2 areas on wafers have often been found with zero micropipes. [44]Figure 7 illustrates an example of the non-uniform micropipe defect density found on a typical 4H-SiC substrate, similar to that reported in the literature.[ 441 While some areas are virtually defect-free, other areas have a large density ofmicropipe defects on the same wafer. Thus, quotes of average micropipe defect density across a wafer may not give a true representation of the substrate quality. Elimination of the micropipes found in bulk SiC is a critical issue for development of SiC power devices and larger-area integrated circuits[ 45] and is expected to happen by the year 2000.

Figure 5. Polished 38 mm diameter, <0001>-oriented 4H-SiC boule grown via seeded sublimation technology shown prior to slicing. Photo courtesy of Northrop Grumman.

192

Wide Bandgap Semiconductors

Figure 6. Examples of sliced and polished SiC substrates. Note that SiC is optically transparent and that substrate color is dependent upon doping type, concentration, and material polytype. Photo courtesy of Northrop Grumman. It should be noted that at least three different corporations in the United States (Northrop Grumman, Cree Research, and ATMI/Sterling Semi.) are currently producing SiC wafers via seeded sublimation, with other companies in Russia, Japan, and Europe also producing wafers (see Table 1). A brief, non-inclusive listing of other outstanding references to bulk growth of SiC are listed for the interested reader.[21][25][43t-[59]Work by H. M. Hobgood et al., [47] D. L. Barrett et a1.,[521[551 and G. Augustine et al. [44] provide superb discussions of SiC bulk growth using a sublimation-source

Processing of Silicon Carbide for Devices and Circuits

193

physical vapor transport system at Northrop Grumman, with results comparable to that of Cree's. Notable achievements include production of 6HSiC boules up to 60 mm (2.36") in diameter. It is estimated that 100 mm, high-quality wafers of reasonable cost will be required for high-power commercial SiC device production, while 50-75 mm wafers should suffice for low-power commercial products. [43]Growth was done at ~2300~ while the oriented SiC seed crystal was held at a lower temperature (~2200~ The major crystalline defects reported in the 4H-SiC substrates grown by physical vapor transport were micropipes (10 cm -2 on best wafers) and dislocations (104 cm -2 range). Room-temperature electrical conductivity of the substrates could be varied from less than 1 x 10 -2 f~-cm, n-type, to insulating (> 1015 ~-cm).

Figure 7. Example of the non-uniform micropipe defect density on a typical 4H-SiC substrate. Photo courtesy of Northrop Grumman.

194

Wide Bandgap Semiconductors

Other recent research has focused upon the use of tantalum coated crucibles and containers for sublimation growth in place of the traditional graphite crucibles. For example, growth rates of 1.5 mm/h have been achieved using Ta container material, with growth temperatures ranging from 1600-2100~ When comparing Ta and graphite crucibles for growth (using polycrystalline SiC source material and (0001) 6H-SiC Lely grown platelets as seed), it was found that at low temperature gradients (<30 K/cm), the growth rate of material in Ta containers exceeded that of material grown in graphite containers. The development of semi-insulating SiC substrates is of extreme importance for microwave applications.[ 58][591 Single crystal 6H-SiC substrates grown by PVT have demonstrated high resistivity (1-30 l~.cm), semi-insulating (> 100 k~.cm), and even insulating (1011-1012 k~.cm) type behavior. Achieving semi-insulating 6H-SiC has been accomplished by using undoped and vanadium-compensated substrates. The resistivity of the undoped 6H-SiC semi-insulating substrates has a strong temperature dependence dominated by a single activation (ionization) energy of-~0.3 5 eV identified as the B acceptor level.[59]The temperature dependence of the vanadium-compensated 6H-SiC material is also strong but more complex, with at least three activation energies identified resulting from residual boron and the complex behavior of vanadium. Vanadium has been proposed as both a donor (ED =1.35 eV) and an acceptor (EA= 0.8 eV). These temperature dependencies should obviously be taken into account when examining limits to high temperature operation of high frequency SiC devices fabricated on semi-insulating SiC substrates.

2.2

Doping of Silicon Carbide

Epitaxial Growth of SiC. Doping in SiC for device fabrication is accomplished via epitaxially controlled doping and hot ion implantation. Temperatures required for diffusion are too high (greater than 1800~ for standard device processing because of the very high bond strength possessed by SiC. The two most common dopants used in SiC are nitrogen (n-type) and aluminum (p-type), as discussed previously, although boron is sometimes used for p-type implants. In the absence of diffusion, epitaxial and ionimplanted control of dopants are critical for the development of devices and IC's. Numerous high-quality publications on epitaxial growth processes exist (see for example, Refs. 22-23, 40, and 60-75). Silane and propane are typical source gases of Si and C, respectively. Typical growth rates for 6H-SiC homoepitaxy layers on Si-face n-type substrates are--3 ~tm/hour. For

Processing of Silicon Carbide for Devices and Circuits

195

example, a prototype horizontal flow epitaxial growth reactor used at Northrop Grumman Science and Technology Center for homoepitaxial 4H-SiC and 6H-SiC growth is shown in Fig. 8. Increasing the growth rate while maintaining polytype control, good surface morphology, uniform thickness, and accurate control of dopant levels is critical in producing thick (50-100 ~tm) blocking layers necessary for high-voltage (>5 kV) power devices. Growth rates of up to 6 ~tm/h have been demonstrated in vertical low pressure CVD systems manufactured by EMCORE, Inc.[TMIn this system, high speed rotation of the wafer carrier was employed to stabilize the gas flow in the reactor at a growth temperature of-1500~ The resulting epitaxial layers were characterized to have breakdown strengths of 2 MV/cm, electron mobility of 700 cm2/V.s in lightly doped 4H-SiC epitaxy, and a background concentration of approximately 2 x 1014 cm-3 (boron). Doping levels were controlled over a magnitude order range of three. Other work using hot-wall CVD reactors has produced impressive results, with 45 ~rn thick epitaxial layers grown and used in 100 ~tm diameter 4.5 kV diodes. [74]

Figure 8. A horizontal Vapor Phase Epitaxy (VPE) growth system used to grow SiC epitaxy at rates of~3 mm/h at temperatures of 1500-1600~ Photo courtesy of Northrop Grumman.

For homoepitaxial growth of 6H-SiC, the substrates used are normally silicon-face, 3.5 degree off-axis, while for homoepitaxial growth of 4H-SiC, silicon-face 8 degree off-axis substrates are generally used. The reason for the difference in off-axis angle for the different polytypes is based upon the greater step height in 4H-SiC. Using on-axis or non-optimized

196

Wide Bandgap Semiconductors

off-axis wafers or non-optimized growth conditions for growth of epitaxy results in the formation of triangular morphological defects, caused by inclusions of 3C-SiC which interrupt the step flow mode of growth. These defects have been shown to cause high leakage currents in diodes. These and other surface morphology defects such as Si-droplet formation, faceting, and hillocks are reported in greater depth elsewhere.[43][73] Nitrogen and aluminum doping have been investigated for many years; however, residual doping levels were too high for many devices. The discovery of the site-competition effect[ 611for Si-face SiC epitaxy enabled the reduction of residual doping levels to 1014c m "3 and intentional incorporation over the entire range of possible concentrations. An example of controlling the nitrogen doping of SiC via gas Si/C gas ratio during growth is seen in Fig. 9,[75] for the same reactor as pictured in Fig. 8. This discovery has opened the way for much of the device results shown in subsequent sections. The site-competition effect works by adjusting the Si/C source gas ratio in the growth reactor to control the amount of dopant incorporated into substitutional SiC crystal lattice sites.[63] The model is based upon N and C competition for C sites, with AI and Si competition for Si sites in the SiC lattice. This effect has also been observed for boron and phosphorous doping of SiC.[61]Growth on Cface substrates has behaved quite differently, and more work is still needed to fully understand all growth mechanisms[ 611and tie together behavior on both faces. Numerous industrial and university laboratories now produce homoepitaxial, device-quality growth of SiC.

~I - - - l x 1 0 ]u

E t-

.s ~ c-

o

o

lx10

18

I xlO 11

1• lx10

L,.

]~ 15 ,

,--,_ 9,9 lx1014 o

J

,h.

[] i

, / / ' 6 "~'-

lx10

0

0.2

0.4

0.6

0.8

1

h pu t SJ IC Ratio

Figure9. Controlof doping demonstratedby varyingthe Si/Cgas ratio. TakenfromRef. 75.

Processing of Silicon Carbide for Devices and Circuits

197

Ion Implantation. Ion implantation is also proving to be a vital component of device processing in the absence of a usable diffusion process. Ion implantation has several critical roles in device fabrication. First, high-dose, shallow depth implants are often used in order to reduce the contact resistance to SiC. This type of implant is of particular importance when forming a contact to a lightly doped layer. A common example would be to use implantation to form n § drain and source regions in an n-channel SiC MOSFET. Nitrogen implants are typically used to form n § regions while aluminum implants are normally done to form p+ regions. A second use for ion implantation would be to form the p-wells in CMOS and planar power MOSFET's. Obtaining a deep implant into SiC is particularly difficult. Typical implant systems have energy limits of around 400 keV, which is insufficient for implants of >0.8 ~tm depth. Higher energy implants using MeV range energies offer the promise of deeper implants, but the amount of crystal damage may be very high, and impossible to remove. Numerous papers may be found in the literature regarding dopant ion implantation into SIC.[ 76]-[94]Finding the optimum high-temperature implant (eg., 500~ to 1000~ and subsequent high-temperature anneal (eg., 500~ to 1700~ which will fully activate the dopants, prevent amorphization of SiC, and not cause damage to SiC epitaxial layers (remembering that SiC epitaxial layers are typically grown at temperatures <1700~ has proven challenging. A summary of the best reported sheet resistivity values for post-implanted n-type and p-type SiC surfaces is shown in Figs. 10 and 11, respectively. Two typical examples of recent work which contain information on nitrogen, boron, and aluminum implants are found in Refs. 77 and 88, although several other comparable papers exist. Of particular concern has been the p-type implant. Boron has proven easier to implant for deeper implants ( >0.5 Ixm) than A1 because A1 has greater atomic mass than B and causes much more lattice damage. High-temperature ion implantation is preferred, since room temperature ion implantation results in amorphized material, while high-temperature implants do not. In Ref. 87, A1 and N implants were performed at 850~ and 700~ respectively, into Si-face, (100) 6H-SiC, followed by a 10-45 minute anneal at 1100~ to 1650~ in Argon. All implants were single energy implants ranging from 50 keV to 3 MeV. SIMS (secondary ion mass spectroscopy) analysis revealed no N redistribution after anneal, although a slight out-diffusion of A1 was reported. Also, due to its higher atomic mass, A1 implants caused more lattice damage than N implants as confirmed from RBS (Rutherford backscattering) data. Aluminum, because of its deep acceptor level and large mass, was not fully activated at anneal temperatures less than 1500~ indicating

198

Wide Bandgap Semiconductors

that higher anneal temperatures are necessary for A1. Si and C co-implantations were also performed in attempts to increase A1 activation, but C had no effect and Si co-implantation decreased dopant activation.[S7H ss]

101 '

al.

' J . Gar(Iha', et

' ...... w

.......

J. Bec. Ma1:.,25,885[1996) im plant t mz p. = 700 *(2 adnneed temp. = 161)0 *C anneal l i m e : 15 rain. 9 j \ j i i

-

C

i '5

io o

T. Kimot o, etal. j . R ec. M~., ~ , 2 3 ~ 1995) implant t r a p . = 23 *C ormeed temp. = 1500 anneal t i m e - 3Omin.

3

!

.- ........................... , ........................... .~. ...... ~ ...... . ................

_

mm

4-1 m

-

i

-

.

:

.

~

:

lib

41o

*

mi

N. mnoue,:~t d .

-'

m

.

er~

I

Inst. Phys. C m f . S

3

"

No. H2, 525 (199q implant tem p = 5 0 0 - 860 "C anneed tmzp = 1500 anneal tim e = 30 rain. , , , , ,,,I

1 0 "i

!

1 0 is

l

I

l

I

I Ill

l

I

l

1014 Implant

I

i

I Ill

1 0 ts Dose

I

1 0 le

l

I

l

l

l

Ill

1017

(cm -z)

Figure 10. Best reported sheet resistivity in n-type SiC as a function of implant dose. Implant and anneal temperatures (including references) are given. Figure courtesy of J. Cooper, Purdue University.

The effect of implant temperature is still not fully understood, but studies investigating it have revealed some key information. In one such study, [89] N implants at various temperatures into 4H-SiC and 6H-SiC ptype epi layers were examined. When the total dose exceeded 4 x 10 ]5 cm2, room temperature implants resulted in the creation of a totally amorphous layer, which remained heavily damaged even after high-temperature (1500~ anneals. Under the same high-dose conditions, the damaged regions could be annealed back to almost perfect crystallinity if a hightemperature implant was used. However, to achieve full activation of

Processing of Silicon Carbide for Devices and Circuits

199

implanted N, annealing temperatures in excess of 1500~ should be used. When comparing high-temperature implants from 500-950~ it was also found that the surface of the 950~ implanted material was Si depleted; graphitic C-C bonds were observed, while the 800~ implanted material retained the stoichiometric composition as verified by XPS (x-ray photoelectron spectroscopy). This is consistent with earlier research which has found that the surface of SiC at high-temperatures undergoes a preferential desorption of Si atoms resulting in graphitic layers formed on the surface. Thus, after ion implant and/or anneal temperatures which exceed-~950~ a sacrificial oxidation is normally performed to consume the (often conductive) graphitic surface. Another method to prevent loss of Si and maintain surface morphology has been to anneal in a silane (silicon) ambient. This has been demonstrated using an epitaxy reactor by ABB, Cree Research, and Mississippi State University with extremely promising results, as shown in Fig. 12.

1 0s

'

'

'

'

'

''''I

'

.

,

,

,

If,,|

.

,

.

,

li,,]

,

i

i

,

,,ij.

-"

." 9

T. K i ~ t o , ~ al.

-

~

J. Elec. Mat..25.879(1996) -

-

!

.=

!

i 1 i

C

1 0a

.............................

implant temp. = 23 ~ anneal tonnp. = 1500 ~

~ ............. .ann~..I

:

...

i i i

i9

:.

9 ...

:9 i

i :

:9

'5 m m

.m

o,m

~

: : i

i

.m

t ~ e = 30 min.

O

m .

"

:"9

-

O

101

!. . . . . . . . . .i.. .. .. . . . . . .j........ . . . . . . . . . . . . . . .!. . . . . . . . . .

3

.=

.."

9 9

10 0 1 0 ~s

, i ,,,,,,I

I , ,,,,,,I

1014

101~

Impl an t D o s e

I

I

I

I

IIIII

1016

I

I

I

I

IIII

1017

(cm- z )

Figure 11. Best reported sheet resistivity in p-type SiC as a function of implant dose. Implant and anneal temperatures (including references) are given. Figure courtesy of J. Cooper, Purdue University.

200

Wide Bandgap Semiconductors

N, c m "3

10 ~8

:....

N, cm "3

.:..

. . . . . . . . . . . . . .

. . . . . . . . . . . . .

:-..-..

::: : . . : .

-..::

9::..

:

.... ::

.:.....-.

10'*

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

10 '7 10 ~T I 0 ~6

10 ~=

t4

2.5

3.0

3.5

4.0

4.5

Depth, l~m

10 0.25

i

0.30

0,35

0.40

Depth, I~m

Figure 12. Implant activation of Al-implanted (p-type) SiC is demonstrated by CV extracted dopant density before and after anneal. Anneal was performed at 1600~ in an epitaxy reactor with silane overpressure. (Figure courtesy of Dr. S. Saddow, Mississippi State University.)

Other work has looked at implantation of hydrogen and boron into SiC. Hydrogen is important to study because it is present in high concentrations during the epitaxial growth process, and because it also plays a significant role in affecting the SiC/SiO 2 interface. In Ref. 94, boron was implanted at 700~ 350 keV and a dose of 1014 cm "2, and then activated by a 1700~ anneal in a SiC container with an argon atmosphere. The resulting junction depth (at-1018 cm -3) was 0.6 ~tm. Unlike N implantation, the activation anneal can result in appreciable diffusion of the boron implant species. Implanting hydrogen at the same dose (T = 80 K, energy of 80 keV) yielded a concentration of 5 x 1018cm -3 at the same junction depth, and subsequent low temperature photoluminescence has indicated the presence of defects incorporating implanted boron. Little has been reported on the use of MeV implants for the formation of p-type wells in planar CMOS and power device structures. However, the issue ofrecrystallization in MeV implanted 6H-SiC has been examined experimentally in Ref. 93. Wafers of 6H-SiC (0001) were implanted using a tandem accelerator at 160~ with 8 MeV Si 3+ ions (dose = 1 x 1016 c m "2) at a 7 ~ tilt from the [0001 ] direction to prevent channeling. The goal of this implant was to obtain a 1 ~tm deep amorphous SiC layer, and the implant was followed by a 30 minute, 1000~ implant anneal in nitrogen for recrystallization. Cross-sectional TEM micrographs revealed amorphous layers

Processing of Silicon Carbide for Devices and Circuits

201

between 2.6 and 3.4 ~tm from the substrate surface before anneal. Surprisingly, from diffraction pattems, the amorphous layer was completely recrystallized after the anneal, indicating a much lower temperature anneal is required than for keV implants which typically require anneals in excess of 1450~ Also noted was the presence of 3C-SiC material in the midst ofthe recrystallized 6H-SiC, which apparently grows epitaxially in the recrystallized region. Finally, while layer-by-layer growth of 6H-SiC occurred initially during the recrystallization, columnar growth of 6H-SiC followed along with epitaxial growth of 3C-SiC during the recrystallization process. The resulting mismatch between layer by layer growth and columnar growth resulted in stacking fault formation in the columnar 6H-SiC. To summarize SiC doping techniques, both ion implantation and epitaxial controlled doping should both prove vital processes for SiC device fabrication. Ion implantation has the advantage of selective doping, which is important for complementary logic structures, power device termination, and isolation. Ion implantation also results in a more planar device topology as compared to strict epitaxially controlled doping techniques, which have benefits in terms of photolithographic resolution, uniform final passivation coverage, and similar issues. Epitaxial doping has the advantage of not inducing lattice damage which eliminates the need for a high~ temperature anneal and the requisite creation of defects. Also, micropipe defects may affect ion implanted doping profiles by channeling the implanted species deeper into the SiC. Epitaxial regrowth after ion implantation, or annealing in a silane ambient to protect surface morphology, shows the vital link between implant and epitaxy required for advanced device fabrication. Al-implanted 4H-SiC in an epitaxy reactor with silane overpressure is shown in Fig. 12.

2.3

Etching of Silicon Carbide

Wet etching of SiC has not proven to be feasible from a practical device processing standpoint, as it requires molten salts (for example, NaOH-KOH at 350~ to be used at high temperatures. The difficulty encountered in etching SiC results from the high bond strength, a property which makes SiC useful for high-temperature operation, but a hindrance in fabrication. Nonetheless, numerous dry etches (primarily focused on Reactive Ion Etching processes) have been developed for the various polytypes of SiC. [96]-[120] Electron Cyclotron Resonance (ECR) etching has also been employed. [1111-[1141Most published RIE etches all make use of fluorinated gases (typically SF6, CHF3, CBrF3, CF4, NF3) to etch SiC, although etch rates

202

Wide Bandgap Semiconductors

of 1900 A/minute have been obtained in 6H-SiC using a chloride-based (C12/SIC14/O2) etch with SiO2 mask.J1~ RIE etch rates of 6H-SiC and 4H-SiC are typically slow in comparison to Si (300 A/minute to 2000 A/minute), with residue-formation problems commonly found, although not prevalent in all etches. A good overview of SiC etching is also found in the text, Properties of Silicon Carbide. [95] Etching of silicon in fluorinated gas has been found to occur by the reaction mechanism below: Eq. (2)

Si + 4F ---~SiF4

The removal of C has been debated in the literature, with some works indicating that addition of oxygen removes the C, as illustrated by Eqs. 3 and 4. Others claim that the C removal is via physical bombardment (see Eq. 5) or reactive chemistry between the fluorine and carbon. Depending upon the etch chemistry and process parameters (RF power, chamber pressure, electrode area and spacing, etc.), any one of the three mechanisms may actually dominate. Three of the most common fluorinated RIE mechanisms for removal of C proposed in the literature are listed below in Eqs. 3-5: Eq. (3)

C + xO -->CO o r

Eq. (4)

C + xF --~CF4 or CF 2

Eq. (5)

Physical (Ionic) Bombardment

CO 2

Several examples of residue-free RIE of 6H- and 4H-SiC can found in the literature.[96]-[99][l~176176176 For RIE with high etch rates and low surface damage, using high pressure (in comparison to the normal 10 to 100 mT pressures used in RIE) etch recipes seems to be prevalent. For example, in Ref. 97, high etch rates (-~1500 A/minute) were obtained using only NF 3 in a high-pressure (225 mT) RIE recipe. Using a commercial DryTek Quad 400 system, the RF power was 275 W, and the self-induced bias was only 25-50 V. A SEM of a typical sample after RIE is shown in Fig. 13.[97] Another group, using a Plasmatherm Series 790, experimented with 190 mT etch recipes using SF 6 or CF4, and N 2 and 0 2 additives in etching 6H-SiC.[ 1181The peak etch rate was found to be ---2600 A/minute for the following parameters: pressure: 190 mT; SF6 gas flow: 40 seem; 02 gas flow: 10.6 seem; and power: 300 W. Additionally, from atomic force microscopy, a 150 s etch was sufficient to actually remove residual damage leR from the as polished surface. Others have also reported high etch rates of up to 2200 A/minute at 82%O2/18~ r e m o t e plasmas at 330~ [1~176 This etch

Processing of Silicon Carbide for Devices and Circuits

203

produced minimal surface damage since the SiC sample was etched downstream (about 50 cm) from the microwave generated (400 W @ 2.45 GHz) plasma and the pressure was nearly 1 T. This particular etch was performed on the carbon face of 6H-SiC using an evaporated 500 nm thick aluminum mask. When performing the same etch on silicon-face 6H-SiC, higher percentages of 0 2 (~80%) were required to achieve the same high etch rates.

,i

t (a)

(b)

Figure 13. Scanning Electron Microscope(SEM) photographwith (a) 18.1 kX magnification of a 4H-SiC sample, and (b) 3 kX magnification of an etched profile for a 6H-SiC sample. Taken from Ref. 97.

Obtaining a very low etch rate is also important in many device fabrication unit steps. For example, in the fabrication of a bipolar transistor, when making contact to the very thin base region (as thin as 10 nm), it is critical to have a very slow and well-controlled etch recipe. Several etch recipes utilizing R/E at lower pressures (--10-100 mT) and high selfinduced bias (>300 V) can also be found in the literature.[961199]These etch processes often produce more surface damage,Ilia] but are critically needed in some processes. Perhaps because of the high self-induced bias, aluminum micromasking, which may be a result of aluminum being etched from the powered electrode or the mask material, is more common in these etch recipes. Covering the electrode with graphite and/or adding hydrogen to the fluorinated gas chemistry has been shown effective in preventing residue formation.[961[1~176

204

Wide Bandgap Semiconductors

While higher pressure etches have been used to obtain low surface damage, residue-free etching, and high etch rates, other approaches have also been utilized towards those goals. ECR etching has been shown to produce very smooth etched surfaces, and the degree of etch anisotropity can be controlled by the substrate bias.[ 113]ECR etching has the advantages of higher density plasma and independent control of the substrate bias in comparison to RIE etching. When comparing similar RIE and ECR processes using a CF4/O 2 mixture, it was found that the RIE process significantly damaged the SiC surface and also left A1, F, and O impurities on the sidewalls. The ECR process did not.[ 114]The surface damage was measured by comparing the Schottky barrier height of Pd Schottky diodes on ECR etched, RIE etched, and unetched SiC surfaces. However, this particular RIE process has a very high self-induced bias (500 V) in comparison to other SiC RIE processes;[99][ 118]this comparison may vary for different RIE processes. The ECR etching of single crystalline 4H-SiC cantilever structures using a combination of high and low substrate bias has been demonstrated.[ 113]Using a sputtered indium-tin oxide mask, results of ECR etching of 4H-SiC and 6H-SiC in a 1 mT, 17%O2/83%CF 4 plasma indicated a strong dependence of etch rate and etching isotropy on the position of the sample relative to the ECR plasma and the substrate bias. High substrate bias (100 V) produced smoother surfaces and higher etch rates (700 A/minute), but also exhibited trenching (deep trenches at the bottom comers of the etched patterns). Another form of etching has also recently been reported for SiC utilizing Inductively Coupled Plasma (ICP) etching. ICP creates a much denser plasma than traditional RIE by using two RF power supplies instead of one. ICP etching has shown to offer low surface damage, residue-free etching, and fairly high etch rates. Plasma densities in the 1 x 1011 to 1 • 10 ~2ions cm -3 range can be obtained, with the incident ion energy controlled independently by decoupling plasma generation power and bias power. Preliminary ICP etching of 6H-SiC in CF 4 and 0 2 has been investigated using a commercial PlasmaTherm ICP etching system.[ 1191 Using ITO masks, peak etch rates of 1500 A/minute were obtained at 900 W. Sample bias was varied from 10-60 V, with higher etch rates obtained for higher bias values. Using Schottky diodes, the ideality factor (1.1) was the same for both ICP etched and non-etched 6H-SiC. Additionally, the leakage currents of both sets of diodes was similar, indicating that the surface damage was minimal. A cross-sectional SEM of trenches etched into SiC using ICP etching is shown in Fig. 14.

Processing of Silicon Carbide for Devices and Circuits

205

Figure 14: SEM of SiC trenches etched using an Inductively Coupled Plasma (ICP) etch process. Photo courtesy of J. Zhao, Rutgers University. Photoelectrochemical etching (PEC) of SiC has produced very high etch rates. Several examples can be found on PEC as well, but ICP, RIE, and ECR etching currently appear more compatible with small-feature device processing (for example see Ref. 117). However, etch rates of up to 25 lam/min, for n-type 6H and 3C-SiC have been reported with dopantselective etch stops possible,[ 12~ which may be of use in deep trench or via etching. The dopant-selective etch stop could also be of great importance when etching back to make contact to very thin epitaxial layers of SiC, such as the base layer of a SiC bipolar junction transistor. The etch process is briefly described as a hole-catalyzed surface dissolution, with the holes supplied from bulk p-type SiC or from I_W photogeneration in n or p-type SiC. The transfer of holes into the electrolyte promote oxidation of the surface. The oxide is removed with the HF acid contained in the electrolyte.

206

Wide Bandgap Semiconductors

2.4

Oxidation and Passivation of SiC

The ability of SiC to oxidize and form SiO 2 has allowed compatibility with standard silicon-based fabrication. Obtaining a high-quality oxide with low interface state and oxide trap densities has proven to be challenging because of the carbon on the surface, as well as off-axis epitaxial layers which often have rough surface morphologies. Thermal oxidation rates are considerably slower than that of silicon, so typical oxidation temperatures for SiC are often higher than those ofSi (-~1050-1200~ Growth rates of SiO 2 on p-type 6H-SiC are shown in Fig. 15 [124] for both wet and dry oxidation, as well as silicon-face and carbon-face material. Growth on carbon-face SiC is much higher (-5 X) than the growth on silicon-face SiC. The slower oxide growth on Si-face material has been attributed to an interracial oxide layer (Si4C4.xO2 with X -2), identified by angle resolved xray photoelectron spectroscopic analysis, which forms on the Si-face 6HSiC surface and retards the oxidation rate.[1221[142]6H-SiC was polished and cleaned by HF and R/E, and then oxidized for 15 min. at 1100~ in dry O 2. The interracial layer, which has a thickness of about 1 nm situated between SiC and SiO2, is believed to be a reaction product of molecular oxygen bonded peroxidically to a SiC double layer. This interfacial layer can react with oxygen to form Si4C404, SiO2, and CO.[ 122]No interracial layer was found on the carbon-face material, but both silicon-face and carbon-face material had layers of graphite, CxHyOz, and Si4C404 formed on the surface of the oxide. The above discussion of oxide growth and quality is true for oxides grown on standard silicon-face (0001) and carbon-face (000]-) surfaces. The oxide growth rate and quality also varies considerably for other crystal orientations such as (1120) and (1]-00) a-axis orientations.[ 123]Generally, the growth rate on a-axis orientations is 3-5 times higher than that of the silicon-face c-axis SiC material, and the interface state densities are 4-10 times higher as well. This has important ramifications for UMOSFETs, VMOSFETs, and other trench gate MOSFETs in SiC. As an example, a cross-sectional SEM from a 4H-SiC UMOSFET is shown in Fig. 16. The thin white line along the edge of the U trench is SiO2, and as can be seen, this oxide is thicker along the sidewalls than on the bottom of the trench. The higher sidewall Dit also is detrimental to device performance by limiting channel mobility and thus increasing the on-resistance of the device. The high fields present in the device often cause breakdown of the gate oxide via field crowding at the comers of the UMOS trench. The vertical

Processing of Silicon Carbide for Devices and Circuits

207

sidewalls problems relating to variation of oxide growth and quality are also well documented elsewhere.[127][133][16~

2,5 1126 "C

1200 9

E =l. 2,0 c

1090 ~

1,5

Jg .o_

1050 ~

.~_ 1,0

1125 "C

X 0

1090 ~

0,5 1050 "C 0'00

100

200

300

400

oxidation time [hi

! O~ E ~2

Q

i' 0

l I ...

0

1125 "C ~ - - f ~ .

20

40

1090 '~:

1060"C

.....

4 ....

60

80

,c

100

120

oxidation time [hi

Figure 15. Oxidation rates in dry (upper) and wet (lower) 0 2 environments for p-type 6HSiC. Open symbols are for C-face and solid symbols are for Si-face. Taken from Ref. 124.

208

Wide Bandgap Semiconductors

Figure 16. Cross-sectional SEM of a completed 4H-SiC UMOSFET.The thin white region is identified as the gate oxide, with a thicker polysilicon layerdeposited over the gate oxide.

Photo courtesy of Northrop Grumman.

Numerous studies have been published about cleaning and oxidation procedures.[ 121]-[160]The goal of low (comparable with silicon) interface state densities is close to fulfillment. One cause attributed to high interface state density of oxides on SiC is that of C. The problem of C at the SiC-SiO 2 interface is compounded by the high temperatures required for reasonable oxidation rates. When exposed to these high temperatures, the surface of SiC often undergoes a preferential loss of Si, resulting in a graphitic surface layer increasing the carbon concentration.[ 155] Three distinct temperature regimes have been identified for 6H and 15R-SiC. Below 900 K (627~ no graphitization of the SiC surface is observed. Between 900 K (627~ and 1300 K (1027~ both Si-face (0001) and C-face (000T) SiC surfaces are terminated in a surface graphite layer. Above 1300 K (1027~ the C-face SiC graphitizes at a rate higher than the Si-face SiC. Both surfaces exhibit massive graphitization above 1027~ as a result of Si sublimation from the SiC surfaces.[ 155] Because of this problem, SiC wafers are otten loaded into oxidation fumaces at lower temperatures (<950~ in an oxidizing environment, then ramped to a higher temperature (1100-1200~ to obtain the desired oxide thickness. Finally, they are ramped back down to a lower temperature in an oxidizing environment for annealing and/or unloading of the

Processing of Silicon Carbide for Devices and Circuits

209

wafers. This technique has proven effective in lowering the interface trap density on Si-face 6H-SiC to near 1 x 1011per (cm2/eV).[121][13~ In SiC, unlike Si, it is important that C-V measurements be performed at high temperatures (>300~ in order to see the response time of interface states lying deep (>0.6 eV) in the bandgap of SiC. Standard roomtemperature C-V analysis will not reveal the true interface state density. Care must also be exercised because of the large surface potential fluctuations present in SiC C-V measurements.[ 138]J. N. Shenoy et al.[ TM]reported interface state densities (Dit) near the 1 x 1011 per (cm2/eV) order of magnitude, and fixed oxide charge (Qox)of 1 x 1012 (C/cm 2) in thermally grown SiO 2 on 6H-SiC. Another method of reducing Bit and Qo, has been to perform post oxidation anneals in oxidizing environments.[ 156]On p-type (Al-doped) silicon-face 6H-SiC epitaxial layers,Dit of 1 x 1011 cm 2/eV and Qox of 1.0 x 1012 cm-2 was measured on the best oxide, which used a 1050~ oxidation (25 h) and a 3 h post-oxidation anneal in 0 2 at 950~ The interface trap density was extracted using the conductance technique, with measurements taken at 250~ Post-oxidation anneals in oxygen may have a secondary benefit, which would be densification of the SiO 2 to increase oxide reliability and strength. This is evidenced by the high breakdown strength of these oxides (up to 11.5 MV/cm). Time-dependent dielectric breakdown results on these oxides indicated lifetimes of 700 years at 2 MV/cm and 350~ when testing under a negative bias on p-type 6HSiC.[ 158]However, under positive bias on n-type 6H-SiC, the lifetime was significantly less, although no specific numbers were reported. Positive bias stressing was not performed on oxides grown on p-type 6H-SiC, due to the difficulty in forming inversion. The apparent unreliability of oxides under positive bias can also be supported by the finding of hole traps in oxides on SiC (which were not treated to the same annealing conditions).[157] The hole traps (in oxides grown at 1120~ in dry 0 2 with standard 1 h Ar anneal) were found to be related to oxygen vacancies in the oxide, and no direct contribution of impurities such as A1 or C to the trapping was observed. Additionally, thermally unstable interface states were generated during hole injection at the SiC/SiO 2 interface, similar to that of SiO 2 on Si. All of the reliability issues are further supported by the evidence of Fowler-Nordheim tunneling across oxides grown o n SiC. [133][160] That FN injection is more severe for the SiC MOS system is not surprising considering the wide bandgap and reduced transport barrier of the SiC MOS system as compared to silicon. For example, Fig. 17 compares the bandgap, electron affinity, and subsequent barrier for tunneling of silicon, 6H-SiC, and 4HSiC MOS systems via an energy band diagram. In 4H-SiC, the barrier for

210

Wide Bandgap Semiconductors

electrons in the conduction band is only 2.70 eV as compared to 3.15 eV in Si. Thus, the Fowler-Nordheim tunneling probability (exponentially dependent upon the electron barrier height) is much higher for SiC, especially at elevated temperature and bias where SiC has many targeted applications. In fact, for a 4H-SiC n-channel power MOSFET, it has been recommended that the maximum positive applied bias be limited to 1.5 MV/cm at room temperature[1601 which in tum limits the maximum gate bias applied under the on-condition as well as placing constraints on the doping of the p-type channel. Type 6H-SiC is also preferred for MOSFET devices since 4H-SiC has a higher Dit at the conduction band edge.

Z l

3.15

1.1,Si

2.95

T

2.70eV

,,. ~ - E o

? 2.85

3.25

'

,

,

i

3.2

3.05

6H-SiC 4H-SiCI 9.0 eV s!c2

4.75

Figure 17. The energy band offsets of 6H-SiC and 4H-SiC with respect to SiO 2. The barrier for electron injection from 4H-SiC into SiO 2 conduction band is significantly lower than silicon. Taken from Ref.4.

Other high-temperature, high-power operation and reliability issues remain to be addressed. Alternative insulators (such as nitrides and oxynitrides) are also being pursued for high-temperature device applications, just as in the silicon industry. One example involves thermal oxidation of SiC in N 2 0 . [152] Diffusion of CO through the oxynitride was attributed to be the limiting factor in oxidation, and Qoxwas reported to be on the order of 1 x 1012C/cm2. The oxidation rate was found to be initially parabolic with time, eventually switching to linear after some

211

Processing of Silicon Carbide for Devices and Circuits

time, just as the case for Si. It is believed that the introduction of nitrogen into the oxide (or oxynitride) should reduce the Dit, although no specific numbers were mentioned in this work. Other work involving the annealing of oxides with N20 has not proven successful initially, although anneals in NO have shown promise for improvement of oxides on p-type SiC. [159] A quantitative and practical assessment of using thermally grown gate oxides in SiC devices has also been done by researchers at General Electric. [154]Because doping of SiC devices is via implantation or epitaxial growth, most MOSFET devices are fabricated using a non self-aligned process. In fact, the drain/source implantation' s are normally done prior to gate oxide growth. Since the structure is non self-aligned, the gate oxide will partially overlap the implanted drain/source regions. However, thermally grown SiO 2 on implanted regions of SiC has a breakdown strength of only 1-6 MV/cm as compared to 10 MV/cm on non-implanted regions. Premature breakdown of SiO 2 grown on implanted 6H-SiC is shown graphically by Fig. 18. Premature breakdown of the gate oxide in SiC MOSFETs fabricated using this type of process is thus to be expected.

dlO

T = 380~ C 30

1

20

10

0

1

2

3

4

8

8

T

tl

g

10

S (My/era)

(a) Figure 18. (a) BreakdownofSiO2on implantedSiC surfaces.ReproducedfromRef.154. (b) Breakdown of S i O 2 o n non-implantedSiC surfaces. Reproducedfrom Ref.154.

212

4~

Wide Bandgap Semiconductors

--

T m 3110"C

w

~0"

i.i 10

0 0

1

2

$

4

5

8

?

II

I)

E (UV/~)

(b) Figure 18. (Cont'd.)

2.5

Ohmic Contacts to Silicon Carbide

Low resistivity ohmic contacts are essential for high-frequency operation. Additionally, high-temperature and high-power requirements maintain that the contacts must be reliable under extreme conditions. Once seen as one of the primary impediments to SiC technological development, the ohmic contact area has now matured rapidly.[161]-[189] An area still requiring extensive work, however, is that of ohmic contacts to p-type SiC. The p-type contact to SiC is made difficult because ofthe wide bandgap of SiC. As illustrated in Fig. 19 bythe thermal equilibrium metal-semiconductor band diagram (q~M < q~sic), the Schottky barrier (OB) to majority carrier transport must be reduced as much as possible in order to provide for an ohmic contact. Since the bandgap and electron affinity (c) of SiC are fixed, the remaining options for reducing the ~B are to choose a metal with a large work function (OM), and also to dope the ptype SiC as heavily as possible. P-type ohmic contacts to SiC often use some variation of A1/Ti alloys. A contact with specific contact resistance measured at 1.5 x 10-5 f~.cm 2 on Al-doped samples (NA = 2 x 1019 cm -3)

Processing of Silicon Carbidefor Devices and Circuits

213

has been reported by J. Crofton et al.[161]using an A1/Ti alloy. The specific contact resistance is a strong function of doping (seen clearly by Fig. 20, where the specific contact resistance is plotted as a function of doping for the A1/Ti alloy contact). Although A1 melts at ~660~ a 90:10 AI:Ti alloy (by weight) is a mixture of solid and liquid phase at temperatures of 9501150~ which are typical anneal temperatures used in the formation of ohmic contacts to SIC.[161]-[163] More recent experiments using a 90:10 AI:Ti alloy has yielded specific contact resistance ranging from 5 x 10 -6 to 3 x 10 "5 f2ocm 2 on p-type 6H-SiC doped at 1.3 x 1019 cm -3. On the same material, pure Ti (with a 1 minute, 800~ anneal) was also used to form ohmic contacts with specific contact resistance ranging from 2-4 x 10-5 f2.cm 2. Etching off the metals after annealing revealed that the Al-based contact spiked into the SiC, evidenced by pits in the SiC surface up to 2600 A deep, while the Ti contact exhibited little interfacial reaction. Thus, although Al-based contacts can yield exceptionally low specific contact resistances, the contact can suffer from poor reproducibility and aluminum oxidation during annealing (A1203). [164]

Eo

metal

~

p-type 4H-SiC '! Z = 2.70 eVNo

....... l !l

. . . . . . . . . . . . . . . . . . . . . . . . .

E F

Figure 19. Band diagram of metal and p-type 4H-SiC before and after making contact ignoring any effects of Fermi-level surfacepruning.

214

Wide Bandgap Semiconductors

10 -1

...... ' "

'--"

""-"1

9 '"-'~;;:"~

0 I

.....

w

. . . . .

"I

~ ..... " = " - ' ~ " ' ~

ClUW p - t l r p o 6 H - S t C

A

Net

.... "'"';""

H

10 - t

'

o.av ev

Zb -

e o 9

~,o

10

-3

0

p.

IIle ~ aIp

10

9 ,~

...-4

m

'~

O-S

0 0

10 "e 10 is

9

a A 9149

.... J_

10 is

__J__ 9

~__

9

10 ~7 doptnft

9A s---J

9

9~ . 9

10 TM lout 3

101o

9 A ,L . * * -

10 zo

)

Figure 20. Specific contact resistance as a function of doping on p-type 6H-SiC using an AI/ Ti alloy for ohmic contact. Taken from Ref. 161.

Because A1 is easily oxidized, other contacts such as boron carbide (for example, B4C), Ta, Ti, cobalt silicides[ 162]and Mo may prove to be superior for high-temperature ohmic p-type contact. As an example, refractory metal boride ohmic contacts to p-type 6H-SiC (doped at 1.3 x 1019 cm -3 with A1) have recently been reported.[ 1651Short anneals (2-10 min.) at 1100~ in vacuum (5 x 10-7 torr) yielded specific contact resistance of 8.2 x 10-5 f2.cm 2 for CrB 2 and 5.8 x 10-5 f2.cm 2 for W2B. Experiments with TiB 2 were also performed, but with varying results. Longer anneals lowered the specific contact resistance values further. Recent studies have indicated that formation of cobalt silicide (CoSi2) may prove to be a thermally stable, low-resistance ohmic contact with a specific contact resistance of <4 x 10-6 f2.cm 2 reported on very heavily doped (2.0 x 1019 cm -3) p-type 6H-SiC.[ ]62] This particular contact was formed by depositing

Processing of Silicon Carbide for Devices and Circuits

215

sequential layers of Co (50 nm) and Si (160 nm) followed by a two-step anneal at 500~ for 5 h then 900~ for 2 h. N-type contacts have been formed from a variety of silicides and carbides (such as Ni2Si and CoSi2). To our knowledge, the n-type contact based on nickel silicide (Ni2Si)[167]currently has been reported to have the lowest specific contact resistance (<5 x 10-6 f2.cm2 on 6H-SiC doped 7-9 x 1018cm-3), and has actually been reported to have lower contact resistance at 500~ than at room temperature. Other ohmic contacts using TiC to n-type 6H-SiC have also shown promise, but with somewhat higher contact resistances.[ 168]Numerous other metal contacts to n-type SiC have been tried, including Re, Pt, Ta, and titanium nitride.[174][181]Reliability studies are only now taking place.[ 169]Although preliminary results are encouraging, the best ohmic contact for high-temperature operation is still in question. Packaging issues and composite contact delamination at high temperatures must still be addressed. N-type ohmic contacts using NiCr may prove advantageous from contact delamination, reliability, and wirebonding issues.[ 182]Other work which may be advantageous from the point of postoxidation device processing includes research on low-temperature annealing cycles, or even ohmic contacts which are ohmic as deposited.[ 17~ High-temperature anneals (for instance of 1000~ [167] could degrade the interface quality and reliability of thermally grown oxides.[13~ Thus, slowly ramped anneals held to lower temperatures for longer time periods are sometimes preferred for annealing devices with MOS structures.

2.6

Schottky Contacts

Increasing the Schottky barrier height is extremely important for elevating the maximum operating temperature for Metal Semiconductor Field Effect Transistors (MESFETs) and Schottky rectifiers. While MESFET's are discussed below, some recent work on Schottky diodes[189][207]is quite worthy of further review. One of the primary uses of Schottky contacts on SiC, other than that of a gate metal in MESFETs, is for use in efficient, solid state, high-voltage switching applications,[ 1971which could prove to be an important market for SiC electronics. Taking advantage of SiC's high critical electric field for breakdown, it is projected that SiC Schottky diodes will have specific on-resistances much (up to 200 times) lower than that of Si. Metals used for Schottky contacts to SiC include Ni, NiCr, Au, Pt, Ti, Mg, Co, A1, and Pd. Actual Schottky barrier heights (SBHs) depend upon the surface quality of material, metal deposition process, polytype, conductivity

216

Wide Bandgap Semiconductors

(n-type or p-type), and whether the surface is Si-face or C-face. In addition, the measured SBH also depends upon the measurement technique (with CV measurements often yielding higher numbers than I-V measurements). Since C-V and I-V measurements often depend upon several factors, the actual SBH may lie somewhere between these measured values. Internal photoemission spectroscopy has also been used since this method is not dependent upon leakage current, series resistance, and variable ideality factors.[ 19~ On Si-face, 6H-SiC, Au and Ni possess two of the highest Schottky barriers of ~1.47 eV and 1.27 eV, respectively.[ 195] Tabulated values of SBH for Ti, Ni, and Au on 4H-SiC (both C-face and Si-face) measured by C-V, I-V, and internal photoemission spectroscopy are shown in Table 4 (reproduced from Ref. 190). Table 4. Schottky Barrier Heights of Au, Ni, and Ti on N-type 4H-SiC

Metal

Au Ni Ti

Surface

q)BI-V

~OBC-V

q)BIPE

Polarity

(v)

CO

6O

C

1.80

2.10

2.07

1.20

Si

1.73

1.85

1.81

1.08

C

1.60

1.90

1.87

1.08

Si

1.62

1.75

1.69

1.02

C

1.16

1.30

1.25

1.04

Si

0.95

1.17

1.09

1.03

IPE:intemal photoemission

As mentioned above, high SBH is desirable for high-temperature device operation because it has an exponential effect on the reverse bias current density. This can be seen from the thermionic emission expression for current density (J) as a function of applied bias shown below in Eq. 6, which assumes no series resistance: Eq. (6)

ex(q 1 where ,Is = A * T "2 exp(-q~s/kT)

Processing of Silicon Carbide for Devices and Circuits

217

The saturation current (Js), in addition to its strong dependence upon temperature (T), is exponentially dependent upon the SBH. The ideality factor (n) is normally measured to determine the quality of the Schottky diode. Effective Richardson constant (A*) has been defined as 72 A/ cm2K2 in alpha SiC. [166] Although this expression predicts the reverse leakage current for Schottky diodes to be constant (-10 -14 A/cm 2 for a SBH of 1 eV), actual leakage currents in SiC Schottky diodes have been much higher (indicating the presence of other leakage mechanisms). In fact, high reverse leakage currents prior to breakdown in SiC Schottky diodes are a result of mechanisms such as thermionic field emission and field emission.[ 1921 Since these are distinct mechanisms, the tunneling probability as a function of energy through a reverse biased Schottky diode has been approximated and compared to experimental data. The image force lowering of the barrier height was also taken into account. This is extremely important for high-voltage SiC Schottky diodes since the image force lowering of the barrier height is proportional to Emax~ (where Em~ is the maximum electric field at the metal-SiC interface).[ 192] From these calculations, for a SBH of 1.0 eV on 6H-SiC doped at 1 x 1017 cm -3, increasing the temperature from room temperature to 300~ resulted in a projected increase in leakage current by five orders of magnitude; an explanation for the very high observed leakage current in many highvoltage Schottky diodes. A variety of Schottky diodes have been fabricated on SiC. One of the more impressive results to date has been the 1750 V blocking voltage obtained with a Ti Schottky diode on C-face (5 ~ off-axis) n-type 4HSiC.[ TM] The SBH was approximately 1.17 eV for this diode, with an ideality factor of 1.03. The use of C-face material is sometimes desired since for Ni, Ti, and Au, the SBH has been found to be 0.2-0.3 eV higher for C-face than for Si-face material[ 19~ (as shown in Table 4). Ten ~tm thick n-type drift layers were grown via step-controlled epitaxy earned out in Sill 4 (0.30 seem), C3H8 (0.22 seem), and H 2 (3.0 slm) using an atmospheric chemical vapor deposition system at 1560~ 191]Doping density obtained from C-V measurements ranged from 7.0 x 101s cm-3 to ---2.0 x 1016 cm -3. Boron implantation performed at room temperature (energy~ 30 keV, d o s e ~ l . 0 x 1016 cm-2) and annealed at 1050~ was used for edge termination. The Schottky contact physically overlapped the boron implanted layer by about 15 ~tm. Although leaky, (these devices were unpassivated so surface leakage may have been a problem) the devices with boron edge termination exhibited blocking voltages up to 1750 V at room temperature; whereas those devices without edge termination broke down at 760 V.

218

Wide Bandgap Semiconductors

Many metals such as Ni and Ti, can form both ohmic and Schottky contacts,[167][189][2~ with short-term, high-temperature anneals (900~ converting the Schottky contact to ohmic contact. Since Ni and Ti are commonly projected for use in MESFETs, SITs, and Schottky diodes at high temperatures up to 300~ concerns over reliability of these Schottky contacts are well founded. Reliability data are just now beginning to surface, but Ni Schottky diodes on n-type 6H and 4H-SiC have been tested (unbiased) for 9000 hours at 300~ [193]In general, the SBH decreased (by about 0.2 eV) in all samples after 9000 hours, and the ideality factor increased in the majority of cases. This finding is of particular importance since high temperature annealing of Ni to form nickel silicide is the basis for ohmic contact formation on n-type SiC. Additionally, Rutherford backscattering spectroscopy analysis revealed no discernible silicide formation for Ni Schottky contacts after 4000 hours at 300~ Increased leakage currents were also observed after thermal aging, but argon implantation of the surface reversed this trend More work is necessary to understand the mechanism(s) behind this.

3.0

S U R V E Y OF SiC DEVICES

A wide array of devices in 3C, 6H, and 4H SiC have been demonstrated, including: the Bipolar Junction Transistor (BJT) [2081-[212][227], Insulated Gate Bipolar Transistor (IGBT),[ 213] MESFET,[214]-[223] the JFET, [7][41][223]-[232]the Metal Oxide Semiconductor Field Effect Transistor (MOSFET)[233]-[239] (which has been operated up to 923 K), [239] the Static Induction Transistor (SIT),[218][240][241] power MOSFETs,[ 242]-[248] thyristors,[221][2481-[252] Gate Turn-Off thyristors (GTO' s),[41[252] pn-junction diodes, [237][253]-[257] and Schottky diodes. [191][194][197][203][205][258][274] Additionally, heterojunction devices such as the Heterojunction FET (HFET) and Heterojunction Bipolar Transistors (HBTs) have been proposed and fabricated. Examples include: 3C-SiC/Si HBTs, [259]-[261] a 6H-SiC/3CSiC HBT,[ 2~ and a GaN/6H-SiC HBT.[ z62]A snapshot summary of fabricated SiC devices along with the highest recorded temperature of operation and comments are summarized in Table 5. While impractical to discuss each of the above mentioned devices in depth, a cursory treatment of selected devices will now be covered. As the technology matures (both from a materials and processing standpoint), SiC device performance records have been changing dramaticallyover a short time frame. No doubt, this

Processing of Silicon Carbide for Devices and Circuits

219

trend will continue; nonetheless the material presented in Table 5 and following is relevant to discuss early SiC device development as well as provide a snapshot in time of SiC device maturity. While no clear subdivisions of SiC devices are defined, low power SiC devices, high frequency and high power SiC MESFETs, SiC power devices, and the SiC static induction transistor will all be discussed as separate categories.

Table 5. Silicon Carbide Device Performance Snapshot (May 1997) Device

! Tmax(*C ) ,

DMOS (4H)

300

Ref. #

Comments

244

900 V, 5 mA devices

243

760 V, 5 mA devices; geffup to 26 cma/Vs; Ro, , = 66 mf)ocm2

252

700 V blocking

MESFET (4H)

222

65.7 % peak PAE; 3.3 W/mm at 850 MHz

MESFET (4H)

217-218

fmax = 42 Ghz; 5.1 dB gain at 20 Ghz; 3.125 W/mm at 10 GHz

PN Diode (6H)

256

4500 V blocking

Schottky Diode

202

1750 V blocking; Ti on C-face 4H-SiC

DMOS (6H)

GTO

[4H]

400

SIT

200

218,241

120 W at 2.7 Ghz (1 cm periphery)

Thyristor (6H)

300

251

5200 A/cm2 pulse switched current density and < 100 ns fall time

Thyristor (4H)

500

248

700 V, 6 A and 900 V, 2 A devices

UMOS (4H)

300

248

260 V, Ron = 18 m ~ ' c m 2 175 V, 2 A devices; (2.65 V forward drop)

UMOS (4H)

300

244

1100 V, Ro, , = 74 mF2.cm2 at 100 ~

220 3.1

Wide Bandgap Semiconductors Low Power SiC Devices

MOSFETs have been fabricated in 3C, 6H, and 4H-SiC with hightemperature operation demonstrated up to 650~ Aside from MOS reliability issues discussed earlier in the chapter, obtaining theoretical mobility values in the channel has proven difficult. Extracted mobility (/aeff) in the channel of a 6H-SiC depletion-mode MOSFET is shown versus temperature in Fig. 21 .[235] Channel mobility is lower than bulk mobility, and the difference is attributed to interface states between the oxide and semiconductor, material defects, mobility anisotropy, and other factors. Enhancement-mode SiC MOSFETs have had much lower effective channel mobility, presumably a result of deep-level interface traps near the oxide-SiC interface. The low room-temperature mobility of vertical SiC power MOSFETs with positive threshold voltages has been a problem of extreme importance as discussed previously and elsewhere. [242]-[248]Channel mobility in 4H-SiC JFETs with 9 ~tm gate length have been reported at 340 cm2/ V.s at 300K, [226] with a maximum transconductance (gin) of 15 mS/mm. Enhancement-mode JFETs have also been fabricated,[ 228] opening up the possibility of complementary JFET-based logic in SiC, possibly reducing the need for a high-quality gate oxide which can withstand high-temperature, and possibly high-power and high-radiation environments. In the enhancement-mode JFET, the pn junction space charge region at zero gate bias fully depleted the drain-source channel, causing the device to be normally off. The extrapolated threshold voltage (VrH) was measured to be 0.81 V at 300 K, with g~ of 1.5 mS/mm for a device with a gate length of 5 ~tm.[22s] The maximum gate voltage, before forward biasing the gate pn junction diode, fell from---2.5 V at 300 K, to ~1.9 V at 723 K. A good discussion of los s versus temperature can be found in the literature as well.[22s11229]Highvoltage JFETs have also been fabricated, with 6H-SiC JFET's capable of 450 V forward blocking voltage, although pinchoff-voltage was -400 V.[TM] Ro~ was found to vary from 20 mW.cm2 to 80 mW.cm2. BJTs in SiC typically have low DC current gain (-~5-15) in the common emitter configuration, primarily as a result of minority carrier base recombination in npn structures.[2~ Bandgap narrowing with very heavily doped emitters could also play an extremely important role in degrading the DC current gain. Heterostructure devices using wide bandgap emitters or narrow bandgap bases of different SiC polytypes or III-V nitrides will probably be needed to increase DC gain and improve the high-frequency response of the devices.

Processing of Silicon Carbide for Devices and Circuits

221

"%, Comparable Si MOSIO'I~

200 ~ t

""'....(Thmrea~) - T"~'s .

%%%,1 1~/

~'15oi- W/L= lOOjm#4j.tm ";.. .~ i vmfo.1 v; vssfO.OV "'--.. 9

|

..............

275

iC MOS~

It ............................

375

il

. . . . . . .

475

Tempenmre (K)

t

575

.............

.I

675

Figure 21. Effective channel mobility (meff)versus temperature for a 6H-SiC depletionmode MOSFETcomparedto a similar, theoretical device in silicon. Taken from Ref. 235.

3.2

SiC MESFETs

The use of SiC MESFETs for high-power, high-frequency applications has been the focal point for investigation because of the possibility of power densities in excess of four times that of silicon and gallium arsenide. The small-signal RF performance of SiC MESFETs fabricated by Northrop Grumman has been found to have 5.1 dB gain at 20 Ghz, with extrapolation of Mason's unilateral gain at 6 dB/octave resulting in afmaxvalue of 42 Ghz.[217] Impressive high power operation of X-band 4H-SiC MESFETs has been demonstrated as we11,[2181with power outputs of 6 W at 10 Ghz obtained from devices of 1.92 mm periphery and 0.5 ~tm gate length. The results were obtained under class AB conditions with gate biased at 10% loss. Even higher power output (6.3 5 W or 3.3 W/mm) was obtained at a higher input drive level, VD = 45 V, under 10% duty cycle of 100 ~ts width. Gate-todrain breakdown was approximately 100 V for these devices. Impressive SiC MESFET performance has also been demonstrated by a Motorola/Cree team with 65.7% peak power added efficiency and an extremely high power density of 3.3 W/mm at 850 MHz.I252]

222 3.3

Wide Bandgap Semiconductors SiC Power Devices

Power MOSFET's with 260 V blocking capability and Ro, of 18 mf2.cm2 have been obtained [248] for vertical 4H-SiC UMOSFETs. Also, larger area devices, such as 175 V MOSFETs with a rated current of 2 A (J= 200 A/cm 2) at a forward drop of 2.65 V, were also demonstrated. Operation of similar devices had previously been demonstrated at temperatures up to 300~ with short lifetimes because of gate oxide failure, but no high-temperature data was reported for these later developments.[ 248] Because of SiC's superior electric field strength, Ro, has been affected by channel resistance, and not simply by the resistance of the lightly-doped drain-drift region, which is much thinner than that of comparable Si power MOSFETs. Improvement of channel mobility to lower Ronis critical. Poor channel mobility in SiC Power MOSFETs (eg.,--13 cm2/V.s in 4H-SiC UMOSFET)[ 248] has thus far been generally attributed to poor interface quality between oxide and SiC in the channel region. Additionally, long-term stability of the insulator has been shown to be the limiting factor in SiC power UMOSFET structures. Use of a p+ polysilicon gate reduces Fowler-Nordheim injection from the gate electrode, thereby increasing breakdown voltage. For an excellent discussion of these issues, see Ref. 246. In an attempt to increase the channel mobility as well as blocking voltage, the first Double Implanted MOS (DIMOS or DMOS) FET fabricated in 6H-SiC was reported by a group at Purdue University.[ 243]Blocking voltage operation up to 760 V was demonstrated with 10 ~tm blocking layers doped at n-type at 1.7 x 1016 and 6.5 x 1015 cm-3. A peak inversion layer channel mobility of 26 cm2/V.s was obtained, showing some improvement over the UMOSFETs. Later, 4H-SiC UMOSFETs were fabricated by Northrop Grumman with a 12 ~tm thick drain-drift layer (ND --2 x 1015 cm -3) and a 90 nm gate oxide. The gate oxide consisted of a thin layer of thermally grown SiO 2 followed by a thicker layer of deposited and densified SiO 2 to ensure a more uniform gate oxide across the trench. [244]Nickel was used for drain/ source metals, with TLM measurements yielding a specific contact resistance of 5-7 x 10-6 f2.cm 2. Channel length was 4 ~tm. Blocking voltages of 1.0-1.2 kV were measured at room temperature by probing bare die under Fluorinert TM. In Fig. 22, the blocking voltage for a typical device is shown. Under a moderate gate bias of 32 V (3.55 MV/cm) at room temperature, the effective channel mobility (/Xeffl was only 1.5 cm2/V.s, while increasing the temperature to 100~ increased kteffto 7 cm2/V's, at a gate bias of only 26 V (2.9 MV/cm). Specific on-resistance was 74 mf).cm 2 at 100~ Increasing channel mobility with temperature may be explained by deep

Processing of Silicon Carbide for Devices and Circuits

223

level interface traps located at the SiC-oxide interface which are thermally ionized at higher temperatures, resulting in increased conduction for a given bias. This explanation would account for the typically observed increase in IDS with temperature which opposes the expected T -3/2temperature dependence of ~teffresulting from increased acoustic phonon scattering. 4H-SiC DMOS structures were also reported on in the same time frame[2441 using a lightly doped 10 ~m thick drain-drift layer and a 50 nm thick oxide/nitride gate dielectric. Multiple ion implants (high-dose nitrogen and aluminum implants for source/body contact regions, and highenergy aluminum or boron implants for formation of the p-wells) were used to fabricate a planar power MOSFET in SiC. Initial findings have resulted in 900 V devices being demonstrated (also see Fig. 22), although the Pegstill shows the s a m e temperature effects of interface traps as was found for the UMOSFET case. 2.0x10-3 .

VG=20V VG = 10 V

1.5x103"~

DMOS 01 W A F E R 3 P3B OR THO 20

T=IO0~

1.0x10-3"~. ,=

03 =,-E3

t(

5.0x10 -~-

0.0xl0~

VG=5V

C

VG=OV ,

''''I''''I''''I''''I''''

-5.0x10 .4

200

400

,

600

800

1000

VDs (v)

5x10 -3 . . 4x10-3-~ (V G = 60 V

3x10 -3t'

o~ 2X10 "3 E3

--

1x10_3-

4H-SiC UMOS

P 13 ORTHO 20 VG = 5 5 V

VG =50V

VG =0V

,,,

0xl0~ .lx10-3 "1 ' ' ' 0

I'' 200

'I ''' I'' 400 600

'I' '' I '' i 800 1000 1200

VD s (V) Figure 22. Output characteristics for a 4H-SiC Double implanted MOSFET (DMOS) exhibiting 900 V blocking voltage at 100~ (above) and similar characteristics for a I 100 V 4HSiC UMOSFET at room temperature (below). Taken from Ref. 244.

224

Wide Bandgap Semiconductors

Other high-temperature, high-power devices, such as thyristors have been demonstrated to 500~ [248] At room temperature, this same device had a blocking voltage of 600 V with a 1.8 A forward current at a voltage drop of 3.7 V. A 10 A, 200 V thyristor with a forward voltage drop of-3.6 V, current density of 620 A/cm 2, and Ro, of 1.2 m~.cm 2 was also demonstrated.[ 248] Other thyristors with 100 V blocking voltage, pulse switched current density of 5200 A/cm 2 (1.8 A) have also been fabricated.[ TM] The turn-on rise time was 43 ns and a fall time of less than 100 ns was measured. Silicon carbide (4H-SiC), asymmetrical Gate Turn-Off thyristors (GTOs) have also been fabricated and tested with respect to forward voltage drop (VF), forward blocking voltage, and turn-off characteristics. [252] The GTO structure (shown in Fig. 23) has advantages over the conventional thyristor in that it can be turned offwith a gate pulse, allowing it to be used in DC power switching since it does not require the negative AC cycle for turn-off. The asymmetrical GTO structure also has the advantage of easier turn-off in comparison to symmetrical devices. Additionally, the asymmetrical GTO structure is less susceptible to open base transistor breakdown which generally determines the breakdown voltage in conventional thyristors. Devices were tested from room temperature to 350~ in the DC mode. Forward blocking voltages ranged from 600-800 V at room temperature for the devices tested. VF of a typical device at 350~ was 4.8 V ata current density of 500 A/cm 2. Turn-off time was less than 1 ~ts. Although no beveling or advanced edge termination techniques were used, the blocking voltage represented approximately 50% of the theoretical value when tested in an air ambient. Also, four GTO cells were connected in parallel to demonstrate 600 V, 1.4 A performance. Anode

Anode

~rJ///J.//./l

l.,'~JJJ././././A Vjf/ff/~rjfA

P

N .

e

P + Buffer Layer N Buffer Layer 4H- N SiC Substrate Cathode

Figure 23. Cross-sectionof a 4H-SiCasymmetricalGTO. Reproducedfrom Ref. 4.

Processing of Silicon Carbide for Devices and Circuits

225

The p-type buffer layer shown in Fig. 23 was epitaxially deposited in order to form a punch-through structure and to allow easier turn-off of the device by reducing the injection efficiency of the bottom npn bipolar transistor. The blocking voltage is supported across a lightly doped (~5 x 10:4 cm -3) 6.0 ~tm thick p-type base layer. Finally, n-type gate and p-type anode epitaxial layers were grown. Interdigitated anode and gate fingers were of equal lengths and widths for these non-optimized, first-generation devices. Typical finger lengths and widths were 250 ~tm and 15 ~tm, respectively. The top view of one of these devices is shown in Fig. 24, and typical I-V characteristics obtained using a curve tracer in the pnp mode is shown in Fig. 24, as well.

Figure 24. (a) Top-view of a 4H-SiC GTO with interdigitatedgate and anode fingers. The cathode (backside) is not shown. Photo courtesy of Northrop Grumman. (b) Typical I-V characteristics of same device showing off-state blocking,turn-on, and forward conducting states. Courtesy of Northrop Grumman.

226

Wide Bandgap Semiconductors

.4

'r, r ~i

':

f

~~

Figure 24. (Cont 'd.)

This blocking voltage corresponded to a maximum electric field

(Era)of approximately 1.2 MV/cm across the blocking layer for a nominal 700 V blocking device when assuming negligible voltage drop across the anode, substrate, contacts, and buffer layers. This value ofE mis about 50% of the theoretical critical electric field possible in 4H-SiC. Emissions of light at the periphery of the device structure in breakdown were observed under high-power (50X) magnification in the dark, consistent with edgebreakdown. Further improvements in edge termination should improve the breakdown characteristics of the devices. For comparison purposes, it is notable that for punch-through structures in silicon such as a p-i-n diode, a 45 ~tm thick blocking layer of equal doping would be necessary to achieve the same 600 V blocking as with the 6 ~tm SiC blocking layer used here. The dramatic reduction in the required blocking layer thickness for this device structure utilizing SiC's high electric breakdown strength should allow for much faster power switching than conventional silicon devices.

Processing o f Silicon Carbide for Devices and Circuits 3.4

227

SiC Static Induction T r a n s i s t o r

A promising device in SiC for use in a combination of high-power, high-frequency, and high-temperature applications is the SIT. The device cross-section is shown in Fig. 25. A 225 W UHF 4H-SiC SIT was reported,[ 24~ with the maximum output power of 225 W at 600 MHz, and a power added efficiency (PALE) of 47% and a gain of 8.7 dB for the packaged device. The maximum blocking voltage was 200 V, and maximum channel current density was 1 A/cm of source periphery. Maximum transconductance was 75 mS/mm, and afro~ of 4 GHz was measured on these devices.

Source

Source

~ ~ '

]" 4 um

1

i ur.

urn-.,-2

N Epi

urn

ND=I(}6 "3cm

N + 4H-SiC

Drain Figure25. Schematic cross-section of a SiC Static Induction Transistor. Taken from Ref. 4. A SEM of the completed SIT devices is shown in Fig. 26, as taken from. Ref. 240. Aside from the impressive performance, also noteworthy was that 1.5 cm periphery devices are practical despite the micropipes present in the material, and that eleven good devices with output power of 225 W were obtained on the die shown in Fig. 26. For digital TV transmission, SITs have been used in 2 kW peak power 850 MHz (UHF) modules. The component transistors in this module were UHF SiC SITs, with the

228

Wide Bandgap Semiconductors

power output density of the parallel cells at 165 kW/cm 2. The projected peak power density of SiC MESFETs/SITs is approximately four times that possible in silicon or gallium arsenide.[ 218]The SiC SIT device was the primary active device component in a 400 W UHF TV transmitter module used for the first live HDTV television broadcast in April 1996, at the National Association of Broadcasters conference in Las Vegas, NV. Later developments on the SIT have included demonstrated operation from UHF to S-band (3.0 Ghz). [241]Mixed-mode SITs have been developed for use at high frequency. Mixed-mode operation refers to the MESFET/SIT modes of operation used to optimize maximum gain and power density. Packaged (3 cm periphery) mixed-mode SITs have delivered 38 W of output power at 3.0 Ghz with 9.5 dB gain and 43% power added efficiency. Packaged SITs capable of delivering 20 W of power at 2.7 Ghz (1 cm periphery) have also been demonstrated.[218] Development of the SiC static induction transistor has been advanced to the point of pilot production within the Northrop Grumman Corporation.[ 218]

Figure 26. Microphotograph(top-view) of completed 1.5 cm periphery SiC Static Induction Transistors (SITs). Gate and source bonding tabs can be seen. Taken from Ref. 240.

Processing of Silicon Carbide for Devices and Circuits 4.0

SiC C I R C U I T S AND S E N S O R S

4.1

Operational Amplifiers

229

The first SiC-based operational amplifier was assembled as a hybrid circuit on an alumina ceramic substrate, using thick-film resistors, gold conductor traces, and die containing 6H-SiC depletion-mode MESFET pairs.[263][264] The packaged circuit was demonstrated in 1992, with measured gain greater than 60 dB from 25~ to 350~ and a gain bandwidth of 1.08 MHz at 25~ and 0.91 MHz at 350~ The first monolithic op-amp was reported in 1994[265]by researchers at General Electric making use of depletion-mode, n-channel, 6H-SiC MOSFET technology. The conservatively designed, packaged op-amp had a gain of 49 dB at room temperature, and a gain bandwidth of 724 kHz at 25~ falling to 269 kHz at 350~ An enhancement-mode NMOS op-amp, with 12 V power supply, an input offset of 56 mV and open loop gain of 103 (40.3 dB) has also been fabricated,[2661 and was operated successfully to 350~ Monolithic SiC CMOS op-amps have been recently reported with a room-temperature open loop voltage gain of 10,000, [267] and hybrid, 4H-SiC JFET op-amps are currently under development.[268]

4.2

Digital Circuits

Another focus of SiC electronics research is to develop a hightemperature, high-speed logic family. Researchers at Purdue University were the first to demonstrate these capabilities,[269]demonstrating fabrication and operation ofinverters, NAND, NOR, and XNOR gates, D-latches, RS flip-flops, binary counters, and half-adders with successful operation from room temperature to over 300~ Seventeen-stage ring oscillators have also been fabricated, with 40% yield for NMOS circuits on 30 mm (1.182 inch), p-type 6H-SiC substrates, and 25% yield for PMOS circuits on 35 mm (1.379 inch), n-type 6H-SiC substrates. [266]

4.3

Sensors

A 6H-SiC, hybrid temperature sensor was designed, fabricated, and tested from -150~ to 500~ at Auburn University, utilizing two 6H-SiC

23.0 Wide Bandgap Semiconductors JFETs configured in a series-series feedback network, and a 6H-SiC diode formed using the gate-to-source junction in a third JFET.[ 27~ Linear output with respect to temperature was measured from 500~ to about -50~ where carrier-freezeout in the diode began to affect performance. The completed assembled hybrid temperature sensor is shown prior to final wirebonding in Fig. 27. The SiC JFETs are mounted on an alumina ceramic substrate with thick film gold conductor traces and thick film resistors fired from commercially available pastes. Other temperature sensing applications have also been envisioned.[ TM ] A SiC thermistor was demonstrated as early as 1973,[ 272] with sustained operation at 1050~ (1323 K) for over 2500 hours. The thermistors were packaged in stainless steel housings, with logarithmic temperature response to temperature from -200~ to 700~

Figure 27. Photograph of a 6H-SiC, hybrid temperature sensorwhich operated from 500~ to approximately -150~ Three 6H-SiC JFET's (one used as a diode) shown brazed on a thick-film gold paste with thick-film screen-printed resistors. Photo was taken prior to wirebonding. Taken from Ref. 270.

SiC UV photodiodes have also been demonstrated. [1][273][274] SiC UV detectors have the advantages of operation in hostile environments, insensitivity to longer wavelengths, and very small dark currents leading to

Processing of Silicon Carbide for Devices and Circuits

231

a sensitivity four orders of magnitude greater than that of Si detectors.[ 2271 Flame-indication sensors utilizing 6H-SiC transistors[ 111275]have been fabricated and tested at temperatures over 500~ A flame indication sensor with an integral amplifier was earlier operated for over 600 hours in simulated engine environments with no degradation. Flame indication sensors designed for production were then built, qualified, and field tested before being installed in gas turbine power plants. This work represented the first commercialized turbine control application to use SiC electronic devices.[ 275] Other applications for LW detectors include satellite-based missile plume detection, gas sensing (automobile exhaust systems, etc.), and UV dosimetry. SiC hydrocarbon sensors have been pursued as well, with a potentially large market in the automotive industry. [276][277]SiC MOS oxygen sensors, also with a potential large market development, have been developed with promising results.[ 278] A 3C-SiC accelerometer[ 2791fabricated via micromachining of3C-SiC membranes on a silicon substrate has been demonstrated by Daimler-Benz. Free-standing 3C-SiC membranes were obtained by anisotropic etching of the Si wafer. The 3C-SiC membranes were able to withstand much higher temperature and pressure, as well as more corrosive environments than comparable Si membranes, indicating a much more robust accelerometer could be fabricated. Although not a high-temperature application, high-speed Non-Volatile Random Access Memory (NVRAM) in SiC[28~ is worth mentioning, because SiC's wide bandgap and subsequent long charge storage times (> 10 min. at 623 K and projected to be over a million years at 300 K) may give rise to room-temperature memory chips which need no refreshing, thus eliminating hard drives in computers. This could open the commercial personal computer market to SiC, allowing economies of scale to occur and allow the cost of SiC to fall rapidly.

5.0

CONCLUSIONS

The interested reader should look to other reviews found in press and in the literature. In addition, proceedings of past conferences relating to SiC research will also prove helpful.[17][95][281]-[3~176 A brief overview of the electronic properties of SiC as relating to high-temperature electronics, as well as significant achievements involving high-temperature SiC electronics prior to 1996 has been given.

232

Wide Bandgap Semiconductors

A broad overview of some of the high-temperature applications using SiC electronics has also been discussed. Excellent progress has been demonstrated by SiC devices and circuits to date. Still, despite impressive individual device performances, some key barriers remain. First, insulators capable of reliable high-temperature operation with high breakdown field strength and low interface trap densities need further development. Of particular concern is the uniform growth/deposition of insulators on the different surfaces of SiC exposed in a typical vertical power UMOSFET configuration. Second, material defects (in particular, micropipes) and material cost must be decreased, while wafer-size must be increased substantially in order to facilitate commercial production of SiC based electronics. Much progress has been made towards these goals, as evidenced by Fig. 28,[ 1~] and should continue, making SiC an attractive semiconductor choice in the high-temperature electronics marketplace. Devices such as high-frequency MESFET's, which do not rely on large area or a high-quality dielectric may be the first SiC electronic devices (excluding optical devices) to reach the commercial market. In 1997, SiC static induction transistors have reached the stage pilot production, a tremendous advancement considering that less than 10 years earlier SiC substrates were not commercially available.[ 2181Also, in 1996, the first reported use of SiC devices in commercial turbine control systems was announced, which was another indication of a rapidly maturing technology.

/

-...

i~.4Cod per em2 s . 4 ldlcroPJpes per c ms

" ,9, . %%

400

"on,,,,

' MJ~terlxl C a s t

sad Mlcroplpe

A~d~ds ~ ' , , , ~ ~o,,, J ~ , O

UllU

1991

lUUl

"

l

u

1992

- j'

l

1993

A

u

i,,~ . •

i...

1994

~

, _

9 n

1995

"-

~---

1996

Yqmul"

Figure 28. Micropipe defect density and cost in U.S. dollars per cm 2 as a function of time since the first commercial introduction of bulk SiC substrates. Taken from Ref. 300.

P r o c e s s i n g o f Silicon Carbide f o r Devices a n d Circuits

233

REFERENCES 1. Oasack,D. E., Glasheen, W. M., and Steglich, H. R., Trans. of 2ndInternational High Temperature Electronics Conference (HiTEC), Charlotte, N.C., USA, Session III, pp. 17-22 (June 5-10, 1994) 2. Reinhardt, K. C., and Marciniak, M. A., Trans. of 3rd International High Temperature Electronics Conference (HiTEC), Session I, pp. 9-15 (June 914,1996) 3. Dreike, P. L., Fleetwood, D. M., King, D. B., Sprauer, D. C., and Zipperian, T. E., IEEE Trans. on Comp., Hybrids, Man. Tech. -A, 17(4):594-609 (Dec., 1994) 4. Agarwal, A. K., Augustine, G., Balakrishna, V., Bran&, C. D., Burke, A. A., Chen, L. S., Clarke, R. C., Esker, P. M., Hobgood, H. M., Hopkins, R. H., Morse, A. W., Rowland, L. B., Seshadri, S., Siergiej, R. R., Smith, T. J., Jr., and Sriram, S., Tech. Dig. Int'l. Elect. Dev. Meeting, pp. 9.1.1-9.1.6 (Dec., 1996) 5. Tajima, M., Trans. of 2"d International High Temperature Electronics Conference (HiTEC), Charlotte, N.C., USA, Session I, pp. 29-34 (June 5-10, 1994) 6. Fleetwood, D. M., IEEE Trans. Nucl. Sci., Vol. N S-35, pp. 1099-1112 (1988) 7. Palmour, J. W., and Carter, C. H., Jr., Proceedings of 1993 International Semiconductor Device Research Symposium, Omni Charlottesville Hotel, pp. 695-696 (Dec. 1-3, 1993) 8. Trew, R. J., Yan, J. B., and Mock, P. M., Proceedings ofthe IEEE, 79:598-620 (May, 1991) 9. McGarrity, J. M., McLean, F. B., DeLancey, W. M., Palmour, J. W., Carter, C. H., Jr., and Edmond, J. A., IEEE Trans. on Nuclear Science, 39(6): 1974-1981 (1992) 10. Cree Research, Inc., Durham, NC 27713 11. Advanced Technology Materials, Inc., 7 Commerce Drive, Danbury, CT 06810-4169, and Sterling Semiconductor 12. George, T., Pike, W. T., Khan, M. A., Kuznia, J. N., and Chang-Chien, P., J. of Electronic Materials, 24(4):241-247 (Apr., 1995) 13. Yim, W. M., Stofko, E. J., Zanzucchi, P. J., Pankove, J. I., Ettenberg, E., and Gilbert, S. L., J. Appl. Phys., 44(1):292-296 (Jan. 1, 1973) 14. Edmond, J., Kong, H., Dmitriev, V., Bulman, G., and Carter, C., Jr., Inst. Phys. Conf. Ser. No. 137, pp. 515-518, IOP Publishing, Ltd. (1994) 15. Round, H. J., Electrical World, 49:309 (1907) 16. Shockley, W., Proceedings of the First International Conference on Silicon Carbide, Boston, MA, 1959, (J. R. O'Connor and J. Smiltens, eds.), Pergamon, New York (1960)

234

Wide Bandgap Semiconductors

17. Pensl, G., and Choyke, W. J., Physica B, 185:264-283 (1993) 18. Dubrovskii, G. B., Soviet Phys. Solid State, 13:2107-109 (1972) 19. Humphreys, R. G., Bimberg, D., and Choyke, W. J., Solid State Comm., 39:163-167(1981) 20. Tachibana, T., Kong, H. S., Wang, Y. C., and Davis, R. F., J. Appl. Phys., 67:6375-6381 (1990) 21. Shields, V., Fekade, K., and Spencer, M. G., Inst. Phys. Conf. Ser. No. 137, pp. 21-24, IOP Publishing, Ltd. (1994) 22. Golecki, I., Reidinger, F., and Marti, J., Workshop on High Temperature Power Electronics for Vehicles, Fort Monmouth, N. J., USA (April 26-27, 1995) 23. Anthony Powell, J., and Larkin, D. J., NASA Tech. Briefs, pp. 58-59 (May, 1995) 24. Kong, H. S., Glass, J. T., and Davis, R. F.,Appl. Phys. Lett., 49:1047 (1986) 25. Tischler,M. A, Hamaguchi, N., Choi, S., Powell, A., and Dobrilla, P., Workshop on High Temperature Power Electronics for Vehicles, Fort Monmouth, N.J. (April 26-27,1995) 26. Madelung, O., Data in Science and Technology: Semiconductors Group IV Elements and Ill- V Compounds, p. 101, Springer-Verlag, Berlin (1991) 27. Gaberstroh, C., Helbig, R., and Stein, R. A., J. of Appl. Phys., 76:509-513, (July, 1994) 28. Muller, R. S., and Kamins, T. I., Device Electronics for Integrated Circuits, Second Edition, Wiley, p. 54 (1986) 29. Muench, W. V., and Pfaffeneder, I., J. Appl. Phys., 48:4831-4833 (1977) 30. Muench, W. V., and Petterpaul, E., ~ Appl. Phys., 48(11):4823-4825 (Nov., 1977) 31. Baliga, B. J., Springer Proceedings in Physics, 71:305-313 (1992) 32. Morelli, D., Hermans, J., Becto, C., Woo, W. S., Harris, G. L., Taylor, C., Inst. Phys. Conf. Set. No. 137, p. 313-5, IOP Publishing, Ltd. (1994) 33. Slack, G. A., J. Appl. Phys., 35:3460-6 (1964) 34. Son, N. T., Chen, W. M., Kordina, O., Konstantinov, A. O., Monemar, B., Janzen, E., Hofman, D. M., Volm, D., Drechsler, M., and Meyer, B. K., Applied Physics Letters, 66(9): 1074-1076 (Feb. 27, 1995) 35. Schaffer, W. J., Negley, G. H., Irvin, K. G., and Palmour, J. W., MRS Symposium Proceedings, 339:595-600, MRS, Pittsburgh, PA (1994) 36. Jacob, C., Nishino, S., Mehregany, M., Polk, J. A., and Pirouz, P., Inst. of Phys. Conf. Ser. No. 137, pp. 247-250, IOP Publishing, Ltd. (1994)

Processing of Silicon Carbide for Devices and Circuits

235

37. Yoshida, S., Properties of Silicon Carbide, (G.L. Harris, ed.), p. 74, INSPEC, London, UK (1995) 38. Materials for High-Temperature Semiconductor Devices, p. 19, National Academy Press, National Academy of Sciences, Washington, DC (1995) 39. Yamanaka, M., Daiman, H., Sakuma, E., Misawa, S., Yoshida, S., J. Appl. Phys., 61(2):599-603 (Jan. 15,1987) 40. Larkin, D. J., Neudeck, P. G., Powell, J. A., and Matus, L. G., Inst. Phys. Conf. Ser. No. 137, p. 51-54, lOP Publishing, Ltd. (1994) 41. Flatresse, P., and Ouisse, T., Solid-St. Electron., 38(5):971-975 (1995) 42. Patrick, L., and Choyke, W. J., Physical Review B, 2(6):2255-2256 (15 Sept., 1970) 43. Tsvetkov, V. F., Allen, S. T., Kong, H. S., and Carter, C. H., Jr., Inst. Phys. Conf. Ser. No. 142, pp. 17-22, IOP Publishing, Ltd. (1996) 44. Augustine, G., Hobgood, H. M., Balakrishna, V., Dunne, G., and Hopkins, R. H., Phys. Stat. Solidi (submitted Sept., 1997) 45. Neudeck, P. G., andPowell, J. A., IEEE Elect. Dev. Lett., 15:63-5 (Feb., 1994) 46. Giocondi, J., Rohrer, G. S., Skowronski, M., Balakrishna, V., Augustine, G., Hobgood, H. M., and Hopkins, R. H., MRS Symposium Proc., 423:539-544 (1996) 47. Hobgood, H. M., Barrett, D. L., McHugh, J. P., Clarke, R. C., Sriram, S., Burke, A. A., Greggi, J., Brandt, C. D., Hopkins, R. H., and Choyke, W. J., J. Crystal Growth, 137:181-186 (1994) 48. Udding, A., Mitsushashi, H., and Uemoto, T., Jpn. J. Appl. Phys., 33:L908-L911 (1994) 49. Tairov, Y. M., and Tsvetkov, V. F., J. Cryst. Growth, 43:209-212 (1978) 50. Tairov, Y. M., and Tsvetkov, V. F., J. Cryst. Growth, 52:146-150 (1981) 51. Ziegler, G., Lanig, P., Theis, D., and Weyrich, C., IEEE Trans. Elect. Dev., vol. ED-30, pp. 277-281 (1983) 52. Barrett, D. L., Seidensticker, R. G., Gaida, W., Hopkins, R. H., Choyke, W. J., J. Cryst. Growth, 109:17-23 (1991) 53. Davis, R. F., Carter, C. H., Jr., Hunter, C. E., U.S. Patent No. 4,866,0005 (Sept. 12, 1989) 54. Nakata, T., Koga, K., Matsushita, Y., Ueda, Y., Niina, T., Proc. of 2 ndInt'l. Conf. on Amorphous and Crystalline Silicon Carbide H & Related Mat., Santa Clara, CA, Dec. 15-16, 1988, Springer Proc. Phys., 43:26-34 (1989) 55. Barrett, D. L., McHugh, J. P., Hobgood, H. M., Hopkins, R. H., McMullin, P. G., Clarke, R. C., and Choyke, W. J., J. Cryst. Growth, 128:358-362 (1993)

236

Wide Bandgap Semiconductors

56. Takahashi, F., Kanaya, M., and Fujiwara, Y., J. Cryst. Growth, 135:61-70

(1994) 57. Hofinann, D., Karpov, S. Y., Makarov, Y. N., Mokhov, E. N., Ramm, M. G., Ramm, M. S., Roenkov, A. D., and Vodakov, Y. A., Inst. Phys. Conf. Ser. No. 142, pp. 29-32, IOP Publishing, Ltd. (1996) 58. Hobgood, H. M., Glass, R. C., Augustine, G., Hopkins, R. H., Jenny, J., Skowronski, M., Mitchel, W. C., and Roth, M., Appl. Phys. Lett., 66(11):1364-1366 (Mar. 13, 1995) 59. Gradinaru, G., Sudarshan, T. S., Gradinaru, S. A., Mitchell, W., and Hobgood, H. M., Appl. Phys. Lett., 70(6):735-737 (Feb. 10, 1997) 60. Wang, Y. C., Davis, R. F., and Edmond, J. A., Journal of Electronic Materials, 20:289-294 (Apr., 1991) 61. Larkin, D. J., Inst. Phys. Conf. Ser. No. 142, pp. 23-28, IOP Publishing, Ltd. (1996) 62. Larkin, D. J., Sridhara, S. G., and Devaty, R. P., Journal of Electronic Materials, 24:289-294 (Apr., 1995) 63. Larkin, D. J., Neudeck, P. G., Powell, J. A., and Matus, L. G., Appl. Phys. Lett., 65:1659 (1994) 64. Burke, A. A., Jr., Rowland, L. B., Agarwal, A. K., Sriram, S., Glass, R. C., and Brandt, C. D., Inst. Phys. Conf. Set. No. 142, pp. 201-204, IOP Publishing, Ltd. (1996) 65. Kimoto, T., Matsunami, H., Inst. Phys. Conf. Set. No. 137, p. 55-58, IOP Publishing, Ltd. (1994) 66. Kim, H. J., and Davis, R. F., J. Electrochem. Soc., 133:2350-2357 (Nov., 1986) 67. Ikeda, M., Matsunami, H., and Tanaka, T., Phys. Rev. B, 22(6):2842-2845 (Sept. 15, 1980) 68. Kordina, O., Henry, A., Bergman, J. P., Son, N. T., Chen, W. M., Hallin, C., and Janzen, E., Applied Phys. Lett., 66(11):1373-1375 (Mar. 13, 1995) 69. Hoh, A., Akita, H., Kimoto, T., and Matsunami, H., Appl. Phys. Lett., 65(11):1400-1402 (Sept. 12, 1994) 70. Kong, H. S., Glass, J. T., and Davis, R. F., d. Appl. Physics, 64(5):26722679 (Sept. 1, 1988) 71. Davis, R. F.,J. Cryst. Growth, 137:161-169 (1994) 72. Rupp, R., Lanig, P., V'lkl, J., and Stephani, D., Mat. Res. Soc. Syrup. Proc., 423:253-263(1996) 73. Burke, A. A., Jr., Rowland, L. B., Augustine, G., Hobgood, H. M., and Hopkins, R. H., Mat. Res. Soc. Syrup. Proc., 423:275-280 (1996)

Processing o f Silicon Carbide f o r Devices and Circuits

237

74. Janzen, E., and Kordina, O., Inst. Phys. Conf. Ser. No. 142, pp. 653-658, IOP Publishing, Ltd. (1996) 75. Rowland, L. B., Burke, A. A., Jr., Clarke, R. C., Siergiej, R. R., Sriram, S., Augustine, G., Hobgood, H. M., and Driver, M. C., Proc. IEEE/Cornell Conf. Advanced Concepts in High Speed Semiconductor Devices and Circuits (Aug. 7-9, 1995) 76. Rao, M. V., Griffiths, P., Holland, O. W., Kelnar. G., Freitas, J. A., Jr., Simons, P. S., and Ghezzo, M., dr.Applied Physics, 77(6):2479-2485 (15 Mar., 1995) 77. Kimoto, T., Itoh, A., Matsunami, H., Nakata, T., and Watanabe, M., dr. Electronic Mat., 24(4):235-240 (1995) 78. Heft, A., Wendler, E., Bachmann, T., Glaser, E., and Wesch, W., Materials Science & Engineering B, 29:142-146 (1995) 79. Ahmed, S., Barbero, C. J., and Sigmon, T. W., Appl. Phys. Lett., 66(6):712-714 (Feb. 6, 1995) 80. Ahmed, S., Barbero, C. J., Sigmon, T. W., and Erickson, J. W., Appl. Phys. Lett., 65(1):1-3 (July 4, 1994) 81. Wesch, W., Heft, A., Wendler, E., Bachmann, T., and Glaser, E., Nuclear Instruments and Methods in Physics Research B, 96:335-338 (1995) 82. Ghezzo, M., Brown, D. M., Downey, E., Kretchmer, J., and Kopanski, J. J., Appl. Phys. Lett., 63(9):1206-1208 (Aug. 30, 1993) 83. Ghezzo, M., Brown, D. M., Downey, E., Kretchmer, J., Hennessy, W., Polla, D. L., and Bakru, H., IEEE Elect. Dev. Lett., 13(12):639-641 (Dec., 1992) 84. Du, H., Libera, M., Yang, Z., Lai, P. J., Jacobson, D. C., Wang, Yu. C., and Davis, R. F., Appl. Phys. Lett., 62(4):423-425 (Jan. 25, 1993) 85. Edmond, J. A., Withrow, S. P., Wadlin, W., and Davis, R. F., Mat. Res. Soc. Symp. Proc., 77:193-198 (1987) 86. Edmond, J. A., Kim, H. J., and Davis, R. F., Mat. Res. Soc. Symp. Proc., 52:157-164 (1986) 87. Dalibor, T., Peppermiiller, C., Pensl, G., Sridhara, S., Devaty, R. P., Choyke, W. J., Itoh, A., Kimoto, T., Matsunami, H., Inst. Phys. Conf. Ser. No. 142, pp. 517-520, IOP Publishing, Ltd. (1996) 88. Rao, M. V., Gardner, J., Griffiths, P., Holland, O. W., Kelner, G., Ghezzo, M., Simons, D. S., and Chi, P. H., Inst. Phys. Conf. Ser. No. 142, pp. 521524, lOP Publishing, Ltd. (1996) 89. Inoue, N., Itoh, A., Kimoto, T., Matsunami, H., Nakata, T., and Watanabe, M., Inst. Phys. Conf. Ser. No. 142, pp. 525-528, lOP Publishing, Ltd. (1996) 90. Freitas, J. A., Jr., Griffiths, P., Gardner, J., and Rao, M. V., Inst. Phys. Conf. Set. No. 142, pp. 529-532, IOP Publishing, Ltd. (1996)

238

Wide Bandgap Semiconductors

91. Kawase, D., Ohno, T., Iwasaki, T., and Yatsuo, T., Inst. Phys. Conf. Ser. No. 142, pp. 513-516, IOP Publishing, Ltd. (1996) 92. Heera, V., Pezoldt, J., Ning, X. J., and Pirouz, P., Inst. Phys. Conf. Ser. No. 142, pp. 509-512, IOP Publishing, Ltd. (1996) 93. Harada, S., Ishimaru, M., Motooka, T., Nakata, T., Yoneda, T., and Inoue, M., Appl. Phys. Lett., 69:3534-3536(2 Dec., 1996) 94. Peppermiiller, C., Helbig, R., Rottner, K., and Sch6ner, A., Appl. Phys. Lett., 70(8): 1014-1016 (Feb. 24,1997) 95. Harris, G. L., Properties of Silicon Carbide, pp. 131-149 INSPEC, London, UK(1995) 96. Yih, P. H., and Steckl, A. J., J. Electrochem. Soc., 142:2853-2860 (Aug., 1995) 97. Casady, J. B., Luckowski, E. D., Bozack, M., Sheridan, D., Johnson, R. W., and Williams, J. R.,J. Electrochem. Soc., 143(5): 1750-1753 (May, 1996) 98. Casady, J. B., Luckowski, E. D., Bozack, M., Sheridan, D., Johnson, R. W., and Williams, J. R., Inst. Phys. Conf. Ser. No. 142, pp. 625-628, IOP Publishing, Ltd. (1996) 99. Yih, P. H. and Steckl, A. J.,J. Electrochem. Soc., 142:312-319 (Jan., 1995) 100. Luther, B. P., Ruzyllo, J., and Miller, D. L.,Appl. Phys. Lett., 63:171-173 (1993) 101. Yih, P. H., and Steckl, A. J., J. Electrochem. Soc., 140:1813-1824 (June, 1993) 102. Steckl, A. J., and Yih, P. H., Appl. Phys. Lett., 60:1966-1968 (Apr. 20, 1992) 103. Pan, W. -S., and Steckl, A. J., J. Electrochem. Soc., 137:212-220 (Jan., 1990) 104. Sugiura, J., Lu, W. -J., Cadien, K. C., and Steckl, A. J., ,I. Vac. Sci. Tech. B, 4:349-354 (Jan./Feb., 1986) 105. Yih, P. H., and Steckl, A. J., Inst. Phys. Conf. Ser. No. 137, pp.317-320, IOP Publishing, Ltd. (1994) 106. Palmour, J. W., Davis, R. F., Astell-Burt, P., and Blackborow, P., Mat. Res. Soc. Symp. Proc., 76:185-190 (1987) 107. Kelner, G., Binari, S. C., and Klein, P. H., J. Electrochem. Soc., 134:253-254 (Jan., 1987) 108. Niemann, E., Boos, A., and Leidich, D., Inst. Phys. Conf. Ser. No. 137, pp.695-698, IOP Publishing, Ltd. (1994) 109. Palmour, J. W., Williams, B. E., Astell-Burt, P., and Davis, R. F., J. Electrochem. Soc., 136(2):491-495 (Feb., 1989)

P r o c e s s i n g o f Silicon Carbide f o r Devices a n d Circuits

239

110. Palmour, J. W., Davis, R. F., Astell-Burt, P., and Blackborow, P., Ceramic Transactions, 2:491-500 (1989) 111. Flemish, J., Xie, K., and Zhao, J., Appl. Phys. Lett., 64:2315-2317 (1994) 112. Flemish, J., Xie, K., Buchwald, W., Casas, L., Zhao, J. H., McLane, G., and Dubey, M., Mat. Res. Soc. Symp. Proc. 339:145-150, MRS, Pittsburgh, PA (1994) 113. Flemish, J. R., and Xie, K., J. Electrochem. Soc., 143(8):2620-2623 (Aug., 1996) 114. Xie, K., Flemish, J. R., Zhao, J. H., Buchwald, W. R., and Casas, L., Appl. Phys. Lett., 67(3):368-370 (July 17, 1995) 115. Shor, J. S., Zhang, X. G., Kurtz, A. D., and Osgood, R. M., Proc. 4th Int'l. Conf. on SiC and Related Mat., Santa Clara, CA, pp. 355-361 (Oct. 9-11, 1991) 116. Collins, D. M., Harris, G. L., Wongchotigul, K., Zhang, D., Chen, N., and Taylor, C., Inst. Phys. Conf. Ser. No. 142, pp. 617-620, IOP Publishing, Ltd. (1996) 117. Saxena, V., Steckl, A. J., and Bordonaro, G., Electrochemical Soc. Proceedings, 95-21:247-249 (1995) 118. Wolf, R., andHelbig, R.,J. Electrochem. Soc., 143(3):1037-1042 (Mar., 1996) 119. Cao, L., and Zhao, J. H., IEEE Elect. Dev. Lett. (submitted, 1997) 120. Shor, J. S., Kurtz, A. D., Grimberg, I., Weiss, B. Z., and Osgood, R. M., J. Appl. Phys., 81(3):1546-1551 (1 Feb., 1997) 121. Shenoy, J. N., Chindalore, G. L., Melloch, M. R., and Cooper, J. A., Jr., J. Elect. Mat., 24(4):303-309 (1995) 122. Hometz, B., Michel, H. -J., and Halbritter, J., d. Vac. Sci. Technol. A, 13(3):767-771 (May/June 1995) 123. Shenoy, J. N., Das, M. K., Cooper, J. A., Jr., Melloch, M. R., and Palmour, J. W., g. Appl. Phys., 79(6):3042-3045 (Mar. 15, 1996) 124. G61z,A., Horstmann, G., Stein von Kamienski, E., and Kurz, H., Inst. Phys. Conf. Ser. No. 142, pp. 633-636, IOP Publishing, Ltd. (1996) 125. Akita, H., Kimoto, T., Inoue, N., and Matsunami, H., Inst. Phys. Conf. Ser. No. 142, pp. 725-728, IOP Publishing, Ltd. (1996) 126. Tokura, N., Hara, K., Takeuchi, Y., Miyajima, T., and Hara, K., Inst. Phys. Conf. Ser. No. 142, pp. 637-640, IOP Publishing, Ltd. (1996) 127. Bano, E., Ouisse, T., Lassagne, P., Billon, T., and Jaussaud, C., Inst. Phys. Conf. Ser. No. 142, pp. 729-733, IOP Publishing, Ltd. (1996) 128. Stein von Kamienski, E., Portheine, F., G/51z,A., and Kurz, H., Inst. Phys. Conf. Ser. No. 142, pp. 64 1-644, IOP Publishing, Ltd. (1996)

240

Wide Bandgap Semiconductors

129. Sridevan, S., Misra, V., McLarty, P. K., Baliga, B. J., and Wortman, J. J., Inst. Phys. Conf. Ser. No. 142, pp. 645-648, IOP Publishing, Ltd. (1996) 130. Shenoy, J. N., Chindalore, G. L., Das, M. K., Cooper, J. A., Jr., MeUoch, M. R., Palmour, J. W., and K.G. Irvine, Inst. Phys. Conf. Ser. No. 142, pp. 745-748, IOP Publishing, Ltd. (1996) 131. Ramberg, C. E., Spear, K. E., Tressler, R. E., and Chinone, Y., J. Electrochem. Soc., 142(11):L214-L216 (Nov., 1995) 132. Zetterling, C. M., Harris, C. I., t3stling, M., Afanasev, V. V., Inst. Phys. Conf. Ser. No. 142, pp. 605-608, IOP Publishing, Ltd. (1996) 133. Bano, E., Ouisse, T., Lassagne, P., Billon, T., and Jaussaud, C., Inst. Phys. Conf. Ser. No. 142, pp. 733-736, IOP Publishing, Ltd. (1996) 134. Pan, J. N., Cooper, J. A., Jr., and Melloch, M. R., J. Appl. Phys., 78:572-574 (July 1, 1995) 135. Karlsteen, M., Baranzahi, A., Spetz, A. L., Willander, M., and Lundstr/Sm, I., J. Elect. Mat., 24(7):853-861 (July, 1995) 136. Brown, D. M., Ghezzo, M., Kretchmer, J., Downey, E., Pimbley, J., and Palmour, J., IEEE Trans. Elect. Dev., 41 (4):618-620 (Apr., 1994) 137. Neudeck, P., Kang, S., Petit, J., and Tabib-Azar, M.,J. Appl. Phys., 75(12): 7949-7953 (June 15, 1994) 138. Bano, E., Ouisse, T., DiCioccio, L., and Karmann, S., Appl. Phys. Lett., 65 (21):2723-2724 (Nov. 21, 1994) 139. Friedrichs, P., Burte, E. P., and Sch/Srner, R., Appl. Phys. Lett., 65(13): 1665-1667 (Sept. 26, 1994) 140. Alok, D., Baliga, B. J., and McLarty, P. K., IEEE Elect. Dev. Lett., 15(10):424-426 (Oct., 1994) 141. Alok, D., McLarty, P. K., and Baliga, B. J., Appl. Phys. Lett., 65(17):2177-2178 (Oct. 24, 1994) 142. Hometz, B., Michel, H. -J., and Halbritter, J.,J. Mater. Res., 9(12):30883094 (Dec., 1994) 143. Raynaud, C., Autran, J. -L., Balland, B., GuiUot, G., Jaussaud, C., and Billon, T., J. Appl. Phys., 76(2):993-997 (July 15, 1994) 144. Schmitt, J., and Helbig, R., 3'. Electrochem. Soc., 141(8):2262-2265 (Aug., 1994) 145. Tyagi, R., Ghezzo, M., Chow, T. P., and Norton, J. F., J. Electrochem. Soc., 141(8):2188-2191 (Aug., 1994)

P r o c e s s i n g o f Silicon Carbide f o r Devices a n d Circuits

241

146. Powell, J. A., Petit, J. B., Edgar, J. H., Jenkins, I. G., Matus, L. G., Choyke, W. J., Clemen, L., Yoganathan, M., Yang, J. W., and Pirouz, P., Appl. Phys. Lett., 59(2):183-185 (July, 1991) 147. Shinohara, M., Yamanaka, M., Misawa, S., Okumura, H., and Yoshida, S., Jpn. J. Appl. Phys., 30(2):240-243 (Feb., 1991) 148. Zheng, Z., Tressler, R. E., and Spear, K. E., J. Electrochem. Soc., 137(3):854-858 (Mar., 1990) 149. Zheng, Z., Tressler, R. E., and Spear, K. E., J. Electrochem. Soc., 137(9):2812-2816 (Sept., 1990) 150. Zaima, S., Onoda, K., Koide, Y., and Yasuda, Y., J. Appl. Phys., 68(12):6304-6308 (Dec. 15, 1990) 151. Tang, S. M., Berry, W. B., Kwor, R., Zeller M. V., and Matus, L. G., J. Electrochem. Soc., 137(1):221-225 (Jan., 1990) 152. DeMeo, R. C., Wang, T. K., Chow, T. P., Brown, D. M. and Matus, L. G., J. Electrochem. Soc., 141 ( 11):L 150-L 152 (Nov., 1994) 153. Petit, J. B., Edgar, J. H., Jenkins, I. G., Matus, L. G., Choyke, W. J., Clemen, L., Yoganathan, M., Yang, J. W., and Pirouz, P., Appl. Phys. Lett., 59(2):183-185 (July, 1991) 154. Brown, D. M., Downey, E., Ghezzo, M., Kretchmer, J., Krishnamurthy, V., Hennessy, W., and Michon, G., Solid-State Electronics, 39( 11): 1531-1542 (Nov., 1996) 155. Muehloff, L., Choyke, W. J., Bozac, M. J., and Yates, J. T., Jr.,,/. Appl. Phys., 60(8):2842-2853 (15 Oct., 1986) 156. Lipkin, L. A., and Palmour, J. W., J. Electronic Materials, 25(5):909-915 (1996) 157. Afanas' ev, V. V., and Stesmans, A.,Appl. Phys. Lett., 69(15):2252-2254 (Oct. 7,1996) 158. Lipkin, L. A., Palmour, J. W., and Suehle, J. S., Proc. of3rd lnt'l. Conf. on High Temp. Elect. (HiTEC), 2:XIV-21-28 (June 9-14, 1996) 159. Dimitrijev, S., Li, H., Harrison, H. B., and Sweatman, D., IEEE Electron Device Lett., 18(5):175-177 (May, 1997) 160. Agarwal, A. K., Siergiej, R. R., Seshadri, S., White, M. H., McMullm, P. G., Burke, A. A., Rowland, L. B., Brandt, C. D., and Hopkins, R. H., Proc. oflnt'l. Symp. on Power Semiconductor Devices and ICs, pp. 119-122 (May 20-23, 1996) 161. Crofton, J., Barnes, P. A., Williams, J. R., and Edmond, J. A., Appl. Phys. Lett., 62(4):384-386 (Jan. 25, 1993)

242

Wide B a n d g a p S e m i c o n d u c t o r s

162. Lundberg, N., and Ostling, M., Solid-St. Elect., 39(11):1559-1565 (Nov., 1996) 163. Crofton, J., Beyer, L., Williams, J. R., Luckowski, E. D., Mohney, S. E., and DeLucca, J. M., submitted to Solid-St. Elect. (1997) 164. Porter, L. M., and Davis, R. F., Trans. 3ra Int'l. High Temp. Elect. Conf. (HiTEC), Session VII, pp. 3-8 (June 9-14, 1996) 165. Oder, T. N., Williams, J. R., Mohney, S. E., and Crofton, J., submitted to J. Elect. Mat. (1997) 166. Petit, J. B., Neudeck, P. G., Salupo, C. S., Larkin, D. J., and Powell, J. A., Proc. of 5thInt 'l. Conf. on SiC and Related Mater., pp. 679, IOP Publishing (1994) 167. Crofton, J., McMullin, P. G., Williams, J. R., and M.J. Bozack, J. Appl. Phys., 77:1317-1319 (Feb. 1, 1995) 168. Chaddha, A. K., Parsons, J. D., and Kruaval, G. B., Appl. Phys. Lett., 66:760-762 (Feb. 6, 1995) 169. Liu, S., Severt, C., Reinhardt, K., and Scofield, J., Inst. Phys. Conf. Ser. No. 142, pp. 589-592, IOP Publishing, Ltd. (1996) 170. Teraji, T., Hara, S., Okushi, H., and Kajimura, K., Inst. Phys. Conf. Ser. No. 142, pp. 593-596, IOP Publishing, Ltd. (1996) 171. Baud, L., Jaussaud, C., Madar, R., Chen, J. S., and Nicolet, M. A., Inst. Phys. Conf. Ser. No. 142, pp. 597-600, IOP Publishing, Ltd. (1996) 172. Parsons, J. D., Kruaval, G. B., and Chaddha, A. K., Appl. Phys. Lett., 65(16):2075-2077 (Oct. 17, 1994) 173. Shor, J. S., Weber, R. A., Provost, L. G., Goldstein, D., and Kurtz, A. D., J. Electrochem. Soc., 141(2):479-481 (Feb., 1994) 174. Chen, J. S., B~ichli,A., Nicolet, M. -A., Baud, L., Jaussaud, C., and Madar, R., Materials Science and Engineering, B29:185-189 (1995) 175. Alok, D., Baliga, B. J., and McLarty, P. K., IEDM, pp. 691-694 (Dec., 1993) 176. Chen, J. S., Bachli, A., Nicolet, M. -A., Ruiz, R. P., Baud, L., Jaussaud, C., and Madar, R.,J. Appl. Phys., 76(4):2169-2175 (Aug. 15, 1994) 177. Chen, J. S., B~ichli, A., and Nicolet, M. -A., Trans. 2nd Int'l. High Temp. Elect. Conf. (HiTEC), Charlotte, N.C., pp. 203-208 (June 5-10, 1994) 178. Chen, J. S., Kolawa, E., Nicolet, M. -A., Ruiz, R. P., Baud, L., Jaussaud, C., and Madar, R.,J. Mater Res., 9(3):648--657 (Mar., 1994) 179. Chen, J. S., Kolawa, E., Nicolet, M. -A., Baud, L., Jaussaud, C., Madar, R., and C. Bernard, J. Appl. Phys., 75(2):897-901 (Jan. 25, 1993) 180. Dmitriev, V. A., Irvine, K., Spencer, M., and Kelner, G., Appl. Phys. Lett., 64(3):318-320 (Jan. 17, 1994)

Processing of Silicon Carbide for Devices and Circuits

243

181. Glass, R. C., Spellman, L. M., and Davis, R. F., Appl. Phys. Lett., 59(22):2868-2870 (Nov. 25, 1991) 182. Luckowski, E. D., Williams, J. R., Bozack, M. J., Isaacs-Smith, T., and Crofton, J., Material Res. Soc. Syrup. Proc., 423:119-124 (1996) 183. Grimberg, I., Weiss, B. -Z., Eizenberg, M., and Levit, M., Material Res. Society Spring Meeting, Symposium E, San Francisco, CA (April 8-12, 1996) 184. DeLucca, J. M., and Mohney, S. E., MaterialRes. Soc. Symp. Proc., Vol. 423 (1996) 185. Geib, K. M., Wilson, C., Long, R. G., and Wilmsen, C. W., J. Appl. Phys., 68(6):2796-2800 (Sept. 15,1990) 186. Liu, S., Smith, S. R., Adams, S., Severt, C., Leonard, J., Trans. 2~eInt 7. High Temp. Elect. Conf. (HiTEC), Charlotte, NC, Session XIII, pp. 9-14 (June 510,1994) 187. Porter, L. M., Davis, R. F., Bow, J. S., Kim, M. J., Carpenter, R. W., Trans. 2nd Int'l. High Temp. Elect. Conf. (HiTEC), Charlotte, NC, Session XIII, pp. 3-8 (June 5-10,1994) 188. Lundberg, N., and Ostling, M., S o l i d - S t a t e 38(121):2023-2028 (1995)

Electronics,

189. Goesmann, F., and Fetzer, R. S., Semicond. Sci. Technol., 10:1652-1658 (1995) 190. Itoh, A., O. Takemura, Kimoto, T., and Matsunami, H., Inst. Phys. Conf. Ser. No. 142, pp. 685-688, IOP Publishing, Ltd. (1996) 191. Itoh, A., Kimoto, T. and Matsunami, H., IEEE Elect. Dev. Lett., 17(3):139-141 (Mar., 1996) 192. Crofton, J., and Sriram, S., IEEE Trans. on Electron Dev., 43(12):2305-2307 (Dec., 1996) 193. Luckowski, E. D., Isaacs-Smith, T., and Williams, J. R., 1997 Space Technology and Applications Int'l. Forum, AlP Conf., Proc. 387, p. 851 (1997) 194. Bhatnagar, M., Nakanishi, H., McLarty, P. K., Baliga, B. J., Patnaik, B., and Parikh, N., 1EDM Tech. Dig., pp. 789-792 (1992) 195. Waldrop, J. R.,J. Appl. Phys., 75(9):4548-4550 (1 May, 1994) 196. M6nch, W., Control of Semiconductor Interfaces, (I. Ohdomari, M. Oshima, and A. Hiraki, eds.) (1994) 197. Raghunathan, R., Alok, D., and Baliga, B. J., IEEE Electron Dev. Lett., 16(6):226-227 (June, 1995) 198. M6nch, W., Surface Science, 299/300:928-944 (1994)

244

Wide Bandgap Semiconductors

199. Waldrop, J. R., and Grant, R. W., Appl. Phys. Lett., 62:2685-2687 (May 24,1993) 200. Waldrop, J. R., Grant, R. W., Wang, Y. C., and Davis, R. F., J. Appl. Phys., 72(10):4757-4760 (Nov, 1992) 201. Lundberg, L., and 0stling, M., Appl. Phys. Leg., 63(22):3069-3071 (Nov., 1993) 202. Waldrop, J. R., and Grant, R. W., Appl. Phys. Lett., 56:557-559 (Feb 5., 1990) 203. Su, J. N., and Steckl, A. J., lnst. Phys. Conf. Ser. No. 142, pp. 697-700, IOP Publishing, Ltd. (1996) 204. Shenoy, P., Moki, A., Baliga, B. J., Alok, D., Wongchotigul, K., and Spencer, M., Int 'l. Electron Dev. Meet. (IEDM) Tech. Digest, pp. 411-4 14 (Dec. 1114, 1994) 205. Urushidani, T., Kimoto, T., and Matsunami, H., Inst. Phys. Conf. Ser. No. 137, pp. 471-474, IOP Publishing Ltd. (1994) 206. Porter, L. M., Davis, R. F., Bow, J. S., Kim, M. J., and Carpenter, R. W., Inst. Phys. Conf. Ser. No. 137, pp. 581-584, IOP Publishing, Ltd. (1994) 207. Syrkin, A. L., Andreev, A. N., Lebedex, A. A.,J. Appl. Phys., 78:5511-5514 (Nov. 1, 1995) 208. Ruff, M., Mitlehner, H., and Helbig, R., IEEE Trans. Elect. Dev., 41(6):1040-1054 (June, 1994) 209. Wang, Y., Xie, W., Cooper, J. A., Jr., Melloch, M. R., and Palmour, J. W., Inst. Phys. Conf. Ser. No. 142, pp. 809-812, IOP Publishing, Ltd. (1996) 210. Von Muench, W., and Pfaffeneder, I., Int'l. Electron Dev. Meet. (IEDM) Tech. Digest, pp. 337-339 (1977) 211. Chelnokov, V. E., Materials Science and Engineering, B 11:103-111 (1992) 212. Liou, J. J., and Kager, A., lEE Proceedings, 140(4):289-293 (Aug., 1993) 213. Chow, T. P., and Ghezzo, M., Mat. Res. Soc. Syrup. Proc. 423:9-21 (1996) 214. Weitzel, C. E., Inst. Phys. Conf. Ser. No. 142, pp. 765-768, IOP Publishing, Ltd. (1996) 215. Siergiej, R. R., Sriram, S., Clarke, R. C., Agarwal, A. K., Brandt, C. D., Burke, A. A., Jr., Morse, A., and Orphanos, P. A., Inst. Phys. Conf. Set. No. 142, pp. 769-772, IOP Publishing, Ltd. (1996) 216. Sriram, S., Clarke, R. C., Burke, A. A., Jr., Hobgood, H. M., McMullin, P. G., Orphanos, P. A., Siergiej, R. R., Smith, T. J., Brandt, C. D., Driver, M. C., and Hopkins, R. H., IEEE Elect. Dev. Lett., 15(11):458-459 (Nov., 1994) 217. Sriram, S., Augustine, G., Burke, A. A., Jr., Glass, R. C., Hobgood, H. M., Orphanos, P. A., Rowland, L. B., Smith, T. J., Bran&, C. D., Driver, M. C., and Hopkins, R. H., IEEEElect. Dev. Leg., 17(7):369-371 (July, 1996)

P r o c e s s i n g o f Silicon Carbide f o r Devices a n d Circuits

245

218. Morse, A. W., Esker, P. M., Sriram, S., Hawkins, J. J., Chen, L. S., Ostop, J. A., Smith, T. J., Davis, C. D., Barton, R. R., Clarke, R. C., Siergiej, R. R., and Brandt, C. D., IEEE MTT-S (Int'l. Microwave Symp.), Denver, CO (June 8-13, 1997) 219. Weitzel, C. E., Palmour, J. W., Carter, C. H., Jr., and Nordquist, K. J., IEEE Elect. Dev. Lett., 15(10):406-408 (Oct., 1994) 220. Tsap, B., Solid-State Electronics, 38(6):1215-1219 (June, 1995) 221. Palmour, J. W., Carter, C. H., Jr., Weitzel, C. E., and Nordquist, K. J., Diamond, SiC and Nitride Wide Bandgap Semi. Mat. Res. Soc. Syrup. Proc., Vol. 339, MRS, Pittsburgh, PA, USA, pp. 133-134 (1994) 222. Moore, K. E., Weitzel, C. E., Nordquist, K. J., Pond, L. L., III, Palmour, J. W., Allen, S., and Carter, C. H., Jr., IEEE Elect. Dev. Lett., 18(2):69-70 (Feb., 1997) 223. Palmour, J. W., H.-S. Kong, and Carter, C. H., Jr., Int'l. Semi. Device Research Symp., pp. 491-4, Charlottesville, VA USA (Dec. 4-6, 1991) 224. Neudeck, P. G., Petit, J. B., and Salupo, C. S., Trans. 2nd Int'l. High Temp. Elect. Conf. (HiTEC), Charlotte, NC, Session X, pp. 23-28 (June 5-10, 1994) 225. Dohnke, K. O., Peters, D., Rupp, R., Volkl, J., and Stephani, D., Supplemental Trans. 2nd Int '1. High Temp. Elect. Conf. (HiTEC), Charlotte, NC (June 5-10, 1994) 226. Ivanov, P. A., Savkina, N. -S., Samsonova, T. P., Panteleev, V. N., and Chelnokov, V. E., Inst. Phys. Conf. Ser. No. 137, pp. 593-595, IOP Publishing, Ltd. (1994) 227. Palmour, J. W., Edmond, J. A., Kong, H. S., and Carter, C. H., Jr., Physica B, 185:461-465 (1993) 228. Rupp, R., Dohnke, K., V61kl, J. and Stephani, D., Inst. Conf. Ser. No. 137, pp. 503-6, IOP Publishing, Ltd. (1994) 229. McLean, F. B., Tipton, C. W., and McGarrity, J. M., Inst. Conf. Ser. No. 13 7, pp. 507-510, IOP Publishing, Ltd. (1994) 230. Dohnke, K., Rupp, R., Peters, D., V61kland, J. and Stephani, D., Inst. Phys. Conf. Ser. No. 137, pp. 625-627, IOP Publishing, Ltd. (1994) 231. Alok, D., and Baliga, B. J., Inst. Phys. Conf. Ser. No. 142, pp. 749-752, IOP Publishing, Ltd. (1996) 232. Casady, J. B., Sheridan, D. C., DiUard, W. C., and Johnson, R. W., Mat. Res. Soc. Symp. Proc. 423:105-110 (1996) 233. Krishnamurthy, V., Brown, D. M., Ghezzo, M., Kretchmer, J., Hennessy, W., Downey, E., and Michon, G., Inst. Phys. Conf. Ser. No. 137, pp. 483-486, IOP Publishing, Ltd. (1994)

246

Wide Bandgap Semiconductors

234. Sheppard, S. T., Melloch, M. R., and Cooper, J. A., IEEE Trans. on Elect. Dev., 41:1257-1264 (July, 1994) 235. Casady, J. B., Cressler, J. D., Dillard, W. C., Johnson, R. W., Agarwal, A. K., and Siergiej, R. R., Solid-State Electronics, 39(6):777-784 (June, 1996) 236. Kong, H. S., Edmond, J. A., Palmour, J. W., Glass, J. T., and Davis, R. F., Springer Proc. Phys., Vol. 34, Springer-Verlag (1989) 237. Palmour, J. W., Kong, H. S., and Davis, R. F.,J. Appl. Phys., 64(4):2168-2177 (1988) 238. Palmour, J. W., Kong, H. S., and Davis, R. F., Appl. Phys. Lett., 51(24):2028-2030 (1987) 239. Palmour, J. W., Kong, H. S., and Waltz, D. G., Edmond, J. A., and Carter, C. H., Jr., Trans., 1st Int'l. High Temperature Electronics Conf. (HiTEC), Albuquerque, NM USA, pp. 229-236 (June 16-20, 1991) 240. Siergiej, R. R., Clarke, R. C., Agarwal, A. K., Brandt, C. D., Burke, A. A., Jr., Morse, A., and Orphanos, P. A., Tech. Digest, Int'l. Elect. Dev. Meeting (IEDM), pp. 353-356 (Dec., 1995) 241. Siergiej, R. R., Morse, A. W., Esker, P. M., Smith, T. J., Bojko, R. J., Rowland, L. B., and Clarke, R. C., IEEE 55 th Annual Device Research Conference (June 23-25, 1997) 242. Bhatnagar, M., Alok, D., and Baliga, B. J., Inst. Phys. Conf. Ser. No. 137, pp. 703-6, IOP Publishing, Ltd. (1994) 243. Shenoy, J. N., Cooper, J. A., Jr., and Melloch, M. R., IEEE Electron Dev. Leg., 18(3):93-95 (Mar., 1997) 244. Casady, J. B., Agarwal, A. K., Rowland, L. B., Valek, W. F., and Brandt, C. D., IEEE 55 th Annual Device Research Conference (June 23-25, 1997) 245. Bhatnagar, M., and Baliga, B. J., IEEE Trans. on Elect. Dev., 40(3):645-655 (Mar., 1993) 246. Agarwal, A. K., Siergiej, R. R., White, M. H., McMullin, P. G., Burke, A. A., Rowland, L. B., Brandt, C. D., and Hopkins, R. H., Mat. Res. Soc. Syrup. Proc., 423:87-92(1996) 247. Palmour, J. W., Edmond, J. A., Kong, H. S., and Carter, C. H., Jr., Proc.28th Intersociety Energy Conversion Engineering Conf., American Chemical Society, pp. 1.249-1.254 (Aug., 1993) 248. Palmour, J. W., Allen, S. T., Singh, R., Lipkin, L. A., and Waltz, D. G., Inst. Phys. Conf. Set. No. 142, pp. 813-816, IOP Publishing, Ltd. (1996) 249. Xie, K., Buchwald, W. R., Zhao, J. H., Flemish, J. R., Burke, T., Kingsley, L., Weiner, M., and Singh, H., Int'l. Electron Dev. Meet. Tech. Digest, pp. 415-418 (Dec. 11-14, 1994)

P r o c e s s i n g o f Silicon Carbide f o r Devices a n d Circuits

247

250. Xie, C. K., Flemish, J. R., Burke, T., Buchwald, W. R., Mat. Res. Soc. Symp. Proc., Vol. 423 (1996) 251. Xie,K., Zhao, J. H., Flemish, J. R., Burke, T., Buchwald, W. 1L, Lorenzo, G., and Singh, H., IEEE Electron Dev. Lett., 17(3): 142-144 (Mar., 1996) 252. Agarwal, A. K., Casady, J. B., Rowland, L. B., Seshadri, S., Valek, W. F., and Brandt, C. D., IEEEElectron Dev.Lett. (submitted Feb. 11, 1997) 253. Pelaz, L., Orantes, J. L., Vincente, J., Bailon, L. A., and Barbolla, J., IEEE Trans. on Elect. Dev., 41 (4):587-590 (Apr., 1994) 254. Edmond, J. A., Das, K., and Davis, R. F., J. Appl. Phys., 63:922-929 (Feb., 1988) 255. Neudeck, P. G., Larkin, D. J., Salupo, C. S., Powell, J. A., and Matus, L. G., Inst. Phys. Conf. Ser. No. 137, pp. 475-478, IOP Publishing, Ltd. (1994) 256. K/Srdina, O., Bergman, J. P., Henry, A., Janzen, E., Savage, S., Andre, J., Ramberg, L. P., Lindefelt, U., Hermansson, W., and Bergman, K., Appl. Phys. Leg., 67:1561-1563 (1995) 257. Neudeck, P. G., and Fazi, C., IEEE Electron Dev. Lett., 18(3):96-98 (Mar., 1997) 258. Bhatnagar, M., McLarty, P. K., and Baliga, B. J., IEEE Elect. Dev. Lett., 13(10):501-503 (Oct., 1992) 259. Emons, C. H. H., Koster, R., Paxman, D., and Theunissen, M. J. J., Proc. of lEEE Bipolar/BiCMOS Circuits & Technology Meeting (BCTM), pp. 72-75 (Dec., 1994) 260. Sugii, T., Yamazaki, T., and Ito, T., IEEE Trans. Elect. Dev., 37(11):2331-2335 (Nov., 1990) 261. Shaft, Z. A., Post, I. R. C., Whitehurst, J., Wensley, P., Ashburn, P., Moynagh, P. B., and Booker, G. R., Proc. of IEEE Bipolar/BiCMOS Circuits & Technology Meeting (BCTM), pp. 67-70 (1991) 262. Chang, S. S., Pankove, J., Leksono, M., and VanZeghbroeck, B., 53ra Dev. Research Conf. Digest, pp. 106-107 (June 19-21, 1995) 263. Tomano, M., Johnson, R. W., Jaeger, R. C., and DiUard, W. C., IEEE Trans. Comp. Hybrids, andMan. Tech., 16(5):536-542 (Aug., 1993) 264. Tomano, M., Johnson, R. W., Jaeger, R. C., and Palmour, J. W., Proc. of the 42nd Elect. Comp. and Tech. Conf., pp. 157-161 (May 18-20, 1992) 265. Brown, D. M., Ghezzo, M., Kretchmer, J., Krishnamurthy, V., Michon, G., Gati, G., Trans. 2nd lnt'l. High Temp. Elect. Conf. (HiTEC), Charlotte, N.C., Session XI, pp. 17-22 (June 5-10, 1994) 266. Slater, D. B., Jr., Lipkin, L. A., Johnson, G. M., Suvorov, A. V., and Palmour, J. W., Inst. Phys. Conf. Ser. No. 142, pp. 805-808, IOP Publishing, Ltd. (1996)

248

Wide Bandgap Semiconductors

267. Palmour, J. W., private communication (Feb. 21, 1996) 268. Dillard, W. C., Casady, J. B., Jaeger, R. C., Sheridan, D. C., and Johnson, R. W., Trans. 3ra Int'l. High Temp. Elect. Conf. (HiTEC), Albuquerque, NM, Session VIII, pp. 15-20 (June 9-14, 1996) 269. Xie, W., Cooper, J. A., Jr., and Melloch, M. R., IEEE Elect. Dev. Lett., 15(1):455-457 (Nov., 1994) 270. Casady, J. B., Dillard, W. C., Johnson, R. W., and Rao, U., IEEE Trans. Comp., Pack., andMan. Tech.-A, 19(3):416-422 (Sept., 1996) 271. Dezauzier, C., Becourt, N., Amaud, G., Contreras, S., Ponthenier, J. L., Camassel, J., Robert, J. L., Pascual, J., Jaussaud, C., Sensors and Actuators, A: Physical, 46:71-75 (Jan.-Feb., 1995) 272. Campbell, R. B., Silicon Carbide-1973, (R. C. Marshall, J. W. Faust, Jr., and C. E. Ryan, eds.), pp. 611-617, Univ. of South Carolina Press, Columbia, SC (1974) 273. Brown, D. M., IEEE Trans. Elect. Devices, 40:325-333 (1993) 274. Edmond, J. A., Kong, H. S., and Carter, C. H., Jr., Physica B, 185.453-460 (1993) 275. Brown, D. M., Downey, E., Kretchmer, J., Michon, G., Shu, E., and Schneider, D., Trans. 3rd Int'l. High Temperature Elect. Conf. (HiTEC), Volume 1, Session X, pp. 23-28 (June, 1996) 276. Shields, V. B., Ryan, M. A., and Williams, R. M., Inst. Phys. Conf. Ser. No. 142, pp. 1067-1070, IOP Publishing, Ltd. (1996) 277. Hunter, G. W., Neudeck, P. G., Knight, D., Liu, C. C., and Wu, Q. H., Inst. Phys. Conf. Ser. No. 142, pp. 817-820, IOP Publishing, Ltd. (1996) 278. Arbab, A., Spetz, A., ul Wahab, Q., Willander M., and Lundstrom, I., Sensors and Materials, Vol. 4, (MYU, Tokyo), pp. 173-185 (1993) 279. Kr6tz,G., Wagner, C., Legner, W., Sormtag,H., M611er,H., and MOiler, G., Inst. Phys. Conf. Ser. No. 142, pp. 829-832, lOP Publishing, Ltd. (1996) 280. Cooper, J. A., Jr., MeUoch, M. R., Xie, W., Palmour, J. W., and Carter, C. H., Jr., Inst. Phys. Conf. Ser., No. 137, pp. 711-714, IOP Publishing, Ltd. (1994) 281. Morkoc, H., Strite, S., Gao, G. B., Lin, M. E., Sverdlov, B., and Bums, M., J. Appl. Phys., 76(3):1363-1398 (Aug 1., 1994) 282. Davis, R. F., Physica B, 185:1-15 (1993) 283. Neudeck, P. G., J. Elect. Mat., 24(4):283-288 (1995) 284. Perkins, J. F., Hopkins, R. H., Brandt, C. D., Agarwal, A. K., Seshadri, S. and Siergiej, R. R., ASME Meeting-Turbo Expo, Bkmingham, UK (July, 1996)

Processing of Silicon Carbide for Devices and Circuits

249

285. Semler, C. E., and Cawley, J. D., Ceramic Transactions Silicon Carbide '87, The American Ceramic Society, Inc. (1989) 286. Marshall, R. C., Faust, J. W., Jr., Ryan, C. E., Silicon Carbide-1973, Proc. of 3rd Int'l. Conf. on Silicon Carbide, Miami Beach, FL, Univ. of South Carolina Press (1974) 287. Harris, G. L., Yang, C. Y.-W., Proc. ofthe 1st Int'l. Conf., Washington, DC, Dec. 10-11, 1987, Springer-Verlag,Berlin (1989) 288. Rahman, M. M., Yang, C. Y., and Harris, G. L., Proc. of the 2nd Int'l. Conf., Santa Clara, CA, Dec. 15-16, 1988, Springer-Verlag,Berlin (1989) 289. Harris, G. L., Spencer, M. G. and C.Y.-W. Yang, Proc. of the 3rd lnt'l. Conf., Howard Univ., Washington, D.C., April 11-13, 1990, SpringerVerlag, Berlin (1992) 290. Yang, C. Y., Rahman, M. M., and Harris, G. L., Proc. of the 4th lnt 7. Conf., Santa Clara, CA, Oct. 9-11, 1991, Springer-Vedag, Berlin (1992) 291. Spencer, M. G., Devaty, R. P., Edmond, J. A., Khan, M. A., Kaplan, R., and Rahman, M. M., Proc. of the 5th Conf., Nov. 1-3, 1993, Washington, D.C., Inst. of Phys. Conf. Ser. No. 137, IOP Publishing, Ltd. (1994) 292. Nakashima, S., Matsunami, H., Yoshida, S., and Harima, H., Proc. of the 6th Conf., Sept. 18-21, 1995, Kyoto, Japan, Inst. of Phys. Conf. Ser. No. 142, IOP Publishing, Ltd. (1996) 293. O'Connor, J. R., and Smiltens, J., Proceedings of the Conference on Silicon Carbide, Boston, MA, USA, April 2-3, 1959, Pergamon Press, New York (1960) 294. Silicon Carbide-1968, Proc. of the International Conf., University Park, Penn. State University, PA, USA, Oct. 20-23, 1968, Pergamon Press, New York(1969)

295. Diamond, Silicon Carbide and Related Wide Bandgap Semiconductors, MRS Symposium, Nov. 27-Dec. 1, 1989, Boston, MA, MRS Proc. Vol. 162, Pittsburgh, PA (1990) 296. Diamond, Silicon Carbide and Nitride Wide Bandgap Semiconductors, MRS Symposium, April 4-8, 1994, San Francisco, CA, MRS Symp. Proc. Vol. 339, Pittsburgh, PA (1994) 297. llI-Nitride, SiC, and Diamond Materials For Electronic Devices, April 8-12, 1996, San Francisco, CA, MRS Symp. Proc. Vol. 423, Pittsburgh, PA (1996)

298. 38 th Electronic Materials Conference, Univ. of California at Santa Barbara, Santa Barbara, CA, USA (June 26-28, 1996) 299. International Conference on SiC, III-Nitrides, and Related Materials, Stockholm, Sweden, Aug. 31-Sept. 5 (1997)

300. Casady, J. B., and Johnson, R. W., Solid-State Electronics, 39(10):14091422 (Oct., 1996)

6 Plasma Etching of III-V Nitrides Randy J. Shul

1.0

INTRODUCTION

Interest in GaN and related III-V nitride materials continues to grow as demonstrations of blue, green, and ultraviolet (UV) light emitting diodes (LEDs), blue lasers, and high temperature transistors are reported.[l]-[221 Many of these advances may be attributed to the recent progress in material growth technology. Although further improvements in material properties can be expected, enhanced device performance will also require improved process capabilities, including plasma etching. Laser facet fabrication is especially dependent upon plasma etch pattern transfer, since the majority of epitaxially grown III-V nitrides are on sapphire substrates which inhibit cleaving the sample with reasonable yield. Plasma etching has had a significant impact on the fabrication of several photonic and electronic devices to improve critical dimension control, reproducibility, and anisotropy over wet chemical etching. Although many advances in plasma etch technology have transpired, the rapid development of state-of-the-art devices, sub-0.5 gm features, and complex material structures (including the III-V nitrides), has increased the requirements for etch processes. Plasma etch processes often demand highly anisotropic profiles, high etch rates, smooth morphologies, and low-damage. Etching III-V nitrides is further complicated due to their inert chemical nature 250

Plasma Etching of lll- V Nitrides 251 and their strong bond energies, as compared to other compound semiconductors. GaN has a bond energy of 8.92 eV/atom, InN, 7.72 eV/atom, and A1N, 11.52 eV/atom as compared to GaAs which has a bond energy of 6.52 eV/atom. Due to high bond energies and inert chemical behavior, the III-V nitrides resist etching in standard, room temperature wet chemical etchants. Therefore, essentially all device patterning has been accomplished using plasma etching technology. For example, commercially available LEDs from Nichia are fabricated using a C12-based RIE to expose the n-layer of the heterostructure.[1][2][5]The first GaN-based laser diode was also fabricated using RIE to form the laser facets.[71 The etched sidewalls were somewhat rough and may have contributed to scattering loss and a high lasing threshold, thus demonstrating the need for improved plasma etch capability. Perhaps the most significant advancement in plasma etching of the III-V nitrides has been the utilization of high-density plasmas. High-density plasma etch systems typically yield higher etch rates with less damage than more conventional reactive ion etch (RIE) systems. This is due to plasma densities which are 2 to 4 orders of magnitude greater and the ability to effectively decouple ion energies and plasma density. Etch profiles also tend to be more anisotropic due to lower process pressures which result in less collisional scattering of the plasma species. Plasma etching of GaN has been reported using several techniques including: RIE, electron cyclotron resonance (ECR), inductively coupled plasma (ICP), magnetron reactive ion etch (MIE) systems, and chemically assisted ion beam etching (CAIBE). This chapter examines the advances made in patterning III-V nitrides. Etch data are reviewed for wet chemical etching, RIE, ion beam etching, and high-density etch systems. Etch rates, profiles, plasma-induced-damage, near-surface stoiehiometry, and surface and sidewall morphology will be discussed. Several excellent reference sourees[Z3H251 are available to provide a thorough background of plasma science.

2.0. E T C H TECHNIQUES

2.1

Wet Chemical Etching

Wet chemical etching with acid or base solutions is often used to transfer patterns into compound semiconductors. The process involves

252

Wide B a n d g a p S e m i c o n d u c t o r s

either the oxidation or reduction of the semiconductor surface followed by removal of the soluble reaction product. For a diffusion-limited reaction, rates are limited by either the diffusion of reactive species to the surface or diffusion of the soluble reaction products from the surface. Diffusionlimited reactions are often difficult to control due to the flow dynamics involved. Alternatively, for a reaction-limited etch, chemical reaction at the substrate surface is the rate-limiting-step. These processes are much easier to control and are highly temperature dependent as shown in the equation: Eq. (1)

Etch rate = Ke'ea/kr

where K is the temperature dependent constant, E a is the activation energy, k is the Boltzmann constant, and T is the temperature of the solution. Wet etching is typically highly selective with virtually no damage to the material. The etch is often crystallographic or isotropic, difficult to control, and has poor resolution, thereby limiting the critical dimension of the etched features. Due to their inert chemical nature and high bond strengths, the III-V nitrides resist etching in common compound semiconductor wet chemical etchants at room temperature. Very slow etching of GaN has been reported in hot alkalis or electrolitically in NaOH.[26][27] InN etch rates of 300-600 A/min have been reported in aqueous KOH and NaOH at 60~ 29] For amorphous or polycrystalline A1N, wet chemical etching has been reported in several different solutions including H3PO4, HF/H20, HNO3/HF, and dilute NaOH at relatively low rates (<500 A/min).[3~ 36] Mileham et al. compiled etch results for binary III-V nitrides in a series of wet chemical etchants.[ 3711381GaN and InN did not etch in either acid or base solutions below--80~ However, single crystal A1N etched in strong KOH- and NaOH-based solutions at room temperature. Figure 1 shows etch rates as a function of solution temperature for three different A1N samples in a KOHbased photoresist developer solution (AZ400K). The etch was strongly dependent upon etchant concentration and temperature, but was not responsive to agitation. This implies a reaction-limited etch regime. The data are also displayed as an Arrhenius plot in Fig. 2. The data designated by triangles represent the polycrystalline A1N grown on GaAs, while the two other samples are single-crystal A1N grown on A120 3. The polycrystalline sample etches much faster than the single-crystal sample as can be seen in Figs. 1 and 2. The data designated by circles has a double crystal x-ray diffraction peak width of 400 arc sec while the data designated by squares

Plasma Etching of lll- VNitrides 253 have a peak width of--200 arc see. Higher etch rates for lower crystalline quality material may be attributed to more defects or dangling bonds with which OH- ions in the etch solution can interact. The etching process has an activation energy of-15.45 kCal/mol. A1N etch rates are well controlled, highly selective to GaN and InN, isotropic, temperature dependent, and strongly dependent on material quality.

12,000

I'"

I

!

- - ~ Al203HiQu' gh ,dityIklNon I

lO,O00

-(~ -/kiNon AI203 - -~-- AINon GaAs

~

9 -2

!

I

8000

i

iiiii

o

6000

=9 u.l

4000

i

,

A

I

2000

A~ !

!

0

20

(3"" |

~

40

60

I

I

(3

.80

100

Temperature (~ Figure 1. A1Netch rates for threedifferentqualitysamplesversussolutiontemperaturefor a KOH-basedphotoresistdeveloper(AZ400K).

Photoenhanced wet chemical etching has recently been reported for GaN.[39][4~ Photoenhanced wet chemical etching incorporates the photogeneration of electrons and holes in the material to enhance chemical etching. Minsky et al. reported etch rates for unintentionally doped n-type GaN, ranging from a few 100 A/rnin in HC1 solutions to several 1000 A/ rain in KOH solutions using a He-Cd laser.[39] Youtsey et al. recently reported photoenhanced etching of GaN at rates of-200 A/rain in KOH using broad-area Hg lamp illumination.[4~ Etching was observed for both n+ and unintentionally doped GaN while no etching was observed for p-GaN.

254

Wide Bandgap Semiconductors

I

104

I

I

_ AIN

1000

I

I

~AIN on GaAs

_

on(~)Al203

HighQ u a l i t y ~

-- ~

"

m

100 2.6

,

%

2.8

3

3.2

3.4

3.6

3.8

1000/T (K) Figure 2. Arrenhius plot of AIN etch rate for three different samples etched in AZ400K developer solution. The activation energy (Ea) is 15.45 + 0.44 kCal/mol.

2.2

RIE Etching

High quality plasma etching of the III-V nitrides has been developed due, in part, to the shortcomings of wet chemical etching. In general, plasma etching involves two general mechanisms: (i) a chemical based mechanism where reactive species formed in the plasma interact with the substrate and (/0 a physical based mechanism where energetic ions formed in the plasma sputter the substrate surface. The plasma etch process consists of the following series of steps: (a) Production of reactants in the plasma (b) Transport of reactants to the substrate surface (e) Adsorption of reactants onto the substrate surface (d) Chemical reaction (e) Desorption of the etch products from the surface Etch processes dominated by chemical mechanisms tend to be isotropic with virtually no physical ion bombardment of the substrate surface, thus minimizing plasma-induced-damage. However, this can lead to a loss of critical dimensions, thereby reducing the utility of the process for device fabrication. Alternatively, physically dominated etch mechanisms initiate sputter desorption of the substrate and/or etch products formed on

Plasma Etching of llI-V Nitrides 255 the surface by high energy ions (> 100 eV). The etch is characteristically anisotropic but can result in significant damage, rough surface morphology, and a non-stoichiometrie surface (thus minimizing device performance). A well-designed etch process can balance the chemical and physical components of the etch mechanism to yield high resolution features with minimal damage. The purely physical component of an etch mechanism is exhibited by ion milling. Pearton et al.[411reported ion milling rates for GaN, InN, A1N, and InGaN using 100-500 eV Ar + ions. The ion milling rates were approximately a factor of 2 less than that for GaAs or InP due to stronger bond energies of the III-V nitrides. The milled surfaces were quite smooth with virtually no change in the near-surface stoichiometry. RIE utilizes both chemical and physical components of an etch mechanism. The plasma is typically generated at a radio frequency (rf) of 13.56 MHz between two parallel electrodes in a reactive gas (see Fig. 3). The substrate is placed on the powered electrode and is etched by chemical and physical interactions between the reactive and ionized species formed in the plasma and the substrate surface. The etch is enhanced by sputtering of the surface by impinging ions which typically have energies of a few hundred volts. Ion bombardment energies are defined as the energy with which ions cross the plasma sheath formed just above the substrate. This physical component of the etch mechanism enhances the anisotropy of the etch, independent of the crystallographic orientation of the substrate. RIE typically runs at relatively low pressures ranging from a few mtorr up to 200 mtorr, further enhancing the anisotropy of the etch by minimizing collisional scattering of the ions. Several reports of RIE etching of the III-V nitrides show that it has been quite successful with GaN etching in chlorine-[421[43] and bromine-[441 based plasma chemistries. The first RIE etching of GaN was reported by Adesida and co-workers in SIC14, SiC14/Ar, and SiC14/SiF4 plasmas.[421Etch rates increased monotonically with increasing de-bias, exceeding 500 A/min at -400 V. Etch rate was independent of pressure (20 to 80 mtorr) and plasma chemistry. Etches were smooth with an overcut profile due to the physical component of the etch mechanism. Lin et al. reported similar results in BC13 and SiC14 plasmas with increasing GaN etch rates as the rfpower was increased.[43]Etch rates as high as 1050 A/min were reported in the BC13 plasma, while etch rates in SiC14 were slower. Ping et al. reported GaN etch rates in HBr, HBr/Ar, and HBr/H 2 exceeding 600 A/min at -400 V de-bias.[44]The addition of either Ar or H Eto the HBr plasma lowered the

256

Wide Bandgap Semiconductors

etch rates under all plasma conditions studied. GaN etching was also reported in CHF3/Ar and C2C1Fs/Ar RIE plasmas with rates ranging from 60-470 A/min.[ 45] Etch rates increased with rf-power and decreased with pressure. For R/E, the best GaN etch results were obtained under high ion energy conditions where the GaN bonds could be more easily broken and the sputter desorption mechanism was more efficient.

Showerhead Gas D i s t f i r'--'-.-" ~'

b

."r': "-''~~' T I ' :

u

~

..... ~ . ' - ' ~ " ' - : ~ 7 1

Powered .................. ~ Electrode- ~ . ~ _ . . . _ ~ .~.

! ...

-- q ~ m n l a

. . . . . r,--,

13.56 _-MHz To Pumping System Figure 3.

Schematic diagram of a reactive ion etch (RIE) system.

Improved etch rates and anisotropic profiles achieved with RIE, as compared to wet chemical etching, may be partially attributed to the acceleration of energetic ions from the plasma to the wafer. However, this energetic ion bombardment of the surface can damage the sample and degrade both electrical and optical device performance. Attempts to minimize such damage by reducing the ion energy or increasing the chemical activity in the plasma often results in a loss of etch rate or anisotropy. This significantly limits critical dimensions and reduces the utility of the process for device applications requiting vertical etch profiles. Therefore, it is necessary to develop plasma etch processes which yield anisotropic profiles, high etch rates, and low-damage for optimum device performance.

Plasma Etching of lll-VNitrides 257

2.3

High-Density Plasma Etching

Low-damage, high-density plasmas including ECR, ICP, and MIE, have shown improved etch results for the III-V nitrides as compared to R/E. Ion densities often exceeding 5 x 1011 cm -3 are produced in the highdensity systems, which increases the potential etch rate due to higher ion and neutral reactant flux. A schematic diagram of a typical low profile ECR system is shown in Fig. 4. [46]-[48] Due to the magnetic confinement of electrons within the microwave source (2.45 GHz), high-density plasmas are formed at low pressures with low plasma potentials and ion energies. The sample is located downstream from the microwave source to minimize exposure to the intense discharge and to reduce the physical ion bombardment component of the etch. Therefore, less damage than that produced by RIE has been observed during ECR etching of III-V materials.[ 49]-[53] Highly anisotropic etching can be achieved by superimposing an if-bias (13.56 MHz) on the sample and employing low pressure conditions (< 5 mtorr), minimizing ion scattering and lateral etching. With rf-biasing, energetic ions are accelerated from the plasma to the sample with the potential for kinetic damage to the surface. This is differentiated from RIE in that the incident ion energy is decoupled from the ion current density or plasma power.

2.45 GHz Microwave

~

Upper Magnets

[

.!Plasma i ,'~. '7 9

..... : .....

.

...

.

j,

Load Lock

, 9 .. 9

Sample ~-

:

. . .

~ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ ~

/

Collimating Magnets

13.56 MHz

]_--

Figure 4. Schematic diagram of an electron cyclotron resonance (ECR) etch system.

258

Wide Bandgap Semiconductors

The first ECR etching of GaN, InN, and A1N, was reported by Pearton and co-workers using low pressure CH4/H2/Ar , BCI3/Ar, C12, and C12/H2 plasmas.[ 54] The etch rates for all three materials increased with increasing de-bias. As H 2 was added to the C12 plasma, etch rates also increased and the surface morphology became smoother, implying equirate removal of the group-III and N etch products. The etch rates ranged from 100-400 A/min at 1 mtorr and - 150 V de-bias for C12/H2 plasmas. Slower rates (typically < 100 A/min) were observed in CH4/H2/Ar plasmas. Higher de-biases were necessary to initiate etching. Etch rates were fiarther improved as the plasma density or ECR source power was increased.[ 551Etch rates of 1100 A/min for A1N and 700 A/min for GaN at -150 V de-bias in a C12/H2 plasma and 3 50 A/min for InN in a CH4/H2/Ar plasma at -250 V debias were reported. The etched surfaces remained stoichiometric and profiles were anisotropic over a wide range of plasma conditions. Several additional reports have been published for ECR etching of GaN with etch rates as high as 1.3 ~tm/min at <-275 V de-bias. [56]-[64] ICP offers an alternative high-density plasma technique where plasmas are formed in a dielectric vessel encircled by an inductive coil into which rf-power is applied,[47][48] as shown in Fig. 5. The electric field produced by the coils in the horizontal plane induces a strong magnetic field in the vertical plane, trapping electrons in the center of the chamber and generating a high-density plasma. At low pressures (< 20 mtorr), the plasma diffuses from the generation region and drifts to the substrate at relatively low ion energy. Thus, ICP etching is also expected to produce low damage with high etch rates. As with ECR etching, anisotropic profiles are obtained by superimposing a rf-bias on the sample to independently control ion energy. The general belief is that ICP sources are easier to scale-up than ECR sources and are more economical in terms of cost and power requirements. ICP does not require the electromagnets or waveguiding technology necessary in ECR. Additionally, automatic tuning technology is much more advanced for rf-plasmas than for microwave discharges. The first ICP etch results for GaN were reported in a C12/H2/Ar ICP-generated plasma with etch rates as high as 6875 7k/min.[65][66]Etch rates increased with increasing de-bias and etch profiles were highly anisotropic with smooth etch morphologies. MIE is similar to RIE with the addition of a magnetic field which confines the plasma electrons close to the powered electrode, and therefore to the sample.[67][ 68] Under these conditions, electron wall loss is minimized and ionization efficiency, generation of reactive species, and etch

Plasma Etching of lll-VNitrides

259

rates are increased. Higher plasma densities can be sustained at much lower and more controlled de-biases as compared to RIE. Therefore, ion bombardment energies and plasma-induced-damage are lower while etch rates remain high. Etch rates of 3500 A/min were achieved in MIE using BCI 3 plasmas at de-biases less than -100 V and 5 mtorr pressure.[ 69] The etch was fairly smooth, with anisotropic profiles. Auger electron spectroscopy (AES) showed etched surfaces with lower Ga:N ratios up to a depth of-100 A.

Showerhead Gas Distribution..~, iiilili

J

1

I i 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Alumina 0 Chamber,,,,,,~[i i

- i-

rf Power Supply

i:

iPlasma I

O!. L o ad L ock

"........ ~ ; . ~ ; L Z S ; ; ; ; . . , ~ - ~

~,\\\\\\\\\\\\\\~~\\\\\\\\\~%~"~

Sample

J l l l l l l l ' l , ~ l l l l l t l l l l l l l ~ Ill

Powered Electrode

9 =" MHz

Figure 5. Schematic diagram of an inductively coupled plasma (ICP) etch system.

2.4

Additional Plasma Etch Platforms

Ion beam techniques, such as chemically assisted ion beam etching (CAIBE) and reactive ion beam etching (R/BE), typically use ECR or ICP sources to generate high-density plasmas. Samples are separated from the ion source by one or more grids which are used to accelerate ions to the surface at highly controlled energies. Ion energies are ordinarily higher (300-2000 eV) than those generated in RIE, ECR, or ICP. Due to the acceleration of the ions and the low process pressures used (< 5 mtorr), etch profiles are highly anisotropic, however the potential for plasmainduced-damage exists. In CAIBE, reactive gases are added to the plasma

260

Wide Bandgap Semiconductors

downstream of the acceleration grids to enhance the chemical component of the etch mechanism; whereas in RIBE, reactive gases are introduced in the ion source. Using CAIBE, GaN etch rates as high as 2100 A/rain were reported with a 500 eV Ar + ion beam directed onto the sample in either a C12 or HC1 arnbient.[7~ 71] Etch rates increased with beam current, reactive gas flow rate, and substrate temperature. Anisotropic profiles with smooth etch morphologies were observed. An alternate, low-damage etch technology which has recently stimulated interest is low energy electron enhanced etching (LE4).[ 72]-[751Electrons at low energies (- 15 eV) arrive at the semiconductor surface with reactive species at thermal velocity. The electrons have virtually no momenturn and therefore do not induce damage to the semiconductor as observed by other plasma etch techniques. The electrons appear to enhance reactions between the reactive gas and the semiconductor yielding highly anisotropic profiles. LE4 etching of GaN in pure H 2 showed a strong temperature dependence with etch rates ranging from ~70 A/min at 50~ to ~500 A/min at 250~ [74][75]In a pure C12 LE4 plasma, GaN etch rates of 2500 A/min were reported at 100~ Etch profiles were highly anisotropic with smooth morphologies and stoichiometric surfaces. Photoassisted dry etching of GaN has been reported in HC1 using 193 nm ArF excimer laser for illumination.[ 76] At temperatures between 200 and 400~ the GaN etch was smooth with distinct sidewalls and etch rates < 80 A/min. Laser interactions with the GaN and the HC1 resulted in vibrational and electronic excitation which led to bond breaking and desorption of reactant products.[ 77] The laser energy enhanced the reaction between HC1 and GaN, yielding GaC1x products which thermally desorb while N is removed as either molecular N 2 or NH x.

3.0

PLASMA CHEMISTRY

Several different plasma chemistries have been used to etch compound semiconductor materials, including halogen- and hydrocarbon-based plasmas. Review articles are available Which summarize etch results for GaAs and InP under a variety of plasma conditions.[781[79] Table 1 shows the boiling points of several possible etch products formed when compound semiconductor materials are exposed to halogen- or hydrocarbonbased plasmas. For halogen-based plasmas, compound semiconductor etch

Plasma Etching of lll-VNitrides 261 rates are typically limited by the volatility of the group-III halogen etch product. Therefore, chlorine-based chemistries are preferred to etch Gacontaining materials due to the high volatility of GaC13 as compared to fluoride-based etch products. In a chlorine-based plasma, GaAs etch rates are typically fast (> 2 ~tm/min), well controlled, anisotropie, and smooth. The etch rates tend to be faster in pure C12 due to the strong chemical component of the etch mechanism and the high concentration of reactive C1 available in the plasma. However, the surface morphology can be rough. The addition of BC13 to the plasma chemistry reduces the etch rate due to less available reactive C1, but the etch tends to be smoother and more anisotropie (due to a higher physical component of the etch mechanism). Smooth anisotropie profiles have also been obtained for Ga-eontaining compound semiconductors etched in CH4/Hz-based plasma chemistries at much slower rates as compared to Cl-based plasmas. This is unexpected based on the information in Table 1 (the volatility of the Ga(CH3) 3 etch product is much higher than that for GaC13 and demonstrates the complexity of the etch process where redeposition, polymer formation, and gasphase kinetics can influence the etch rates). Etch rates for In-containing species in a RIE-generated C12 plasma at room temperature tend to range from a few hundred to-1000 A/min. The surfaces are typically rough with an overcut profile due to the low volatility of the InC13 etch products and preferential loss of the group-V species. However, at elevated temperatures (>130~ the volatility of InC13 etch product increases and the etch rates and surface morphology improve.[8~ 831 Recently, high rate InP etching (1.0 to 3.8 ~tm/min) has been reported in C12/Ar ECR-generated plasmas at 25~ [84][85]These results suggest that using a high-density plasma etch platform enables sufficient sputter desorption of the low volatility InC13 etch products before the surface is passivated. However, for room temperature etching of In-containing films, CHa/H2-based plasmas have often been preferred due to the formation of volatile (CH3)3In etch products.[ 55][86]-[89]InP etch rates obtained in RIEand ECR-generated CHa/H2 plasmas range from--500 to 1000 A/min with anisotropic profiles and smooth etch surfaces. GaN typically etches at much slower rates than GaAs. The high volatilities of the Ga- and the nitrogen-based etch products shown in Table 1 implies that the etch rates are not limited by desorption of the etch products. However, due to the strong bond energies of the III-V nitrides, the initial breaking of the group-III-N bond, which must precede the etch product formation, may be the rate limiting step.[ 9~ This topic will be discussed in more detail later in this chapter.

262

Wide Bandgap Semiconductors

T a b l e 1. Boiling Points for Possible Etch Products o f C o m p o u n d S e m i c o n ductor M a t e r i a l s E t c h e d in H a l o g e n - or CHa/H2-based P l a s m a s

Etch Products

3.1

Boiling Points (~

AICI3

183

AIF3

na

AII3

360

A1Br3

263

(CH3)3AI

126

GaCI 3

201

GaF 3

1000

GaI 3

sublimes 345

GaBr3

279

(CH3)3Ga InCl 3

55.7 600

InF 3

>1200

InI 3

na

InBr3

sublimes

(CH3)3In

134

PC13

76

PF 3

-101

PI 3

decomposes

PBr 3

173

PH 3

-88

AsC13

130

AsF 3

-63

AsI 3

403

AsBr3

221

AsH 3

-55

Gas Mixtures A l t h o u g h fast G a N etch rates h a v e b e e n observed in c h l o r i n e - b a s e d

plasmas, the source o f reactive C1 as well as the use o f additive gases h a v e not been discussed. V e r y often gas m i x t u r e s are used in a p l a s m a to increase etch rate, i m p r o v e anisotropy, increase selectivity, or p r o d u c e

Plasma Etching of lll-V Nitrides 263 smoother etch morphologies by adjusting the chemical:physical ratio of the etch mechanism. For example; 02, H 2, or N 2 have been added to halogenbased plasmas to modify the concentration of reactive species in the plasma or initiate the formation of sidewall polymers to enhance anisotropy. Additionally, inert gases such as Ar or He are often used to stabilize or dilute the plasma. Fluorine-containing gases are often added to improve etch selectivity of GaAs films over A1GaAs films due to the formation of non-volatile A1F3 etch products. The use of gas mixtures to etch the UI-V nitrides in high-density plasma reactors has had varying success and is summarized below.

3.2

Hydrogen

The addition ofH 2 to chlorine-based plasmas typically results in slower etch rates for GaAs and GaP since H 2 acts as a scavenger of reactive C1 and forms HC1. In Fig. 6, GaN etch rates are plotted as a function of %H 2 concentration for ECR- and ICP-generated C12/H2/Ar plasmas. ECR etch conditions were: 1 mtorr, 850 W ECR source power, 150 W rf-cathodepower with a corresponding de-bias of-190 + 20 V, 25 sccm C12/H2, and 5 sccm Ar. ICP etch conditions were: 2 mtorr, 500 W ICP source power, 95115 W rf-cathode-power with a constant dc-bias of-250 + 10 V, 25 sccm C12/H2, and 5 sccm Ar. Samples etched in the ECR were grown by metalorganic molecular beam epitaxy (MO-MBE), whereas samples etched in the ICP were grown by metal-organic chemical vapor deposition (MOCVD). GaN etch rates in the ECR and ICP increased slightly as H 2 was initially added to the C12/Ar plasma (10% H2). Using quadrupole mass spectrometry (QMS) in the ECR discharge, the C1 concentration (indicated by m / e = 35 peak intensity) remained relatively constant at 10% H 2. As the H 2 concentration was increased further, the C1 concentration decreased and the HC1 concentration increased as the GaN etch rates decreased in both plasmas, presumably due to the consumption of reactive C1 by hydrogen. GaN etch profiles showed a strong dependence on the %H 2 in the C12/H2/Ar ECR plasma (Figs. 7a-c). The etched surface was quite rough (Fig. 7a) in the C12/Ar plasma, possibly due to preferential removal of the GaC13 etch products or micromasking of the surface. The foot observed at the edge of the etched feature may be attributed to mask-edge erosion due to the aggressive attack ofphotoresist by reactive C1. As the H 2 concentration was increased to 20%, the etch became smooth and very anisotropic (Fig. 7b). At 60% H 2, the etch remained smooth and anisotropic with a slight foot at the base of the feature (Fig. 7c). For GaN samples etched in

264

Wide Bandgap Semiconducwrs

the C12/H2/ArICP-generated plasma (Figs. 7d-0, the features were anisotropic and smooth, independent of%H 2.

5000

4OOO

ICP

3000

2000

ECR

I000

0

t

i 0

i 20

i 40

! 60

~

i 80

i 100

% H 2 Concentration in CI2/H 2 Plasma

Figure 6. GaN etch rates as a function of%H 2 for ECR- and ICP-generated CI2/H2/Arplasmas.

I

ttt~/

ECR

(a)

(b)

(c)

|CP

(d)

(e)

(f)

Figure 7. SEM micrographs of GaN samples etched in either an ECR or ICP CI2/H2/Arplasma (a, d) 0 % HE, (b, e) 2 0 % HE, and (c, f) 6 0 % H E. T h e photoresist m a s k has b e e n r e m o v e d .

at

Plasma Etching of lll-VNitrides 265 In Fig. 8, BCI 3 was substituted for C12 and was used to etch GaN in both the ECR and ICP reactors. The increase in etch rate observed at 10% H 2 concentration in the ECR-generated BC13 plasma correlated with an increase in the reactive C1 concentration as observed by QMS. As the H 2 concentration was increased further, the C1 concentration decreased, the HC1 concentration increased, and the GaN etch rates decreased due to the consumption of reactive C1 by hydrogen. In the ICP reactor, GaN etch rates were quite slow and decreased as H 2 was added to the plasma, up to 80% where a slight increase was observed. The GaN etch rates were consistently faster in the C12-based plasmas as compared to BC13, due to the generation of higher concentrations of reactive C1.

2000

.=. E -2

i

i

i

1500 ECR 1000

Z

ICP

500

0 !

I

I

I

I

I

I

0

20

40

60

80

100

% H 2 Concentration in BCI3/H 2 P l a s m a

Figure 8. GaN

3.3

etch rates as a function o f % H 2 for ECR- and ICP-generated BCI3/H2/Ar plasmas.

Sulfur Hexafluoride

In Figs. 9 and 10, GaN etch rates are shown for ECR- and ICPgenerated C12/SF6/Ar and BC13/SF6/Ar plasmas as a function of %SF 6. ECR plasma conditions were: 1 mtorr, 850 W ECR source power, 150 W rf-cathode-power with a corresponding dc-bias o f - 190 + 20 V, 25 sccm C12/SF 6 or BC13/SF 6, and 5 seem Ar. ICP plasma conditions were: 2 mtorr, 500 W ICP source power, 95-115 W rf-cathode-power with a constant debias of-250 • 10 V, 25 seem C12/SF 6 or BC13/SF 6, and 5 seem Ar. (The pressure was increased to-~3 mtorr in the ICP reactor at > 60% SF6). With

266

Wide Bandgap Semiconductors

the substitution of SF 6 for H 2 in the C12-based plasma, the GaN etch rates were typically slower, by a factor of 2. In general, as the concentration of SF 6 was increased, the etch rates decreased independent of etch technique. As the %SF 6 was increased from 0 to 20 in the ECR, the C1 concentration (m/e = 35) decreased but remained significant; faster GaN etching at 20% SF 6 might be expected, based on the C1 concentration alone. However, formation of SC1 (m/e = 67) was observed at 20% SF6 which may be responsible for the reduced GaN etch rate due to consumption of the reactive C1 by S. At 30 and 40% SF6, the C1 concentration was greatly reduced and slow GaN etch rates resulted.

5000

,, I

I

I

,

,

I ~

~ECR 4000

ICP

....,

E 3000

2000 t.t.l t,.3

I000

-

,_, I

0

. /

.

20

I

I

40

60

0

% S F 6 C o n c e n t r a t i o n in CI2/SF6Plasm,'~

Figure 9. GaN etch rates as a function of %SF 6 for ECR- and ICP-generated CI2/SFe/Ar plasmas.

In Fig. 10, the opposite trend was observed for BC13, where the GaN etch rates were typically greater when SF 6 was substituted for H 2 in both the ECR and ICP. The GaN etch rate increased up to 20% SF 6 in the ICP and 30% SF 6 in the ECR, and then decreased sharply. In the ECR, the C1 concentration (m/e = 35) increased as the SF 6 increased to 30% and then decreased at 40%. This implies that at low SF 6 concentrations, the SF 6 enhanced the dissociation of BC13 resulting in higher concentrations of reactive C1 and faster etch rates. However, above --30% SF6, the sulfur appeared to consume reactive chlorine as the SC1 concentration increased and the etch rate decreased.

Plasma Etching of lll- V Nitrides 267

5000

,

l

i

"'C

i ~

4000 E

-2

,,...,

3000

2000

t.rd

1000

0

1

1

1

I

0

20

40

6()

_

i

~0

% SF Concentration ill BCIJSF Plasnia 6

Figure 10. GaN etch rates as a function o f % S F 6 for ECR- and ICP-generated BC13/SF6/Ar plasmas.

3.4

Nitrogen

GaN etch rates were also obtained for C12/N2/Ar and BC13/N2/Ar plasmas under the following ECR and ICP conditions: 2 mtorr, 850 W ECR source power or 500 W ICP source power, 110 to 170 W rf-cathode-power with a constant de-bias of-200 + 10 V, 25 seem C12/N2 or BC13/N2, and 5 seem Ar. Figure 11 shows GaN etch rates as a function of %N 2 concentration in both ECR and ICP C12 plasmas. As the %N 2 increased in the C12 plasma, the GaN etch rates decreased due to less available reactive C1. However, as shown in Fig. 12, GaN etch rates increased significantly as N 2 was added to the ECR- and ICP-generated BC13 plasma. Etch rates increased up to 40% N 2 and then decreased at higher N 2 concentrations. This trend was similar to that observed in ECR and ICP etching of GaAs, GaP, and !n-containing materials.[ 91]-[93]Ren et al. observed peak etch rates for In-containing materials in an ECR plasma at 75% BC13- 25% N 2. As N 2 was added to the BC13 plasma, Ren observed maximum emission intensity for atomic and molecular C1 at 75% BC13 using optical emission spectroscopy (OES). Correspondingly, the BC13 intensity decreased and a BN emission line appeared. It was suggested that at 75% BC13, N 2 enhanced the dissociation of BC13 resulting in higher concentrations of reactive C1 and C1 ions and higher etch rates. This may explain faster GaN etch rates observed as N 2 was added to the BC13/Ar plasmas. In this study, however,

268

Wide Bandgap Semiconductors

higher concentrations of reactive C1 and BN emission were not observed using OES in the ICP.

7000

I

I

I

I

I

-~

6ooo

ECR [

" - ~ ICP I "~

5000 4OOO

.~

3O00

2ooo r~ 1000 I

I

I

I

I

0

20

40

60

80

I

100

% N 2 Concentration in Cl2/N2Plasma Figure 11. GaN etch rates as a function of %N 2 for ECR- and ICP-generated CI2/N2/Ar plasmas.

4000

I~.~ i

ECR

3500

ICP

3000 *~

2500 2000 1500

z

r~

lO~ 500 "

I

I

I

I

I

I

0

20

40

60

80

100

-

% N 2 Concentration in BCl3/N2Plasma Figure 12. GaN etch rates as a function of %N 2 for ECR- and ICP-generated BC13/N2/Ar plasmas.

Plasma Etching of lll-VNitrides 3.5

269

Argon

Small amounts of Ar are often added to halogen-based plasmas to stabilize the plasma or to improve the sputter desorption efficiency of etch products from the substrate surface (thus increasing etch rate and anisotropy). [93]-[95] GaN etch rates are shown in Figs. 13 and 14 as a function of %Ar in ICP- and ECR-generated C12 and BC13 plasmas. The plasma etch conditions were: 2 mtorr chamber pressure, 30 sccm total flow rate,-~-250 V dc-bias, 25~ substrate temperature, and 500 W ICP source power or 850 W ECR source power. As Ar was added to either ICP or ECR C12 discharges, GaN etch rates decreased due to less available reactive C1 in the plasma. When BC13 was substituted for C12, etch rates were as much as 6 times slower than those obtained in the C12/Ar plasma due to lower concentrations of reactive C1 (see Fig. 14).

7000

i

t

I

!

t'

I -El- ~c~ I 6000 "~

5000 4000

t~ .m

3000

m Z

2000 IOO0

-

I

I

I

I

I

0

20

40

60

80

t

=

100

% Ar Concentration in Cl2/ArPlasma Figure 13. GaN etch rates as a function of%Ar for ECR-and ICP-generatedCI2/Arplasmas.

Based on the preceding observations, GaN etch rates as a function of plasma chemistry can be summarized as follows: (a) comparable in ECR and ICP reactors; (b) faster in C12-based plasmas as compared to BC13based plasmas; (c) sensitive to the gas mixtures and gas ratios; and (d) highly dependent upon the concentration of reactive C1 in the plasma.

270

Wide Bandgap Semiconductors 2000

I

I

I

I

L~

15oo

/

ICP

E ~

1000

Z

500

I

r~

I

0

I

I

I

I

20

40

60

80

II 100

% Ar Concentration in BCI3/Ar Plasma

Figure 14. GaN etch rates as a functionof%Ar for ECR-and ICP-generatedBCI3/Arplasmas.

3.6

Additional Halogens

Additional halogen-containing plasmas have been used to etch GaN including IC1/Ar and IBr/Ar. In Fig. 15, GaN etch rates are shown as a function of if-cathode-power in ECR-generated IC1 and IBr plasmas. Etch conditions were: 1.5 mtorr pressure, 1000 W ECR source power, 4 seem IC1 or IBr, and 4 seem Ar. The GaN etch rates increased with increasing if-cathodepower at similar rates up to 150 W. However, at 250 W if-cathode-power, the GaN etch rate in ICI/Ar increased to > 1.3 ~tm/min whereas the IBr etch rate decreased slightly. This is the fastest etch rate reported to date for GaN. Near-surface AES showed no loss of N in the Ga:N ratio at low rfcathode-powers. The high volatility of the GaI 3 etch products may have increased the chemical etch mechanism of the Ga and thereby minimized preferential loss of the lighter N atoms and maintained the stoichiometry of the as-grown film. [58][96][97]

3.7

Hydrocarbons

Pearton and coworkers were the first to etch GaN, A1N, and InN in ECR-generated CH4/H2/Ar plasmas.[ 55] Etch rates were < 400 A/min at 1 mtorr pressure, 1000 W ECR source power, and ~ -250 V de-bias. As stated earlier, CH4/H 2 plasmas have been used to etch compound semiconductor materials at etch rates much slower than those obtained for C1based plasmas. In Fig. 16, GaN etch rates are plotted as a function of rf-

Plasma Etching of l l l - V Nitrides 271 cathode-power for C12/Ar and CH4[H2/Ar ECR- and RIE-generated plasmas. These experiments were performed in the same chamber with either 1000 or 0 W of applied ECR source power. The pressure was held constant at 1.5 mtorr and the plasma chemistry was 5 seem C12 and 10 sccm Ar or 5 seem CH4, 15 seem H 2, and 10 sccm Ar. GaN etch rates were significantly faster in C12/Ar, independent of etch technique, possibly due to more efficient formation and sputter desorption of the GaC13 etch products as compared to Ga(CH3) 3. Redeposition or polymer formation on the etched surfaces may also be initiated by CHa/H 2 plasmas, thereby reducing the etch rates. Faster GaN etch rates in the high-density ECR were attributed to higher plasma densities, which enhanced the GaN bond breaking and sputter desorption of the etch products.

14,ooo 12,ooo

I

I

i

I

I

--~ -~cll

10,000 ~

8000

-~

6000

z

4ooo

r~

2000

0

I

I

I

I

I

50

100

150

200

250

300

if-cathode-power (W) Figure 15. GaN etch rate as a function of if-cathode-power for ECR-generated ICI/Ar and IBr/Ar plasmas.

In Fig. 17, InN etch rates are plotted as a function of if-cathodepower for C12/Ar and CH4/H2/Ar ECR- and RIE-generated plasmas. These experiments were performed under the same conditions discussed above. InN etch rates obtained in the ECR were fast (independent of plasma chemistry) and increased with if-cathode-power, due to more efficient bond breaking and/or sputter desorption of the etch products at high de-bias. In RIE, the etch rates were quite low and comparably independent of plasma

272

Wide Bandgap Semiconductors

chemistry. GaN etched much faster than InN in the RIE-generated C12/Ar plasma due to lower volatility oflnC13 as compared to GaC13.

7000 - - ~ - RIE Clz/Ar ECR CI21Ar ""A" "RIECH4/H2/Ar - ~ ECR CH4/I-12/Ar

6000 o,,.~

E

5000

i

4000 .=

;~

3000

S

..Q

.(3 s

2ooo

.

..~

..... ,., O..., .,, ........~jC. .....

1000 .z~ -<

0

A

~

.

-

-

.-A ......

4

....

-& . . . . . .

i

i

I

!

1~

2~

3~

400

A

5~

if-cathode-power (W) Figure 16. GaN etch rates as a function of rf-cathode-power for ECR- and RIE-generated Cl2/Ar and CH4/H2/Ar plasmas.

I

8000 .,.q

6000 .m

Z

m

I

I

_ ~ R I E Clz/Ar ECR Cl2/Ar " "&- - RIE CH4/H2/Ar ECR CH4/Hz/Ar

4000

.... 4 , 2o00

-0- ....., 0

100

.-..-..-~-.--,, .--Jr. . . ., 200

300

400

O 500

if-cathode-power (W) Figure 17. InN etch rates as a function of rf-cathode-power for ECR- and RIE-generated C12/Ar and CH4/HE/Ar plasmas.

Plasma Etching of l l l - V Nitrides 273 4.0

PRESSURE

Plasma conditions change quite dramatically as a function of pressure; in particular the mean free path decreases and the eollisional frequency increases as the pressure is increased. This results in changes in both ion energy and plasma density which strongly influence etch properties. Etching GaN in a Cl2-based plasma ordinarily results in faster etch rates as pressure is increased, suggesting a reactant limited regime at lower pressures. However, as pressure is increased further, etch rates often decrease due to higher ion energies, lower plasma densities, redeposition, or polymer formation on the substrate surface. Additionally, at higher pressures, the etch frequently becomes more isotropie and rougher due to lateral etching of the sidewall. In Fig. 18, GaN etch rates are shown as a function of pressure for ECR and ICP plasmas. ECR plasma conditions were: 10 seem C12, 15 seem H2, 3 seem CH4, 10 seem Ar, 850 W ECR source power, and 170~ electrode temperature. ICP plasma conditions were: 22.5 seem C12, 2.5 seem H2, 5 seem Ar, 500 W ICP source power, and 25~ electrode temperature. During these runs, the rf-eathode-power was held constant at 150 W, which resulted in an increase in de-bias (-190 to -450 V for the ECR and -180 to -375 V for the ICP) as the pressure was increased (1 to 10 mtorr). Higher de-biases at higher pressures were attributed to increased eollisional recombination which decreased the plasma density at higher pressures. Despite the differences in plasma conditions, the etch rate trends were similar. Etch rates increased for GaN as the pressure was increased from 1-2 mtorr for the ECR and from 1-5 mtorr for the ICP suggesting a reactant limited regime at low pressure. As the pressure was increased (> 2 mtorr for ECR and > 5 mtorr for ICP), GaN etch rates decreased due either to higher ion energies, lower plasma densities, redeposition, or polymer formation on the substrate surface.

5.0

I O N E N E R G Y AND P L A S M A D E N S I T Y

Etch characteristics show a strong dependence on ion energy and plasma density. Ion energies influence the physical component of the etch, whereas plasma density can effect both the physical and chemical components. In general, etch rates increase as the ion energy increases due to improved sputter desorption of etch products from the surface as well as

274

Wide Bandgap Semiconductors

more efficient breaking of the group-III-N bonds. Anisotropy also typically improves due to the perpendicular ion energy which maintains straight wall profiles. Etch rates may decrease under high dc-bias conditions due to sputter desorption of reactive species from the surface before reaction occurs indicating an adsorption limited etch regime. As a function of increasing plasma density or source power, etch rates typically increase due to: (1) higher concentrations of reactive species which increase the chemical component of the etch mechanism, and/or (2) higher ion flux which increases the sputter desorption component of the etch mechanism. However, trends have been observed where etch rates stabilize or decrease at high plasma densities due either to saturation of reactive species at the surface or creation of an adsorption limited regime. The effects of ion energies and plasma densities are more obvious for high-density plasma systems, since ion energies and plasma densities can be more effectively decoupled as compared to RIE.

8000 7000

E

6000

o

~

4000

~

3000

~ r~

2~

_]+

ICP

i'

1000 0

0

I

I

I

I

I

2

4

6

8

10

Pressure (mTorr) F i g u r e 18. GaN etch rates as a function o f pressure in an ICP-generated C12/H2/Ar plasma at 25~ and in an ECR- generated C12/HE/CH4/Ar plasma at 170~

The III-V nitrides typically etch at much slower rates than more conventional compound semiconductors films, including GaAs and InP. As stated earlier, due to the high volatility of etch products formed in Cl-based plasmas and the strong bond energies of the III-V nitrides, the GaN bond breaking may be the rate limiting step of the etch.[ 9~ These tendencies may

Plasma Etching of l l l - V Nitrides 2 75 be observed in Fig. 19 where GaN etch rates are plotted as a function ofrfcathode-power for ECR, ICP, and RIE discharges. Plasma etch conditions were: 1 mtorr pressure, 10 seem C12, 15 seem H 2, 3 sccm CH4, 10 seem Ar, 1000 W ECR source power, and 500 W ICP source power. For ECR and ICP, the rf-cathode-power ranged from 1-450 W with corresponding dc-biases o f - 1 0 to -400 + 25 V. RIE rf-powers ranged from 50-450 W with corresponding de-biases of-270 to -950 + 50 V. Independent of etch technique, the GaN etch rate increased as the rf-cathode-power or ion energy increased, due to more efficient bond breaking of the GaN and/or more efficient sputter desorption of the etch products. GaN rates were-- 5 to 10 times faster in the high-density ECR and ICP etch systems as compared to R/E, due to a two step process directly related to the plasma flux. Initially, the high-density plasmas increase the bond breaking mechanism allowing the etch products to form and then produce efficient sputter desorption of the etch products. The etched surface morphology or rootmean-square (rms) surface roughness was evaluated and quantified using atomic force microscopy (AFM) as a function of plasma etch technique. It remained relatively constant and smooth (< ---3 nm), independent of ion energy and etch platform.

7000

I

I

--~

I

- ECR

6OOO .N

E

5ooo

.

- -A,- - ICP

/

v

4OOO

-~

3OOO

Zea

2OOO

J

.§

/

o 9 f

o~ .f

1000 0 I

I

I

I

0

1~

2~

3~

I

4~

5~

if-cathode-power (W) Figure 19. GaN etch rates as a function of rf-cathode-power as etched in ECR-, ICP-, and RIE-generated CI2/H2/CH4/Arplasmas.

276

Wide Bandgap Semiconductors

In Fig. 20, GaN etch rates and rms roughness are shown as a function of ICP-source power. During these runs, the if-cathode-power was held constant at 150 W, resulting in a decrease in de-bias (-285 to - 130 V) as the ICP source power was increased (250 to 1000 W). Lower debiases were attributed to increased plasma density and lower mean free path at higher ICP source powers. GaN etch rates increased as the ICP source power was increased from 250-500 W, due to higher concentrations of reactive species in the plasma or higher plasma flux which resulted in more efficient bond breaking and sputter desorption of the etch products. As the ICP source power was increased further, the GaN etch rates decreased, due either to lower ion energies or sputter desorption of the reactants at the surface prior to reaction. The rms roughness increased to greater than 65 nm at 750 W ICP source power (---150 V dc-bias), possibly due to preferential sputtering of N at higher plasma densities.

7000

I

I

I

I

70

I

m / 6000

60

/ /

E

5000

~

4000

-~

3000

so

I -

2ooo

40

30 ~ 0

m

z

~"

-

20 ~

/

r~

I

rrns

10

I

1000 I

I

I

I

I

200

400

600

800

1000

ICP Power (W) Figure 20. GaN etch rates and rms-roughness as a function o f l C P source power in an ICPgenerated CI2/H2/Ar plasma.

6.0

TEMPERATURE DEPENDENCE

As stated earlier, etch rates and profiles can be influenced by the volatility or desorption rate of etch products. Using a variety of clamping techniques along with backside heating or cooling procedures, the

Plasma Etching of llI-VNitrides 277 temperature at which the sample is maintained during the etch can be controlled. The substrate temperature can effect the desorption rate of etch products, the gas-surface reaction kinetics, and the surface mobility of reactants which earl influence etch results. Several reports are available which present temperature dependent etching of GaAs and InP. [80]-[82][9811991 GaAs etch rates generally increase at substrate temperatures above -~150~ due to one of the mechanisms stated above; high temperature etching of Incontaining species has been quite successful using C12-based chemistries, due to the higher volatilities of In-chlorides at temperatures above--130~ Etch rates for GaN and A1N are shown in Fig. 21 as a function of temperature. The GaN samples were etched in C12/I-I2/CH4/Arand C12/H2/ Ar plasmas while the A1N samples were etched only in C12/HJCH4/Ar plasmas. The etch rates for A1N showed a monotonic decrease of a factor of--2 as the temperature was increased from 30-170~ The GaN etch rate in the CI2/Hz/CH4/Ar plasma was relatively constant up to ~125~ and then increased slightly as the temperature was increased to 170~ When CH 4 was removed from the plasma chemistry, the rates decreased by---2040% and showed a monotonic increase as a function of temperature. Faster GaN etch rates at 170~ may be attributed to either higher volatility and desorption rates of etch products, or changes in the gas-surface reaction kinetics. In Fig. 22, the InN etch rates are shown as a function of temperature for the same plasma chemistries. The InN etch rate decreased by > 60% as the temperature was increased to 150~ for CI2]Hz/CH4/Ar plasma chemistry, however, the etch rate increased rapidly above 150~ A similar trend was observed in the CI2/H2/Ar plasma chemistry, with etch rates 20-50% slower than those obtained with CH 4 in the plasma chemistry. Faster GaN and InN etch rates observed with CH 4 present in the plasma chemistry, may be attributed to the formation of the group III-methyl etch product. The initial InN etch rate decrease observed in the C12]I-I2/CH4/Ar plasma chemistry may be due to competitive reactions to form either InC13 or (CH3)3In. As the temperature was increased above 150~ either the volatility or desorption rate of one of the etch products increased or the reaction kinetics became dominated by one of the surface reaction mechanisms. Figure 23 shows characteristic Auger spectra for GaN samples before and after ECR etching at 30 and 170~ in a C12/H2/CH4/Ar plasma. Prior to exposure of the GaN to the plasma, the Auger spectrum showed normal amounts of carbon and native oxide on the GaN surface. Following exposure to the plasma, within experimental error, there was no change in the stoichiometry of the GaN surface, with some residual atomic C1 present. Similar results were observed for the InN and AIN samples.

278

Wide Bandgap Semiconductors

2500

i

i

GaN with CH4

2000

-~N--~ rrl,

1500 GaN without CH 4 1000

-

-|

......... |

AIN with CH 4 500

-.|

-

0 0

I

I

I

50

100

150

200

Temperature (~ Figure 21. Etch rates of GaN and AIN as a function of temperature for ECR-generated C12/ H2/CH4/Ar or CI2/HE/Ar plasmas.

2500

i

i

i

InN

?

2000

-2 1500

t~ 9 ut _CH 4

1000

500 0

I

I

I

50

100

150

Temperature

200

(~

Figure 22. Etch rates of InN as a function of temperature for ECR-generated CI2/H2/CH4/ Ar or C12/HE/Ar plasmas.

Plasma Etching of llI-VNitrides 279

GaN

I

N

O

I

.... GaI

C

(b) 30~ Etch

I

i

Ud

1

Ga

c (c) 170~ Etch

J

CI

I C

I 200

N

I Ga

I

I

400

600

I 800

I 1000

I 1200

Electron Energy (eV) Figure 23. AES surface scans ofGaN (a) before exposure to the plasma, (b) at 30~ at 170~ in an ECR-generated CI2/H2/CHa/Ar plasma.

7.0

GROWTH

and (c)

TECHNIQUE

Since several GaN samples discussed in this chapter were grown by different techniques, it is important to identify any etch dependence on growth technique. In Fig. 24, GaN ECR etch rates are shown as a function of rf-cathode-power for samples grown by MO-MBE, rf-MBE, and MOCVD. The ECR plasma conditions were: 2 mtorr pressure, 22.5 sccm C12, 2.5 sccm H2, 5 sccm Ar, 30~ electrode temperature, 1000 W microwave power, and rf-cathode-powers ranging from 1-450 W with corresponding dc-biases of-25 to -275 + 25 V. GaN samples were etched simultaneously. As the rf-cathode-power and ion energies increased, the

280

Wide Bandgap Semiconductors

GaN etch rates increased (due to more efficient bond breaking of the GaN and/or more efficient sputter desorption of the etch products), independent of growth technique. Etch rates approaching 9000 A/min were obtained for the MO-MBE and rf-MBE GaN samples at 450 W rf-cathode-power (--275 V de-bias). A trend where the MO-MBE GaN etched faster than the rf-MBE GaN (which was faster than the MOCVD GaN) was observed. Faster etch rates correlated with higher rms-roughness for the as-grown GaN samples. The rms-roughness for the as-grown GaN samples were: 19.38 + 0.44 nm for MO-MBE, 3.12 + 0.84 nm for rf-MBE, and 1.76 + 0.29 nm for MOCVD.

10,ooo -~

--

I

I

I

I'

""

I

8000

o,,.

E ~"

6000 oO

J

e-_

2

4(~o

z

~

2000

0

IO0

200

300

400

500

rf-cathode-power (W) F i g u r e 24. GaN etch rates as a function o f if-cathode-power for M O - M B E , M O C V D , and If-MBE grown GaN in a CI2/H2/Ar ECR-generated plasma.

8.0

E T C H P R O F I L E , M O R P H O L O G Y , AND STOICHIOMETRY

Three of the more critical parameters used to evaluate the value of an etch process are etch profile, surface morphology, and the chemical composition of the etched surface. Maintaining surface stoichiometry and smoothness are critical for subsequent process steps including the formation of metal contacts, deposition ofinterlevel dielectric films, passivation, or regrowth. Maintaining etch profile is essential for dimensional control and reliability of metal step coverage.

Plasma Etching of l l l - V Nitrides 281 Figure 25 shows SEM micrographs of GaN and InN samples etched in an ECR-generated C12/H2/CHa/Ar, at 850 W of ECR source power, 1 mtorr, 170~ and 150 W rf-cathode-power. The GaN etch was approximately 5800 A deep and was anisotropic with reasonably smooth sidewalls and surfaces. The high anisotropy of the etch may be attributed to the possible formation of a sidewall polymer involving the methane, as previously reported by Constantine et al. with this plasma chemistry.[ 8~ A trench was observed at the base of the GaN feature which may have occurred due to the Si3N4 mask-edge erosion. The InN etch was somewhat rough with a sloped sidewall, also due to erosion of the mask-edge. The InN was etched approximately 1.12 lam deep, which was approximately 1000 A into the GaAs substrate. The surface roughness may have been due to etching GaAs under high temperature C12 plasma conditions or to preferential etching of the InN.

GaN I LIIII

....

InN LIIIi

(a)

(b)

Figure 25. SEM micrographsof (a) GaN and (b) InN etchedat 150 W rf-cathode-powerin an ECR-generatedC12/H2/CH4/Arplasma. Highly anisotropic, smooth GaN etching has been achieved in a C12/ H2/Ar ICP-generated plasma as shown in Fig. 26. The GaN, which was grown by MOCVD, was overetched by approximately 15%. The plasma conditions were: 5 mtorr pressure, 500 W ICP source power, 22.5 seem C12, 2.5 seem H 2, 5 seem Ar, 25~ electrode temperature, and 150 W

282

Wide Bandgap Semiconductors

if-cathode-power with a corresponding dc-bias of-280 • 10 V. Under these conditions, the GaN etch rate was -~6880 A/min with highly anisotropic, smooth sidewalls. The vertical striations observed in the sidewall were due to striations in the photoresist mask which were transferred into the GaN feature during the etch. The sapphire substrate was exposed during the overetch period and showed significant pitting possibly due to defects in the substrate or growth process. With optimization of the masking process, these etch parameters may yield profiles and sidewall smoothness which improve etched facet laser performance.

Figure 26. SEM mierograph ofMOCVD GaN etched in an ICP-generated CI2/H2/Arplasma.

As discussed earlier, surface roughness can be quantified using AFM. The rms roughness and AFM images for GaN and InN etched in an ECR-generated C12/H2/CH4/Ar plasma are shown in Figs. 27 and 28, respectively, as a function of if-cathode-power. The GaN rms roughness remained relatively constant as the if-cathode-power was increased from 0-150 W; however, as the if-cathode-power was increased further to 275 W, the rms roughness increased to ~85 nm. The data suggest that at high if-cathode-power preferential sputtering or micro-masking occurred which roughened the surface. The rms roughness for InN was greatest at 65 W if-cathode-power. This implies that the ion-bombardment energy was critical to balance the chemical and sputtering effect of this plasma chemistry in order to maintain smooth surfaces and reasonable etch rates. In general, III-V nitride rms-roughness was smooth over a wide range of etch parameters, but appeared to be very sensitive to high-density plasma conditions and high ion energies.

Plasma Etching of lll-V Nitrides 283

100

I

I

I

| I I I

80 E

~ ~:~

60

I

InN/~

///

40

0

1,-,

~

2o

!

!

i

0

100

200

300

if-cathode-power (W) Figure 27. RMS roughness for GaN and InN as a function of rf-cathode-power in an ECRgenerated CI2/H2/CH4/Ar plasma.

In many cases, preferential etching of the substrate can cause significant variations in the stoichiometry of the material (which in tum can effect device performance or post-etch process steps). Plasma species may also be adsorbed or implanted into the substrate, which can also reduce device performance. In Fig. 29, Auger spectra were taken to evaluate the near-surface stoichiometry of GaN samples (a) before and after ECR etching at 850 W applied microwave power and rf-cathodepowers of (b) 65 W, and (c) 275 W. Prior to exposure of the GaN to the plasma, the Auger spectrum showed a Ga:N ratio of 1.5 with normal amounts of adventitious carbon and native oxide on the GaN surface. Following exposure to the plasma, a general tendency for the Ga:N ratio to increase with rf-cathode-power was observed. The Ga:N ratio increased from 1.8 to 2.3 as the rf-cathode-power was increased from 65-275 W at a microwave power of 850 W. Additionally, the Ga:N ratio also increased as a function of ECR microwave power. Within experimental error, these trends imply that the GaN film was being depleted of N, perhaps due to preferential etching or sputtering of the lighter N-atoms due to higher ionbombardment energy and higher ion density. Auger spectra were also taken to determine the near-surface stoichiometry of GaN as a function of plasma chemistry following exposure to ECR-generated C12/H2/Ar, C12/

284

Wide Bandgap Semiconductors

SF6/Ar , BC13/H2/Ar , and BC13/SF6/Ar plasmas. In general, the Ga:N ratio increased as the %H 2 or %SF 6 concentration increased in either BC13 or C12. These trends imply that the GaN films were depleted of N perhaps due to preferential chemical etching of the N atoms with the addition ofH 2 or SF 6 to the plasma.

GaN

InN

a)

As-grown

3.2u,,, gm

lain

~.3Ilnl

b) rf=65

9.5r,,,,

W

lain

Jam

-~ t m'rl

c)

rf = 2 7 5 W

85nm

~m

.m

39nm

Figure 28. AFM micrographs for (a) GaN and InN as-grown, (b) GaN and InN etched at a if-cathode-power of 65W, and (c) GaN and InN etched at a if-cathode-power of 275 W in an ECR-generated C12/H2/CH4/Ar plasma. The Z-scale is 100 nm/division.

9.0

PLASMA INDUCED DAMAGE

As stated in earlier sections, the fabrication of high-density integrated circuits and optoeleetronie devices often requires pattern transfer with

Plasma Etching of lll- V Nitrides 285 highly anisotropic profiles and smooth surface and sidewall morphologies. These requirements are often achieved using plasma etch techniques where energetic ions are accelerated from the plasma to the wafer. However, when these energetic ions strike the sample, surface damage as deep as 1000 A can occur (100), causing degradation of electrical and optical properties of the device. Plasma-induced-damage can include defects or dislocations in the lattice, formation of dangling bonds on the surface, implanted etch ions, or deposition of material on the sample. Attempts to minimize the damage by reducing the ion energy below the damage threshold for compound semiconductors (< 40 eV) [49] or by increasing the chemical component of the etch often results in more isotropic profiles, significantly limits minimum dimensions, and reduces the etch rate.

As-grown

100 :

(a)

--

.

,

~.0

100

Si

65 W if-cathode-power

~

.

-

200 I

(b)

O v

C,:N-1.8__,!.]j

.

.

.

.

.

0

Si

O

IGa m

w

I-

-

200 275 W 200

-

if-cathode-power -

I-- . . . . .

-

-

I

__'

o

(c)

Ga:N =2.3

I|

o .,',4

t

Si

il

O II 500

I

"'' ,

I

Ga

I--I

1000

Kinetic Energy (eV) Figure 29. AES surface scans ofGaN (a) before exposure to the plasma, (b) at 65 W, and (c) 275 W rf-cathode-power, 1 mtorr, 170~ and 850 W microwave power in an ECRgenerated CI2/H2/CH4/Ar plasma.

286

Wide Bandgap Semiconductors

Since GaN is more chemically inert than GaAs and has higher bonding energies, higher ion energies may be used during the etch process with potentially less damage to the material. However, reports of plasmainduced-damage of the III-V nitrides have been limited. Pearton and coworkers have reported plasma-induced-damage results for InN, InGaN, and InA1N in an ECR-generated plasma (the damage increased as a function of ion flux and ion energy).[l~ In the following section, ICP and ECR plasma-induced-damage of GaN is evaluated using photoluminescence (PL) measurements as a function of rf-cathode-power and source power. Pure Ar plasmas were used to simulate the ion bombardment conditions created during plasma etching of the III-V nitrides. It is important to realize that the use of a pure Ar plasma creates a worse case scenario for plasma-induced-damage due to the lack of chemical interactions. With the introduction of reactive gases to the plasma for a given plasma power and density, the damage will be reduced when compared to a sputter mechanism, since damaged material is typically being removed at a higher rate, leaving a shallower damage depth. Prior to the plasma exposure, a 2 inch GaN wafer (grown by MOCVD) was mapped out to examine the uniformity of the PL emission. In Fig. 30, the PL spectrum at the center of the wafer taken with the low resolution grating is shown. The spectrum consisted of two distinct features. The dominant near band-edge resonance was seen at 3.472 eV. Emission resonances in this spectral region have been identified with recombination of a neutral-donor-bound exciton.[l~ 1~ The free exciton resonance, expected at an energy of~3.485 eV (102), was not clearly resolved. The broad spectral feature centered at approximately 2.21 eV was associated with emission from deep level impurities. In Fig. 31, the peak intensity of the near band-edge emission is plotted as a function of radial position on the two inch GaN wafer. The intensity dropped by --15% at a radial position of 0.5 inches and ---30% at a radial position of 0.8 inches. Toward the edge of the wafer, the intensity drop was significantly more rapid. The samples that were used in the etch studies were taken from the center of the wafer. The PL spectra were taken before and after etching for each sample. In Fig. 32, the % change in the peak PL intensity versus rf-cathodepower is plotted for both ECR and ICP etching. GaN samples were exposed to ICP- and ECR-generated Ar plasmas for 1 minute under identical plasma conditions while the rf-cathode-power was increased. The de-bias was --10 to 65% higher in the ECR under comparable conditions. For the ICP case, at relatively low rf-cathode-powers (1 and 50 W), the PL intensity slightly degraded, and as the rf-cathode-power was increased up to 250 W, increasing degradation in PL intensity was seen. Depth profiling

Plasma Etching of l l l - V Nitrides 28 7 of similar films at a rf-cathode-power of 1 W (-~ - 10 V de-bias) revealed no detectable removal of GaN; whereas, at 250 W etch (-300 V de-bias)-~770 A of GaN was lost during a 1 minute exposure. Distinctly different results were obtained for etching in the ECR plasma system. Etching with very low rf-cathode-power (1 W) resulted in > 80% increase in the PL intensity and virtually no sputter loss of GaN. Etching at higher rf-cathode-powers also improved the PL intensity, but to a lesser degree as the rf-cathodepower was increased. The highest rf-cathode-power (150 W) etch resulted in a very slight decrease in PL intensity and a GaN sputter rate of--820 A/min.

9

!

,

0

,

,

!

,

,

9

,

,

..,d

2.00

2.50

3.00

3.50

Energy (eV) Figure 30. PL spectrum from the GaN film at T=10 K. The effect of plasma density on the peak near band-edge PL intensity was also studied. GaN samples were exposed to ICP- and ECRgenerated Ar plasmas for 1 minute while the de-bias was held at -~ -65 V. The data were more scattered than the rf-cathode-power data for both ICP and ECR conditions. The ICP showed virtually no change in PL intensity at 250 W source power and then decreased by 30% as the ICP source power was increased to 750 W. The PL intensity decreased by only 10% at 1000 W ICP source power, which was an improvement of almost 20% over 750 W. In the ECR, an increase of--115% in PL intensity at 250 W ECR source power was observed. Similar to the trend observed as a function of rf-cathode-power, the PL intensity also improved at higher ECR source powers but at a lower rate. Sputter rates for GaN were <10 A/min at 250 W source power and--225 A/min at 1000 W.

288

Wide Bandgap Semiconductors

2000 "

i . ~

I

~

'

.

I

i

,

,

!

.

,

..

l

,

|

,

!

9

,

.

.

,

!

.

,

,

I

t

|

r

.,,,q

1500

"~ 1000

..~

500

e~

0

9

,

i

0

0.2

I

0.4

,

0.6

0.8

Radial Position on Wafer (inches) Figure 31. Variation of peak PL intensity as a function of radial position on the two inch wafer.

100

50

=

.,..~

0

ECR

cD

(-; -50 ICP - l O 0

I

0

.

.

.

.

I

50

i

.

,

,

I

,

100

,

t

,

I

.

150

.

.

.

I

.

200

.

.

.

I

250

.

,

.

i

~00

rf-cathode-power (W) Figure 32. Percent change in the peak PL intensity as a function of if-cathode-power for both ECR and ICP Ar plasmas.

The PL emission e f f i c i e n c y o f other III-V bulk semiconductors (e.g., GaAs) has been shown to be strongly dependent on surface c o n d i t i o n s

Plasma Etching of l l l - V Nitrides 289 (specifically the presence of a native oxide), the surface recombination velocity, as well as band bending at the surface.[ 1~176 While the exact nature of surface oxides and surface states in GaN are not well understood, the data show that the PL efficiency can be strongly affected by exposure to an Ar plasma and are highly sensitive to plasma conditions. A common result for both plasma environments was the decrease in PL intensity with increasing rf-cathode-power (suggesting that plasma-induced-damage can occur in GaN films under moderate ion energies). The data suggest that GaN surfaces are considerably more sensitive to process-induced changes than is widely recognized. Brief low bias exposures to ECR discharges almost doubled the band-edge PL, suggesting that native oxide has a strong effect on the surface recombination velocity.

10.0 P L A S M A E T C H A P P L I C A T I O N S The principal criteria for any plasma etch process is its utility in the fabrication of a device. The two most obvious applications of plasma etching oflII-V nitride devices have been the fabrication of blue LEDs and laser diodes, as shown by Nakamura.[ 1][2][5]-[7] In Fig. 33, the laser diode structure used by Nakamura is shown schematically.[ 7] A C12 RIE plasma was used to etch down into the n-GaN layer. The roughness of the facet surface was -500/1, and required a highly reflective coating to reduce the threshold current. The laser showed a relatively high threshold voltage and some light scattering, which may be attributed to the roughness of the etch. As plasma etch and masking technologies are refined, improved laser performance is expected. Alternative optical devices may also have applications in future photonic or optoelectronic circuits, including the whispering-gallery-mode microdisk laser.[1~ 1~ In Fig. 34, a schematic diagram of the fabrication process for a microdisk GaN-based laser is shown. A GaN/InGaN quantum well was grown on an A1N buffer by MO-MBE. Circular posts were patterned using a Si3N4 mask. Plasma etching was performed in an ECR C12/H2/CH4/Ar plasma at 1 mtorr, 170~ 850 W ECR source power, and 150 W rf-cathode-power with a corresponding de-bias of ~-175V. The etch was anisotropic and smooth with rates of-2200 A/min for GaN and -1000 A/min for A1N. The A1N was then selectively etched in a KOHbased etchant for--30 min at 85~ There was no measurable etching of the GaN or InGaN. A SEM micrograph showing the dry and wet etch steps is displayed in Fig. 35. Approximately 500 A of the sapphire substrate was etched in the ECR plasma.

290

Wide Bandgap Semiconductors

p-GaN p-AIGaN p-GaN

p-AIGaN

I

InGaN MQW

~-~--------n-GaN n-AIGaN n-InGaN

(a)

n-GaN

sapphire substrate

Ni/Au ,

i

(b) Ti/AI ~'

.

v

I

sapphire substrate

Figure 33. Schematic diagram ofa III-V nitride laser diode (Nakamura, Ref. 5) showing the (a) growth structure and (b) the device after RIE etch and metallization.

Plasma etching has also been prominent in the fabrication of III-V nitride based electronic devices. This is especially critical for devices where active channels are close to the etch region, thus requiring well controlled, low-damage conditions. In Fig. 36, the process sequence for a GaNjunction field-effect transistor (JFET) [22]is shown schematically. The process required the following steps: (a) selective area implant of the n-channel and p-gate (b) refractory metal gate deposition, gate definition, and selective area implant of the source and drain (c) a timed plasma etch to remove the p-GaN in the source and drain regions (d) formation of ohmic source and drain contacts

Plasma Etching of lll-V Nitrides 291

InGaN

(a) Mask

~////////'//>/;~/>,/'///~/;~/',/'/,//J ~

sapphire

GaN Bu ffer

]

Mask InGaN ~

b.\\\\\\\\\\\

Co)

- GaN AIN

///.////.//,/////////////~

~

G aN Buffer

ECR Etch sapphire

InGaN ~

K\\\\\\\\\\"~I "~,.,.. GaN

~,,~,,,~.~,~,4,,...

(c) GaN Buffer

Wet Etch

sapphire

Figure 34. Schematic diagram of the process steps for the fabrication o f a GaN microdisk laser showing (a) the epitaxial growth structure, (b) ECR plasma etch, and (c) selective wet etch of the AIN.

The following ECR conditions were incorporated to remove-500 to 1000 A of the p-GaN: 1.0 mtorr pressure, 850 W ECR source power, 150 W if-cathode-power with a corresponding de-bias of~- - 185 V, 15 seem BC13, 20 seem H 2, and 5 seem Ar. The etch process must be well controlled and repeatable with smooth etch morphology and minimal damage for subsequent metallization of the source and drain ohmic contacts.

292

Wide Bandgap Semiconductors

a)

/

GaN/InGaN A1N

GaN/InGaN b)

A1N

Figure 35. SEM micrograph of the microdisk laser showing (a) the non-selective ECR plasma etch and (b) the selective wet etch of A1N.

11.0 C O N C L U S I O N S In summary, the utilization of high-density ECR and ICP chlorinebased plasmas has resulted in high rate (> 1 ~m/min), smooth, anisotropie etching of the III-V nitrides. The source of reactive C1 (C12, BC13, IC1, etc.) and the use of additive gases (Ar, H2, N2, and SF6)have several effects on the etch characteristics of III-V nitride films. Using Cl2-based plasmas typically results in high concentrations of reactive C1 which increases the chemical component of the etch mechanism and the GaN etch rate. The addition of H 2, N 2, or SF 6 to C12- or BC13-based plasmas appears to effect the chemical removal of the N atoms from the GaN as well as the concentration of reactive C1 in the plasma (which directly correlates to etch rate). Very smooth pattern transfer has been obtained for a wide range of plasma chemistries and conditions. The bond breaking mechanism for the

Plasma Etching of l l l - V Nitrides 293 III-V nitride bonds appears to be critical and perhaps the rate limiting step in the etch mechanism. The use of high-density plasmas results in improved etch results possibly due to a two step process directly related to the plasma flux. Initially the high-density plasmas increase the bond breaking mechanism allowing the etch products to form and then produce efficient sputter desorption of the etch products. ICP etching of the III-V nitrides in C12/H2 plasmas yields etch profiles and sidewall morphologies which may improve the yield and performance of etched facet lasers. Although work has been significant in this area over the past few years, identifying plasma conditions which maintain the stoichiometry of the as-grown films and minimize plasma-induced-damage are critical to the fabrication of high performance III-V nitride devices and must be further developed and better understood.

GateImplant

photoresist (a)

"~ISI_GaN ~ ~ ~ ~ photoresist

Ohmic Implant

Co)

Gatecontact ~

Gatecontact "Dryetch of p-GaN"

(c)

Ti/AIohmiccontacts

~txxXxXxl

LX%XXXX~I

(d)

I'

'1

Figure 36. Schematic diagram of the process steps for the fabrication of a GaN JFET (Zolper, Ref. 20) showing (a) selective area implant of the n-channel and p-gate, (b) refractory metal gate deposition, gate definition, and selective area implant of the source and drain, (c) a timed plasma etch to remove the p-GaN in the source and drain regions, and (d) formation of ohmic source and drain contacts.

294

Wide Bandgap Semiconductors

ACKNOWLEDGMENTS The author would like to acknowledge S. J. Pearton, C. R. Abemathy, J. C. Zolper, M. Hagerott Crawford, R. D. Briggs, C. B. Vartuli, R. F. Karlicek, Jr., C. Tran, M. Schurman, C. Constantine, C. Barratt, K. P. Killeen, J. Han, R. J. Hickman, and P. A. Barnes for their collaboration on this work. The author would like to also thank P. L. Glarborg, A. T. Ongstad, and L. Griego for their technical support. This work was supported by the United States Department of Energy under contract DEAC04-94AL85000. Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy.

REFERENCES 1. Nakamura, S., Mukai, T., and Senoh, M., Jpn. J. AppL Phys., 30: L 1998 (1991) 2. Nakamura, S., Mukai, T., Senoh, M., and Iwasa, N., Jpn. J. Appl. Phys., 31:L139(1992) 3. Foresi, J. S., and Moustakas, T. D.,Appl. Phys. Lett., 62:2859 (1993) 4. Binari, S. C., Rowland, L. B., Kruppa, W., Kelner, G., Doverspike, K., and Gaskill, D. K., Electron. Lett., 30:1248 (1994) 5. Nakamura, S., Mukai, T., and Senoh, M., Appl. Phys. Lett., 64:1687 (1994) 6. Nakamura, S., Senoh, M., Iwasa, N., and Nagahama, S., Jpn. J. Appl. Phys., 34:L797(1995) 7. Nakamura, S., Senoh, M., Nagahama, S., Iwasa, N., Yamada, T., Matsushito, T., Kiyoku, H., and Sugimoto, U., Jpn.. J. Appl. Phys., 35 :L74 (1996) 8. Matsuoka, T., Sasaki, T., and Katsui, A., Optoelectronic Devic'es and Technologies, 5:53 (1990) 9. Amano, H., Kito, M., Hiramatsu, K., and Akasuki, I., Jpn. J. Appl. Phys., 28:L2112 (1989) 10. Strite, S., andMorkoc, H.,J. Vac. Sci. Technol. B, 10:1237 (1992) 11. Kahn, M. A., Kuznia, J. N., Van Hove, J. M., Olson, D. T., Krishnankutty, S., and Kolbas, R. M., Appl. Phys. Lett., 58:526 (1991) 12. Tansley, T. L., and Egan, R. J., Phys. Rev. B, 45" 1094 (1993) 13. Khan, M. A., Bhattarai, A. R., Kuznia, J. N., and Olson, D. T., Appl. Phys. Lett., 63:1214(1993) 14. Pearton, S. J., Abemathy, C. R., Wisk, P., Hobson, W. S., and Ren, F., Appl. Phys. Leg., 63:1143 (1993)

P l a s m a E t c h i n g o f l l l - V Nitrides 15. 16. 17. 18. 19. 20. 21. 22. 3.

24. 25.

26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40.

295

Davis, R. F., Proc. IEEE 79:702 (1991) Nakamura, S., Senoh, M., and Mukai, T., Jpn. J. Appl. Phys., 30:L 1708 (1991) Akasaki, I., Amano, H., Kito, M., and Kiramatsu, K., Lumin., 48/49:666 (1991) Nakamura, S., Senoh, M., and Mukai, T.,Appl. Phys. Lett., 62:2390 (1993) Khan, M. A., Chen, Q., Shur, M. S., Dermott, B. T., Higgins, J. A., Burm, J., Schaff, W., and Eastman, L. F., Electron. Lett., 32:257 (1996) Khan, M. A., J. N. Kuznia, Bhattarai, A. R., and Olson, D. T., Appl. Phys. Leg., 62:1248 (1993) Khan, M. A., Shur, M. S., and Chen, Q., Electron. Lett., 31:2130 (1995) Zolper, J. C., Shul, R. J, Baca, A. G., Wilson, R. G., Pearton, S. J., and Stall, R. A.,Appl. Phys. Leg., 68:2273 (1996) Manos, D. M., and Flamm, D. L., Plasma Etching, Academic Press, San Diego (1989) Chapman, B., Glow Discharge Processes, Academic Press, New York (1980) Dautremont-Smith, W. C., Gottscho, R. A., and Schultz, R. J., Semiconductor Materials and Processing Technology Handbook, (G. E. McGuire, ed.), pp. 191, Noyes, Park Ridge, NJ (1988) Chu, T. L., J. Electrochem. Soc., 119:1200 (1971) Pankove, J. I.,J. Electrochem, Soc., 119:1118 (1972) Guo, Q. X., Kato, O., and Yoshida, A.,J. Electrochem. Soc., 139:2008 (1992) Pearton, S. J., Abemathy, C. 1L, Ren, F., Lothian, J. R., Wisk, P. W., and Katz, A., J. Vac. Sci. Technol. A, 11:1772 (1993) Sheng, T. Y., Yu, Z. Q., and Collins, G. J.,Appl. Phys. Lett., 52:576 (1988) Pauleau, T.,J. Electrochem. Soc., 129:1045 (1982) Taylor, K. M., and Lenie, C.,J. Electrochem. Soc., 107:308 (1960) Long, G., and Fuster, L. M.,J. Am. Ceram. Soc., 42:53 (1959) Barrett, N. J., Grange, J. D., Sealy, B. J., and Stephens, K. G., J. Appl. Phys., 57:5470(1985) Aita, C. R., and Gawlak, C. J., J. Vac. Sci. Technol. A, 1:403 (1983) Kline, G. R., and Lakin, K. M.,Appl. Phys. Lett., 43:750(1983) Mileham, J. R., Pearton, S. J., Abernathy, C. 1L, MacKenzie, J. D., Shul, R. J., and Kilcoyne, S.P.,J. Vac. Sci. Technol. A, 14:836(1996) Mileharn, J. IL, Pearton, S. J., Abemathy, C. R., MacKenzie, J. D., Shul, IL J., and Kilcoyne, S. P., Appl. Phys. Lea., 67:1119 (1995). Minsky, M. S., White, M., and Hu, E. L., Appl. Phys. Lett., 68" 1531 (1996) Youstey, C., Adesida, I., and Bulman, G., Electron. Letts., 33:245 (1997)

296

Wide Bandgap Semiconductors

41. Pearton, S. J., Abemathy, C. R., Ren, F., and Lothian, J. R., J. Appl. Phys.,

76:1210,(1994) 2, Adesida, I., Mahajan, A., Andideh, E., Asif Khan, M., Olsen, D. T., and 3, 4. 5.

46. 47. 48.

49. 50. 51.

52. 53. 54. 55. 56. 57. 58. 59. 0.

61.

Kuznia, J. N., Appl. Phys. Lett., 63:2777 (1993) Lin, M. E., Zan, Z. F., Ma., Z., Allen, L. H., and Morko~, H., Appl. Phys. Lett., 64:887 (1994) Ping, A. T., Adesida, I., Asif Khan, M., and Kuznia, J. N., Electron. Lett., 30:1895 (1994) Lee, H., Oberman, D. B., and Harris, J. S., Jr., Appl. Phys, Lett., 67:1754 (1995) Asmussen, J.,J. Vac. Sci. Technol. B, 5:328 (1987) Popov, O. A., High-density Plasma Sources, Noyes Publications, Park Ridge, NJ (1996) Lieberonan, M. A. ,and Gottscho, R. A., "Plasma Sources for Thin Film Deposition and Etching," (M. H. Francombe, and J. L. Vossen, eds.), Physics of Thin Films, Vol. 18, Academic Press, San Diego (1994) Constantine, C., Johnson, D., Pearton, S. J., Chakrabarti, U. K., Emerson, A. B., Hobson, W. S., and Kinsella, A. P.,J. Vac. Sci. Technol. B, 8:596 (1990) Pearton, S. J., Chakrabarti, U. K., Kinsella, A. P., Johnson, D., and Constantine, C., Appl. Phys. Lett. 56:1424 (1990) Murrell, A. J., Grimwood, R. C., O'Sullivan, P., Gilbert, M., Vanner, K., Ruddell, F., Davies, I., Hilton, K., Bland, S., and Spear, D., Technical Digest, Proc. 1992 GaAs IC Symposium, pp. 173 Cheung, R., Lee, Y. H., Lee, K. Y., Smith, T. P., III, Kern, D. P., Beaumont, S. P., and Wilkinson, C.D.W.,J. Vac. Sci. Technol. B, 7:1462 (1989) Ko, K. K., and Pang, S. W.,J. Electrochem. Soc., 141:250(1994) Pearton, S. J., Abemathy, C. R., Ren, F., Lothian, J. R., Wisk, P. W., Katz, A., and Constantine, C., Semicond. Sci. Technol., 8:310 (1993) Pearton, S. J., Abemathy, C. R., and Ren, F., Appl. Phys. Lett., 64:2294 (1994) Zhang, L., Ramer, J., Zheng, K., Lester, L. F., and Hersee, S. D., Mat. Res. Soc. Proc., 395:763 (1996) Humphreys, B., and Govett, M., MIJNSR 1 (1996) Vartuli, C. B., Pearton, S. J., Lee, J. W., Hong, J., MacKenzie, J. D., Abemathy, C. R., and Shul, R. J.,Appl. Phys. Lett., 69:1426 (1996) Vartuli, C. B., MacKenzie, J. D., Lee, J. W., Abemathy, C. R., Pearton, S. J., and Shul, R. J., J. Appl. Phys., 80:3705 (1996) Pearton, S. J., Abemathy, C. R., and Ren, F., Appl. Phys. Lett. 64:3643 (1994) Shul, R. J., Kilcoyne, S. P., Hagerott Crawford, M., Parmeter, J. E., Vartuli, C. B., Abemathy, C. R., and Pearton, S. J.,Appl. Phys. Lett., 66:1761 (1995)

Plasma Etching of llI-VNitrides

297

62. Shul, R. J., Howard, A. J., Pearton, S. J., Abemathy, C. R., Vartuli, C. B., Barnes, P. A., and Bozack, M. J., J. Vac. Sci. Technol. B, 13:2016 (1995) 63. Vartuli, C. B., Pearton, S. J., Abemathy, C. R., Shul, R. J., Howard, A. J., Kilcoyne, S. P., Parmeter, J. E., and Hagerott Crawford, M., J. Vac. Sci. Technol. A, 14:1011 (1996) 64. Shul, R. J., Ashby, C. I. H., Rieger, D. J., Howard, A. J., Pearton, S. J., Abemathy, C. R., Vartuli, C. B., and Barnes, P. A., Mat. Res. Soc. Symp. Proc., 395:751 (1996) 65. Shul,R. J., McClellan, G. B., Pearton, S. J., Abemathy, C. R., Constantine, C., and C. Barratt, Electron. Lett., 32:1408 (1996). 66. Shul, R. J., McClellan, G. B., Casalnuovo, S. A., Rieger, D. J., Pearton, S. J., Constantine, C., Barratt, C., Karlicek, R. F., Jr., Tran, C., and Schurman, M., Appl. Phys. Lett., 69:1119 (1996) 67. McLane, G. F., Meyyappan, M., Cole, M. W., and Wrenn, C., J. Appl. Phys., 69:695 (1991) 68. Meyyappan, M., McLane, G. F., Lee, H. S., Eckart, E., Namaroff, M., and Sasserath, J.,J. Vac. Sci. Technol. B, 10:1215 (1992) 69. McLane, G. F., Casas, L., Pearton, S. J., and Abemathy, C. R., Appl. Phys. Lett., 66:3328 (1995) 70. Adesida, I., Ping, A. T., Youtsey, C., Sow, T., Asif Khan, M., Olson, D. T., and Kuznia, J. N., Appl. Phys. Lett., 65:889 (1994) 71. Ping, A. T., Adesida, I., and AsifKhan, M.,Appl. Phys. Lett., 67:1250 (1995) 72. Gillis, H. P., Choutov, D. A., Martin, K. P., and Song, L., Appl. Phys. Lett., 68:2255(1996) 73. Gillis, H. P., Choutov, D. A., Steiner, P. A., Piper, J. D., Crouch, J. H., Dove, P. M., and Martin, K. P., Appl. Phys. Lett., 66:2475 (1995) 74. Gillis, H. P., Choutov, D. A., and Martin, K. P., JOM, 50 (1996) 75. Gillis, H. P., Choutov, D. A., Martin, K. P., Pearton, S. J., and Abemathy, C. R., d. Electrochem. Soc., in oress. 76. Leonard, R. T., and Bedair, S. M., AppL Phys. Lett., 68:794 (1996) 77. Chuang, T. J., Laser Microfabrication: Thin Film Processes and Lithography, (D. J. Ehrlich and J. Y. Tsa, eds.), p. 87, Academic, San Diego (1989) 78. Ashby, C. I. H., Properties ofGaAs, 2nd edition INSPEC, London, Ch. 20, 653-682 (1990) 79. Hayes, T. R., Indium Phosphide and Related Materials: Processing, Technology, and Devices, (A. Katz, ed.), Artech House, Boston, Ch. 8, 277-306(1992) 80. Constantine, C., Barratt, C., Pearton, S. J., Pen, F., and Lothian, J. R., Appl. Phys. Lett. 61:2899 (1992)

298

Wide Bandgap Semiconductors

81. Constantine, C., Barratt, C., Pearton, S. J., Ren, F., and Lothian, J. R., Electron. Lett. 28:1749 (1992) 2, Lishan, D. G., and Hu, E. L., Appl. Phys. Lett., 56, 1667 (1990) 83. Donnelly, V. M., Flamm, D. L., Tu, C. W., and Ibbotson, E., J. Electrochem. Soc., 129:2533 (1982) 4,, Thomas, S., III, Ko, K. K., Pang, S.W.,J. Vac. Sci. Technol. A, 13:894 (1995) 85. Lee, J. W., Hong, J., and Pearton, S. J.,Appl. Phys. Lett., 68:847 (1996) 86. Niggebrugge, U., Klug, M., and Garus, G., Inst. Phys. Conf. Ser., 79:367 (1985) 87. Keskinen, J., Nappi, J., Asonen, H., and Pessa, M., Electron. Lett., 26:1371 (1990) 88. Hayes, T. R., Dreisbach, M. A., Thomas, P. M., Dautremont-Smith, W. C., and Heimbrook, L. A., dr. Vac. Sci. Technol. B, 7:1130 (1989) 89. Cheung, R., Thoms, S., Beaumont, S. P., Doughty, G., V. Law, and Wilkinson, C. D. W.,Electron. Lett., 23:857 (1987) 0. Pearton, S. J., and Shul, R. J., III-Nitrides, Academic Press, in press. 91. Ren, F., Lothian, J. R., Kuo, J. M., Hobson, W. S., Lopata, J., Caballero, J. A., Pearton, S. J., and Cole, M. W., Jr. Vac. Sci. Technol. B, 14:1 (1995) 2. Ren, F., Hobson, W. S., Lothian, J. R., Lopata, J., Caballero, J. A., Pearton, S. J., and Cole, M. W., Appl. Phys. Lett., 67:2497 (1995) 3. Shul, R. J., McClellan, G. B., Briggs, R. D., Rieger, D. J., Pearton, S. J., Abemathy, C. R., Lee, J. W., Constantine, C., and Barratt, C., J. Vac. Sci. and Technol. A, in press (1996) 94. Shul, R. J., Howard, A. J., Vartuli, C. B., Barnes, P. A., and Seng, W., J. Vac. Sci. Technol. A, 14:1102 (1996) 5. Shul, R. J., Baca, A. G., Rieger, D. J., Hou, H., Pearton, S. J., and Ren, F., Mat. Res. Soc. Syrup. Proc., 421:245 (1996) 6. Vartuli, C. B., Pearton, S. J., Lee, J. W., MacKenzie, J. D., Abemathy, C. R., and Shul, R. J., J. Vac. Sci. and Technol B, 15:98 (1997) 7. Vartuli, C. B., Pearton, S. J., MacKenzie, J. D., Abemathy, C. R., and Shul, R. J.,J. Electrochem. Soc., 143:L246 (1996) 98. Contolini, R. J.,J. Electrochem. Soc., 135:929 (1988) 99. Pearton, S. J., and Hobson, W. S.,J. Appl. Phys., 66:5018 (1989) 100. Pang, S. W.,J. Electrochem. Soc., 133:784 (1986) 101. Pearton, S. J., Lee, J. W., MacKenzie, J. D., Abernathy, C. R., and Shul, R. J., Appl. Phys. Lett., 67, 2329 (1995) 102. Shan, W., Schmidt, T. J., Yang, X. H., Hwang, J., Song, J. J., and Goldenberg, B., Appl. Phys. Lett., 66:985 (1995)

P l a s m a E t c h i n g o f l l I - V Nitrides

299

103. Chen, G. D., Smith, M., Lin, J. Y., Jiang, H. X., AsifKhan, M. and Sun, C. J., Appl. Phys. Lett., 67:1653 (1995) 104. Gottscho, R. A., Preppemau, B. L., Pearton, S. J., Emerson, A. B., and Giapis, K. P.,J. Appl. Phys., 68:440 (1990) 105. Aydil, E. S., and Gottscho, R. A., Mat. Sci. For., 148/149:159 (1994) 106. McCall, S. L., Levi, A. F. J., Slusher, R. E., Pearton, S. J., and Logan, R. A., Appl. Phys. Lett., 60:289 (1992) 107. Abemathy, C. R., Pearton, S. J., MacKenzie, J. D., Mileham, J. R., Bharatan, S. R., Krishnamoorthy, V., Jones, K. S., Hagerott Crawford, M., Shul, R. J., Kilcoyne, S. P., Zavada, J. M., Zhang, D., and Kolbas, R. M., Solid-State Electron., 39:311 (1996) 108. Lee, J. W., Vartuli, C. B., Abemathy, C. R., MacKenzie, J. D., Mileham, J. R., Pearton, S. J., Shul, R. J., Zolper, J. C., Hagerott Crawford, M., Zavada, J. M., Wilson, R. G., and Schwartz, R.N.,J. Vac. Sci. Technol. B, 10:3637 (1996)

7 Ion Implantation Wide Bandgap Semiconductors

in

John C. Zolper

1.0

INTRODUCTION

,\

The wide bandgap semiconductors, group-Ill Nitrides (INN, GaN, and A1N), SiC, and diamond have long been recognized as being ideal materials for short wavelength light emitters or detectors and for highpower, high-temperature electronics.[1]-[ 3] The group III nitride material system has generated considerable excitement in the semiconductor research community with the success in fabricating high performance light emitting diodes (LEDs), lasers, and transistors.J4] -[12] The interest in SiC has mostly been fueled by its potential in the electronics arena duc to its high breakdown field and high thermal conductivity making it useful for high-power or high-temperature operation.[13]-[ 16] Diamond also is expected to bc most applicable to electronic applications (transistors or detectors) due to its predicted excellent transport and thermal properties which arc superior to the group-HI Nitridcs and SIC.[1Ill7] Details of the attractive properties of these materials arc described in other chapters of this book. The wide bandgap II-VI semiconductors (e.g., ZnSe, ZnS, etc.) are not included in this chapter since their low melting point (-300~ makes them unattractive for electronic device applications where implantation is expected to have the most impact. Their low melting point will also make 300

lonlmplantation 301 it difficult to perform the required thermal annealing to activate implanted dopants in the II-VI materials. For this reason, little work has been reported on implantation in the wide bandgap II-VI semiconductors.[ 18] In this chapter, the status of ion implantation doping and isolation in group-III Nitride, SiC, and diamond semiconductors is presented. Ion implantation is a process whereby doping or compensating impurities are injected into a semiconductor by a high energy accelerator.[ 19]Implantation has come to be the dominant doping technology in silicon and gallium arsenide microelectronics, although there was significant technological development required to achieve this success. This success was due to the ability to precisely control the doping concentration and profile as well as the ability to minimize the processing thermal budget. Therefore, with improved wide bandgap semiconductor starting materials and improved understanding of the ion implantation process, this technology can be expected to play a critical enabling role in the maturation of these promising materials into sophisticated, manufacturable devices. In the following sections, the material and process technologies related to ion implantation in wide bandgap semiconductors are presented. First, the use of implantation to produce high resistivity, isolating regions is discussed. Second, ion implantation doping to achieve n- and p-type material is described. Third, results are given for the redistribution or diffusion of implanted impurities that occurs during the high temperature implantation activation annealing process. Fourth, a review is given of implantation-induced crystal defects and the removal of implant damage by thermal annealing. Fifth, examples are presented of devices in each material system that have employed implantation isolation or doping. Finally, areas for future work are outlined and conclusions are drawn. In each of the sections, a discussion is given for the three materials considered in the chapter: group-III nitrides, SiC, and diamond.

2.0

IMPLANTATION ISOLATION

Implant isolation has been widely used in compound semiconductor devices for inter-device isolation (as in transistor circuits) or to produce current channeling in lasers.[2~ 22] The implantation process can compensate the semiconductor layer either by a damage or a chemical mechanism. For damage compensation, the resistance typically goes through a maximum

302

Wide Bandgap Semiconductors

with increasing post-implantation annealing temperature as the damage is annealed out and hopping conduction is reduced. At higher temperatures, the defect density is further reduced below that required to compensate the material, and the resistivity decreases. For chemical compensation, the post-implantation resistance again increases with annealing temperature, showing a reduction in hopping conduction; but it then stabilizes at higher temperatures as a thermally stable compensating deep level is formed. Typically, there is a minimum dose (dependent on the doping level of the sample) required for the chemically active isolation species to achieve thermally stable compensation.[ 23] Thermally stable implant isolation has been reported for O-implanted n- and p-type A1GaAs, where an A1-O complex is thought to form. A C-N complex is postulated for N-implanted C-doped GaAs and A1GaAs.[23]-[25] With this background, implant isolation properties of wide bandgap semiconductors are reviewed.

2.1

Group IIl Nitrides

GaN. The first report of the use of implantation to reduce the free carrier concentration in GaN was by Khan, et al. Their work involved Be and N implants in GaN and A1GaN to achieve compensation of native donor defects to enhance Schottky barrier formation.j26][ 27] It was shown that shallow implants into samples with a high concentration of native donors reduces the donor concentration and allows the formation of improved Schottky contacts. Since the samples were annealed at 1100~ after implantation, the compensation was most likely due to activated acceptors in the case of Be and perhaps due to a reduction in the concentration of N-vacancies that acted as donors (in the case of N-implantation). A study on the thermal characteristics of compensation via Nimplantation in n- and p-type GaN was first reported by Pearton.[ 28] As shown in Fig. 1, N-implantation effectively compensates both p- and n-type GaN when implanted to achieve an implanted N density of-4 x 10 TMcm -3. For both doping types, the resistance first increased with annealing temperature, then reached a maximum before demonstrating a significant reduction in resistance (after a 850~ anneal for n-type and a 950~ anneal for p-type GaN). This behavior is typical of implant-damage compensation. The defect levels estimated from Arrhenius plots of the resistance/temperature product are 0.83 eV for initially n-type and 0.90 eV for initially p-type GaN.[ 28] These levels are not at midgap, but are sufficiently deep to realize a sheet resistance > 109 ~/EI. The implantation has also been reported to

Ion Implantation 303 effectively isolate n-type GaN, with the material remaining compensated to over 850~ 29] Interestingly, H-implant compensation of n-type GaN was reported to anneal out a t - 4 0 0 ~ with an anomalous dependence on implant energy.[ 291The reason for this is presently not known. In light of this result, however, H-implantation in GaN will require further study since H is often the ion of choice for photonie device isolation applications that require deep isolation schemes. Moreover, both He- and N-implant isolation appear to rely solely on implantation damage without any chemical compensation effects analogous to those in the O/A1GaAs case.J2~ 24] However, the implantation-induced defects in GaN are more thermally stable than other III-V semiconductor materials, such as GaAs or InP, where damage levels begin to anneal out below 700~ 2~ This may be a result of the higher bandgap of GaN or the more polar nature of the lattice causing more stable defects. Furthermore, recent results for very low dose (10 l~ to 1011 cm -2) N-implantation in lightly n-type GaN (as-grown ---5 x 1016 em-3) suggest that, for low doses, N-implantation may actually increase the free electron concentration by increasing the number of N-vacancies or interstitials.[ 3~ Clearly, further work is required to fully understand the nature of implantation damage in GaN.

1010 - I F p-type 109

n-type

10 8 107 10 6 105 , ~ , , , ~ 300 400 500 600 700 800 900 1000 annealing temperature (~ Figure 1. Sheet resistance versus annealing temperature for N-implanted initially n- and ptype GaN. The N was implanted at multiple energies to give an approximatelyuniform ion

concentration of 4 x 10Is cm-3 across ~500 nm (after Ref. 27).

304

Wide Bandgap Semiconductors

InGaN. Implant isolation of the In-containing nitrides (INN, InGaN, and InA1N) was first reported using F-implantation.[ 31] That work showed that InN did not demonstrate significant compensation while the temaries (InGaN and InA1N) increased in sheet resistance by roughly an order-ofmagnitude after a 500~ anneal, Data from a more extensive study of InxGal.xN implant isolation for varying In-composition using N- and Fimplantation is summarized in Fig. 2.[ 32] The InGaN temaries only realize a maximum of a 100 fold increase in sheet resistance independent of ion species after a 500~ anneal. Pure InN shows a higher increase of 3 ordersof-magnitude, but still only achieves a maximum sheet resistance of 104 f2/EI. This may be high enough for some photonic device current-guiding applications, but is not sufficient for inter-device isolation in electronic circuits. The damage levels created by N-implantation are estimated from an Arrhenius plot of the resistance/temperature product to be a maximum of 390 meV below the conduction band.J32] The defect level is high in the energy gap, not near midgap as is ideal for implant compensation. The position of the damage level is analogous to the defect position reported for implant compensated n-type InP and InGaAs, but different from the damageassociated, midgap states created in GaAs and A1GaAs.[23][24][33]

108 , , , , , . , , . , . , . , . , , , , 107 106

91

105

D

104 r

Q.

103

102 l 01 10~

--0- as-grown - t ~ F-isolation N-isolation ,

,

I

20

,

,

,

I

,

,

,

I

t

40 60 percent In

i

,

i

80

,

,

,

i

100

Figure 2. Maximum sheet resistance versus percent In for InGaN either as-grown or implanted with F or N and annealed at the temperature for maximum compensation for each composition (ion concentration - 5 x 1019cm-3) (after Ref. 34).

lonlmplantation 305 InAIN. As shown in Fig. 3, In0.75A10.25N, in contrast to InGaN, can be highly compensated with N- or O-implantation with over a three orderof-magnitude increase in sheet resistance after a 600 to 700~ anneal while F-implantation produces only one order-of-magnitude increase in sheet resistance.[a1][34][35] The compensating level in InA1N is also high in the bandgap (with the deepest level estimated from Arrhenius plots as being 580 meV below the conduction band edge in high dose N-isolated material). However, it is sufficiently deep to achieve highly compensated material.[34][35] The enhanced compensation for N- and O-implantation in InA1N suggests some chemical component to the compensation process (as compared to F-implantation). For N-implantation, a reduction in N-vacancies (thought to play a role in the as-grown n-type conduction), may explain the enhanced compensation. For O-implantation, the enhanced compensation may be the result of the formation of an O-A1 complex as is thought to occur in O-implanted A1GaAs.[23][24]

10 lo 10 9 10 8 [Z]

ck

10 7 10 6

105 10 4 103 200

I

I

400

!

I

600

I

I

I

800

.

1000

annealing temperature (~

Figure3. Sheet resistance versus annealing temperature for O-, N-, or F-implanted Ino.7sAlo.25N (ion concentration- 5 x 10 ts cm "3) (after Ref. 34).

306

Wide Bandgap Semiconductors

Figure 4 schematically summarizes the present knowledge of the position in the bandgap of the compensating implanted defect levels in group-III nitride materials and compares these to those in GaAs and InP. Although the levels are not at midgap, as is ideal for optimum compensation as occurs in GaAs and p-type InP, the levels are sufficiently deep to produce high resistivity material (with the exception of InGaN).

n,p GaN n-InGaN (47% InN)

n-InAIN (75% InN) n: 830 meV (He 760 meV)

n&p-GaAs

n&p_InP

~-390meV

~~580meV

EC

l p: 900 meV EV Eg =

1.42 eV

1.35 eV

-2.5 eV

-2.5 eV

3.39 eV

Figure 4. Schematic representation of the position in the energy gap of compensating defect levels from implant isolation in GaAs, InP, Ino.47Gao.53N,Ino.7sAlo.25N,and GaN.

SiC. Implant isolation of SiC layers is also of technological importance to allow the fabrication of planar electronic devices and circuits (limited work has been reported in this area). Nadella reported on the properties of H-implanted, n-type 4H-SiC and achieved resistivities as high as 2 x 106 ~_cm.[ 36] The H-implanted samples demonstrated a temperature activated conduction with characteristic energy of 0.41 eV suggesting that the compensating defect is in the upper half of the bandgap. This may limit this approach for devices that need to operate at elevated temperatures since carriers will be thermally activated out of the levels. Chemical compensation has been achieved in SiC by doping with vanadium either during epitaxy or by implantation. [371-[39]Vanadium was shown to act as a deep donor 1.35 eV below the conduction band edge that effectively compensated residual boron in bulk SiC.[38]V-implantation into

Ion Implantation 307 initially p-type (boron doped) SiC achieved resistivities as high as 1012 ~')cm after annealing up to 1500~ For N-doped, n-type SiC, the same Vimplantation scheme resulted in resistivities of 106 ~-cm that, although lower than the p-type starting material, is still sufficiently high for most device applications. In contrast to the trend for the resistivities, the activation energy for conduction was estimated to be 0.76 eV for the n-type starting material, and 0.1 eV for the p-type starting material. The lower activation energy for the p-type material was not consistent with a 1.35 eV V-donor level and was attributed to alternative leakage paths, such as on the mesa edge.[ 39]This may indeed be the case since V-implantation is now being applied to electronic devices as discussed in Sec. 6.2.[ 4~ Diamond. To date, the primary challenge for diamond devices has been realizing n- and p-type doping. Therefore, little work has been done on implantation isolation in this material. One study by Kalish and coworkers examined the compensation properties of He- and H-implants on p-type diamond.[41] Type IIa diamond films were doped with boron with a multiple energy implantation scheme followed by a two step anneal at 1050~ and 1350~ 42] The films were then implanted with He at an energy of 320 keV or H at an energy of either 30 or 320 keV. Figure 5 shows the effect on resistance versus implanted dose for the three implantation schemes. All three approaches effectively compensated the film for a sufficiently high dose. The dose dependence is explained by the difference in defect profile generation for the different ions and energies. Figure 6 shows resistance versus damage density for the different implants and demonstrates that complete compensation occurs at roughly the same damage level (within a factor of two for all three approaches). This is evidence that damage profile is the key element in compensation and not a chemical compensation by H or He. The damage could be annealed out, and the initial conductivity restored, by annealing the samples at 1350~ but not at lower temperatures (600~ This work also suggests that p-type diamond may be susceptible to radiation damage that will alter the electrical properties of active devices.[ 41]

308

Wide Bandgap Semiconductors

10 TM

~ 101o

o~ Z

109

VHe

ni

u') rY

I0 e

H

107 ,

101~

10"

10 '2

DOSE

10 I~

,

, IA,,~

10"

10 '5

(ions/cm2)

Figure 5. Resistance of B-doped diamond layers under irradiation with 320 keV He, 320 keV H, or 30 keV H. The resistance of a undoped layer measured under the same conditions is shown for comparison (after Ref. 41).

1012

" "Tl'"al" ' ' 'i .... I

' ' '1 .... I

' ''i""l

' '~i .... I

1011

E o

10 l~

Ld 0 z

109

Unirradioted

j

U3 m 10e

9320keY 9320

He

"

keY H

0 undoped

9 30

10 7 101~ DAMAGE

101.

1015

DENSITY

1016

keY H 1017

1010

( v a c a n c i e s / c m 3)

Figure 6. The data of Fig. 5 replotted after converting the dose into damage density expressed as vacancies/cm 3. The damage density required to completely remove the conductivity due to the B dopant is about 5 x 1016/cm3, which is comparable to the concentration of acceptors in the B-doped layer (after Ref. 41).

lonlmplantation 309 3.0

IMPLANTATION DOPING

3.1

Carrier Ionization in Wide Bandgap Semiconductors

Before reviewing the work on implantation doping in GaN, SiC, and diamond, it is constructive to review the physics of free carrier ionization. The assumption often taken in silicon and GaAs, where ionization energies are typically < 20 meV, of compete carrier ionization at room temperature does not apply for wide bandgap semiconductors due to the high ionization energies. Table 1 contains the accepted values for ionization energies for the most commonly used dopants in GaN, SiC, and diamond. The relationship between these ionization energies and the free carrier concentration is reviewed in the following section.

Table 1. Summary of Ionization Energies of the Most Common Dopants in GaN, SiC, and Diamond

GaN a

3C-SiC b

4H-SiC b

6H-SiC b

Diamond c

donors

Si (25), 0(29)

N(28)

N (45, 100)

N (80, 130)

N (1700)

acceptors

Ca (169), Mg (170), Zn (230)

B (350), Ai (200)

B(350), Al (200)

B (350), Al (200)

B (370)

a j. C. Zolper, et al., J. Electron. Mat., 25:839 (1996); J. C. Zolper, et al.,Appl. Phys. Lett., 68:1945 (1996); S. Strite and H. Morkoc, J. Vac. Sci. Tech. B, 10:1237 (1992) b G. Pensl and W. J. Choyke, Physica B, 185:264 (1993) c R. J. Farrer, Solid State Corn, 7:685 (169); A. T. Collins and A. W. Willimas, J. Phys. C.

Solid St. Phys., 4:1789 (1971)

310

Wide Bandgap Semiconductors The free electron density (n) can be expressed as: [43]

n=N ex~_(E _Ef)]kT

Eq. (1)

and the free hole density (p) can be expressed as:

Eq. (2)

p= N ex~-(Ekf E)]

Ec(v)

where is the conduction (valence) band minimum (maximum) energy. For n-type material, the position of the Fermi level can be solved for from the following expression for the density of ionized donors:

(Ef)

N~d -Nd 1-

Eq. (3)

1

1 (E~.Z.~" ~

l + -geXp~

kT )

while for p-type material the density of ionized aceeptors is:

Eq. (4)

N~=N

1 1+ g e x p ( E :

f,]

and the conduction (valance) band density-of-states, No(v),is defined as" 3/2

Eq. (5)

Mc

Nc(~)-212zcmhz(h)kT ) M~

where is the number of equivalent minima in the conduction band and mde is the density-of-states effective mass. The electron or hole ground

Ion Implantation 311 state degeneracy is expressed by g. The donor (acceptor) ionization energy (Ed(a)) is listed in Table I. Other terms in Eqs. 3-5 have their usual meaning. To simplify the discussion, we assume that the material does not contain significant compensating impurities. The key observation from the above equations is that the free electron and hole concentrations are exponentially dependent on the carrier ionization energy as it determines the position of the Fermi level. To illustrate this point, a simple exponential dependence or Boltzmann statistics of free carrier concentration (n,p .~ N d , a exp (Ed,a/kT)) at room temperature (23~ and 300~ versus carrier ionization energy is shown in Fig. 7. Nd, a w a s set to 100 meV for Fig. 7 so that the result can be displayed as percent ionized. For an ionization energy of 170 meV, as is reported for Mg acceptors in GaN, only---0.14% of the substitutional Mg will yield free holes at room temperature.[ 4][35] This increases to -~3% at 300~ This relationship should be kept in mind when considering the effectiveness of implantation doping in these materials since proper lattice occupation of the dopant (i.e., substitutionality) alone does not control the measured free carrier concentration.

100

i

I

\-.

o~

i

1

-

I L

~23

v

--

10

~

"300 ~

,irN ,,==,,

o o,=,~ r ro

%

L_

Q.

0.1

%

0.01

,

0

i

,

50

100

,.l

~i

150 200 250 300 350 400

ionization energy (meV)

F i g u r e 7. Percent of ionized carriers versus ionization energy at 23 or 300~ simplified exponential or Boltzmann statistics; n , p - Nd, a exp (Ed, a/kT).

based on a

312

3.2

Wide Bandgap Semiconductors

Implantation Activation Temperature

Selective area ion implantation doping can be used to form highly doped contact regions in lasers and FETs or to create precisely doped transistor channels. The ability to precisely control the doping level and spatial location enables many high-performance device designs. To achieve activated implanted dopants, the ability of the material to withstand the required annealing process must be accessed. Table 2 compares the melting point of wide bandgap semiconductors with more mature semiconductors (Si and GaAs), to the temperatures commonly reported for achieving activation of implanted dopants in these materials.[ 44]For Si and GaAs, the activation temperature is roughly two-thirds of the melting temperature. However, for the higher melting point materials (SiC and GaN), this temperature is closer to 50% of the melting point. The temperature reported for dopant activation in diamond is at an even lower fraction of the melting point (--0.4). While this may suggest that SiC and GaN are actually more stable at the activation temperature, this is not the case since both materials sublime well below their melting point. In the following sections, the status of activating implanted dopants in the wide bandgap semiconductors is reviewed. Details of various annealing schemes are also presented.

Table 2. Comparison of Semiconductor Melting Points (Tmp) to the Temperature Required to Activate Implanted Dopants (Tact) (after Ref 44)

T,.p(~ GaSb InP GaAs Si SiC c GaN c diamond

707 a 1057 a 1237 a 1410 a 2797 a 2518 b 4000

T,a (~ 500--600 700-750 750-900 950-1050 1300-1600 --1100 .-1300 c

rac~/Tmp 0.71-0.85 0.66-0.71 0.61-0.73 0.67-0.74 0.46-0.57 0.44 0.33

a Robert C. Weast, Handbook of Chemistry and Physics, CRC Press, Boca Raton, FL, p. E92-93 (1983) b j. A. Van Vechten, Phys. Rev B, 7:1479 (1973) r May include defect and impurity conduction[ TM

Ion Implantation 313 3.3

GaN

In the early 1970's, Pankove and co-workers used ion implantation to characterize the photoluminescence spectra from an array of dopant species in GaN.[ 45] In this work, the energy levels of the common III-V semiconductor acceptors (C, Be, Mg, Zn, and Cd) were first determined, as summarized in Table 3. Magnesium was reported as having the shallowest acceptor level (-~240 meV), while Zn (with an energy level of~580 meV) had the strongest luminescence intensity.[46] To remove the implantation damage, these samples were annealed for 1 hr in flowing NH 3. No electrical properties of the implanted species were reported. These layers were most likely compensated by H that was generated from the decomposition of the NH 3 present during growth and during the annealing process. As discussed later, annealing in a non-hydrogen containing ambient is required to achieve electrical activity of acceptors on GaN.

Table 3. Photoluminescence Peak Position, Distance from the Conduction Band Edge, and the Rank Order of Luminescence Intensity for GaN Implanted with the Species Listed and Annealed in NH 3 at 1050~ for lh. (Data after Pankove in Ref. 45.)

Species

Peak Position

Bandedge-Peak

(eV)

(meV)

3.21

240

B

3.2

250

P

2.88

570

Zn

2.87

580

Cd

2.70

800

As

2.58

920

Ca

2.50

1000

ng

2.43

1070

C

2.17

1280

Be

2.16

1290

Mg

Rank Order of PL Intensity

314

Wide Bandgap Semiconductors

Figure 8 shows the evolution of sheet resistance versus annealing temperature for Si-implanted (200 keV, 5 x 10 TM c m -2) and unimplanted GaN.[ 28] The samples were annealed for 10 s in flowing N 2 in a SiC coated graphite susceptor. It is critical to use a hydrogen free ambient to avoid hydrogen passivation of the dopants and thus to achieve electrical activity. This discovery is in contrast to earlier annealing studies that used an NH 3 ambient to stabilize the GaN surface but allowed the generation of atomic hydrogen to passivate dopant species. [45][46] As seen in Fig. 8, electrical activity starts to occur at 1050~ as evident by the drop in sheet resistance, and further increases at 1100~ The ionization levels of implanted Si have been estimated from Arrhenius plots of either the sheet resistance or the carrier density to be 25 and 62 meV, respectively.[ 35] These values are in the range reported for epitaxiai doping of GaN with Si. [47]

10 6

105

D t~

a.

104 unimplanted --tl- Si-implanted

103 600

I

700

I

800

I

900

I

I

1000 1100 1200

annealing temperature (~ Figure 8. Sheet resistance versus annealing temperature for GaN either unimplanted or implanted with 28Si (200 keV, 5 x 1014cm-2). The implanted sample demonstrates enhanced ntype conduction after a 1050~ anneal with further improvements at 1100~ (aider Ref. 27).

Figure 9 shows an Arrhenius plot of the sheet electron concentration versus annealing temperature for Si-implanted GaN (200 keV, 5 x 1014 cm2).

lon lmplantation 315 A carrier activation energy (not to be confused with the ionization energy) can be estimated as 6.7 eV from the region of increasing electron concentration by examining Fig. 9.[35]This high activation energy is not consistent with a simple hopping process whereby the Si atoms occupy the nearest Ga vacancies to become electrically active. Such a hopping process is thought to occur during GaAs and InP implantation activation with activation energies in the range of 0.4 to 1.9 eV. [48] The high activation energy for Siimplantation in GaN can be explained in terms of a substitutional diffusion process. In this case, the implanted Si would occupy substitutional Ga-sites at relatively low temperatures, but remain electrically inactive (compensated) by implantation induced point defects. The Si would then become electrically activated only when these defects diffuse away and re-occupy an appropriate lattice site. An example of such a defect reaction is given in Eq. 6: Sio a _ N +Oa ]0 ~ Sioa + NON + Ga 0Ga

Eq. (6)

1014 r J

o I::I O

9 i,,,,,i

1013 a

= 6.7 eV

o I::I

1012 m I

lO 11 . . . . 0.7

~ .... 0.8

I I

0.9

1

,

1.1

1000/T (K l ) Figure 9. Arrhenius plot of the sheet electron concentration for Si-implanted (200 keV, 5 x 10TM cm-2) GaN versus annealing temperature. The extracted activation energy for donor formation is 6.7 meV which is consistent with an inter-diffusion process (after Ref. 34).

316

Wide Bandgap Semiconductors

Therefore, the activation energy corresponds to the inter-diffusion coefficients of Ga and N in GaN, not the energy for Si to become substitutional. The energy of 6.7 eV, estimated from Fig. 9, is in the range reported for Ga and As inter-diffusion in GaAs and, therefore, is consistent with an inter-diffusion process.[ 49] Ion implantation has also been used to estimate the ionization energy of O in GaN.[ 5~ Determining the electrical nature of O-impurities in GaN is of interest, since O has long been suspected as playing a contributing role in the background n-type conductivity of GaN.[51][52] Figure 10 shows an Arrhenius plot of the resistance/temperature product for unimplanted and O-implanted (70 keV, 5 x 1014cm -2) GaN after annealing. The O-sample is n-type and displays an estimated ionization energy of-~28 meV while the unimplanted sample remains highly resistive with an effective ionization energy of 335 meV.

1013

, '

'

'

!

'

'

'

i

'

'

I

'

'

'

I

'

'

'

I

'

'

unimplanted, 1100 ~ - I - O-implanted, 1050 ~

1012

= 335 meV

O4

~:

'

lO ll

v

~.~

1010

Q.

109 I

108

J

2

,

i

I

4

i

i

i

i

6

,

,

,

I

8

,

,

,

I

,

10

,

,

I

,

12

,

,

t4

1000/T ( K ~) Figure 10. Arrhenius plot of the resistance/temperature product for unimplanted, annealed (1100~ and O-implanted (70 keV, 5 x 10TM cm-2), annealed (1050~ GaN. The extracted ionization level for O is 28 meV while that for the unimplanted sample is 335 meV (after Ref. 47).

Figure 11 shows the evolution of sheet resistance versus annealing temperature for Mg (180 keV, 5 x 1014cm-2), Mg+P (180/250 keV, both

_Ion Implantation 317 5 x 1014cnl2), and unimplanted GaN.[ 28] The Mg-only samples remained n-type up to 1100~ while the Mg samples co-implanted with P converted from n-top type after a 1050~ anneal. The effect of the P co-implantation may be explained by a reduction of N-vacancies, or an increase in Ga-vacancies, leading to a higher probability of Mg occupying a Ga-site. Co-implantation of P has also been shown to be effective in enhancing activation and reducing diffusion for p-type implantation in GaAs. [53][54][55]The ionization levels of implanted Mg has also been determined, from an Arrhenius plot of carrier density, to be 171 meV and is consistent with the value reported for epitaxial Mg-doped GaN.[35][4]

10 7

106

D

a- ~

10 5

104 600

unimplanted ~Mg "-II-- Mg+P(n) Mg+P(p) I

700

I

800

I

900

I

1000 1100 1200

annealing temperature (~ Figure 11. Sheet resistance versus annealing temperature for GaN either unimplanted or implanted with 24Mg (180 keV, 5 x 10TM cm-2) or 24Mg+alp (180 keV/250 keV, both 5 • 1014 cm2). The Mg+P implanted sample converts from n-to-p type after a 1050~ anneal while the Mg-only sample remains n-type even after a 1100~ anneal (after Ref. 27).

Since the ionization level of Mg in GaN is much greater than kT, less than 1% of the Mg-aeeeptors will be ionized at room temperature as described in Sect. 3.1. Therefore, it would be desirable to identify an aeceptor species with a smaller ionization energy. Since Ca has been

318

Wide Bandgap Semiconductors

theoretically, suggested to be a shallow acceptor in GaN, ion implantation was used to determine the ionization energy of Ca in GaN.[5~ Figure 12 shows the evolution of sheet resistance versus the annealing temperature of Ca (180 keV, 5 x 1014 cm-2), Ca+P (180/130 keV, both 5 • 1014 cm-2), and unimplanted GaN. Both the Ca-only and the Ca+P samples converted from n-to-p type after a 1100~ anneal with a further increase in p-type conduction after a 1150~ anneal. The fact that P co-implantation is not required to achieve p-type conductivity with Ca can be understood based on the higher mass of the Ca-ion, as compared to Mg, generating more implantation damage and therefore more Ga-vacancies. This explanation is supported by the higher activation temperature required for conversion from n-to-p type for the Ca-implanted samples compared to the Mg+P implanted samples (1100 versus 1050~ The ionization level of Ca was estimated (from the Arrhenius plot shown in Fig. 13) to be 169 meV, which is equivalent to that of Mg.[ 5~ Although the ionization level of Ca is not smaller than that of Mg, Ca may be preferred for forming shallow implanted p-regions in GaN due to its heavier mass and resulting smaller projected range and straggle than Mg for a given energy.

107 _ ,

..... , . . . . .

, ....

, ....

, ....

10 6 n-to-p

V1 105 unimplanted ---{S}-Ca: n-type Ca+P: n-type Ca: p-type Ca+P: p-type

104

103 950 .

.

.

.

1

.

1000

,

,

~

I

.

1050

.

.

.

I

.

1100

.

.

.

I

.

1150

.

.

.

1200

annealing temperature (~

Figure 12.

Sheet resistance versus annealing temperature for GaN either unimplanted or implanted with 4~ (180 keV, 5 x 10 TMcm -2) or 4~ (180 keV/130 keV, both 5 x 10 TMcm2). The implanted samples convert from n-to-p type after a 1100~ anneal (after Ref. 47).

lonlmplantation 319

1013

'

.,

'

'

I

'

'

"

i

'

'

'

i

'

'

+--

0

o

1012 E = 169 meV

0

""'~

a

o O a= o o r ~lU

II

, , ,

2.8

I

3

,

,

,

I

,

,

,

3.2

,

. . . .

3.4

3.6

1000/T (K 1) Figure 13. Arrheniusplot of the sheet hole concentration for Ca-implanted(180 keV, 5 x 1014cm-2)GaN annealedat 1150~ The extractedionization level for Ca is 169 meV (after Ref. 47).

3.4

SiC

Although reports of implantation doping of SiC date back at least to 1969,[57]-[59] significant advances in understanding of implantation damage accumulation and dopant activation occurred with work by Edmond (circa 1988)on the use of elevated temperature implantation.[6~ The key result of Edmond's research was that elevated temperature implantation markedly enhanced activation of implanted dopants by limiting the buildup of implantation damage via in-situ annealing. This has led to elevated temperature implantation being widely employed for doping of SiC, as discussed in this section. Implantation of the most common donor species in SiC, N, will be presented first, followed by discussion of the most widely used p-type dopants, B and A1. Hiranoet al. reported on the activation of implanted N in 3C-SiC layers as a function of implant temperature and N-dose.[ 62]As seen in Fig. 14, for a sample annealed at 1200~ for 20 min (furnace anneal, FA), the electron concentration was significantly enhanced for N-implantation (N2 at 30 keV and a dose of 3 x 1015 cm -2) by implanting at 400~ as compared to 200~

320

Wide Bandgap Semiconductors

(or at room temperature, RT). This effect was ascribed to less damage accumulation, which is supported by damage measurements by Rutherford Backscattering (RBS) discussed later in this chapter. Hirano also showed that there is a dose saturation level for activation near 3 x 1015cm -2 for their implantation conditions. For their conditions, they achieved an electron concentration near 5 x 1019 cm -3 as determined by C-V measurements. For room temperature (RT) N-implants Kimoto reported a minimum sheet resistance of 770 DJr'l (that corresponded with a saturation in N-donors) at a dose of 8 x 1014 c m -2 for furnace annealing at 1500~ 631Hirano also reported using 10 s, 1100~ rapid thermal annealing for both RT and 400~ implants. He reported an order-of-magnitude (to 1 x 1020 cm-3) increase in the electron concentration for the elevated temperature implant, as determined from C-V measurements.[ 62] Significantly, the RTA sample had a higher electron concentration than the 1200~ FA-sample implanted under the same conditions. The high concentration achieved for the RTA sample was reported to be near the highest n-type doping reported for implanted or epitaxial SiC at the time. The success using RTA processing and the resulting high doping level suggests this will be an effective process technology for device applications. 2

10

I

I I M P L A N T S "3.0 x 1015/cm 2 400~ O 9 ~ 200~ ~ 9 ~ 0 0 ~ 20min

f ~,

.

u

n 0~176176 o 0o00 o

r~

0 (:~

,

\ _ i 03,~

o

Z 101 " o a O

~SIMS

a

aa

"

A

,.-A

o

Oo0

A A

.o

%

0

:~

~"

.

U

9

.-

-

" -

-

g

1

~

-

"I

0

-

17 10 0

I 500 DEPTH

I 1000 (,~)

101 1500

Figure 14. Carrier concentration and mobility profiles for n-type layers formed in 3C-SiC by implanting 30 keV N 2 ions to a total dose of 3.0 x l 015 cm -2 at 200~ (triangles) and 400~ (circles) and by subsequent annealing at 1200~ for 20 min (after Ref. 62).

lon Implantation 321 J. N. Pan et al. also studied elevated temperature N-implantation in SiC and, in particular, the details of the annealing process.[641 As shown in Fig. 15, for implants performed at 650~ the optimum time for the anneal is a strong function of the anneal temperature. The minimum anneal time used in this study was 5 min and was limited by the furnace annealing apparatus. It is interesting to note, however, that the sheet resistance appears to have not yet reached a minimum value at 5 min for the 1200~ samples. This also suggests that a rapid thermal anneal (t < 60 s), as reported by Hirano, may be effective to activate these types of implants.

4000 ....... i ..... 0"

,

......

,._iii

_i.ii.i.

3000

........

C v >.,

.....

2000

]

[

a

o m

n" ,-.9 r O3

1000 900 800 700 ....................................................

60O

500

|

I

,

,

, ,

|l|'

io

,

.

.

.

.

.

.

.

]

ioo

.

.

.

.

.

.

.

.

.

.

.

Iooo

.

.

.

.

Ioooo

Anneal Time (minutes)

Figure 15. Sheet resistivity as a function of anneal time for 900, 1050, and 1200~ anneal temperatures for N-implanted (3.8 x 1015cm-2at 30 keV plus 7.1 x 1015cm-2at 70 keV) 6HSiC (after Ref. 61). P-type implantation doping of SiC has been achieved with both B and A1. One of the early reports was for Al-implantation in n-type SiC by Gudkov.[ 65] Gudkov reported difficulty in maintaining the surface stoichiometry during the activation anneal, limiting the ability to quantify the acceptor activity. Rao made a significant contribution in this area by studying both A1 and B-implantation in 6H- and 3C-SiC. [66] Rao showed the importance of using elevated temperature implantation as developed by

322

Wide Bandgap Semiconductors

Edmond to achieve p-type activation. In this case, A1 was implanted at 850~ and annealed at 1400~ to produce p-type conductivity as determined by Hall and C-V measurements. These layers had--1% of the implanted Al-dose produce ionized acceptors at room temperature, due to the high ionization energy of A1 as discussed in See. 3.1. This means that 100% of the implanted A1 impurities occupied substitutional lattice sites. In this same study, p-type conductivity was not reported for B-implanted SiC. The lack of hole conduction for B was attributed to the still larger ionization energy of B in SiC, compared to A1, or a lack of suitable vacancies being formed during the implantation process for the B to occupy substitutional sites.[66] The difficulty in achieving p-type SiC by B-implantation was also reported by Kimoto, where B-implantation resulted in a highly resistive layer with an indeterminate carrier type. [67] In that same study, Kimoto was successful in producing p-type SiC with room temperature Al-implantation and annealing in an rf-induction furnace at 1500~ The apparent contradiction between Rao's report for the requirement for elevated temperature and that ofKimoto' s room temperature Al-implantation was postulated to be due to differences in the annealing dynamics between Rao's fiwnace anneal and Kimoto' s more rapid annealing scheme.[ 67] Finally, Rao has studied the effect of co-implantation of Si and C on the activation properties of Al-implanted 6H-SiC.[ 68]Co-implantation is often used in compound semiconductors to maintain the local crystal stoichiometry and promote the preferred site occupation of the implanted dopant.J53]-[55]In this case, the co-implantation of either Si or C did not enhance the formation of Al-acceptors. In fact, Si co-implantation resulted in highly resistive layers, either due to compensation by resulting carbon vacancies or other impurity/ defect complexes.[ 68]This area of co-implantation may warrant further study since there is often a dose dependence between the dopant and co-implantation species to achieve optimum electrical results.

3.5

Diamond

Diamond has long been identified as a promising semiconductor. However, its use has been limited by the inability to grow semiconductor grade synthetic diamond and by the difficulty in doping this very stable material. Ion implantation has been one of the most successful approaches to doping diamond due to the ability to create non-equilibrium defects during the bombardment process.j69][7~ A process described as cold-implantation-rapid annealing (CIRA) has been developed and applied to p-type,

Ion Implantation 323 boron doping. This involves implanting the sample at low temperature, typically liquid nitrogen temperature (77 K), to "freeze in" point defects that can then interact with dopant species during subsequent annealing. The annealing is done with a two step process, with a first anneal at ~1000~ and a second at ~1300~ 71] This approach has been successful in achieving p-type doping with a maximum hole mobility of 385 em2/Vs.[ 72] Successful boron doping has also been realized by combining the CIRA (77 K implantation and 1373 K rapid thermal annealing) process with carbon co-implantation.[73] In this study, the B-dose was varied between 1 and 10 x 1014c m "2 while the C-dose ranged from 3 to 20 x 1014 e m "2. The co-implanted sample with the optimum annealing sequence achieved a resistivity of 100 ~-cm with an estimated activation energy of conduction for boron of 0.1 eV. Optical data also support a high fraction of the boron being substitutional acceptors.[ 73] A second co-implantation approach to controlling the vacancy distribution and enhancing dopant activation is the use of a low dose preimplantation scheme that has been named "Low-Damage-Drive-In" Implantation (LODDI). In this approach, the non-dopant or co-implantation species creates the required defect distribution for the dopant atom. Although the annealing sequence depends on the dopant species employed, the key aspect in the process is the low dose required of the pre-implant. By combining He pre-implantation and various annealing steps with B-implantation, this approach produced a hole concentration of 4.0 x 1013 c m 2 and a low field mobility of 1953 c m E / V s . [69] Figure 16 shows results for a Hedamage implant (5 x 1010 cm -2) followed by a 600~ anneal, then an elevated temperature (400~ doping B-implant. During the B-implant, the B is thought to diffuse interstitially into the undamaged, underlying diamond layer. Finally, a carbon amorphizing implant is performed to allow a subsequent etch removal of the amorphous damaged region that is above the interstitially B-doped diamond. At this point, the diamond was subjected to a series of anneals at 1550~ Hall characterization of this sample gave a hole concentration of 4.0 x 1013 c m -3 with a low field mobility of 1953 cmE/Vs.[69] A similar approach for P-implantation in diamond using C as the damage implant species compared it to an Ar-implanted sample that should have the same damage profile. The results, shown in Fig. 17, demonstrated a lower resistivity for the P-implanted samples as compared to the Ar-sample, with the highest dose P-sample having the lowest resistivity. [69] The P-samples were implanted at 100 and 400~ at doses of 5 or 50 x 1016 cm -2 at each

324

Wide Bandgap Semiconductors

temperature. The Ar-implant was implanted at the sample temperatures and a dose of 4.2 x 1016 cm -2 (the same damage profile as the lower dose P-sample should be evident).[ 69]The temperature variation of the resistance for these Pimplanted samples was indicative of a highly compensated, low density of dopant. The activation energy for conduction was estimated at 0.3 eV for a P-dose of 1 x 1016 cm -2 and 0.25 eV for a dose of 1 x 1017 cm -2 suggesting that the P ionization energy is not more than 0.25 eV below the conduction band.[ 69] Although this P-implantation doping result is encouraging, significantly more work is needed to understand the actual doping and conduction mechanisms resulting from this process.

6.5 o o

ffl

o

E

o

cO

o o

W

0 0

Z < I--

0

m_ 5.5

0 0

0

UJ

0 0 o

holes=4.0x1013

0 o

UJ LIJ 7"

cm-3

0

/

O0 ~

mobility = 1953 cm 2 N - s

o _J

4.5

9

1.6

I

2

I

2.4

I ,

2.8

!

3.2

INVERSE T E M P E R A T U R E (I 000/3)

Figure 16. Sheet resistance of a diamond layer doped by B ion implantation using the LODDI process. This layer had a room temperature Hall mobility of 1953 cm2/Vs (after Ref. 66).

Ion Implantation 325

15

A

o"

E tO

Argon ions "

o Z <: I-(/3

o

o

I'I!1

w

o

o

o

o

o

o

09 Ill

-r0o

O

13

Ill

O

O

O

0 O 0 0

oo

/

11

0

o

0

o

O

0

On no ++++

+

0

+ 4-

4-

+ +

m~ _ ~ + + + + + +C!1:11:3+ + +

OOO O 0

..J

P'\ Phosphorus ions

.

1.6

.

.

.

i

2

.

.

.

.

.

i

2.4

.

.

.

i

2.8

.

.

.

.

i

3.2

INVERSE TEMPERATURE (1000/T) Figure 17. Sheet resistance of diamond layers implanted with P or Ar using the LODDI process. The P-samples had two different doses (5 or 50 x 1016 cm -2) implanted at 100 and 400~ with the Ar-sample used as a control sample to compare conduction due to implantation damage to conduction due to P-doping (after Ref. 66).

4.0

IMPURITY REDISTRIBUTION

4.1

GaN

When applying ion implantation doping to device structures, it is important to know how the impurities redistribute during the activation anneal. Initial studies by Wilson and coworkers on the redistribution of implanted impurities in GaN was limited to temperatures up to "--800~ [74][75] That work showed negligible redistribution of all elements studied with the exception of S which had appreciable diffusion even at 600~ (see Fig. 18). When it became apparent, however, that temperatures of 1100~ or higher are required to achieve activated dopants in GaN, the question of redistribution

326

Wide Bandgap Semiconductors

was revisited.[ 76] Figure 19 shows the Secondary Ion Mass Spectroscopy (SIMS) profile for 28Si in GaN as-implanted and annealed (1050~ Despite the interference in the mass 28 SIMS signal from 2SN2, the annealed Si profile demonstrates no measurable redistribution. Using a conservative estimate of 20 nm for the resolution of the SIMS measurement, an upper limit of 2.7 x 10 "13 cm/s can be set on the diffusivity of Si in GaN at 1050~ The redistribution of O has also been studied, as shown in Fig. 20. Here again, no measurable redistribution is seen and an upper limit of 2.7 x 10 -13 cm/s can be set on the diffusivity of O in GaN at 1125~

1020

,,

-

9425-26OO6

,

I

I

I

.

-

S in G a N 200 keV 1.0 x 10 TM cm -2 10 TM

A

?

E (.1

>. i-. z 1018

600~

UJ

t2

NO A N N E A L

=i

O

i.=

<

1017

10 TM 0.0

I

I

i

0.5

1.0

1.5

2.0

DEPTH (l~m)

Figure 18. SIMS profile of S-implanted (1 x 1014c m

after a 600~ anneal (after Ref. 71).

"2

at 200 keV) GaN as implanted and

lonlmplantation 327

1020 i

,

,

,

!

,

,

as implanted

,

- - "1050 ~ 019

~

10 Is

01: rr = lli 1017

"~ ~~' IIir

1016[i................... 0

0.2

0.4

0.6

0.8

depth (~tm) Figure 19. SIMS profile of 2gSi-implanted (100 keV, 5 x 1014 cm -2) GaN as-implanted and annealed at 1050~ for 15 s. No measurable redistribution is seen at the peak of the annealed sample (after Ref. 73).

1 0 20

~ . , , ..... /t~

9

~

,

.

!

,

,

,

|

9

,

,

as-implanted

. . . . . 1125 o ~ '1=

1019

0

~

10 Is

o 0

9

1017

1016 0

................... 0.2 0.4

0.6

0.8

depth (lain) Figure 20. SIMS profile of 180-implanted (70 keV, 5 x 10 TM cm "2) GaN as-implanted and annealed at 1125~ for 15 s. No measurable redistribution is seen at the peak of the annealed sample (after Ref. 47).

328

Wide Bandgap Semiconductors

The lack of Si-redistribution at the implant activation temperature is consistent with the behavior of Si in other compound semiconductors; however, acceptor species are generally more susceptible to diffusion at high temperatures.[ 77] Figure 21 shows the SIMS profiles for Mg, asimplanted and after a 1150~ 15 s anneal. After annealing, the Mg-profile shows a slight movement towards the surface that is estimated to be 50 nm near the peak of the profile. Based on a 50 nm diffusion length and a 15 s anneal, an upper limit of 6.7 x 10-13 cm/s can be set on the diffusivity of Mg in GaN at 1150~ Profiles for Mg co-implanted with P demonstrated a similar amount of redistribution that is somewhat in contrast to the need for co-implantation to achieve acceptor activity. The Mg-only sample should have more Mg in non-active, interstitial sites that should act as fast diffusers as they do in other compound semiconductors.[ 54]This potential conflict has not yet been resolved, however, similar diffusion results have been reported for Mg co-implanted with P by Edwards, confirming the initial result.[ 78]

102 0

9 9 9, . . . . . . . . ,.7/~

~,

0 ~9

[ ~

9 9 . ,

9 , .'

as-implanted

[ ..... l lsO~

1 o

10Is ~D 0 O r ca

1017

1016

.................... 0 0.2 0.4 0.6 0.8

I

depth (l.tm)

Figure 21. SIMS profileof24Mg-implanted (100 keV, 5 x 1014 cm -2)G a N as-implanted and annealed at 1150~ for 15 s.The profilepeak shows an approximately 50 n m shifttowards the surface a1~erthe anneal (afterRef. 73).

Figures 22 and 23 show the SIMS profiles for as-implanted and annealed Be and Zn, respectively.[ 79]Zn has played a role as a color center in GaN LEDs, while Be may be of interest as an alternative acceptor

Ion Implantation 329 species.[5][8~ 831Neither Be nor Zn shows measurable redistribution after annealing, which suggests an upper limit of 2.7 x 10-13 cm/s for the diffusivity of Be and Zn in GaN at 1125~ Finally, as shown in Fig. 24, implanted Ca also shows no measurable redistribution even after a 1125 ~ anneal.[ 5~ This Ca result was confirmed by Edwards. [78] The lack of significant redistribution of all the acceptor and donor species studied suggests that ion implantation will be a viable technology for controllable doping of GaN. Furthermore, due to the lack of diffusion, external source diffusion appears not to be practical in GaN. It should also be noted that, with the exception of Be and to a lesser extent O, none of the as-implanted profiles had significant tailing due to channeling. Therefore, implantation of Si, Mg, Ca, and Zn in GaN can be used to define shallow, abrupt doping profiles. This is particularly important when these implants are applied to FET structures such as JFETs or MESFETs.[ 54][84]

10 20 . . . . . . .

, 9 9 9, . . . . . . as-implanted

1019 r o

~ 1018 t.J

o

~

101~

1016

9

0

.

.

!

0.2

,

.

,

i

0.4

.

,

,

|

,

.

0.6

.

|

0.8

9

9

9

1

depth (~tm)

Figure 22.

SIMS profile o f 9Be-implanted (45 keV, 5 x 10 TMcm -2) GaN as-implanted and annealed at 1125~ for 15 s. No measurable redistribution is seen at the peak of the annealed sample (after Ref. 79).

330

Wide Bandgap Semiconductors

1020

9

,

9

i

,

,

+ 'N

9

I

9

9

9

i

,

9

9

i

,

,

[ __-~_..as-implantedl 125 ~

1019

o

~

1018

u Q

~

N

m

1017

|!

i

,, i

1016 0

................... 0.2 0.4

0.6

0.8

depth (~tm) Figure 23. SIMS profile of 64Zn-implanted (260 keV, 5 x 1014cm -2) GaN as-implanted and annealed at 1125~ for 15 s. No measurable redistribution is seen at the peak of the annealed sample (after Ref. 79).

1 0 20

9

,

, - T

- - , - - =

|

,

-.....

9

9

!

9

9

9

I

9

as-implanted 25 ~

o

1019 o

o

o o o

1018

1017 . . . . . . . . . . . . . . . . . . . 0 0.2 0.4 0.6

0.8

depth (j.tm) Figure 24. SIMS profile of 4~ (180 keV, 5 x 1014cm -2) GaN as-implanted and annealed at 1125~ for 15 s. The slight shift of the profile into the bulk for the annealed sample is within the resolution of the SIMS measurement and probably results from experimental variation in measuring the sputtered depth (after Ref. 47).

lon Implantation 331 4.2

SiC

Redistribution of implanted dopants in SiC has also been studied. Kimato reported SIMS profiles as shown in Figs. 25 and 26 for multiple energy A1 or B implants in 6H-SiC either as-implanted or annealed (1500~ 30 min).[ 67] This work supports the results of others that A1 and B display limited redistribution at 1500~ 681 Implanted N in SiC has also been shown to not significantly diffuse, up to 1500~ [63][85] However, in an earlier work, Ryu reported that both B and N started to diffuse towards the surface at 1600~ and diffused completely out of the sample during annealing at 1800~ 86]Ryu' s results were reported at a time (circa 1989) when the SiC material had large micropipe (vertical defects with hollow cores) densities that may have contributed to enhance diffusion. However, that work suggests that as implantation activation temperatures are pushed higher, impurity redistribution may become more significant.

""/U22E

1

1

I

I i

1 I

J I

I

1

!

AI + -~ 6 H - S i C 180keY, 2.7x1015cm -2 --= lOOkeV, 1.4x1015cm -2 = 50keV, 9.0x1014cm -2 -

1 0 21 _ r I

E 0 Z

020 ~ , , ~ ' ~ ..v

0

~

1019_ Z

I-Z LU

o10

18

Z

0 ~0._1017 <

"

annealed 9 (1500%) 1016..

0

Figure 25.

I

I

I

! I I l,. I I i 0.2 0.4 DEPTH (# m)

I

-

I.. I0.6

SIMS profiles for Al-implanted 6H-SiC at the conditions shown either asimplanted or after a 1500~ 30 s anneal (after Ref. 63).

332

Wide Bandgap Semiconductors

10 ~

10 21 I

~ ' , ,

E

B + --* 6H-SiC 100keV, 2.8x1015e~n-2 -_ 60keV, 1.4x1015cm -2 30keV, 8.0• -2 "

o,~ 1020 z

0

e

1019 " I-Z LU 1018 = Z 0 0 1017: rn

1016~ 0

..-~ .=

...........

as-implanted 9

~

annealed (1500"C)

" 0.2

DEPTH

0.4

-

0.6

(/z m)

Figure 26. SIMS profiles for B-implanted 6H-SiC at the conditions shown either as

implanted or after a 1500~ 30 s anneal (after Ref. 63).

4.3

Diamond

Comprehensive results for the depth distribution and range parameters for a large number of elements implanted into single-crystal diamonds and chemically vapor-deposited polycrystalline diamond films have been reported.[ 87] No results were identified, however, for the redistribution of these impurities during annealing. Based on the nature of impurity redistribution in the other wide bandgap semiconductors discussed in this chapter which also have high bond energies, redistribution of impurities in diamond can be expected to be very low. This is supported by the lack of success in external source diffusion in diamond.[ 69][7~

lon Implantation 333 5.0

I M P L A N T A T I O N D A M A G E : C R E A T I O N AND REMOVAL

During the ion implantation process, damage is introduced into the semiconductor crystal via electronic and nuclear interactions. Typically, this damage is largely removed during the implant activation annealing process. However, if the amount of damage is sufficient to amorphize the semiconductor, complete recovery of the crystal lattice is often not possible. This is particularly true for compound semiconductors such as GaAs and InP where amorphization must be avoided if electrical activity of implanted dopants is desired.[ 88]Therefore, it is important to determine the implantation conditions that amorphize the wide bandgap semiconductors.

5.1

GaN

The first work on implantation induced damage in GaN was reported by Tan and eoworkers for 90 keV Si-implantation at 77 K.[89]Implantation at low temperatures limits the dynamic annealing that can occur and tends to produce amorphization at a lower dose than if the sample was at room temperature. That is, from an amorphization perspective, low temperature implantation represents a worst ease scenario. As shown in Fig. 27, Tan and coworkers demonstrated that at 77 K, even for a high dose of 7.2 • 1015 cm -2, the GaN sample was not amorphized while a GaAs sample, also implanted at 77 K, was amorphized at a dose two orders-of-magnitude lower (8 x 1013cm-2).[9~ This result suggests that ion implantation can be used to achieve very high Sidoping levels in GaN, since this dose corresponds to an estimated peak Si concentration of--5 x 1020em -3 without the sample being amorphized. Such high doping levels can be expected since epitaxial Si-doping of GaN has achieved donor concentrations in excess of 1020 cm-3.[47][91]In fact, recent results for high-dose (1 x 1016 cm -2) Si-implantation in GaN have demonstrated that such high donor levels are achievable with implantation.j92][ 93J While the high amorphization level of GaN is encouraging, the study of damage removal during annealing must also be addressed. Figure 28 shows RBS spectra for Si-implanted GaN, now implanted at room temperature, either as-implanted or after a 1100~ 30 s anneal. Although there is a reduction in the backseattering yield after annealing that might be construed as a reduction in the damage created by the implantation, when changes in surface dechanneling are accounted for, no measurable

334

Wide Bandgap Semiconductors

decrease in the buried damage level occurred due to the anneal.[ 94] This suggests that higher annealing temperatures may be needed to optimize the electrical transport in the implanted layer.[ 95]However, it has also been shown that complete removal of the implant damage in GaN is not a requirement for successful activation of the implanted dopants, although further damage reduction may be needed to optimize the transport properties of the implanted layer.[ 95]This is in contrast to other compound semiconductors where damage removal and dopant activation are serial processes.[ 96]

60

1

-

50 "0 I,,,=,,I o

40

"0 o N

30

ot,~l

E ~O

~_, ~(,4k,=

I

unimplanted

I

-" 7.2x10 ~s cm "a

[~ 8.4xl 014 cm-2 ~ 2.4xl 018 cm-2 ] --~--2.2x1015 cm "2 --xT-- rand~

20 10

0.5

1

1.5

2

energy (MeV)

Figure 27. Ion channeling spectra (2 meV He+ ions) illustrating the build up of disorder in 90 keV Si-implanted GaN at 77 K. The dose for each spectra is shown in the legend (after Ref. 90).

5.2

SiC

Significant work has been done on implantation-induced damage in SiC. Some of the early work focused on how implantation could be used to alter the mechanical properties ofSiC.[97][98] Willams showed that SiC was amorphized by a Cr-implant dose of 2.9 x 1014 cm -2 at 260 keV, while lighter N-ions required a dose of 2.7 x 1015 cm-2 at 62 keV to achieve a similar damage level.[ 97] In that work, it was also reported that SiC swells when amorphized with step heights on-the-order-of 50 nm for a N-dose of

Ion Implantation 335 --2 x 1016 c m -2. Interestingly, the step height was not linear with dose but displayed a dramatic step near the crystalline-to-amorphous dose regime. The authors attributed these results to in-situ annealing during the implant, making the damage unstable during bombardment until the amorphous zone was formed. Dose levels for amorphization were also established for 100 keV Ar (4 x 1014 c m -2) and Xe (~8 x 1013 c m -2) implants and are consistent with the Cr and N results based on the atomic mass of these ions. [99]

60

.

.

.

.

I

~i~

'

'

'

i

'

'

'

I

.

.

.

.

~ unimplanted --'~-- 6x 10 ~5 cm2: as implanted _-l~6x 1015 cm2:1100 ~ 30 s

50 40 N

30

@

20

9 !,,,~

10 0

_

,m

t

0

0.5

1

1.5

energy (MeV) Figure 28. Channeling Rutherford Backseattering (C-RBS) spectra for as-grown (random and aligned, unimplanted) and Si-implanted (90 keV, 6 x 10 Is cm -2) GaN (as-implanted and after a 1100~ 30 s anneal). The implants were performed at room temperature (after Ref. 92).

More recent work examined the damage build-up during the implantation of dopant species in SiC. Chechenin and coworkers reported RBS spectra for room temperature Al-implanted SiC versus dose at 40 keV.[ 1~176 They found the formation of an amorphous region for a dose of 1 x 1015cm-2 (as shown in Fig. 29) with damage below the amorphous regime being effectively removed by annealing at-1800~ for 5 s at 5 atm of Ar. Kimoto et al. reported room temperature implants of A1 and B into 6H-SiC with amorphization levels of 1 and 5 x 1015 cm -2, respectively, for multiple implant energies from 30-180 keV.[ 67]P-type conductivity was measured

336

Wide Bandgap Semiconductors

for the Al-implanted material with a resistivity of 22 kDJ while B-implanted materials had high resistivity that made carrier type identification difficult. The difficulties with B may be due to its high ionization energy as discussed in Sects. 3.1 and 3.4.

321 '

1

I

1

2~

!

AI

R7 R6 j

o

"-9 16 I--

0

~

(

~,5

o

i..l,a

A/+

0 50

100

150

200

250

CHANNEL NUMBER Figure 29. The aligned (A1-A6) and random (R6,R7) C-RBS spectra of unimplanted (A1, R7) and 90 keV Al-implanted SiC with doses of 3 x l013 (A2), 1 x 1014 (A3), 3 x 1014(A4), 1 x 1015 (A5), and 3 x l016 (A6, R6) cm-2 (after Ref. 100).

Although there has been success using room temperature implantation and very high annealing temperatures, significant efforts have focused on the use of elevated temperature implants in SiC to limit the damage accumulation and promote impurity activation. Some of the first work in this area was reported by Edmond, as discussed previously in the doping Sect. 3.4, and been further developed by Rao .[6~ Figure 30 shows RBS spectra from Rao for Al-implanted (200 keV, 8 x 1014 cm -2) SiC done at 850~

Ion Implantation 337 as-implanted and after annealing at 1100 or 1400~ [66]Samples implanted at the same dose and energy, but at room temperature, were amorphous and did not recover their initial crystallinity after annealing at 1400~ However, as seen in Fig. 30, the elevated temperature sample was not amorphizcd for this dose and approached the virgin channeling yield after annealing at 1400~ In this study, p-type conductivity was only achieved for the elevated temperature implanted A1 sample and not for those implanted at room temperature. The lack of p-type conductivity in the room temperature samples was attributed to residual implantation damage compensating the Al-acceptors.[ 66]

DEPTH (microns) 1.25

10 a

1.00

0.75

0.50

AI+(200 keV, 8xl0Zac'm-Z, 850~

0.25

0.00

SiC b u l k RANDOM

m

v

102

6 +..

',

10 l

++ 4~" 9

~ +

l 10 o

L

AS-IMPIANTED ANNEALED ! 400oc

* ANNEALED ll00~ i

0.4

I

I

0.6

I

|

I

I

0.B

1.0

1.2

SCATTERED ENERGY (MeV) Figure 30. C-RBS spectra for Al-implanted (200 keV, 8 x 1014cm "2) 6H-SiC performed at

850~ before and after annealing at the temperatures shown (after Ref. 66).

338 5.3

Wide Bandgap Semiconductors Diamond

Ion-beam-induced amorphization of diamond, followed by annealing, has been shown to form a graphite region.[ 1~ Although the amorphous layer can be used as a preferential etching layer (the amorphous layer etches more rapidly than the crystalline diamond), when implantation doping is desirable, amorphization should be avoided. Several reports exist on the dose and energy threshold for amorphization as summarized in Table 4.[101 ]-[105]T h e authors pointed out that the amorphization level depends on the nuclear stopping (S~); and that the product of the critical dose and the nuclear stopping was roughly constant between -~2 and 4 eV i o r g ~ 3 . [101]

Table 4. List of Critical Doses to Amorphize Diamond (after Ref. 101).

Ion

Energy (keV)

Critical Dose, ( D o) (ion/cm 2)

Nuclear energy loss,

Electronic energy loss,

(ev/A)

(s~ (ev/A)

DcSn

Reference

76.7

2.0

98

100

2.5 x 1015

Ar

40

3.7 x 1014

117.1

63.6

4.3

99

Sb

340

1.0 x 10 TM

341.2

135.8

3.41

100

Xe

320

2.0 x 1014

132

370

2.64

101

C

6.0

8.1

DEVICE D E M O N S T R A T I O N S

With the advances in the science and technology of ion implantation in the wide bandgap semiconductors discussed in the previous sections, implantation is now finding application in electronic devices. In this section, key device demonstrations in each material system that employ ion implantation are presented. These results demonstrate the true utility of implantation but are by no means to be considered exhaustive in their scope. Future devices will most likely make even more use of implantation doping to realize more advanced device structures.

lon Implantation 339 6.1

GaN

The first GaN-based devices to use ion implantation were transistors that incorporated implant isolation. This was first done with H-implantation and later with He.[9][29] Such implant isolation allows planar device topologies which facilitates device interconnection and circuit fabrication. More advanced implanted devices have also been demonstrated, with all doping done by implantation in a GaN junction field effect transistor (JFET).[ 111Figure 31 shows a schematic representation of the process flow used to fabricate the GaN JFET in semi-insulating GaN grown by metal organic chemical vapor deposition (MOCVD) on (1000) A1203 .[1~ The key processing steps are as follows: 1. Selective area ion implantation of the n-channel (28Si: 100 keV, 2 x 1014cm-2) and p-gate 4~ (40 keV, 5 x 1014cm-2) 2. Sputter deposition of 300 nm of W gate contact metal 3. Reactive ion etching (RIE) gate contact patteming using an SF 6/Ar plasma 4. Selective area, non-self-aligned 28Si ion implantation of the source and drain regions 5. A 1150~ 15 s rapid thermal anneal to activate the implanted dopants 6. Electron cyclotron resonant (ECR)-plasma etching of ~50 nm of p-GaN from the source and drain regions using a BC13/H2/Ar chemistry [1~ 7. Deposition of Ti/A1 (20 nm/200 nm) ohmic metal 8. 500~

15 s ohmic alloy

This structure minimizes the gate capacitances often associated with JFETs by self-aligning the p-type gate to the gate contact metal.[ 1~ In addition, since the doping was done in selective areas, device isolation was realized via the semi-insulating properties of the GaN substrate. That is, no implant isolation or mesa etch isolation was required to isolate these devices. Figure 32 shows the Ios versus Vos curves for varied gate biases for a ~1.7 ktm x 50 ~tm CmN JFET with a 4 ~tm source-to-&ain spacing. The JFEF demonstrates good modulation characteristics with nearly complete pinch-off at a threshold voltage of approximately -6 V for Vos=--7V. For Vos= 25 V, a maximum transconductance of 7 mS/mm was measured at Vcs = -2.0 V with a saturation current of 33 mA/mm at VGS= 0 V. Four-probe measurements of

340

Wide Bandgap Semiconductors

the source resistance gave R s ~ 500 f~. Although this value o f R s is extremely large, it only accounts for a 20% reduction in the external transconductance with respect to a corrected internal transconductance of 8.5 mS/mm. This high resistance is attributed to the region between the ohmic contact and the channel. Transmission line method (TLM) test structures using the same source and drain implants on C_mNwitness pieces gave a value of the specific contact resistance of -~1 x 10.5 f~-cm2 (-~1 ~')-mm). [77]This access resistance can be substantially reduced by optimizing the source and drain implant and anneal conditions as well as by self-aligning these implants to the gate contact metal. A second possible cause of the low transconductance is low electron mobility in the implanted channel region. If this is the case, optimization of the implant activation process should lead to improved mobilities. In addition, optimization of the epitaxial GaN layers for maximum electron mobility, as has been done for epitaxial FETs, should result in improved JFET performance.[ 1~ This device demonstrated a unity current gain cutoff frequency 0rt) of 2.7 GHz and a maximum oscillation frequency (fm~) of 9.4 GHz. These frequency metrics are comparable to similar gate length epitaxial GaN transistors.[~ ~0] photoresist (a)

n-channeland

28Si + 40Ca

~//~/~ I I.I, I~_.///'~ undoped GaN

(b)

28Si

gate contactdepostionand~_~ definition, ~ source and drain implants, and RTA

n-type

P"~Ype 28Si ~$

~.~ n+ source or drain n-type channel

gate contact (c) plasma etch of p-type GaN from source and drain regions

~ ~ N ~ \ \ \ \p'typr \~

n+ source or drain

n-type channel gate contact \ p-type (d) T~;A! % . 7 Ti/gi ohmic contact deposition ~ ~ , ~ ~ ~ ~ ~ (e.g. Ti/AI) and alloy ~

r--

",,,---

n+ source

ordrain

n-type channel

Figure 31. Schematic representation of the processing steps for fabricating an all ion implanted GaN .fF'ET(after Ref. 44).

lon Implantation 341

30

............................................................................................................................. i................................... i i ~ i

~

!

i

:.

15 10

..........

i

~

j

i

.

~. . . . . . .

i ................

t .....................................

!..............................

.

5 0

0

5

10

15

20

25

VDS (V) Figure 32. Ios versus Vos for a - - 1 . 7 lxm x 50 ~tm G a N JFET. Gate bias starts at 0 V with - 1 V steps (after Ref. 11).

6.2

SiC

Significant progress has been made in the use of implantation in SiC devices. First among these are implanted diodes which were reported by Marsh and Dunlap in 1970.[ 58] Marsh implanted N into Al-doped bulk crystal and achieved rectifying characteristics after annealing at 1100~ that were consistent with a p-i-n structure. The diode behavior was enhanced by higher temperature annealing (up to 1500~ with the reduction of the thickness of the intrinsic layer. Vodakov et al. were also successful in diode fabrication by implanting A1 at room temperature into unintentionally doped n-type SiC and realized mesa diodes with breakdown voltages (defined a t / = 1 ~tA) up to 280 V.[ 111] Edmond reported the use of elevated temperature (600~ implants of A1 into n-type 3C-SiC and N into p-type SiC to achieve diodes with capacitance/voltage characteristics indicative of abrupt junctions.[ 112]The samples were annealed at 1200~ for 30 min. The diodes maintained rectifying properties to 400~ but showed evidence of trap assisted conduction. Improved implanted diode behavior was reported by Ghezzo for both polarities (N-implanted into p-SiC and B-implanted into n-SiC).[ 113][114]

342

Wide Bandgap Semiconductors

For N-implanted diodes, a minimum sheet resistance was achieved for an implantation temperature of 1000~ and an anneal temperature of 1300~ The diodes had leakage currents < 100 nA/cm 2 and room temperature reverse breakdown voltages of 95 V.[ zz3] The 1000~ B-implanted diodes were annealed at 1200~ for 5 hrs and had very low leakage currents (4 nA/cm 2) with a reverse breakdown voltage of 650 V at 10 mA for a 2x2 mm 2 diode. A schematic of an implanted SiC MESFET fabricated by Lam is shown in Fig. 33.[ 4o] This device makes use of elevated temperature (650~ N-implantation to form n+-source and drain regions. Vanadium, which acts as a compensating deep level in SiC, was also implanted to form high resistivity regions outside the active transistor area. The transistor DC characteristics are shown in Fig.. 34 for a 5 pm x 150 pm device. The low knee voltage in Fig. 34 is largely the result of the high source and drain doping achieved with the N-implantation. This work is an extension of earlier work by this group on implanted SiC nMOS devices.[ 115]The diode and transistor results just discussed, clearly demonstrate the utility of implantation doping in SiC and should lead to more advanced device structures (such as thyristors) being realized with implanted dopants. Ni interconnect VG

VS

Ni Schottky VG

b,\\'q

~

K'~ _ k\\\X\\'q

doped

(a)

N+

N-

Ni ohmic VD

N

V

P SiC (5xlO15cm"3)

Ni Schottky Gate

lx\\\\\\\\\\\\\\\\\\'~\\\\\\'q

.. (b)

)v

P SiC (5x1015cm "3) P+ SiC (1.7xl01ecm "3)

AI ohmic

Figure 33. Cross-section along the channel length (a) and along the channel width (b) of a SiC MESFET that employed N-implantation for source and drain formation and Vimplantation for isolation (after Ref 40).

Ion Implantation

250 2 2~

343

Vo.~

W=12011m, L=51~m

"-'200

150

100

a

50 ~ 0

10

20

30

40

50

60

p1.0V 70

80

Drain Voltage (V)

Figure34. los versus lidsfor a 5 ~tmx 120 ~tmSiC MESFET shown schematicallyin Fig 33. Gate bias starts at 1.0 V with -0.5 V steps (after Ref. 40).

6.3

Diamond

Although there remain many issues in material quality and controllable doping of diamond, some preliminary devices have been demonstrated using ion implantation. Figure 35 shows the current/voltage characteristics of a diamond diode formed by ion implanting carbon to form a n-type damage layer in naturally p-type diamond.[7~ 1161The data shown in Fig. 35 show results for two different implantation processes. Diode 1 was implanted with a target temperature of 300~ while diode 2 was formed using the CIRA process (77 K implant followed by a 1200~ anneal) outlined previously in Sect. 3.5. Diode 2 demonstrated a significant decrease in the forward on resistance with a voltage drop of only 4 V for a forward current of 1.5 mA as opposed to ~9 V for the same current in diode 1. Diode 2 also had a reduced reverse current that is ascribed to an improved defect structure resulting from the CIRA process. Furthermore, diode 2 also demonstrated stronger luminescence in the blue region of the spectrum than diode 1, as shown in Fig. 36.[7~ 116]Blue emission from a diode in a semiconductor with a bandgap of 5.45 eV is most likely due to a donor/acceptor recombination process or other defect mitigated recombination. A diamond bipolar transistor was demonstrated for the first time in 1982 and used C-implantation at 320~ to form the emitter and collector in p-type type IIb diamond.[ 1171Transistor action was achieved as shown in

344

Wide Bandgap Semiconductors

Fig. 37. The high collector/emitter turn-on voltage of~l 0 V required to reach the linear portion of the transistor's curves was attributed to high contact resistance.j70][116] Significant improvements in device operation were predicted if the advanced implanted schemes described in this chapter are employed along with improved ohmic contacts on a more optimum device design.

1.5'

< E

1.0

Z UJ n" rr

iiiIII 0.5

,'k / DIODE 1

ii I _

-8

-4

0

4

8

POTENTIAL (VOLT)

Figure 35. Current versus voltagefor diodes in type IIb diamond formed by implantation at 300~ (diode 1) and with the CIRA process (diode 2) (afterRefi 116).

The first diamond field effect transistor to employ ion implantation was reported in 1991 by Zeisse and coworkers.[ lls] This device used multiple energy B-implantation at 80 K into natural type IIa diamond to form the channel and contact regions. The implants were annealed at 1263 K. A SiO 2 gate insulator was deposited by indirect plasma-enhanced chemical vapor deposition upon which a Ti/Au gate electrode was defined by liftoff. The transistor was configured in a concentric ring geometry (a gate ring between the source and drain rings) so that no lateral isolation was required. Saturation and pinchoff were observed with a transconductance of 3.9 ~S/mm. These diamond device results are encouraging with improved device performance expected as further advances in implantation and material technology occur.

lon Implantation 345

lODE 2

m

03

Z iii

I-Z

1

iii

III

217

2.3

3'.1

3.s

PHOTON ENERGY (eV)

Figure 36. Photoluminescence spectra from the same diodes as in Fig. 35. Diode 1 was pulsed at a forward bias of 80 V while diode 2 was operated under a constant bias of 20 V (after Ref. 116).

0.6

IB=5mA

< E z uJ nIT" 0 nO F0 UJ -J .J

~

IB=4mA

0.4

~

IB=3mA IB=2mA

0.2 _

~

IB--lmA

0 o

IB=0mA 20

40

60

80

COLLECTOR-EMITTER- POTENTIAL.(Volt) Figure 37. Collector current (lc) versus collector-emitter voltage (Vce,) for the base current (I~ levels shown for a lateral n-p-n transistor in type lib diamond fabricated with carbonion implantation (after Ref. 117).

346 7.0

Wide Bandgap Semiconductors F U T U R E W O R K AND C O N C L U S I O N S

Although significant progress has been reported for ion implantation doping and isolation of wide bandgap semiconductors, there are still many areas for further research. Areas of interest for implantation doping in GaN include determining the maximum attainable doping levels for various elements, optimizing the activation annealing process to maximize dopant activation and minimize surface degradation, and understanding of the dynamics of the defect generation and removal process. Areas of interest for implant isolation in GaN include the demonstration of truly thermally stable isolation, understanding the nature of hydrogen implant isolation, determination of the behavior of deep levels such as Cr and Fe, and the effect of implant isolation on the properties of waveguide losses. For SiC implantation, continued work is needed to better understand the trade-off between implantation temperature and the activation annealing process. The maximum doping levels should also be more clearly defined and alternative dopants, particularly for acceptors, should be considered. More work is need to optimize implantation isolation in SiC either with a deep impurity such as vanadium or with a damage related energy level. Diamond implantation has the farthest to come of these wide bandgap semiconductors. This is largely due to the metastable nature of the diamond/graphite phases which can be created by the implantation and annealing process. Continued studies of the implantation induced defects in diamond are important, along with exploring novel co-implantation or predamage techniques. While implantation in the wide bandgap semiconductors is far from a mature technology, it also may be the only way to achieve external doping (either of a blanket or selective area nature), due to the low diffusivities of impurities in these materials. Moreover, if the success of implantation in silicon and gallium arsenide technologies is any guide, the possibilities are extensive for the application of ion implantation to devices in the wide bandgap materials. Therefore, continued development of this technology can be expected to yield significant device improvements.

Ionlmplantation

347

ACKNOWLEDGMENT The author would like to thank his collaborators on various aspects of this work, including: C. R. Abemathy, J. Avery, A. G. Baca, M. Hagerott Crawford, J. Escobedo, G. Lopez, S. J. Pearton, R. J. Shul, R. A. Stall, H. H. Tan, J. S. Williams, and R. G. Wilson. Additional thanks is due to colleagues who supplied reprints and figures including: R. F. Davis, M. Melloch, J. F. Prins, M. V. Rao, and R. G. Wilson. Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy under contract #DE-ACO4-94AL85000.

REFERENCES 1. Yoder, M. N., IEEE Trans. Elec. Dev., 43:1633 (1996) 2. Davis, R. F., Special Issue on Large Bandgap Electronic Materials and Components, Proceed. IEEE, vol. 79 (1991) 3. Morkoc, H., Strite, S., Gao, G. B., Lin, M. E., Sverdlov, B., and Bums, M., J. Appl. Phys., 76:1363 (1994) 4. Akasaki, I., Amano, H., Kito, M., and Hiramatsu, K., J. Lumin., 48/49:666 (1991) 5. Nakamura, S., Mukai, T., and Senoh, M., Appl. Phys. Lett., 64:1687 (1994) 6. Nakamura, S., Senoh, M., Nagahama, S., Yamada, T., Matsushita, T., Kiyoku, H., and Sugimoto, Y., Jap. ,I. Appl. Phys., 35 :L74 (1996) 7. Nakamura, S., Senoh, M., Nagahama, S., Iwasa, N., Yamada, T., Matsushita, T., Sugimoto, Y., and Kiyoku, H.,Appl. Phys. Lett., 69:1477 (1996) 8. Nakamura, S., MRS Bulletin, 22:29 (1997) 9. Khan, M. A., Bhattarai, A., Kuznia, J. N., and Olson, D. T., Appl. Phys. Lett., 63:1214 (1993) 10. Binari, S. C., Rowland, L. B., Kruppa, W., Kelner, G., Doverspike, K., and Gaskill, D. K., Elect. Lett., 30:1248 (1994) 11. Zolper, J. C., Shul, R. J., Baca, A. G., Wilson, R. G., Pearton, S. J., and Stall, R. A., Appl. Phys. Letts., 68:2273 (1996) 12. Shur, M. S., and Khan, M. A., MRS Bulletin, 22:44 (1997) 13. Weitzel, C. E., Palmour, J. W., Carter, C. H., Jr., Moore, K., Nordquist, K. J., Allen, S., Thero, C., and Bhatnagar, M., IEEE Trans. Elec. Dev., 43:1732 (1996) 14. Neudeck, P. H., J. Electron. Mater., 24:283 (1995)

348

WideBandgap Semiconductors

15. Casady, J. B., and Johnson, R.. W., Solid-State Electronics, 39:1409 (1996) 16. Capano, M.A., andTrew, R. J.,MRS Bulletin, 22:19 (1997) 17. Trew, R. J., Yan, J. B, and Mock, P. M.,Proceed. IEEE, 79:598 (1991) 18. Some examples of implantation in wide bandgap II-VI semiconductors can be found in: Wilson, R. G., Solid-State Electronics, 39:1113 (1996); Sigmon, T. W., Nucl. Instr. Meth. Phys. Res. B, 7/8:402 (1985); Khalal, K., Chami, A. C., Ligeon, E., Fontenille, J., Hamoudi, A., Berard, G., and Cibert, J., 3`. Appl. Phys., 78:37096 (1995) 19. See for example: Ziegler, J. F., Handbook of Ion Implantation Technology, Elsevier Science Publishers, The Netherlands, pp. 271-362 (1992); Ghandhi, S. K., VLSI Fabrication Principles: Silicon and Gallium Arsenide, Ch. 6, John Wiley and Sons, New York, NY, (1983) 20. Pearton, S. J., Mat. Sci. Rep., 4:313 (1991) 21. Orenstein, M., Stoffel, N. G., Von Lehmen, A. C., Harbiunson, J. P., and Florez, L. T., Appl. Phys. Lett., 59:31 (1991) 22. Lear, K. L., Schneider, R. P., Choquette, K. D., Kilcoyne, S. P., Figiel, J. J., and Zolper, J. C., IEEE Photonic Tech. Lett., 6:1053 (1994) 23. Zolper, J. C., Baca, A. G., and Chalmers, S. A., Appl. Phys. Lett., 62:2536 (1993) 24. Pearton, S. J., Iannuzzi, M. P., Reynolds, C. L., Jr., and Peticolas, L, Appl. Phys. Lett., 52:395 (1988) 25. Zolper, J. C., Sherwin, M. E., Baca, A. G., and Schneider, R. P., Jr., J. Elec. Mat., 24:21 (1995) 26. Khan, M. A., Skogman, R. A., Schulze, R. G., and Gershenzon, M., Appl. Phys. Lett., 42:430 (1983) 27. Khan, M. A., Skogman, R. A., Schulze, R. G., and Gershenzon, M., Appl. Phys. Lett., 43:492 (1983). 28. Pearton, S. J., Abemathy, C. R., Vartuli, C. B., Zolper, J. C., Yuan, C., and Stall, R. A., Appl. Phys. Lett., 67:1435 (1995) 29. Binari, S. C., Dietrich, H. B., Kelner, G., Rowland, L. B., Doverspike, K., and Wickenden, D. K., J. Appl. Phys., 78:3008 (1995) 30. Haase, D., Schmid, M., Kurner, W., Dornen, A., Harle, V., Scholz, F., Burkard, M., and Schweizer, H., Appl. Phys. Lett., 69:2525 (1996) 31. Pearton, S. J., Abemathy, C. R., Wisk, P. W., Hobson, W. S., and Ren, F., Appl. Phys. Lett., 63:1143 (1993) 32. Zolper, J. C., Pearton, S. J., Abemathy, C. R., and Vartuli, C. B., Appl. Phys. Lett., 66:3042 (1995) 33. Pearton, S. J., Abernathy, C. R., Panish, M. B., Hamm, R. A., and Lunardi, L. M., 3". Appl. Phys., 66:656 (1989)

Ion Implantation 349 34. Zolper, J. C., Pearton, S. J., Abemathy, C. R., and Vartuli, C. B., Conf. Proc. MRS Spring Meeting 1995, Symposium B, p. 485, Materials Research Society, Pittsburgh, PA, (1995) 35. Zolper, J. C., Crawford, M. H., Pearton, S. J., Abemathy, C. R., Vartuli, C. B., Yuan, C., and Stall, R. A., J. Electron. Mat., 25:839 (1996) 36. Nadella, R. V., and Capano, M. A., Appl. Phys. Lett., 70:886 (1997) 37. Hobgood, H. M., Glass, R. C., Augustine, G., Hopkins, R. H., Jenny, J. R., Skowronski, M., Mitchel, W. C., and Roth, M., Appl. Phys. Lett., 66:1364 (1995) 38. Jenny, J. R., Skowronski, M., Mitchel, W. C., Hobgood, H. M., Glass, R. C., Augustine, G., and Hopkins, R. H., J. Appl. Phys., 78:3839 (1995) 39. Kimoto, T., Naklima, T., Matsunami, H., Nakata, T., and Inoue, M., Appl. Phys. Lett., 69:1113 (1996) 40. Lam, M. P., Komegay, K. T., Cooper, J. A., Jr., and Melloch, M. R., IEEE Trans. Elec. Dev., 44:907 (1997) 41. Kalish, R., Uzan-Sagay, C., Philosoph, B., Richter, V., and Prawer, S., Appl. Phys. Lett., 70:999 (1997) 42. Prins, J. F., Phys. Rev. B, 44:2470 (1991) 43. Sze, S. M., Physics of Semiconductor Devices, 2nd Ed., pp. 17-24, John Wiley, New York, NY, (1981) 44. Zolper, J. C., and Shul, R. J., MRS Bulletin, 22:36 (1997) 45. Pankove, J. I., and Hutchby, J. A., J. Appl. Phys., 47:5387 (1976) 46. Pankove, J. I., and Hutchby, J. A., Appl. Phys. Lett., 24:281 (1974) 47. Kim, J. G., Frenkel, A. C., Liu, H., and Park, R. M.,Appl. Phys. Lett., 65:91 (1994) 48. Cummings, K. D., Pearton, S. J., and Vella-Coleiro, G. P., J. Appl. Phys., 60:166 (1986); Hiramoto, T., Saito, T., and Ikoma, T., Jpn. J. Appl. Phys., 24:193 (1985); Barrett, N. J., Grange, J. D., Sealy, B. J., and Stephens, K. G., J. Appl., Phys., 56:3503 (1984); Kular, S. S., Sealy, B. J., Ono, Y., and Stephens, K. G., Solid-State Elec., 27:83 (1984); Pearton, S. J., Williams, J. S., Short, K. T., Johnson, S. T., Jacobsen, D. C., Poate, J. M., Gibson, J. M., and Boerma, D. 0., J. Appl. Phys., 65:1089 (1989) 49. Tan, T. Y., and Gosele, U.,Appl. Phys. Lett., 52"1240 (1988) 50. Zolper, J. C., Wilson, R. G., Pearton, S. J., and Stall, R. A., Appl. Phys. Lett., 68:1945 (1996) 51. Chung, B.C., and Gershenzon, M., J. Appl. Phys., 72:651 (1992) 52. Sato, H., Minami, T., Yamada, E., Ishii, M., and Takata, S., J. Appl. Phys., 75:1405 (1994) 53. Patel, K. K., and Sealy, B. J. Appl. Phys. Lett., 48:1467 (1986)

350

Wide Bandgap Semiconductors

54. Sherwin, M. E., Zolper, J. C., Baca, A. G., Drummond, T. J., Shul, R. J., Howard, A. J., Rieger, D. J., Schneider, R. P., and Klem, J. F., J. Elec. Mater., 15:809 (1994) 55. Zolper, J. C., Baca, A. G., Sherwin, M. E., and Shul, R. J., Electron. Letts., 31:923 (1995) 56. 57. 58. 59.

60.

61. 62. 63. 64. 65. 66. 67. 68. 69.

70. 71. 72. 73.

Strite, S.,Jpn. J. Appl. Phys., 33:L699 (1994) Dunlap, H. L., and Marsch, O. J., Appl. Phys. Lett., 15:311 (1969) Marsh, O. J., and Dunlap, H. L., Rad. Effects, 6:301 (1970) Marsh, O. J.,Silicon Carbide 1973, p. 471, (R. C Marschall, J. W. Faust, Jr., and C. E. Ryans, eds.), University of South Carolina Press, Columbia, SC (1974) Edmond, J. A., Withrow, S. P., Wadlin, W., and Davis, R. F., Conf. Proc. Mater. Res. Soc., 77:193, (J. Dow, T. K. Schuller, and J. Hillard, eds.), Material Research Society, Pittsburgh PA (1987) Edmond, J. A., Davis, R. F., and Withrow, S. P., Ceramic Transactions, 2:479 (1989) Hirano, Y., and Inada, T., J. Appl. Phys., 77:1020 (1995) Kimoto, T., Itoh, A., Matsunami, H., Nakata, T., and Watanabe, M., J. Electron. Mat., 24:235 (1995) Pan, J. N., Cooper, J. A., Jr., and Melloch, M. R., J. Electron. Mat., 26:208 (1997) Gudkov, V. A., Krysov, G. A., and Makarov, V. V., Soy. Phys. Semicond., 20:105 (1986) Rao, M. V., Griffiths, P., Holland, O. W., Kelner, G., Freitas, J. A., Jr., Simons, D. S., and Chi, P. H., J. Appl. Phys., 77:2479 (1995) Kimoto, T., Itoh, A., Matsunami, H., Nakata, T., and Watanabe, M., J. Electron. Mat., 25:879 (1996) Rao, M. V., Griffiths, P., Gardner, J., Holland, O. W., Ghezzo, M., Kretchmer, J., Kelner, G., and Freitas, J. A., Jr, J. Electron. Mat., 25:75 (1996) Prins, J. F., "Application of Diamond Films in Electronics," in: "The Physics of Diamond," Proceedings of the International School of Physics, Enrico Fermi, Varenna, 1996, Course CXXXV, IOS Press, Amsterdam (1997) Prins, J. F., Mat. Sci. Rep., 7:271 (1992) Prins, J. F., Phys. Rev. B, 38:5576 (1998) Fontaine, F., Uzan-Saguy, C., Philosoph, B., and Kalish, R., Appl. Phys. Lett., 68:2264 (1996) Sandhu, G. S., Swanson, M. L., and Chu, W. K., Appl. Phys. Lett., 55" 1397 (1989)

Ionlmplantation

351

74. Wilson, R. G., Pearton, S. J., Abemathy, C. R., and Zavada, J. M., Appl. Phys. Lett., 66:2238 (1995) 75. Wilson, R. G., Vartuli, C. B., Abemathy, C. R., Pearton, S. J., and Zavada, J. M., Solid-State Elec., 38:1329 (1995) 76. Zolper, J. C., Hagerott Crawford, M., Pearton, S. J., Abemathy, C. R., Vartuli, C. B., Ramer, J., Hersee, S. D., Yuan, C., and Stall, R. A., Conf. Proc. Material Research Society, Fall 1995, 395:801, (F. A. Ponce, R. D. Dupuis, Nakamura, S., and J. A. Edmond, eds.), Material Research Society, Pittsburgh, PA (1996) 77. See for example: Naik, I. K., J. Electrochem. Soc., 134:1270 (1987); Humer-Hager T., and Zwicknagl, P., Jpn. J. Appl. Phys., 27:428 (1988) 78. Edwards, A., Rao, M. V., Molnar, B., Wickenden, A. E., Halland, O. W., and Chi, P. H., J. Electron. Mat., 26:334 (1997) 79. Zolper, J. C, GaN and Related Materials, Ch. 12, (S. J. Pearton, ed.), Gordon and Breach, New York, NY (1997) 80. Maruska, H. P., private communication. 81. Orton, J. W., Semicond. Sci. Technol., 10:101 (1994) 82. Brandt, O., Yang, H., Kostial, H., and Ploog, K. H., Appl. Phys. Lett., 69:2707 (1996) 83. Salvador, A., Kim, W., Aktas, O., Botchkarev, A., Fan, Z., and Morkoc, H., Appl. Phys. Lett., 69:2692 (1960) 84. Rosenblatt, D. H., Hitchens, W. R., Anholt, R. A., and Sigmon, T. A., IEEE Elec. Dev. Lett., 9:139 (1988) 85. Gardner, J., Rao, M. V., Halland, O. W., Kellner, G., Simons, D. S., Chi, P. H., Andrews, J. M., Kretchmer, J., and Ghezzo, M., J. Electron. Mat., 25:885 (1996) 86. Ryu, J., Kim, H. J., Glass, J. T., and Davis, R. F., J. Electron. Mat., 18:157 (1989) 87. Wilson, R. G., Surface and Coatings Technology, 47:559 (1991) 88. Pearton, S. J., s Mod. Phys., 7:4687 (1993) 89. Tan, H. H., Williams, J. S., Yuan, C., and Pearton, S. J., Conf. Proc. Material Research Society, Fall 1995, 395:807, (F. A. Ponce, R. D. Dupuis, S. Nakamura, and J. A. Edmond, eds.), Material Research Society, Pittsburgh, PA (1996) 90. Tan, H. H., Jagadish, C., Williams, J. S., Zoa, J., Cockayne, D. J. H., and Sikorski, A., s Appl. Phys., 77:87 (1995) 91. Rowland, L. B., Doverspike, K., and Gaskill, D. K., Appl. Phys. Lett., 66:1495 (1995)

352

Wide B a n d g a p S e m i c o n d u c t o r s

92. Zolper, J. C., Crawford, M. H., Tan, H. H., Williams, J. S., Zhou, J., Cockayne, D. J. H., Pearton, S. J., and Karlicek, R. F., Jr., Appl. Phys. Lett., 70, May 15, 1997, 70:2729 (1997) 93. Zolper, J. C., Crawford, M. H., Williams, J. S., Tan, H. H., and Stall, R. A., Nuclear Inst. and Methods in Phys. Res. B, 127/128:467 (1997) 94. Tan, H. H., Williams, J. S., Zou, J., Cockayne, D. J. H., Pearton, S. J., and Yuan, C., Proc. 1st Symp. on III-V Nitride Materials and Processes, 96( 11): 142, Electrochemical Society (1996) 95. Zolper, J. C., Han, J., Biefeld, R. M., Van Deusen, S. B., Wampler, W. R., Pearton, S. J., Williams, J. S., Tan, H. H., Karlicek, R. J., and Stall, R. A., Conf. Proc. Material Research Society, Spring 1997, Vol. 468, (C. R.Abemathy, H. Amano, and J. C. Zolper, eds.), Material Research Society, Pittsburgh, PA, in press 1997 96. Williams, J. S., and Poate, J. M. Ion Implantation and Beam Processing, Academic Press, Sydney (1984) 97. Williams, J. M., McHargue, C. J., and Appleton, B. R., Nucl. Instr. and Meth., 209/210:317 (1983) 98. Spitznagel, J. A., Wood, S., Choyke, W. J., Doyle, N. J., Bradshaw, J., and Fishman, S. G., Nucl. Instr. Meth. Phys. Res. B, 16:237 (1986) 99. Fohl, A., Emrick, R. M., and Carstanjen, H. D., Nucl. Instr. Meth. Phys. Res. B, 65:335 (1992) 100. Chechenin, N. G., Bourdelle, K. K., Suvorov, A. V., and Kastilio-Vitloch, A. X., Nucl. Instr. Meth. Phys. Res. B, 65:341 (1992) 101. Avashi, D. K., Vacuum, 47:1249 (1996) 102. Kalish, R., Reznik, A., Nugent, K. W., Prawer, S., Nucl. Instr. and Meth. in Phys. Res. B, 148:626 (1999) 103. Sato, S., and Iwaki, M., Nucl. Instr. Meth, B, 32:145 (1988) 104. Kalish, R., Bemstein, T., Shapira, B., and Talmi, A., Radiat. Eft., 52:153 (1980) 105. Prawer, S., Hoffman, A., and Kalish, R.,Appl. Phys. Lett., 75:2187 (1990) 106. Yuan, C., Salagaj, T., Gurary, A., Zawadzki, P., Chern, C. S., Kroll, W., Stall, R. A., Li, Y., Schurman, M., Hwang, C.Y., Mayo, W. E., Lu, Y., Pearton, S. J., Krishnankutty, S., and Kolbas, R. M., J. Electrochem. Soc., 142:L163 (1995) 107. Shul, R. J., Howard, A. J., Pearton, S. J., Abemathy, C. R., Vartuli, C. B., Barnes, P. A., and Davies, P., Conf. Proc. Material Research Society, Fall 1995, 395:751, (F. A. Ponce, R. D. Dupuis, S. Nakamura, and J. A. Edmond, eds.), Material Research Society, Pittsburgh, PA (1996) 108. Zolper, J. C., Baca, A. G., Shul, R. J., Howard, A. J., Rieger, D. J., Sherwin, M. E., Lovejoy, M. L., Hjalmarson, H. P., Draper, B. L., Klem, J. F., and Hietala, V. M., IEEE Trans. Elec. Dev., 41:1078 (1994)

lon lmplantation

353

109. Doverspike, K., Wickenden, A. E., Binari, S. C., Gaskill, D. K., and Frietas, J. A., Conf. Proc. Material Research Society, Fall 1995, 395:897, (F. A. Ponce, R. D. Dupuis, S. Nakamura, and J. A. Edmond, eds.), Material Research Society, Pittsburgh, PA (1996) 110. Binari, S. C., Proc. Sym. on Wide Bandgap Semiconductors and Devices, Fall ECS meeting, 1995, p. 136, The Electrochemical Society, Pennington, NJ (1995) 111. Vodakov, Y. A., Demakov, K. D., Kalinina, E. V., Mokhov, E. N., Ramm, M. G., and Kholuyanov, G. F., Sov. Phys. Semicond., 21" 1017 (1987) 112. Edmond, J. A., Das, K., and Davis, R. F.,J. Appl. Phys., 63:922 (1988) 113. Ghezzo, M., Brown, D. M., Downey, E., Kretchmer, J., Hennessy, W., Polla, D. L., and Bakhru, H., IEEE Elec. Dev. Letts., 13:639 (1992) 114. Ghezzo, M., Brown, D. M., Downey, E., Kretchmer, and J., Kopanski, J. J., Appl. Phys. Lett., 63:1206 (1993) 115. Xie, W., Cooper, J. A., Jr., and Melloch, M. R., IEEE Elec. Dev. Lett., 15:455 (1994) 116. Prins, J. F., US patent # US5385762 117. Prins, J. F., AppL Phys. Letts., 41:950 (1982) 118. Zeisse, C. R., Hewett, C. A., Nguyen, R., Zeidler, J. R., and Wilson, R. G., IEEE Elec. Dev. Lett., 12:602, (1991)

8 R a r e E a r t h I m p u r i t i e s in Wide Gap Semiconductors John M. Zavada

ABSTRACT In recent years, optoelectronic materials, both insulators and semiconductors, doped with rare earth atoms have received widespread attention due to their impact on optical communication systems operating at 1.54 gm and 1.3 gin. Optical amplifiers based on Er-doped fibers have demonstrated major improvements in link distance, data rates, and reduced needs for signal regeneration. The intra-subshell transitions of4felectrons in rare earth ions lead to narrow absorption peaks in the ultra-violet, visible, and near-infrared regions of the electromagnetic spectrum. Since it appears that the intensity of the room temperature light emission of these ions depends upon the energy bandgap of the host material, wide gap semiconductors may prove to be the best materials for optoelectronic device applications. Semiconductors doped with rare earth ions offer the prospect of very stable, temperature-insensitive, optical amplifiers, and light emitting diodes operating at wavelengths from the ultra-violet to the near infrared. This paper presents a review of the lurninescence characteristics of wide gap compound semiconductors doped with rare earth atoms. In particular, aspects of rare earth atom incorporation in the semiconductor crystal, photoluminescence properties, and prototype electrolurninescent devices are addressed. 354

Rare Earth Impurities 355 1.0

INTRODUCTION

Crystalline Si has emerged as the primary semiconductor material for modem micro-electronics, largely due to an effective native oxide. With the development of fiber optic communications, the natural choice would have been to use Si as the basic material for optical transmitters and detectors. However, Si is an indirect bandgap material. Consequently, even with the advent of porous Si, efficient light emission has not been possible.[ ~]/2] Therefore, direct bandgap III-V compound semiconductor materials have been synthesized and used effectively for optoelectronic sources and detectors.[3] Materials such as AIGaAs and InGaAIP have been epitaxially grown with great success to form light emitters and detectors operating at the major wavelengths (0.82 lam, 1.3 lam, and 1.5 lam) used for optical fiber communications. Long distance optical communication systems also require amplifiers to boost the transmitted signal and, initially, this amplification was done at repeater stations. At these stations, the optical signal was converted into an electrical signal, amplified, and then optically regenerated. Recently, erbium doped fiber amplifiers (EDFA) have been used to perform this amplification while keeping the signal in the optical domain. This has resulted in a large improvement in the capacity of optical fiber communication systems.[4]-[6] However, there are certain disadvantages to the use of semiconductor lasers and Er doped silica fibers. While semiconductor quantum well lasers have high gain and high quantum efficiency, the lasing wavelength is strongly dependent upon the operating temperature. This is especially true for narrow bandgap InP-based devices. EDFAs, which are based on the 4fintra-subshell transitions of the Er ions, require elaborate optical pumping arrangments.[7][8] If rare earth (RE) atoms can be introduced into an appropriate semiconductor material and effectively activated by electrical means, then it might be possible to achieve nearly ideal, optical emission and amplification at the desired wavelengths for optical communication and display systems. Charge carriers would be electrically injected into a RE-doped semiconductor, exciting the RE ions, and inducing 4fintra-subshell optical transitions. Since the trivalent erbium ion (Er 3§ emits mainly around 1.54 ~tm, which is in the range of minimum loss in silica fibers, Er-doped semiconductors may be ideal source and amplifier materials for modem optical fiber communication systems. There are several excellent summaries of the properties of RE doped semiconductors.J9][ 1~ In this chapter, recent developments concerning the

356

Wide Bandgap Semiconductors

luminescence characteristics of wide gap semiconductors doped with rare earth atoms are reviewed. In See. 2, the basic properties of the rare earth elements and the wide gap semiconductors are discussed. Section 3 is devoted to the different methods that have been used to dope wide gap semiconductors with RE atoms. In See. 4, photoluminescence spectra from RE 3+ ions incorporated in different wide gap semiconductor host materials, mainly the III-V nitrides and SiC, are presented. The dependence of the emission intensity of the RE ions on the bandgap of the host semiconductor and on the material temperature is discussed. Characteristics of prototype electroluminescent devices based on RE-doped wide gap semiconductors are presented in See. 5.

2.0

BASIC CONCEPTS

2.1

Rare Earth Elements

The lanthanide or "rare earth" elements denote the metallic elements having atomic numbers 57 to 71. These elements are designated "rare earths" due to the difficulty in extracting them from materials known as "earths" (e.g., lime and alumina). Actually, rare earth elements constitute about one-fourth of the known metals and are more abundant than gold or silver. The electronic structure of each trivalent rare earth ion (RE 3+) consists of a partially filled 4f subshell, and completely filled, outer 5s 2 and 5p 6 shells. With increasing nuclear charge, electrons enter into the underlying 4fsubshell rather than the external 5d shell[ 11] (see Table 1). The energy states of the RE 3+ are represented as mLj, where L is the total orbital angular momentum, J i s the total angular momentum, and m is the multiplicity of terms in the atomic configuration. Since the filled 5s 2 and 5p 6 subshells screen the 4felectrons, the rare earth elements have very similar chemical properties and are highly reactive with many other elements forming a variety of organic complexes. The screening of the partially filled 4f subshells also gives rise to sharp emission spectra approximately independent of the host material. The intrasubshell transitions of 4felectrons lead to narrow emission peaks in the ultraviolet, visible, and near-infrared regions of the spectra. For example, the Er 3§ exhibits well defined emission lines near 1.54 ~tm, and the trivalent praesodymium ion (Pr 3§ yields emission lines near 1.30 ~tm. In the free (isolated) RE 3+ state, these electric dipole transitions are forbidden to first

Rare Earth lmpurities 357 order because of parity conservation. However, when a rare earth ion is embedded in a host material, such as a glass or a crystal semiconductor, the local electric fields produce a splitting of the energy levels and certain electric dipole transitions are permitted. In Fig. 1, the spin-orbit splitting of the energy levels of Er 3+under different symmetry conditions is illustrated.[ ill Transitions of 4felectrons between these split levels lead to the observed emission spectra. In particular, the transitions between the first excited manifold of levels (4113/2) and the ground level manifold (4115/2) give rise to the emission lines near 1.54 ~tm.

Table 1. Some of the Basic Properties of the Lathanide Rare Earth Elements

Atomic Number

Element

57

Lanthanum

58

Cerium

59

Praseodymium

60

Neodymium

61

Promethium

62

Samarium

63

Europium

64

Gadolinium

65

Terbium

66

Dysprosium

67

Holmium

68

Erbium

69

Thulium

70

Ytterbium

71

Lutetium

RE 3+ Electron Configuration

4f~ 6 4f15s25p 6 4f25s25p 6 4f35s25p 6 4ff5s25p 6 4f55s25p 6 4f65s25p 6 4fTs25p 6 4fS5s25p 6 4f95s25p 6 4flO5s25p6 4fl15s25p6 4f125s25p6 4f135s25p6 4f145s25p6

R E 3+ Ground State

RE 3+ Ionic Radius (,~)

~So

1.061

2F5/2 3H 4

1.034

419/2 5I 4

0.995

6H5/2

0.964

7Fo

0.950

857/2 7F 6

0.938 0.923

6HI5/2

0.908

1.013 0.979

5I 8

0.894

4II5/2 3H 6

0.881

2F7/2

0.858

IS o

0.848

0.869

The optical properties of RE ions in insulating materials have been extensively studied for applications in solid state lasers and optical fiber amplifiers.Jail 12] Solid state lasers, such as Nd3+:YAG, are based on the 4f

358

Wide Bandgap Semiconductors

intra-subshell transitions of Nd 3+ ions, and exhibit a very stable lasing wavelength and minimum temperature dependence. Because of these characteristics, such lasers have found widespread applications in laboratory and military systems. Furthermore, Er-doped silica fibers are being used for amplification of optical communication systems operating at 1.54 ~tm, and Prdoped fibers are being developed for use at 1.3 ~tm. However, these optical components, lasers and amplifiers, have a relatively low cavity gain.

Spin orbital interaction free Er 3"~on

eV 3

ZHe/z

II il II II il

2.5. __.~

1.5

-

3.04

4Fa/z

2.781

4F$/z

2.737

4F7~

2.533

4Hll ~

.

4S3/z 4Fs,, z

2.388 2.276

9

1.890

419/Z

1.533

41111Z

1.255

4113/Z

Er 3~in cubic symmetry (Td) crystal field

.803

~'

Fa~ } lOmeV rs

r6 ) 2o mev

rsra ]~ 42mev

0.5, I I

4115/Z

0.0

~ } '.~rs

74meV

Figure 1. Schematicdiagramof the energylevelsof a free Er3+ion and the splittingof the 4f

subshell levelsin a field of Tdcubic symmetry. [11]

In semiconductor materials, it is only within the past decade that the optical properties of RE ions have been studied in depth. Beginning with the

Rare Earth lmpurities 359 work of Ennen et al.,[ 13] the luminescence of RE ions in III-V compound semiconductors received considerable attention.[ 14]Investigations of RE ions in a variety of different semiconductors were conducted.[9][l~ This includes the III-V, the II-VI, and the IV-IV compounds as well as porous Si. The goal of this work was to develop electrically pumped optical sources for use in optical communications and in full color displays. Due to the importance of the 1.54 ~tm emission for optical communications, Er has been the main RE element to be studied in these semiconductors.

2.2

Wide Gap Semiconductors

Wide gap semiconductors are usually defined as those semiconductor materials having an energy gap (E~ in the visible or UV region (see Fig. 2.)[ 15] While these materials have been studied for many years, only within the last decade have they been under intense scrutiny for electronic and optoelectronic applications. In the electronics area, SiC has been the major wide gap semiconductor. There were a number of impressive demonstrations reported for SiC electronics, including high power metal-oxide semiconductor field effect transistors, microwave devices and high temperature junction field effect transistors.[ 16]In addition, until a few years ago, the only commercially available blue light emitting diodes (LED) were fabricated from SiC wafers. However, as with Si, this material is an indirect bandgap semiconductor. The external quantum efficiency of the blue LEDs was not very high. In the optoelectronics area wide gap II-VI and III-V semiconductors have been the dominant materials. Using ZnSe compounds, Haase et al. were the first group to successfully demonstrate a blue laser diode at room temperature.[ 17] Subsequently, Eason et al. demonstrated both blue and green light emitting diodes based on these wide gap II-VI semiconductors.[ 18]These impressive results caused a dramatic increase in the research activity in wide gap II-VI semiconductors. In 1994, Nakamura et al. demonstrated a high brightness blue LED based on III-V nitride semiconductors.[ 19]Both blue and green LEDs based on single quantum well InGaN structures are now commercially available. The high brightness and long operational lifetime of the InGaN diodes led to their widespread acceptance for use in full-color displays and traffic lights. Nakamura et al. were also the first to demonstrate a blue laser diode based on the GaN material system.[ 2~ Presently, several groups in Japan and the U.S. have reported on room temperature blue laser diodes using InGaN quantum wells. These developments have attracted intense interest for the III-V nitride

360

Wide Bandgap Semiconductors

semiconductors.[ 21] There is also a potential for nitrides to be used in high frequency electronic power devices and in high temperature electronics. --r-m-r-T-r-T--r-t-.t--r- r-, T-'T] m

r

r-1

i-r,

'-'-I"

'- '--r-i-r- j-~ -'-V T - ' - r ,--

AIN O|

I ll,,-v,

~9

iGaO2 5 J-

9 ..-.

l-

Mgb

('~ E

LU 4

Z

L'

Cubic

nO

LLI

Ga:/ii~

t:L

36 ' \ \ H' )~ , . _ SiC( ~ i n N

Gat',l

9 III-V 9 Nmides

/

(~ Others -

L ..........

ZnS

gSe

III ZnSelll CdS 9

GaPII,,,, 9

AlAsT

~2 Z

"~ m

I j

.,~

Basal Plane 1 - Lattice Constants ......

2.5

ZnTe-

--- I~-

GaAsO Si Q

9 CdSe InP -,.

L . . . . . . . . . . . k . . . . . . . . _1............. t . . . . . . . . . . J_. . . . . . . . . . .L~_.,G.._~_ t . . . . . . . .

3.0

3.5

4.0

LATTICE

4.5

5.0

CONSTANT

5.5

6.0

6.5

(A)

Figure 2. Bandgap energies and lattice constants of the principal wide gap compound semiconductors as well as other important optoelectronic materials.[~5]

Other wide gap III-V semiconductors include GaP and AlAs, both of which are indirect materials. GaP doped with N has been extensively used in green LEDs.[::] However, the purity and the brightness of the emission cannot compete with the newer InGaN based green LEDs. The A1GaAs material system has been thoroughly investigated for electronic and optical applications. With low concentrations orAl, the alloy A1GaAs is not a wide gap material. At high A1 concentrations, the alloy does have a wide, indirect bandgap. However, ambient oxidation and deep level defects restrict the use of this material. Porous Si is somewhat of an anomaly. It is formed from Si material, which is an indirect, non-wide gap semiconductor. However, after electrochemical processing, porous Si exhibits room temperature luminescence from the visible to the near infrared, depending upon the processing conditions.[ 2] It has been conjectured that, after processing, direct bandgap quantum wires or quantum dots are formed, resulting in a wide gap semiconductor material. One of the main reasons for studying RE impurities in wide gap semiconductors stems from the work ofFavennec et al.[23]They reported a

Rare Earth Impurities

361

strong dependence of the emission intensity of the E r 3+ ions on the bandgap of the host semiconductor and on the ambient material temperature. Several different semiconductors were implanted with Er § ions and the photoluminescence (PL) emission spectrum was subsequently measured. It was found that the PL intensity decreased at higher temperatures. The results of these experiments are shown in Fig. 3. Each material was implanted with Er § ions to a dose ofl013 cm "2 atan energy of 330 keV. The thermal quenching ofthe emission intensity was more severe for the smaller bandgap materials, such as Si and GaAs. The wide bandgap compounds, such as ZnTe and CdTe, exhibited the least temperature dependence.

ut

IoOL

-

.,.

,. ,,,,, m

..,,

-... ~

,,,.

,

,

c I,,.

,ID L_

~ 10"1 m

0

~ D

m

E §

,.,W'-1 O-Z 0

100 200 temperature. K

300 I e~.o~21

Figure 3. Emission intensity of Er3+ions in various semiconductorhost materials as a function of temperature. The semiconductor materials are: (a) Ino.16Gao.ssAso.s4Po.16;(b) Si; (c) InP; (d) GaAs; (e) Alo.17Gao.s3As;09 ZnTe; (g) CdS. Each material was implanted with Er+ ions to a dose of l013 cm-2 at an energy of 330 keV.[TM

Neuhalfen et al. observed a similar temperature dependence of the intensity of the Er 3+ emission for different composition In ]_xGaxP epilayers

362

Wide Bandgap Semiconductors

grown by MOCVD. [24]Based on their data, they proposed a model to explain the thermal quenching as a function ofbandgap energy. In this model, the Er a+ ions act as radiative recombination centers. With increased temperatures, the electrons become delocalized from these centers into the conduction band and recombine via non-radiative channels. Another explanation for the decrease of thermal quenching with wider band gap semiconductor materials was given by Takahei et al.[25] They argued that with wider gap materials, energy backtransfer was less likely to occur because the energy needed to form the bound exciton, which subsequently recombines without Er a+ excitation, is larger. In general, optical emission in the semiconductors appears not as efficient as in the dielectric materials. It may be that ionic bonds found in dielectrics may be better for forming the required energy levels than the covalent bonds present in most semiconductors. Co-doping the host semiconductor with impurity elements, such as O or F, seems to enhance optical emission. This may be due to the impurity atoms forming ligands with the RE atoms and converting the local bonds into a more ionic state. Favennec et al. [26] and Michel et al. [27] observed a significant enhancement of the E r 3+ emission in Si due to the presence of O impurities. These results suggest that O assists in the optical activation of the Er 3+ ions. In similar experiments, Colon et al.[TMshowed that the addition of O significantly enhances the Er 3+ emission from Er-implanted A1GaAs. There are considerable materials issues associated with each of these wide gap semiconductors. Improved bulk and epitaxial growth methods need to be developed. Ohmic contacts, processing steps (such as wet and dry etching), and doping/isolation procedures need further research to fully exploit these promising material systems for electronic as well as optical applications.

3.0

I N C O R P O R A T I O N O F R E A T O M S IN W I D E G A P SEMICONDUCTORS

A number of different methods have been used for incorporating RE atoms into semiconductor materials. These methods include: epitaxial growth, ion implantation, e-beam sputtering, laser ablation, electrolysis, and diffusion.J9][ l~ There does not seem to be any report of RE atoms incorporated into

Rare Earth lmpurities 363 semiconductor materials during bulk growth. The principal methods developed so far are: ion implantation and doping during epitaxy. Ion implantation has received considerable attention as a method for doping materials, due to its widespread use in processing integrated circuits and optoeleetronies devices.[ 29] Because this method is a non-equilibrium process, introduction of dopants is limited neither by solubility constraints nor by surface chemistry. However, implantation produces crystal damage, especially with high mass atoms such as the RE elements. The damage leads to electronic traps and non-radiative recombination centers in the host material. Consequently, some form of post-implantation annealing is necessary. Ion implantation has been used extensively to dope semiconductors with RE atoms.[g][l~ However, annealing still needs to be performed and higher concentrations of RE atoms will lead to clustering effects and reduced optical activity. Using ion implantation, Wilson et al. showed that it was possible to dope GaN films with a sufficiently high concentration of Er to observe light emission at 1.54 ~tm.[3~ The GaN specimens were implanted with Er + ions at room temperature in the (100) random orientation. Implantation energy was 300 keV and the fluenee was 2 x 1014 cm -2. Figure 4 shows the result of a secondary ion mass spectrometry (SIMS) measurement of the Er depth profile in a typical ion implanted GaN film.J31][32]Under these conditions, the peak Er concentration reached a level of approximately 1019 cm -3 and the projected range Rp was about 0.1 m. The GaN films were also implanted with O + ions at an energy of 180 keV and a fluenee of 10 is cm -2. The ion energies were chosen to make the projected ranges of the two elements coincide. After implantation, the GaN films were furnace annealed at 650-700~ No change in the Er or O depth profile was observed under such annealing. Several types ofepitaxial growth have been successful for doping the III-V semiconductors with RE atoms: liquid phase epitaxy (LPE), molecular beam epitaxy (MBE), metal-organic chemical vapor deposition (MOCVD), and metal-organic molecular beam epitaxy (MOMBE).[9][ 1~ Using LPE, reproducible epilayers have been obtained only for Er doping of In-based compounds.[ 33]This difficulty was attributed to the high chemical reactivity of Er and the tendency of these atoms to form oxides. The epilayers were found to consist of Er precipitates and microparticles.[ 34] With MBE, which is an ultra-high vacuum technique, precise control ofepilayer composition, thickness, and doping profile is possible. However, even with this technique, there have been difficulties incorporating the RE atoms into the epilayers. The solubility limit of RE atoms in these epilayers has been on the order of 1019

364

Wide Bandgap Semiconductors

cm "3 and the occurrence of precipitates and microparticles has also been noted. Nevertheless, high quality GaAs epilayers doped with Er ions have been achieved using MBE. Distinct lines in the emission spectra at 1.54 ~tm have been observed.[35][36]Using MOCVD, high quality, Er doped GaAs and GaP epilayers have been produced with very good luminescence characteristics. [37][38] MOMBE is a combination of the MBE and MOCVD methods. It is a high vacuum technique which uses gaseous as well as solid sources. This method has been especially successful for C doping of GaAs epilayers. Recently, Er-doped III-V nitride films have been successfully grown using MOMBE and MBE.

1020

-

I

I

I

-

"~

1019 in GaNISapphire 300 keY 2X1014 cm "2

-

i e 1018

1017

":

1016

1015

I

0.0

I 0.2

I 0.4

I 0.6

,

"

0.8

DEPTH (jam)

Figure 4. Atomic depth distribution measured by SIMS of Er implanted at room temperature into GaN/sapphire in the (100) random orientation. Implantation energy was 300 keV and the fluence was 2 x 1014 cm-2.[3~

Rare Earth Impurities 365 MacKenzie et al. were the first to show that AIN films could be doped with Er during epitaxial growth. [39] Using MOMBE techniques, controlled Er densities in the 1019to 1020cm -3range were achieved. The Er densities were confirmed using SIMS analysis that was quantified using Er implanted standards. In Fig. 5, the result of a SIMS measurement of the Er depth profile in a GaN film grown by MOMBE is shown.[4~ In this figure, high concentrations of O also appear in the A1N films. What is not known is whether the presence of O enhances the luminescence of the Er ions. In order to achieve significantly higher Er concentrations in the films, effusion cell temperatures greater than 1000~ were required for the solid Er source.The results of the effusion cell temperature on the Er concentration in the film are shown in Fig. 6.[ 39]

,,

t.

|

A~N ~

1

10 7

1

E r - D O P E D AIN/SAPPHIRE BY M O M B E 0

s

10;12

(n i-. 10 s z o o z

I?

uE1021

.9.o

m

(n z uJ D

10 4

7 0 W

Er

glOa)

_" 10 3

10111 10 2

i

i

i

, , , I,

i

.

, , t ,,,

1~

10 w 0.0

0.2

0.4

0.6

0.8

1.0

12.

1.4

DEPTH (lun)

Figure 5. Atomic depth distribution measured by SIMS of Er and O incorporated in an A1N film during MOMBE growth.t4Ol

366

Wide Bandgap Semiconductors

1022

,-

.

9 _ , ....

.-

-

|

"

.

'_

i

IS

I0 2t A

'G lo

g~ @ ~m

a

1019

Q0

r gg O

to l_

10 is

1

6.0x10"4 " 6.5x10"4

A

7.(NIO"4

,

L,,

7.5x10

9

"4

I

8.0x10

.

.,

"4

8.5x10

"4

I/T 1s"I

Figure 6. Er concentration in A1N films as a function of cell temperature during MOMBE growth.[

39]

Diffusion of RE atoms into the semiconductor materials has had very limited use. The low diffusivity of the large RE atoms, the relatively low decomposition temperatures of typical semiconductor materials, and solubility constraints have been major obstacles hindering the use of diffusion. However, Horiguchi et al. showed the diffusion process can be used to dope Si with Er atoms. [41] There does not seem to be a fundamental reason preventing this method from being used for wide gap materials. Diffusion has been successfully used to introduce RE atoms into various dielectric crystals, such as LiNbO 3, which have a lower melting temperature than several wide gap semiconductors.[ 42] Not a great deal is known concerning the exact locations of the RE atoms in the semiconductor lattice. Basic questions remain whether interstitial or substitutional sites are the preferred locations. Rutherford backscattering analysis (RBS) has been useful in obtaining some understanding concerning these locations. Kozanecki et al. studied (100) GaAs wafers implanted with Er. [43] and concluded that the optical activity of the Er 3§ ions disappears once the ions are located on substitutional lattice sites.

Rare Earth lmpurities 367 A similar study was conducted by Tang et al. using Si wafers implanted with Er. [44] Their RBS data also indicated that the lattice positions of the Er atoms depend upon the annealing conditions. After implantation, the Er atoms occupy random sites in the lattice. With annealing, the Er atoms move to substitutional lattice sites. However, even when the Er atoms are on substitutional lattice sites, there is optical activity and distinct luminescence lines near 1.54 lam can be detected. These results are consistent with data obtained by Nakata et al. on Erdoped GaAs epilayers grown by M O C V D . [37]The peak of RBS data in the <110> channeling direction supported their conclusion that the Er ions occupy a displaced tetrahedral interstitial site. Photoluminescence measurements at 4 K on these samples displayed a complex spectrum indicating the presence of at least three kinds of active Er 3+centers. In dielectric matrials, it was established that the RE ions occupy a variety of sites. Witte et al. showed that Er ions occupy at least 4 different sites in LiNbO 3 after diffusion.[ 45] Typically, separation of the optical activity from these different sites was a difficult task. After implantation, the RE atoms are probably in interstitial locations. Following annealing, some of the RE ions are probably on lattice sites. With the A1N samples doped with Er, the RE atoms are likely to be on the group III sub-lattice. Kim et al. have studied the location of Er 3+ ions in GaN by siteselective photoluminescence excitation (PLE) spectroscopy.[ 46] The GaN epilayers were grown on sapphire substrates using MOCVD and implanted with Er + and O+ ions. The PLE spectra displayed several broad, below bandgap, absorption bands which in turn could excite distinct Er3+site-selective PL spectra. Three different Er3+sites were identified. Excitation of two of the bands involved optical absorption by defects, and impurities with subsequent energy transfer to Er 3+centers. Excitation of the third band appeared to involve a bound exciton at an Er-related trap center. Further studies are needed to determine the precise locations of the RE ions and the associated optical properties.

4.0

R E 3+ P H O T O L U M I N E S C E N C E

Photoluminescence spectroscopy has been the principal technique used to characterize the optical emission of the RE ions. This technique involves excitation of the RE ions by means of a laser and measurement of the spectnma of the resulting light emission as a function of intensity and energy. While the exact mechanisms are not fully understood, it appears that in a

368

Wide Bandgap Semiconductors

semiconductor doped with RE ions, the ions may be optically excited either indirectly or directly. In the first case, laser radiation, at an energy level above bandgap energy, leads to the creation of electron-hole pairs which in turn recombine near a RE center. The energy from the electronhole pair is transferred to the RE ion, exciting it to one of its higher energy states. Upon relaxation, the RE ion emits the optical radiation. Defects and other impurities in the material serve as non-radiative recombination centers that can alter the light emission properties. Most of the PL measurements of RE-doped semiconductors involve the use of a pump laser operating at an energy level above bandgap energy. However, with the wide gap semiconductors, use of a blue or UV laser is required. In the second case, the RE ions in the semiconductor may be excited directly by the laser pump radiation provided that the energy of the laser coincides with one of the excited energy levels of the RE ion. Wu et al. have shown that in defected A1N layers, a carrier-mediated process can also be used to excite the RE ions, even when the laser radiation has an energy below that of the bandgap.[ 47] Carrier lifetime of electron-hole pairs in an undoped semiconductor is determined by the band structure and the defect density of the material. Typically, in direct bandgap material, this lifetime is on the order of a few ns. The intra- sub shell 4ftransitions of RE 3+ ions in a semiconductor exhibit a much longer lifetime, on the order of a few ms. Using PL techniques, Klein and Pomerenke measured the time decay of 1.54 ~tm emission of Er 3§ ions in several different semiconductors and showed that the emission decay for GaAs, GaP, InP, and Si is on the order of 1 ms.[ 48] Wilson et al. reported the first observation of light emission of Er 3+ ions incorporated in III-V nitride semiconductors (in particular, GaN and A1N films).[ 3~ The films, which were grown on either GaAs or sapphire substrates, were co-implanted with Er § and O § ions and then annealed at 650~700~ The Er-O doped GaN films, which were grown on sapphire substrates, were optically excited using an Ar + laser at a wavelength of 457.9 nm which corresponds to an energy below the bandgap ofGaN (Eg.~ 3.4 eV). Infrared PL spectra were measured at 6, 77, and 300 K. As shown in Fig. 7, the spectra were centered at 1.54 ~tm and displayed many of the allowed tmnsitiom between the 4113/2 and the 4115/2 manifolds typical of the Er 3§ configuration. In addition, the luminescence intensity was more than half as intense at room temperature as at 77 K. This result provided further evidence that wide bandgap materials, such as ClaN, tend to suppress the temperature d ~ d e n c e of the Er3+luminescence. Since the laser excitation energy was below the GaN bandgap, it was uncertain whether the observed luminescence of

Rare Earth Impurities 369 Er 3+ was due to direct optical excitation o f the Er § ions or some carriermediated process.

300 K

77K

6K

1500

1520

1540

1560

1580

1600

1620

1640

1660

WAVELENGTH (nm)

Figure 7. Photoluminescence spectra of Er3+ions implanted into a GaN/sapphire host as a function of measurement temperature. The sample film was co-implanted with O+ ions and annealed at 700~176

The A1N films, which were grown on GaAs substrates, were coimplanted with Er § and O § ions and then annealed at 650~ Similarly, an Ar + laser at a wavelength o f 457.9 n m was used to optically excite the Er 3§ ions. No PL spectra could be observed at 77 and 300 K. Only at 6 K could a PL spectrum be measured, as shown in Fig. 8. The spectrum was centered at

370

Wide Bandgap Semiconductors

1.54 ~tm, but did not display many of the allowed transitions between the 4113/ 2 and the 4115/2 manifolds that were found in the Er-doped GaN samples.

I

9,1211-a~.o~

ttJ tLI

I 1500

1520

1540 1560 WAVELENGTH (nm)

1580

1600

Figure 8. Photoluminescencespectrum, measured at 6 K, of Er3+ions implanted into A1N/

GaAs sample. The film was co-implantedwith O+ions and annealed at 650 89 3~

Torvik et al. performed a systematic study of GaN films co-implanted with Er § and O § ions and subsequently annealed. [49]The films were grown by chemical vapor deposition on R-plane sapphire. Different combinations of fluences for the implants and annealing temperatures were examined. After each annealing stage, the samples were examined for Er 3+ luminescence using a laser at a wavelength of 980 nm. Figure 9 shows a typical PL spectrum of one of their sample films measured at 77 K. The inset shows the three level excitation process leading to an emission of around 1.54 ~tm. In their experiments, the samples with the highest Er fluence (1015 Er 2+ cm -2) yielded the strongest PL intensity. Silkowski et al. examined the luminescence of Er and Nd implanted into GaN films which were grown by MOCVD on sapphire substrates.[ 5~ The ion implanted films were annealed and then excited using an Ar + laser operating at 514.5 nm. No co-implantation with O was made. Since the films were prepared by MOCVD, it is posible that high levels of O were already present. The Er-implanted samples showed the 4113/2 ~ 4115/2 transitions around 1.54 ~tm and the 4Ill/2 ~ 4115/2 transitions near 1.0 ~tm. Based on these data, it appeared that multiple Er 3§ radiative centers were present in the samples. The Nd implanted samples showed three well resolved manifolds of 4flines near

Rare Earth Impurities 3 71 0.98, 1.14, and 1.46 lam. PL spectra, measured at 2 and 300 K, of Nd 3+ ions (5x1013 cm -2) implanted into GaN are shown in Fig. 10. The narrow lines nearly correspond to transitions from the 4F3/2 ~ 4I 11/2manifolds. __

.......

4111/241

~. 0.8

13/2

-

A

4Itm

0.69

0.4

c

0.2

-

0

I " 0 v--

0 04

0 eO tO

0 ~" tO

I

,.t

I

-I

I

0 U') LO

0 (0 u'J

0 ~ t~

0 O3 u')

0 O~ u~

0 0 (s

Wavelength (nm) Figure 9. Photoluminescence spectrum, measured at 77 K, of Er 3+ ions (1015 cm -2) implanted into GaN/sapphire sample. The film was co-implanted with O + ions (1016 cm -2) and annealed up to 900 89 [49]

The first report of strong Er 3§ luminescence from III-V nitride films doped with Er during epitaxial growth was from Pearton et al.[sl] and Zavada et al.[4~ The A1N films were prepared using MOMBE techniques, as discussed above.[39]The Er-doped epilayers were optically excited using an Ar+laser at 458 nm and strong infrared spectra were measured at 300 K. The spectra were centered at 1.54 gm and were similar to that obtained from the Er-implanted A1N film. However, the PL intensity from the Er-doped AIN layer grown by MOMBE was nearly two orders of magnitude greater than that from the Er-O implanted A1N film. Figure 11 shows a comparison of the PL spectra taken at 5 K, of Er 3§ ions in an A1N layer grown by MOMBE, and of the Er-O implanted A1N film. No post-growth processing of the Er-doped A1N layer grown by MOMBE was required in order to observe the strong luminescence. However, SIMS analysis, as shown in Fig. 5, indicated that the O background was -- 2 x 102~cm -3 in the

372

Wide Bandgap Semiconductors

A1N sample. Sputtering of oxygen from the alumina ECR plasma cup may be the source of this impurity element in the films.

Wavelength (nm) 1200

1180 '

1180

I

'

1140

I

,

I

4F

1120 '

I

- . ) , 41

3/2

300 K

1100 '

1080

I

9

I

1060

'

I

NdS*

1112

~ = 5x1013cm-2

m

]

I

|

i

i

,

1.00

1.04

,

1.08

t

,

1.10

I

i

,

1.12

,

1.14

I

9

1.16

I

1.18

Photon Energy (oV)

Figure 10. Photoluminescence spectra, measured at 2 and 300 K, of Nd 3+ ions (5 x 1013cm -2) implanted into a GaN/sapphire host.[ s~

80000-

.

,

9

,

9

I

9

I

9

70000

i'

9

,"

9

,

9

Er3§

PL

I

9

at

,--

EK

60000 50000 .~

40000

-.

30000

.N qCJ

p-.t

r

20000 10000

~ . ~ . _ _ ~ I

1.48

MOMBE AIN:Er

~ I

[replanted AIN:Er I

I

9

I

I

9

!

_

1.50 1.52 !.54 1.56 1.58 1.60 1.62 1.64 1.66 Wavelength (p.m)

Figure 11. Comparison of the photoluminescence intensity spectum, measured at 5 K, of Er 3+ ions in an A1N film grown by MOMBE with that of an Er-implanted AIN film. Is9]

Rare Earth Impurities 3 73 Wu et al. performed a careful optical characterization of the Er-doped AIN layers grown by MOMBE. [52] Using an optical parametric oscillator (OPO), pumped by a Nd:YAG laser, as the excitation source at 488 nm, infrared PL spectra were measured in the temperature range from 13-300 K. As shown in Fig. 12, the emission spectra were centered at ~ 1.54 ~tm and correspond to the weakly split levels of the 4113/2 and the 4115/2 manifolds. The room temperature spectrum resembles the one shown in Fig. 8 for Erimplanted A1N film. The main difference between these two spectra is the more intense signal from the AIN layer grown by MOMBE. No room temperature spectrum could be measured for the Er implanted AIN film. The integrated Er3+ luminescence is quenched only by a factor of about two between 6 and 300K. This provides further support that wide gap materials, such as AIN (Eg ~ 6.2 eV), tend to suppress the temperature dependence of the RE 3+ luminescence.

Er: AIN (~.ox =488 nm, pulsed)

5 >, I/1 r" ID

_=

=

I

15oo

~

,

I

155o

,

I

18oo

,

I

185o

Wavelength (nm)

Figure 12. Photoluminescence spectra of Er 3+ ions in AIN semiconductor host material as a function of measurement temperature. The sample film was grown by MOMBE.[521

374

Wide Bandgap Semiconductors

Photoluminescence excitation (PLE) spectra were also measured for Er-doped A1N layers grown by MOMBE.[ 47] For these measurements, the same OPO system was used as the source and the PLE signal was recorded as the ratio between the PL intensity detected at 1.54 m and the excitation power. The PLE spectra taken at 15 K from two different spots on the same Er doped A1N sample are shown in Fig. 13. Each spectrum contains one or more sharp lines superimposed on a broad PLE signal. The sharp lines may be due to a direct optical excitation mechanism via an intra-4ftransition (e.g., 4115/2 ~ 2Hll/2).The broad features may be indicative of a photocarrier mediated process. These PLE data give a partial explanation of why it was possible to excite RE doped GaN and A1N films with below bandgap optical radiation. The RE ions can be excited either through direct optical pumping into one of the 4flevels or through a carrier-mediated process.

I 41tS/2->

'

I

4F~2,SS2

'

I

'

I

'

I

,

I

~

I

4111i~ -=" 2HI1/2

4115/2 -> 4F1/2

,

I

450

L

I

,

5oo sso 600 Excitation Wavelength (nm)

I

6so

Figure 13. Photoluminescenceexcitation spectraof Er3+ions from two different spots on an A1N film that was grown by MOMBE.[47]

Rare Earth Impurities 3 75 Time-resolved PL measurements at various wavelengths gave further evidence of two excitation mechanisms occurring for the Er-doped A1N layers grown by MOMBE.[ 47] Two distinct PL decay patterns at 15 K were observed as shown in Fig. 14. The Er 3+PL, due to direct optical excitation at the absorption peaks (494, 525, and 653 nm), was longer-lived than that which occurred in the vicinity of the peaks (497, 537, and 640 nm). The PL decay transients were found to be non-exponential at all excitation wavelengths. Following a fast initial decay with a lifetime of about 50 s, the PL decayed with a long lifetime of about 0.83 ms. The PL, due to direct optical excitation at one of the absorption peaks, contained less fast-decay components than the PL which occurred via the carrier-mediated process. The de-excitation of RE PL depends upon the non-radiative recombination processes associated with the local RE ion environment. Consequently, the difference between the two types of measured PL decay indicates that two different subsets of Er 3+ sites are excited in these experiments. At least two Er 3+ centers are present in the Er-doped AIN layers grown by MOMBE. The long-lived Er 3+sites are excited primarily through direct optical excitation of one of the 4flevels; the shortlived Er 3+ sites are excited through an indirect via the carrier-mediated process.

10

I0

497 nm

e-

"e 10"

C:

10~

c .J

o.o

0.5

1.o

1.5

"rime(ms) Figure 14. Timedecayofthe 1.54mmemissionofEra§ionsin anA1Ngrownby MOMBE.The curves illustratetwo differentlifetimesdependinguponthe wavelengthofthe opticalsource.t47]

376

Wide Bandgap Semiconductors

MacKenzie et al. also succeeded in growing GaN films doped with Er atoms using MOMBE techniques.[ 53]The GaN films were grown on sapphire substrates and Er densities in the 1018 to 1019 cm -3 range (as measured by SIMS) were achieved. The Er doped films were optically excited at 488 nm and infrared spectra were measured at 13 and 300 K, as shown m Fig. 15. The room temperature PL of the Er 3+ ions in the GaN film was not nearly as intense as that from the A1N film shown in Fig. 15. The GaN films contained only a background level of O which may have influenced the lower PL intensity.

5

'

I

'

I

"

I

'

Er: GaN

i

~. =488nm (pulsed) 15K

30 ,

1450

,

I

1500

,

I

~

1550

I

1600

,

1650

Wavelength(nm) Figure 15. Photoluminescence spectra, measured at 15 and 300 K, of Er3+ ions in a GaN/ sapphire host.The sample film was grown by MOMBE.t531

Choyke et al. observed intense PL spectra in the region near 1.54 ~tm from Er-implanted samples of 4H, 6H, 15R, and 3C SiC crystals.[ 54] The samples were implanted to a fluence of about 1013 Er ions/cm 2 using four implant energies. Annealing was done in a SiC cavity at 1700~ The films were excited by an argon laser at 488 nm and PL spectra were measured at 2 K and 300 K (as shown in Fig. 16) for each of the polytypes. The PL spectra at 2 K are quite similar for the 6H, 4H, and 15R polytypes with a main line at 1.534 ~tm and 14 or more smaller peaks at longer wavelengths. The PL spectrum for the 3C sample contains two major peaks (one at 1.528 and the

Rare Earth Impurities 3 77 other at 1.534 ~tm) in addition to the smaller peaks. The integrated PL intensity varies little over this temperature range from 2 to 300 K. Moreover, even at elevated temperatures, strong emission from the Er ions can be detected. Figure 17 shows a comparison of the relative PL intensity spectra of Er 3+ ions in 6H-SiC, measured at 297, 391, and 518 K. This development of the PL spectrum indicates that the Er luminescence response is essentially flat from 2-400 K, after which it drops o f f b y a factor of about ten a t - 500 K.

0.82 l"

0.80

I

'

i

",==-- PHOTON ENERGY (oV) 0.78 0.76 0.82 'i'"

"

0.80

0.78

0.76

I

2K

i ,A,A, [=~

,

9

,..,.,

' i

!

,,

-

i

I

l

,,l,

I

r~

i,k ,iJ z

IJ.I

o z uJ r,.) r,/) uJ

z_

r

i

I

4H SIC i

,

I li

_1

o o

'

3O0K

z

m.,

"

i

15R SIC i

i, - 1

w

.

.

~. " ' - ~

~=

I

Ill 13:

300K

3C SIC o

t

15000 15400 15800 16200

.

.

I

_

l

15O0O 154OO 158OO 152OO

WAVELENGTH (A)

Figure 16. The relative photoluminescence intensity spectra, measured at 2 and 300 K, of Era+ ions in SiC polytypes of6H, 4H, 15R, and 3C. The sample films were implanted with Er+ ions and annealed at 1700 8917654]

3 78

Wide Bandgap Semiconductors

(a) .

.

.

.

(b)

6.sic

.

297 -

~

^

Z (/) UJ L..

T

v

.

zLU

, ,

(~ LU Z

A

'

--

~

--

UJ

C)

. . . . . .

-

"

-

' .....

'

9 0r

0

LU

ffl

j

zW

~

.J

~

_~ uJ

0

(/) Z

->

om'~ 0

100

UJ

Z

m

6H SiC

g3 -

K

~

...........

" ..... BK

N z

UJ

>

W

m 15000

15400 15800 16200 WAVELENGTH (A)

0

,

300

.

.

i

.

L

9

350 400 450 500 TEMPERATURE (K)

Figure 17. (a) Comparison of the relative photoluminescence intensity spectra, measured at 297, 391, and 518 K, of Er 3+ ions in 6H-SiC. (b) The relative integrated intensity of 6H-SiC implanted with Er § ions and heated from 297-518 K.[ s4]

Porous-Si may be another promising host for Er since porous Si appears to have a significantly larger bandgap than bulk Si and is rich in oxygen concentration.[ 2] Wu et al. studied porous-Si samples that were implanted with Er ions.[ sS] The samples were prepared using a wet anodization technique of p-type Si <100> wafers and then implanted with Er + ions to a fluence of about 1015 ions/cm 2 at an energy of 380 keV. Annealing was done at 650~ for 30 minutes. The treated regions of the wafer were excited by an Ar + laser at 488 nm and the overall room temperature luminescence was recorded. As shown in Fig. 18, the PL spectrum consisted of a broad visible emission band and a narrow emission band ranging from the center at-1540 nm. The broad visible and near IR bands originated from the porous-Si host. The narrow band ( a t - 1.54 ~tm) was due to the Er 3+ transitions between the 4113/2 and the 4115/2manifolds. High resolution PL spectra, measured at 15, 150, and 300 K, of the narrow band at-- 1.54 ~tm are shown in Fig. 19.[56] The relatively large linewidth of this signal at 15 K suggested that there existed

Rare Earth Impurities 3 79 multiple Er sites within the porous-Si material. The temperature quenching of the integrated PL intensity in going from 15 K to 300 K decreased by less than a factor of two. This result supported the increased bandgap thesis ofporousSi and the use of wide gap semicondutors as hosts for RE ions. The Er 3+ PL decay transients were measured at 15 and 300 K and found to be nonexponential (see Fig. 20). [56] The decay transients were fitted with a double expontial curve, a fast initial decay (~ 140 s), and a long lifetime component of -1.3 ms. Based on these decay studies, the occurrence of two distinct Er sites within the porous-Si material seems likely. '

,

Er:Porous

9

,

---'--

', '

',

//'-r--

.....

"r-

Si

(n t--

.6 t..

.b,

Sample

A

~ ~ ~ . . _ .

_~

tn riD t~ ...I I:L

Sam ple.~,,~"'---'"'~,~~

!

600

j

I

__~

700

.

I

800

i

//

1

1500

,L . . . .

.L_

llJO0

Wavelength (nm) Figure 18. Room temperature photoluminescence spectrumof porous Si implantedwith Er3+ ions. The visible PL spectrum is fromthe porous-Si host, whereas the spectrumnear 1.54 I~m is from the Era+ions.[5s]

Another approach to RE doping of porous Si was reported by Kimura et a1.[5711581Samples were prepared using anodic etching of p-type Si <100> wafers. Subsequently, these wafers were placed into an ethanol solution containing either ErC13 or YbC13. The RE ions were electrochemically incorporated into the porous Si layer by applying a negative bias. Room temperattwe luminescence was observed for both the Er 3+ions at-~1.54 ~tm and the Yb 3+ ions ate1.0 ~tm, after annealing at room temperatures above 900~ An Ar + laser at 514.5 nm was used to excite the RE3+e ions. As shown in Fig. 21, the PL spectrum from the Y b 3+ ions consisted of a broad emission band with a small peak at --0.98 ~tm.[5s] At 20 K, this peak, which can be attributed to the 2F5/2 --->2F7/2 transition of Yb 3+, becomes much more pronounced. The

380

Wide Bandgap Semiconductors

luminescence from the Yb 3+ ions was enhanced after annealing in H2; whereas, the luminescence from the Er3§ ions required annealing in 0 2.

m

3

U) C:

1500

1550

1600

Wavelength (nm) Figure 19. High resolution photoluminescenee spectra of porous Si implanted with Er+ ions, measured at 15, 150, and 300 K.[ 56]

A

C: c "O m

0.01

0

I

J

I

Time (ms)

Figure 20. Time decay of the 1.54 lam emission of porous Si implanted with Er3+ ions, measured from 15 to 300 K. Is6]

Rare Earth Impurities 381

I

RTA(1100%)

II

I

In H t

A

(n m

=3 .a

20K

m m

(n (l) r

4b,,r e~m

R .T. a.

0.6

0.8

1.0

1.2

1.4

1.6

W avelength (l~m) Figure 21. Photoluminescence spectra, measured at 20 and 300 K, of porous Si doped with Yb + ions. The sample films were annealed at 1100 89 for 30 sec in H2.lssl

Wang and Wessels studied the luminescence of Er 3+ions in GaP films and observed strong emissions from 12-295 K.[381The films were prepared by atmospheric pressure metalorganie vapor phase epitaxy (MOVPE) and doped during growth by sublimating Er(thd)3 and tranporting the vapor to the reaction zone with H 2 gas. The films were excited by an Ar +laser at 488 nm and the luminescence was recorded. Figure 22 shows the high resolution PL spectrum of the Er doped GaP epilayer measured at 12 K. The spectrum consists of a series of emission lines that may be due to weakly split excited levels or to different Er 3+ sites in the semiconductor host. The origin of these peaks was further investigated by Culp et al. [sg]They examined the luminescence properties of Er doped GaP as a function of temperature, applied hydrostatic pressure and excitation wavelength. They concluded that several Er 3+ sites contributed to the luminescence. Additionally, the PL intensity was significantly stronger when the Er doped GaP films were excited with below bandgap radiation. This dependence of the PL spectra on the excitation wavelength is shown in Fig. 23. This result was attributed to a competition between Er 3+excitation and non-radiative deep level recombination.

382

Wide Bandgap Semiconductors

m

w

,

: GoP:Er :

t-*i no5

9

l::,,O.

'

0.77

0.79

O.81

Emission energy (eV)

0.1~

Figure 22. High resolution photoluminescence spectrum of Er doped GaP measured at 12 K. The sample was prepared by MOVPE and doped with Er during growth.t3s]

4113/2--> 4115/2

.

1450

.

.

.

I

"

1500

"

I

I

I

.

1550

9

9

,

I

9

-

9

1600

Wavelength (nm) Figure 23. Photoluminescence spectra of Er doped GaP as a function of excitation wavelength. The sample was prepared by MOVPE and doped with Er during growth.[59]

Rare Earth Impurities 383 5.0

E L E C T R I C A L A C T I V A T I O N O F R E 3+ I O N S

The use of semiconductors as the host material offers the possibility of developing optical amplifiers and light emitting devices based on electrical excitation of the RE 3§ ions. Such a development would lead to a new class of compact optical sources for optical fiber communication and display systems. These sources could be designed to emit light at different wavelengths and to be integrated with electronic circuits. Most experiments concerning R E 3+ ions in semiconductor materials have centered on optical excitation. The few investigations of electrical excitation have dealt primarily with Er doping and have reported low optical emission efficiency. This is a major obstacle facing the development ofelectroluminescent (EL) devices based on RE doped semiconductor materials. A better understanding of the electrical excitation mechanisms could help to improve this situation. A schematic diagram of a typical semiconductor structure used for electrical excitation of the RE 3+ions is given in Fig. 24.[6~ Basically, it consists ofa p-i-n structure with the RE ions (Er in this illustration) incorporated in the undoped layer. In the forward bias condition, minority carriers are injected into the active layer and the RE ions are excited due to electron-hole pair recombination. As the electron-hole pairs recombine in the vicinity of the RE ion, a portion of the recombination energy is available for transfer to the ion. This is the standard configuration that has been used for the Er doped semiconductor light emitting diodes. When the p-i-n structure is operated in the reverse bias condition, it can also function as an EL device. Majority carriers (electrons) are then injected into the active region and the RE ions are directly excited by collisions with the high energy electrons. This process is known as impact excitation. Chang and Takahei reported on the electrical excitation of Er 3+ions in GaAs epilayers grown by MOCVD under both forward and reverse bias conditions.J61 ] When the Er-doped epilayer was excited using above bandgap radiation, a typical PL spectrum consisting of a set of emission lines near 1.54 ~tm was observed. After processing into a device structure and under forward bias conditions, the EL spectrum was very similar. There was little difference between the PL and EL spectra. In both cases, the Er 3+ ions were apparently excited by electron-hole pair recombination and subsequent energy transfer from the GaAs host. When the device was operated under reverse bias conditions, a different spectrum was observed. The different spectra indicate that the Er 3+ centers excited by electron-hole pair recombination may be different than those excited by direct impact of electrons.

384

Wide Bandgap Semiconductors

//

kk

::::::::::::::::::::::::ibium":i::::::::::::::::::::::::: Oxide

"////////////A

Oxide

l,+

|l

i

Figure 24. Schematic diagram of a typical semiconductor structure for electrical excitation of Er 3+ ions which are incorporated in a p-i-n structure.[ 6~

Torvik et al. were the first to demonstrate room temperature operation of a LED emitting at 1.54 I~m based on III-V nitride semiconductors.[62]The device consisted ofa metal/i-GaN/n-GaN (m-i-n) structure. The GaN layers were grown by CVD on an R-plane sapphire substrate. The active region of the device, the i-GaN layer, was co-implanted with Er § and O § ions and annealed (as in their prior studies).[ 491 The topography of the i-GaN top surface consisted of sharp ridges and valleys due to growth conditions.This surface topography permitted high field injection of electrons from the metal, under reverse bias. Apparently, the excitation of Er 3+ centers was by impact of energetic electrons. Typical current-voltage characteristics of the device are shown in Fig. 25. When the device was operated under forward bias, no Er 3+ related luminescence was observed. PL emission from the Er doped layer was obtained using a diode laser at 980 nm. At room temperature, the EL intensity was about 3 times stronger than the PL (see Fig. 26). They estimated that the electrical excitation cross section for Er in GaN was ~ 5x 1016cm-2 which is about 105 greater than the optical excitation cross section for 4115/2 - ~ 4I ~1/2transitions. [63]T h e external quantum efficiency, measured at 126 V and 318 A, was estimated to be ~ 10-7.

Rare Earth Impurities 385

50

L /

-50

~" .1oo ~L v

= -150

/

= -200 U

-250

/

I

Room Temperature

I

~~176 I -350 -150

-100

-50

I

0 50 Voltage (V)

100

150

F i g u r e 25. Current-voltage data, measured at room temperature, of an Er-doped GaN m-i-n light emitting diode.[ 62]

PL (135 mW at 980 nm)

1.5

3X Room Temperature

C

EL (318 I~A)

_J IX

0.5 ,0x

0 ', ": : : : : : , 850

: ' l " I-'-7",. 1050 1250 1450 Wavelength (nm)

: :,"~ 1650

F i g u r e 26. Electroluminescence and photoluminescence spectra, measured at room temperature, o f an Er-doped GaN m-i-n light emitting diode under reverse bias operation.t 62]

386

Wide Bandgap Semiconductors

Yoganathan et al. produced an EL device, based on 6H-SiC material, that showed room temperature operation under forward bias conditions.[ 64] The device consisted of an n-type SiC epilayer grown on a polished slice of a 6H-SiC boule. The epilayer was Er-implanted at four different energies to obtain a fiat Er dopant profile. After annealing the samples at 1700~ a ptype epilayer of 6H-SiC was grown to form an Er-doped p-n junction. The PL and EL spectra near 1.54 lam were measured at temperatures from 2-520 K. As shown in Fig. 27, the PL and EL spectra were nearly identical, indicating that the same Er 3§ centers were responsible for both emissions. There also appeared to be a correlation between the emission intensity and the shallow nitrogen donor concentration.

Wavelength (A) 16200 15800 15400 15000 :

'H' , ' ' '20, A ' resolution '~11~ ' i ' "

-~ u}9

=

6

.

SIC:Er3+ (p/n/n +)A

L

l

~

I

I

l

I

'

I

il

I

I

I

"qlll

6HSIC2E9r53K(pIn+)/n ..1

~

om

EL

0.775 0.800 0.825 Photon Energy (eV).---

Figure 27. Electroluminescenceand photoluminescencespectra, measuredat room temperature, of an Er-doped 6H-SiC p-n light emittingdiode under forward bias operation.t64]

Wang & Wessels fabricated an Er-doped light emitting diode based on GaP material. [65] They observed strong room temperature EL emission at 1.54 ~tm when the p-n diodes were operated under forward bias. The EL intensity was only weakly temperature dependent over a temperature range of 20-300 K. Ford and Wessels continued this work and examined the saturation of the EL intensity as a function of applied current. [66]The Er doped films

Rare Earth Impurities 387 were prepared by atmospheric pressure MOVPE and doped during growth by sublimating Er(thd)3 and transporting the vapor to the reaction zone with H 2 gas. The films were co-doped with Se to increase the net carrier concentration to-~ 1018 cm -3. The p-n junctions were made by diffusing Zn from a ZnP source into the epilayer. Figure 28 shows the room temperature EL spectra of one of the Er-doped diodes at three different current densities. At low current densities, the EL intensity is linearly dependent upon input power. However, at current densities of~ 10 A/cm 2, the EL intensity begins to saturate. While the precise cause is not determined, it may be due to a saturation of the optically active Er 3§ centers.

4oo

Room Temperature Er:GaP diode A

350'

I

1:If = 19 A/cm 2 2: If = 9 A/cm2 3: If = 5 A/cm 2

300 250g)

200"

150" 10C 50

ot 0.7

"'

'

'

I

. . . .

0.75

'I ~ '

0.8

"-~--i

"' '

0.85

'

'

0.9

Emission Energy (eV)

Figure 28. Electroluminescence spectra, measured at three different current densities, of an Er-doped GaP light emitting diode under forward bias operation.t661

In addition to the above efforts, strong EL emission from Er 3+ centers has been reported in a variety of low gap semiconductors. However, with the lower gap semiconductors, such as Si and GaAs, thermal quenching of the luminescence was a serious problem. Use of impurity elements such as O, F, C, and N alleviated this problem and room temperature Er doped Si LEDs have been demonstrated.[67][68][69]Priolo et al. measured the room temperature EL emission from Er doped Si diodes under both forward and reverse bias operation.[ 7~ The EL intensity under reverse bias was nearly 12 times

388

Wide Bandgap Semiconductors

higher than that under forward bias. Apparently, the excitation mechanisms were very different in the two cases. Under forward bias, electronhole recombination seemed to dominate. Under reverse bias, impact excitation was the main excitation process. The discovery of visible emission from higher excited 4f levels of RE ions incorporated in GaN was reported by Steckl and Birkhahn.[ 71] They achieved in-situ Er incorporation during GaN growth by MBE on sapphire substrates. Under UV laser excitation, the GaN:Er films exhibited intense green PL from the 2H 11/2 and 4S3/2 states. EL devices were fabricated on the GaN/Si structures using indium-tin (90%-10%) oxide (ITO) rectifying contacts. Visible EL emission at the wavelengths of 537 and 558 nm was measured. Near-IR EL emission was also observed with three well-defined peaks between 1.51 and 1.56 ~tm. Visible emission from RE doped GaN films may lead to novel optoelectronics applications.[ 72] As these results show, more research is required to better understand the excitation mechanisms for Er 3§ ions in semiconductor materials and to develop more efficient EL devices. However, many different experiments indicate that wide gap semiconductors offer a distinct advantage for reducing the thermal quenching of both PL and EL emissions from the RE 3§ centers.

6.0

SUMMARY

Substantial progress has been made in the past ten years in understanding the optical properties of RE-doped semiconductors. Luminescence of R E 3+ ions in many different semiconductors has been observed. Experiments show that the use of wide gap semiconductors reduces the thermal quenching of the luminescence. However, use of wide gap semiconductors alone may not offer the complete solution. The location of the RE ion in the crystal and its local chemical environment appear to be extremely important in determining the optical properties of the R E 3+ centers. Difficulties remain concerning the incorporation of RE atoms in wide gap semiconductor materials and the proper processing conditions necessary for optical activation. Prototype EL devices have been fabricated based on both low gap semiconductors and wide gap semiconductors. In general, the emission efficiencies need to be greatly improved. As with the PL experiments, the use of wide gap semiconductors leads to less thermal quenching of the electroluminescence. However, O doped Si does lead to state-of-the-art EL devices. Only a few EL devices based on wide gap semiconductors have been fabricated

Rare Earth Impurities

389

and tested. Further research to optimize the design of EL devices, based on both Si and wide gap semiconductor materials, is clearly needed. Due to the intense current research in SiC and III-V nitride semiconductors, improvements in the crystal quality and in the processing technology of these materials are very likely to occur. Such advances will assist efforts to develop EL devices based on wide gap semiconductors doped with RE ions. If the RE 3+related luminescence can be controlled and the efficiency of EL devices increased, then a new class ofoptoelectronic components for optical communications and display systems will be possible.

ACKNOWLEDGMENT The author expresses his appreciation to U. H6mmerich for valuable discussions in preparation of this chapter.

REFERENCES 1. Sze, S. M., Physics of Semiconductor Devices, John Wiley & Sons (1981) 2. Canham, L. T., Appl. Phys. Lett., 57:1046-1048 (1990) 3. Bhattacharya, P., Semiconductor Optoelectronic Devices, Prentice-HaU (1994). 4. Saito, S., Imai, T., and Ito, T., J. Lightwave Technol., 9:161 (1991) 5. Walker, G. R., Walker, N. G., Steele, R. C., Creaner, M. J., and Brain, M. C., d. Lightwave Technol., 9:182 (1991) 6. Mollenauer, L. F., Evangelides, S. G., and Haus, H. A., J. Lightwave Technol., 9:194 (1991) 7. Poole, S. B., Payne, D. N., Meats, R. J., Ferrmann, M. E., and Lanning, R. I., J. Lightwave Technol., 4:870 (1986) 8. Desurvire, E., Simpson, J. R., and Becket, P. C., Opt. Lett., 12:888 (1987) 9. Rare Earth Doped Semiconductors I," Materials Research Society Proc, Vol. 301, (G. S. Pomerenke, P. B. Klein, and D. W. Langer, eds.) (1993) 10. Rare Earth Doped Semiconductors II," Materials Research Society Proc., Vol. 422, (S. Coffa, A. Polman, and R. N. Schwartz, eds.) (1996) 11. Hiifner, S., Optical Spectra of Transparent Rare Earth Compounds, Academic Press (1978) 12. Koechner, W., Solid State Laser Engineering, Springer Verlag, 3rd Ed. (1992)

390

Wide Bandgap Semiconductors

13. Ermen, H., Schneider, J., Pomrenke, G., and Axmarm,A., Appl. Phys. Lett., 43(10):943-945 (1983) 14. Zavada, J. M. and Zhang, D., Solid-St. Electron., 38:1285 (1995) 15. Schetzina, J. F., Mat. Res. Soc. Syrup. Proc., 395:123-134 (1996) 16. See MRS Bulletin, 22(3) (1997) for an overview of SiC materials and devices. 17. Haase, M. A., Qiu, J., Depuydt, J. M., and Cheng, H., Appl. Phys. Lett. 59:1273-1275 (1991) 18. Eason, D. B., Yu, Z., Hughes, W. C., Roland, W. H., C. Boney, C., Cook, J. W., Jr. and Schetzina, J. F., Appl. Phys. Lett., 66:115-117 (1995) 19. Nakamura, S., Mukai, T., and Senoh, M., Appl. Phys. Lett., 64:1687-1670 (1994) 20. Nakamura, S., Senoh, M., Nagahama, S., Iwasa. N., Yamada, T., Matsushita, T., Sugimoto, Y., and Kiyoku, H., Appl. Phys. Lett., 70:868-870 (1994) 21. See MRS Bulletin, Vol. 22, No. 2 (1997) for an overview of III-V nitride materials and devices. 22. Casey, H. C., Jr., and Trumbore, F. A., Mater. Sci. and Eng., 6:69-109 (1970) 23. Favennec, P. N., L'Haridon, H., Salvi, M., Moutonnet, D., and GuiUou, Y. L., Electron. Lett., 25(11):718-719 (1989) 24. Neuhalfen, A. J., and Wessels, B. W., Appl. Phys. Lett., 60(21):2657-2659 (1992) 25. Taguchi,A., Takahei, K., andHoffkoshi, Y., J. Appl. Phys., 76(11):7288-7295 (1994) 26. Favennec, P. N., L'Haridon, H., Moutonnet, D., Salvi, M., and M. Gauneau, Jpn. J.Appl. Phys., 29(4):L524-526 (1990) 27. Michel, J., Benton, J. L., Ferrante, R. F., Jacobson, D. C., Eaglesham, D. J., Fitzgerald, E. A., Xie, Y., Poate, J. M., and Kimerling, L. C., J. Appl. Phys., 70(5):2672-2678 (1991) 28. Colon, J. E., Elsaesser, D. W., Yeo, Y. K., Hengehold, R. L., and Pomrenke, G. S., Mat. Res. Soc. Symp. Proc. 301:169-174 (1993) 29. Pearton, S. J., Mat. Sci. Rep., 4:313 (1990) 3 0. Wilson, R. G., Schwartz, R. N., Abemathy, C. R., Pearton, S. J., Newman, N., Rubin, M., Fu, T., and Zavada, J. M., Appl. Phys. Lett., 65(8):992-994 (1994) 31. Wilson, R. G., Stevie, F. A., and Magee, C. M., Secondary Ion Mass Spectrometry: A Practical Guide for Depth Profiling and Bulk Impurity Analysis, Wiley (1989) 32. See chapter "SIMS Analysis of Wide Bandgap Semiconductors" in this volume.

Rare Earth Impurities 391 33. Smith, R. S., Muller, H. D., Ennen, H., Wennekers, P., and Mailer, M., Appl. Phys. Lett., 50:49-51 (1987) 34. Nakagome, H., Takahei, K., and Homma, Y., J. Cryst. Growth, 85:345 (1987) 3 5. Charasse, M. N., Galtier, P., Huber, A. M., Grattepain, C., Chazelas, J., and Hirtz, J. P., Electron. Lett., 24:1458 (1988) 36. Zhang, T., Sun, J., Edwards, N. V., Moxey, D. E., Kolbas, R. M., and Caldwell, P. J., Mat. Res. Soc. Symp. Proc. 301:257-262 (1993) 37. Nakata, J., Taniguchi, M., and Takahei, K., Appl. Phys. Lett., 61:2665-2667 (1992) 38. Wang, X. Z., and Wessels, B. W.,Appl. Phys. Lett., 64(12): 1537-1539 (1994) 39. MacKenzie, J. D., Abemathy, C. R., Pearton, S. J., H6mmerich, U., Wu, X., Wilson, R. G., Zavada, J. M., and Schwartz, R. N., Appl. Phys. Lett., 69(14):2083-2085 (1996) 40. Zavada, J. M., Wilson, R. G., Schwartz, R. N., MacKenzie, J. D., Abemathy, C. R., Pearton, S. J., Wu, X., and H6mmerich, U., Mat. Res. Soc. Symp. Proc., 422:193-197 (1996) 41. Horiguchi, H., Kinone, T., Saito, R., Kimura, T., and Ikoma, T., Mat. Res. Soc. Syrup. Proc., 422:81-86 (1996) 42. Baumann, I., Beckers, L., Buchal, C., Brinkrnan, R., Dinand, M., Gog, T., Holzbrecher, H., Fleuster, M., Materlik, M., Muller, K. H., Paulus, H., Sohler, W., Stolz, H., vonder Osten, W., and Witte, O., Appl. Phys. A, pp. 33-44 (1997) 43. Kozanecki, A., Chan, M., Jeynes, C., Sealy, B., and Homewood, K., Solid State Commun., 78(8):763-766 (1991) 44. Tang, Y. S., Zhang, J., Heasman, K. C., and Sealy, B. J., SolidState Commun., 72(10):991-993 (1989) 45. Witte,O., Stolz,H., andvonderOsten, W.,J. Phys. D, Appl. Phys., 29:561-568 (1996) 46. Kim, S., Rhee, S. J., Tumbull, D. A., Reuter, E. E., Li, X., Coleman, J. J., and Bishop, S. G., Appl. Phys. Lett., 71:231-233 (1997) 47. Wu, X., H6mmerich, U., MacKenzie, J. D., Abemathy, C. R., Pearton, S. J., Schwartz, R. N., Wilson, R. G., and Zavada, J. M., Appl. Phys. Lett., 70:2126-2128(1997). 48. Klein, P. B., and Pomrenke, G. S.,Electron. Lett., 24(24): 1502-1503 (1988) 49. Torvik, J. T., Feuerstein, R. J., Qiu, C. H., Leksano, M. W., Namavar, F., and Pankove, J. I., Mat. Res. Soc. Symp. Proc., 422:199-204 (1996) 50. Silkowski, E., Yeo, Y. K., Hengehold, R. L., Goldenberg, B., and Pomerenke, G. S., Mat. Res. Soc. Symp. Proc. 422:69-74 (1996) 51. Pearton, S. J., Abemathy, C. R., MacKenzie, J. D., Schwartz, R. N., Wilson, R. G., J. M. Zavada, and R. J. Shul, Mat. Res. Soc. Symp. Proc., 422:47-56 (1996).

392

Wide Bandgap Semiconductors

52. Wu, X., H6mmerich, U., MacKenzie, J. D., Abemathy, C. R., Pearton, S. J., Wilson, 1L G., Schwartz, R. N., and Zavada, J. M., J. Lumin. 72-74:284-287 (1997) 53. MacKenzie, J. D., unpublished. 54. Choyke, W. J., Devaty, R. P., Clemen, L. I., Yoganathan, M., Pensl, G., and Hassler, C., Appl. Phys., 65(13):1668-1671 (1994) 55. Wu, X., H6mmerich, U., Namavar, F., and Cremins-Costa, A. M., Appl. Phys., 69(13):1903-1905 (1996) 56. Wu, X., White, R., H6mmerich, U., Namavar, F., and Cremins-Costa, A. M., J. Lumin., 71:13-20 (1997) 57. Kimura, T., Yokoi, A., Horoguchi, H., Saito, R., and Ikoma, T., Appl. Phys., 65:983-985 (1994) 58. Kimura, T., Hosokawa, I., Nishida, Y., Dejima, T., Saito, R., and Ikoma, T., Mat. Res. Soc. Symp. Proc., 422:149-154 (1996) 59. Culp, T. D., Wang, X. Z., Keuch, T. F., Wessels, B. W., and L. Bray, K., Mat. Res. Soc. Symp. Proc., 422:279-284 (1996) 60. Ren, F. Y. G., Michel, J., Sun-Paduano, Q., Zheng, B., Kitagawa, H., Jacobson, D. C., Poate, J. M., and Kimerling, L. C., Mat. Res. Soc. Symp. Proc., 301:87-95 (1993) 61. Chang, S. J., and Takahei, K., Appl. Phys. Lett., 65:433-435 (1994) 62. Torvik, J. T., Feuerstein, R. J., Pankove, J. I., Qiu, C. H., and Namavar, F., Appl. Phys., 69(14):2098-2100 (1996) 63. Payne, S. A., Chase, L. L., Chase, L. K., Smith, L. K., Kway, W. L., and Krupke, W. F., IEEE J. Quantum Electron., 28:2619 (1992) 64. Yoganathan, M., Choyke, W. J., Devaty, R. P., Pensl, G., and Edmond, J. A., Mat. Res. Soc. Syrup. Proc., 422:339-344 (1996). 65. Wang, X. Z., and Wessels, B. W., Appl. Phys. Lett., 65(5):584-586 (1994) 66. Ford, G. M., and Wessels, B. W., Mat. Res. Soc. Symp. Proc., 422:345-350 (1996) 67. Franzo, G., Priolo, F., Coffa, S., Polman, A., and Camera, A., Appl. Phys. Lett., 64:2235-2237 (1994) 68. Zheng, B., Michel, J., Ren, F. Y. G., Kimerling, L. C., Jacobson, D. C., and Poate, J. M., Appl. Phys. Lett., 64:2842-2844 (1994) 69. Lombardo, S., Campisano, S. U., van den Hoven, G. N., and Polman, A., J. Appl. Phys., 77:6504 (1995) 70. Priolo, F., Coffa, S., Franzo, G., and A. Polman, A., Mat. Res. Soc. Symp. Proc. 422:305-316 (1996) 71. Steckl, A. J., Birkhahn, R., Appl. Phys. Lett., 73:1702 (1998) 72. See MRS Bulletin, 24(9) (1999), for a description ofphotonic applications of rare earth doped materials.

9 SIMS Analysis of Wide Bandgap Semiconductors Robert G. Wilson

1.0

INTRODUCTION

Most semiconductors, including wide bandgap semiconductors, are used in the fabrication of devices for electronics or optics or optoelectronics, or electro-optics. Impurities may be intentionally introduced into the these materials to change their optical or electrical properties. Native or unintentional impurities may also affect the optical or electronic properties of these materials, and, in some cases, cause deleterious or uncontrolled effects. Secondary ion mass spectrometry (SIMS) is used to measure the presence, concentration, and depth distribution of these intentional or unintentional impurities, and changes in their depth distributions with processing. Wide bandgap materials are generally insulators; therefore, wide bandgap semiconductors may actually often be insulators. SIMS analysis of insulators poses a special issue of charging of the dynamic surface being sputtered by the incident primary ion beam. This subject is discussed in Ref. 1, but is summarized generally in the following. The charging issue is 393

394

Wide Bandgap Semiconductors

more serious for instruments with high secondary ion extraction voltage, or voltage placed on the target (sample for analysis), for example a CAMECA sector magnet instrument. Charging is less serious for low extraction voltage instruments such as quadrupole instruments. Both approaches use electron flooding of the target surface (directed beam or free electrons) to compensate the positive charge from the incident ion beam. In both approaches, the sputtering rate used to depth profile insulators is reduced from that used to profile semiconductors or conductors. In addition, a thin (30 nm) layer of a conductor may be deposited on the analysis surface, provided that it does not cause interfering masses for the desired analysis. Automatic variation of the potential placed on the target surface may be used with the CAMECA, to partially compensate for the change in potential caused by the incident ion beam (up to about 125 V). Insulating layers thinner than about 0.1 mm on a semiconducting substrate generally do not cause any difficulty using a CAMECA instrument. Analysis of semiconducting layers on bulk insulators like sapphire generally do cause difficulties using a CAMECA. In the material that follows: SIMS = secondary ion mass spectrometry; RSF = (SIMS) relative sensitivity factor, the calibration factor to convert secondary ion signal (counts/s) to atom density (cm-3), via the matrix signal measured under identical instrument conditions; wrt = with respect to; and E is often used to represent element, O for oxygen, M for a matrix element, etc., when secondary ions are discussed.

2.0

WIDE BANDGAP MATERIALS DISCUSSED HERE

A selected representative group of materials is discussed here, materials that are of interest today in the fields of electronics and optoelectronits. Those materials are diamond (5.4 eV), SiC (2.9 eV), ZnSe (2.7 eV), LiNbO 3, and the group III nitrides, e.g., A1N (6.1 eV) and GaN (3.4 eV).

3.0

S E C O N D A R Y I O N M A S S S P E C T R O M E T R Y (SIMS)

SIMS is a sensitive analytical technique (sputtering) that can detect all elements in all solid materials, with detection limits from 1013to 1016 cm -3,

SIMS Analysis 395 and to depths of many micrometers (which is compatible with the depths of dopants and the thicknesses of electronics and optoelectronics devices and other microelectronics structures). Depth profiles of intentional and unintentional impurities were measured in this work using both sector magnet CAMECA SIMS instruments (4fand 5f) and quadrupole instruments (PHI 6600) at Charles Evans and Associates. Oxygen primary ion bombardment was used to measure positive secondary ions, and cesium primary ion bombardment was used to measure negative secondary ions and, in limited cases, also Cs molecular ions. Additional details of the SIMS analysis techniques carried out in this work are described in Ref. 1. SIMS was used to measure changes in the depth distributions of elements grown in, implanted, or introduced unintentionally during processing steps.

4.0

SIMS ISSUES

Mixing of atoms in the dynamic surface being sputtered by the incident energetic ions and the equilibration/stabilization time/depth are issues that must always be addressed. Depth resolution in SIMS profiles is important for defining the structure of sharp interfaces and superlattices. Several factors affect depth resolution. Interface broadening is caused by surface topography and nonuniformity of layer thickness, and by ion mixing, which occurs within the penetration depth of the primary sputtering ions. If the layer thicknesses are uniform and the surface topography is good, then the SIMS experimental conditions become the determining factors. The mixing thickness decreases with decreasing primary ion energy and increasing angle of incidence (to a point). Reducing the primary ion energy is necessary to achieve the best depth resolution. Of the two most commonly employed primary ion species (oxygen and cesium), oxygen produces less ion mixing. The lowest practical oxygen ion bombardment energy is then used to produce the best depth resolution. For quadrupole instruments, this energy may be less than 1 keV. For CAMECA sector magnet instruments, this energy is often 1.5 keV/O (3 keV for 02). Another factor in achieving good depth resolution is sputtering rate combined with secondary ion collection time. The sputtering rate and the collection time must be adjusted to create enough data points to define a layer or interface accurately.

396

Wide Bandgap Semiconductors

Often, information is desired from SIMS profiling of unwanted impurity species (elements) in various materials. Some of the common impurity species are from the ambient vacuum or heated components of materials growth machines. The same ambient vacuum species exist in a SIMS instrument. The lower the ambient vacuum, the lower the sputtered secondary ion intensities of these species. The higher the sputtering rate during SIMS profiling, the lower the adsorbed density of these ambient species and the lower their sputtered secondary ion intensities. Thus, for the lowest backgrounds of these species, or the best detection limits, the lowest practical vacuum and the highest practical sputtering rate should be employed. Note that improved background and improved depth resolution cannot be achieved simultaneously because they vary in opposite dependence on sputtering rate. Often, separate profiles must be measured at very different sputtering rates to achieve depth profiles with good depth resolution, and with good detection limits or backgrounds for certain species (elements). These elements include H, C, N, O, Si (N 2 and CO can produce signals that interfere with Si), and all elements that may have a molecular interference when any of these elements are combined with the masses of the matrix materials (which may be dozens of masses in some cases). The sputtering rate in SIMS depth profiling of multilayer/ multimaterial structures is again an important issue. For fixed SIMS profiling conditions, the sputtering rates of all materials are different. Thus, the sputtering rate changes whenever an interface between different materials is crossed. The depth scale of a SIMS profile is usually obtained by measuring the crater depth at the end of each profile. This depth, divided by the sputtering time, yields the average sputtering rate. If this rate is applied uniformly to the profile, inaccurate layer thicknesses result if different materials are sputtered. To obtain accurate layer thickness, the sputtering rates of each and every material in the structure must be measured or otherwise known (from other work or published sputtering rates~for similar sputtering conditions). Then the total depth profile must be divided into layers of each different material and the appropriate sputtering rate applied to each layer. This capability is provided in the Charles Evans and Associates replot software used in this work, and by many other CAMECA instrument users.

SIMS Analysis 397 5.0

QUANTIFICATION

SIMS is a powerful analysis technique, but to be quantitative, a quantification system is required that relates the measured relative intensities of the secondary ions to absolute concentrations of the elements in the host matrix, as well as the accurate relative amounts of the matrix elements (host material) themselves. Different elements exhibit different relative secondary ion yields depending generally on their ionization potentials for positive secondary ions, and on their electron affinities for negative secondary ions. In the first case, sputtered atoms must lose an electron as they leave the sputtered surface (i.e., be ionized). In the latter case, they must take on an electron as they leave the sputtered surface, related to their electron affinity. Experimental work has shown that the secondary ion yields are not in all eases simply related to the ionization potential or the electron affinity of the elements; for some elements, there are other special features of the process that effect/alter the values of the secondary ion yields. All of this information is accounted for by determining the relative sensitivity factors (RSFs) for all of the elements in any or all host materials or matrices. This study has been largely completed and data for many materials have been compiled, including the materials discussed here. When RSFs have not been published for specific element-matrix combinations of interest, implanted standards can be prepared specifically for that purpose, or the systematics of RSFs in related materials can be used. Tabulations of SIMS RSFs have been published for diamond,[ 21but not for SiC, ZnSe, LiNbO 3, nor the III-nitrides, the materials chosen for discussion here. We have generated tables of SIMS RSFs for these materials using ion implanted standards, and they will be published in the future. For the purpose of this work, specific RSFs for the element-matrix combinations discussed here fi'om that work have been used, and some of them are listed here.

6.0

DIAMOND

Natural diamonds, synthetic diamonds, and CVD diamond films were characterized using SIMS, with a Cs beam and negative secondary ions for H, B, N, O, Si, and P, and an 0 2 beam and positive secondary ions for B, Fe, Mo, Ta, W, and Re. High mass resolution was used to profile N as 14N12Cbecause of the interference from 13C2-.B and P are intentional dopants for

398

Wide Bandgap Semiconductors

device applications. H, N, and O are unintentionally incorporated during growth from the ambient. Si, Fe, Mo, Ta, W, and Re may be unintentionally introduced from the growth apparatus or substrate. H can be electrically active in diamond films. Impurity densities were determined from ion implanted standards of the subject impurities. Natural IIa and IIb diamonds, HPHT synthetic diamonds, and diamond thin films grown using CVD techniques, undoped and doped with either B or P, obtained from a variety of sources, were analyzed using SIMS. Depth profiling was done using CAMECA sector magnet instruments and 8-keV 02 or 14.5-keV Cs primary beams for positive or negative secondary ions, respectively. The full 128-eV ban@ass was used except for high mass resolution measurements and measurements made using voltage offset. Bulk impurity mass surveys were made using an offset voltage to suppress interfering molecular ions. Quantification standards for all elements studied were implanted using a 400-keV, post acceleration mass separation custom ion mass spectrometer, in which the ion beam is mass separated at full energy using a double focusing magnet. The results of impurity analyses for six thin CVD diamond films are shown in Table 1, two doped with P (A and B), two undoped (C and D), one doped with B (E), one synthetic bulk crystal, and five natural crystals. The significance of these data depends partly on the detection limits (DL) for these elements in diamond, measured using SIMS. All of the CVD films and the synthetic crystal are seen to contain densities of H well above the DL, while natural diamonds 1, 2, and 3, have concentrations less than the DL. The last two natural crystals were selected because they were believed to have high H concentrations, which was verified by these analyses. Natural crystals 1 and 2 were selected because they were believed to have low concentrations of all impurities, including N, which was verified by these analyses as having less than the DL. Natural crystal 3, the synthetic crystal, and all six CVD films have significant N concentrations. The DL for O was high enough that no O concentrations greater than the DL could be detected in any of these diamond materials. Significant Si concentrations were measured in all six CVD films, but not in the synthetic nor natural crystals. The B-doped CVD films contained about 1 and 6xl 019 cm -3 B. The gray natural crystal was seen to contain significant concentrations of Na, A1, K, and Ca, and possibly other impurities. The yellow-brown natural crystal had fewer impurities. Fe was not detected in any of these diamond samples at densities greater than the DL of about 1 • 1014 cm -3 under routine analysis conditions.

Table 1. Impurities in Selected Commercial Diamond Films and Selected Natural Crystals (determined from SIMS analyses using the RSFs fi-omTable 2, determined from implanted standards) Synth.

Nat.

Nat.

Nat.

Nat.

Xtal

Xtal

Xtal

Xtal

Xtal

Xtal

Sumito

IA

IA

IIA

yellow

gray

8E20

9.7E18

4E18

<5E18

1.5E19 7E13

2E18 1.OE19 6E19 4E18 7E19

5E20 5E20 8E17 5E18

1.7E19 C3E18

4.1 E20 <2E18

C6E16 <3E18 2E15 2E17

3E18 3E17

Com’rc’l film A P doped

Com’rc’l film B P doped

Com’rc’l film C undoped

Com’rc’l film D undoped

Com’rc’l film E B doped

H Li Be B

9E20

1.E21

9E20

8E19

N‘ 0 Na A1 si P K Ca Mo Ta W

1.3E17 <2E18

&El 8

Elem.

8.5E17 2.OE20

1.4E17 4E18

>1E18 2.5E20

6E17 <2E18

1.5E20 5E16

1.5E20 5E16

*N measured with mass resolution of 7200 as 1zC’4N-

2E21

(3E16 <2E16

4E15 4E16

<3E16 <2E16 2E15 4E16

4E18 3E15 7E18

Nat.

400

Wide Bandgap Semiconductors

Boron doped CVD films can contain significant Mo, Ta, W, or Re concentrations (from filaments or holders); an example is shown in Fig. 1. B-doped CVD films can contain about 10 21 cm -3 B, and W from a filament can be kept low, as seen in Fig. 2. P-doped CVD films can contain >l 0 2o cm 3 P (as shown in Fig. 3, in which a P implant for SIMS quantification can be seen). As shown in Fig. 4, a B implant into a natural IIa crystal can be used to determine the density of B in the crystal (about 4 x 1016 cm -3 in this crystal) and the variation of density of B within the natural crystal. H densities in CVD films are about 1021 cm -3 as shown in Table 1. Redistribution of natural, film-grown, and implanted impurities can be measured using SIMS. H did not redistribute (diffuse) in diamond films during an 1100~ anneal, but did during a 1350~ anneal, more so for a fine grained film than for a coarser grained film.

10 ~

,. N~.o,..o~o I

=

HF CVD DIAMOND ON SI (15 V OFFSET)

1020

1019 A

E u 1018

a

1017

Ta 101e

:Re Mo

F

1014 0.0

I 0.S

I 1.0

I 1.S

2.0

DEPTH (l~m)

Figure 1. Depth profiles ofMo, Ta, and Re in a hot filament CVD-deposited diamond film.

SIMS Analysis 401 9S3ffO~-O40

1022=

I

I

I

'

' I"

' !

'~

i

e in DOPED LAYER = 2,1E21 cm ~

" TOTAL

1021 ~ CVD DIAMOND RL

l-CALIBRATION IMPLANTII 1011) L- 11B B42933E13 cm"2 II

E ~ookov

II

II

'~_ (RSR: 2.2rE21; 120 mx)

q

~l It

--J

It

:1

I|

.

101e 11

1017

1011) 0.0

0.5

1.0

1.5

2.0

2.5

2.5

DEPTH ~ m )

Figure 2. Depth profile for B implanted in a CVD-grown B-doped diamond film.

1021 _=-

I

I

~

I

~

I

~

I

I

=

CALIBRATION IMPLANT 6 " 1020 E

n

:z

o v

rn z 101 o ill Q X

~ 101e

P

-

P-DOPED CVD DIAMOND FILM ,

1017 0

1 2

~

i 4

!

I

i

6

I

,

,

0

DEPTH ~ m )

Figure 3. Depth profile for a P-doped CVD diamond film.

Implants of some elements in diamond films have been annealed at temperatures from 1000 to 1300~ and their thermal stability has been measured (redistribution) using SIMS. Li was found to redistribute along

402

Wide Bandgap Semiconductors

grain boundaries, whereas H did not, and it was concluded that Li is probably not a suitable dopant in polycrystalline diamond films. Most implanted H is electrically inactive, and the effect of implantation damage appears to be significant for Li, N, and P doping.

1020

' ~

1

'1

....

!

:l

1

1

B ..~ 10 TM

7

E u

z I.L/

2Z 0 i,,~

1018

1017

10 TM 0

! 0.1

I 0.2

t 0.3

0.4

0.5

DEPTH (l~m)

Figure 4. Depth profile of B in a IIb natural diamond crystal showing an implant to quantify the natural B density.

While some SIMS relative sensitivity factors (RSF) for diamond were published previously,[ 2] an updated and expanded version is given in Table 2. Recently, thick, free-standing plates of synthetic diamond have become available, in diameters up to 6 inches and thicknesses typically of 0.5 to 2 ~tm. These plates are grown using a variety of CVD techniques that include microwave plasma, arc-jet, and hot filament plasma. They can be obtained commercially polished on one or both sides. The cost of polishing is a significant fraction of the total commercial costs. The cost is also a function of the thickness of the plates and the quality of the diamond material, which is usually inversely related to the growth rate. That is, the cost of such material increases with the thickness and with the quality of the diamond material. A variety of both variables is available. One application of interest is thermal management of packaged semiconductor devices, especially high power devices and in satellite or avionics environments.

SIMS Analysis 403 Table 2. SIMS Relative Sensitivity Factors (RSF) in cm -3 for Positive (O) and Negative (Cs) Secondary Ions Measured for Elements E. The Matrix Reference is C

E l e m e n t 1 0 RSF

Cs RSF

RSF for EC- Element i O RSF Cs RSF

1H

2.0E23

2.4E23

Fe

1.7E20

3.2E23

Ni

3.3E20

2H

1.0E23

He

1E26

Li

7.9E18

Be

4.3E20

B

2.8E21

4.3E24 3.0E24 3.4E24

C

[1.8E23] [1.8E23]

N

3.2E22

O F

2.5E22

3.9E22

Na

3.6E18

3.3E25

Mg

5.0E19

A1

2.4E19

2.9E25

Si

1 . 5 E 2 1 2.3E23

P

1.7E22

1.8E23

S

1.E23

4E22

Cl

1.6E22

2.1E22

1E21

Zn

2.7E22

5.8E24

2.8E24

9.9E25

2.0E25

Ga

1.0El9

5.4E23

As

5.0E22 3.0E23

Se

1.5E22

4.2E22

2.5E22

Br

i2.3E22

1.5E22

Kr

2.5E23

1.7E23 7.4E24

!

Cu

i

Mo

1.0E20

Ag

1 . 2 E 2 1 1E23

Cd

<4E22

3.8E24

In

6El 8

3.5E23

Te

6.8E22

Cs

2.5E18

Er

4E19

5.7E24

RSF for EC-

1.2E25

1.7E25 1.2E22

Ar

3.7E22

Hf

1.7E21?

K

4.6E18

Ta

6E20

6E23

Ca

1.5E19

W

1.2E21

8E25

Hg

4E21

1.0E26

TI

1.0El 9 >6E25

Sc

1.5E19

Ti

3.9E19

V

6E19

Pb

2.4E20

3.4E23

Cr

5E19

Bi

2.5E21

4.7E23

Th

6E19

1.3E26

U

5E19

1.4E27

7.9E25

Materials from four suppliers using a variety of techniques (one of which was SIMS) were recently studied. SIMS analyses were used to measure the concentrations of all elements in these materials. Some materials were of final commercial quality and some were from progressively developed materials. The complete results of this study will be published in

404

Wide Bandgap Semiconductors

a year or so; some preliminary results are given here in Table 3. Data are shown for three materials from a development study, showing improvement. Typical results are shown for two selected commercially available materials. These diamond plates have a growth side and a substrate side and impurities vary with depth both from growth and from polishing. All of these issues are being studied. Table 3. Impurities in Diamond Materials Related to the Fabrication of Large Plates of Polished CVD Diamond, in PPM

Sample > Element

D square 0 2 SIMS Cs SIMS

D oval

D irreg.

OzSlMS

Cs SIMS

O z SIMS

Cs SIMS

2.7%

3.6%

450

3.6

H

150

2000

1900

4100

Li

0.001

310?

0.0037

0.83

B

0.09

N

90

2.8% 740

100

O

2.4

F Na

28

0.11

4.7

21

0.058

Si

(4.8)

CI

610

35 50

160

2.8%

3.5%

180

1300?

46

3

Ca

0.045

32

220

K

0.021

0.75

39

Cr

0.045

10

Fe

0.19

Ti

1.6

25 11 100

10

30

Co 0.038

100

70

42

7.5 2.3

Br Sr

1400

56

AI

Cu

500 110

11

Mg

Ni

780

0.007

Mo

90

Ba

0.18

0.022

SIMS Analysis 405

Acknowledgments. Some natural crystals were provided by Michael Seal of Drukker International, now of Sigilum. The synthetic crystal was a commercial Sumitomo diamond via Cal Tech. CVD films were provided by various commercial suppliers that will not be named to avoid any comparisons of developmental materials that may not be representative of what can be obtained presently.

7.0

SiC

Ion implantation is often used to dope SiC because diffusion temperatures are very high for SiC or diffusion does not work. SIMS can be used to measure the resulting depth distributions in bulk SiC or in SiC epilayers grown on various substrates, especially SiC of the opposite conduction type. Implantation has been studied in SiC for several applications, a few of which are noted below. One application is surface hardness; B and N[ 3] and Cr [4] have been studied. Doping and annealing is another application. See Ref. 5 for a study of Ga. B and A1 were studied,[ 6] and are the commonly used p dopants in device fabrication. N, P, Sb, and Bi were studied some years ago.[ 7] N is the most commonly used n dopant in device fabrication. Elevated temperature N implants in SiC are reported in Ref. 8. Er may be implanted into SiC to produce stimulated photoemission at 1.54 ~tm.[9] SIMS is the standard technique used to measure the depth distributions of these dopants in SiC. The implantation of B, A1, and Ga as potential p-type dopants, N as the primary n-type dopant, F as a possible isolation species, and Ti to quantify Ti that might diffuse into the SiC epilayer grown on TiC substrates have been studied. Only B and N are small enough to fit into the SiC lattice as substitutional dopants. The interface between a SiC epilayer and the TiC substrate measured using SIMS is illustrated in Fig. 5. In this case, the SiC layer was doped with B at about 7 x 10s8 em-2. The depth profile of A1 implanted in bulk cxSiC is shown in Fig. 6, where high mass resolution SIMS was used to determine carefully whether any I1B160 was present at the mass of 27A1. The depth profile for A1 implanted into a [3SIC epilayer grown on TiC is shown in Fig. 7. Depth profiles comparing B and A1 implanted in a [3SIC epi layer on TiC are shown in Fig. 8. A depth profile for N implanted into a [3SIC epilayer at 4 x 1017 c m -2 is shown in Fig. 9 for the measurement of negative secondary ions. The effect of the formation of SiN or CN at the

406

Wide Bandgap Semiconductors

peak of the distribution is also seen in Fig. 9 via the change in Si and C matrix signal intensities (change in sputtering rate and/or ion yield~the matrix effect). Corresponding results for the measurement of the same sample shown in Fig. 9 but using positive secondary ions and very high mass resolution are shown in Fig. 10. Depth profiles for Ga implanted at three fluences are shown in Fig. 11. A profile for Ti implanted into BSiC at a high fluence to be comparable with Ti atom densities that might diffuse during growth into the SiC epilayer from the TiC substrate measured using various positive secondary ions is shown in Fig. 12.

952s-2e-o43

1021 F

J

i

i

1020~A 1019k . . . . . . . . . .

~ .......

,..,

~: 1018-

' I

! /~

106

---~"/J~" -

105 0

2_ 104 0Z

f

(n

Z MJ

0 9 1017~"-

- 107

~

-

""

>" n-

-

[

- -=1 0 3 < Z 0

~

-

o

-

< 1016

0

/

102 (n'~'

SOn l 1015 1014 0

101

1

2 3 DEPTH (~m)

4

5

100

Figure 5. Depth profile of a SiC/TiC interface.

To summarize the application of SIMS to SiC: N and Ti must be measured using high mass resolution in SiC because of serious interference; the measurement of B, A1, Ga, and F is straight-forward. Important numbers for the use of SIMS analysis in SiC are the SIMS relative sensitivity factors, which have been determined and are given in Table 4.

SIMS Analysis 407

I 1020 ~ Rm = 0.250 lJm

I

i

10 TM

-~,10 TM o

~ 1017 zw t: lO16

=o F

i

os,

1015 ~-- ,,.,..'~. BO + 1014/ 0

I 0.5

~, I 1.0 DEPTH

(gin)

t 1.5

2.0

Figure 6. High mass resolution depth profile of AI implanted in SiC.

a525-26.o42

1020 I

I AI IONS I 200 keV AI 3.0 x 1014 cm-2 I~ SiC EPI SUBSTRATE _ RANDOM ORIENTATION Si (ARB UNITS) ROOM TEMPERATURE

10TM

V

A

?

E

~

IMPL4NT

NO ANNEAL

1018 SIMS: OXYGEN PRIMARIES POSITIVE SECONDARIES _

z ca :i lO17 SIC

TIC

AI_

10TM

lO15

I 0

0.5

1.0 DEPTH (l~m)

Figure 7. Depth profile of AI implanted in SiC.

1.5

2.0

408

Wide Bandgap Semiconductors

g525-~

10TM

I

I A I ~ 10TM E o

I

I

I

B IONS 100 keY AI IONS 200 keY 2.0 x 1013 cm-2 p SIC EPI SUBSTRATE RANDOM ORIENTATION ROOM TEMPERATURE IMPLANT NO ANNEAL SIMS: OXYGEN PRIMARIES DARIES

(n z 1017 w O

_

I

m

10TM

10TM

l

0

I

0.2

I',,

~

0.4 0.6 DEPTH (pm)

I

0.8

1.0

Figure 8. Depth profiles of B and AI implanted in SiC.

1023

~,

o,

X O

or~o-,e~

CI~ ~ N IONS C( e f ' / ~ 150 keV it',ON / N ~ 4.0xl0'Tcm "2 1022 ,4 Si2N t I~SIC EPI SUBSTRATE l/ ~ RANDOM ORIENTATION // I ROOM TEMPERATURE IMPLANT // ~ NO ANNEAL f/ | SIMS: ~/ ~ OXYGEN PRIMARIES 1021 ,,""~, t POSITIVE SECONDARIES ,,'"f ~ 1 HIGH MASS RESOLUTION


.........

"~s . . . . . . . . .

t

, 10TM

0

0.2

C (ARB UNITS)

, ,'vv. vc~ 0.4 0.6 DEPTH (pm)

0.8

Figure 9. Depth profile of a high fluence N implant in J3SiC.

1.0

SIMS Analysis 409

9525-28-O38 1022

A

1021 i F

|

z• tu (2 =S O

I

1020

I l |

~__ I [/

I

I

I

N IONS 150 keY 1.0 x 1016 cm-2 SlC EPI SUBS~A~ N RANDOMORIENTATION _ /~ ROOMTEMPERTUREIMPLANT / \ NO ANNEAL / ~ SIMS: I I CESIUMPRIMARIES / ~ SIN NEGATIVESECONDARIES / \ VERYHIGH MASS RESOLUTION-

'

~

30 Si (ARB UNITS) C (ARB UNITS,

.......",.~...............~ 1 3

10TM

10TM 0

J 0.2

I I 0.4 0.6 DEPTH (pro)

I 0.8

1.0

Figure 10. Depth profile of the sample of Fig. 13 except using positive secondary ions.

0730-00le2

1021

?A E

zuJ a :E

I

A IMPLANT 1020 --/ \ FLUENCE / [ ~ \ 3x15cm'2 1019 _~ ~ / 3 1 1 4 c m "2 I ~ 3 x 13 cm'2 1018

t/ \ \~

I/ | | :E :) 1017 ~..i .J I _ (.-1 10TM O I<[

I

10TM 0

~ ~ ~ ~ \ ~, \ ~ \ \\

"~,,o.s

200keV VARIOUSFLUENCES a SIC SUBSTRATE RANDOMORIENTATION .oo. TE.~.ATU.E,.~L'NTS

~ N

EAL

I 0.5

-_

I 1.0

DEPTH (pm)

Figure 11. Depth profiles of three Ga fluences implanted in SiC.

1.5

410

Wide Bandgap Semiconductors

9730.00-183

1021

I i

4aTi keY IONS 200 9 x 1016 cm -2 SIC SUBSTRATE RANDOM ORIENTATION ROOM TEMPERATURE IMPLANTS _ NO ANNEAL

1 0 20

A

?

E

u

I

1019 _

ol z uJ r~ 1018 - ~ - S ~t M S : OXYGEN P R I N A R I E S =s o _ ~_. POSITIVE SECONDARIES I

Z

1017

N 101e

101S

-

o

~..~

-

I 0.5

! 1 .o

1.5

DEPTH (l~m)

F i g u r e 12. Depth profiles of various molecules for Ti implanted in SiC.

Table 4. SIMS Relative Sensitivity Factors

Element/ion B+ A1+ Ga +

8.0

R S F (cm -3) 6.5E20 7.5E22 3.6E19 3.4E21 2.7E19

M a t r i x ref.

Element/ion

R S F (em -3)

M a t r i x ref.

12C+

N +

28Si+

NSi-

2.5E22 1.1E23

2sSi + 2sSi-

7.4E21 3.6E21

2aSi2sSi +

12C+

28Si + 2sSi +

.

F Ti +

ZnSe

ZnSe is representative of wide bandgap II-VI materials, and has been implanted with a variety of elements in an attempt to achieve electrical and optical activation. One such study was carried out by P. Lowen and K. Jones at the University of Florida, which included implantation with N and with C1 as potential n- and p-type dopants followed by rapid annealing. "Although good acceptor and donor activation were realized optically, no significant electrical activity was observed" was the conclu-

SIMS Analysis 411

sion of those studies. SIMS results showed that the depth profiles of N and C1 do not change with annealing for up to 10 s at 500 and 700~ respectively. No redistribution of N was observed for a 700~ rain anneal. The diffusivity of N in ZnSe was shown to be less than 5 x 10 .]7 cm2/s at 700~ suggesting that N diffuses substitutionally and not interstitially. Figures 13 and 14 show profiles for N and C1 implanted in ZnSe, measured using Cs SIMS, and Fig. 15 shows the profile of2-MeV Si implanted in ZnSe measured using Cs SIMS.

1020

~ o , ~,o N In ZnSe 125 keV 3 x 1014 cm "2

-

0.5

1.0

1019

101e

1017'

10~s 0.0

DEPTH (pm)

Figure 13. Depth profile of N implanted in ZnSe.

1020

-

Ci in ZnSe 200 keV 3 x 10 TM cm "2

1019

1 101a Q 1017

1016 0.0

0.5 DEPTH (l~m)

Figure 14. Depth profile of CI implanted in ZnSe.

1.0

412

Wide Bandgap Semiconductors

1019

I

I

I

I

I

I

SI In ZnSe 2.0 MeV 2 x 1014 cm "2

1018

G" E

Z 1017 UJ Q o 30SI 1018

_

101s 0

I

2

3

4

5

DEPTH (tim)

Figure 15. Depth profile ofMeV Si implanted in ZnSe.

9.0

L i N b O 3 (AND LiTaO3)

Sanford and Robinson[ 1~ briefly mentioned the use of SIMS for analysis of LiNbO 3. Wilson et al.[ 11] discussed the use of SIMS for depth profiling in LiNbO 3, especially hydrogen (SIMS technology described in Ref. 1). Canali et al.[ 12] showed unquantified SIMS data of a depth profile in LiNbO 3, but they concluded that SIMS was not useful because they apparently did not know how to correct for what they called the "matrix effect" between the exchanged and bulk material, or because their SIMS measurement was not entirely charge compensated. (The "matrix effect" can occur when secondary ion signals change when moving from one material to a second material in which the ion yields are different. This potential problem can be corrected for by proper use of SIMS quantification standards~measuring the SIMS RSFs in the two regions, and adjusting the data during quantification through a multi-material depth profile. Inadequate charge compensation in crossing the interface between two materials of differing charging [dielectric] properties may also cause or increase this effect. Well adjusted electron beam flooding of the interface can reduce or even eliminate this effect in certain cases.) Landheer et al.[ 13] used SIMS (and Auger) to study impurities introduced into LiNbO 3

SIMS Analysis 413 during processing. Si and O can be introduced by the use of silica furnaces and Na, from alumina furnaces. Novak et al.[14][151discussed SIMS of H, Li, Nd, and Er in both LiNbO 3 and LiTaO 3. Loni, et al.[ 16] discussed materials analysis and depth profiling ofLiNbO 3 waveguides using SIMS. 9.1

Hydrogen in LiNbO 3

SIMS was used to determine the effective diffusion coefficient of H in LiNbO 3 by implanting a fixed source of H into LiNbO 3, annealing at various temperatures, measuring the redistributed depth profiles using SIMS, and plotting the data as an Arrhenius plot to yield the effective diffusion coefficient. The depth profiles for H in LiNbO 3 exhibited classical "up-hill diffusion," which became another issue that SIMS demonstrated well. Implantation of 2H at 100 keV and a fluence of 5 x 1015 cm -2, and 1H at 1.0 MeV and a fluence of 1.0 x 1016 cm-2 into single crystal LiNbO 3 created a controlled source of H of equal to or less than 1020 cm -3 for diffusion. Then pieces of the implanted materials were annealed at temperatures from 100-400~ by 50~ increments, and the resulting depth distributions (diffused profiles) using secondary ion mass spectrometry (SIMS) with Cs primary ions and negative secondary ion mass spectrometry using a CAMECA instrument as described in Ref. 1 were measured. In one case, 2H was used to improve the detection sensitivity of SIMS for H, by about two orders of magnitude. No redistribution of H was observed in the samples until an annealing temperature of 250~ was reached. The SIMS profiles of H for the 100-keV implant are shown in Fig. 16 for deeper depths, and in Fig. 17 for the shallower region near the peak of the implanted depth distribution. The profile for 400~ was terminated, because of time limitations, before it reached the no-counts limit that would have occurred at some depth greater than 25 mm. Maximum diffused H density in the bulk crystal was in the mid 1018 cm -3 range. Profiles were observed that showed significant "uphill diffusion" in the deeper bulk crystal region for all temperatures from 250-400~ In the region under the peak of the implant distribution (Fig. 17), the H profile was seen to retain the approximate shape of the damage profile created by the implantation of the H as calculated using TRIM92, but to decrease in intensity (atom density) with increasing temperature. This behavior indicated that the H that does not leave this part of the profile to diffuse deeper into the crystal but remains bound to the defects in this distribution and that these defects are not annealed away for temperatures up to 400~

414

Wide Bandgap Semiconductors

1021 E

I

,

,

F

1020 ~.~ NO ANNEAL

~ ~

, 2H IN LINbO3 100 keY 5 x 10ls cm "2

,

=

[

"~

350oC

1019 I]IR

400~

~ u') 10le z IM a X

-

1017

101s

1015 0

5

10 15 DEPTH (~m)

20

25

30

Figure 16. Depth profiles of 2H implanted into LiNbO3 and annealed (deep).

1021

I

I NO ANNEAL ~ 300oC , ~ 5 0 o c 102~ - ~

, , 2H IN LINbO 3 100 key 5 x 1015 cm'2

~'E-"10191 ~ / ~3'0/~ ~ ~

1

iiiiI 1015L 0

I 1

I I 2 3 DEPTH (pm)

I 4

5

Figure 17. Depth profiles of 2H implanted into LiNbO3 and annealed (shallow).

SIMS Analysis 415 The diffusion depths for the 250, 300, and 350~ profiles were used to calculate values of diffusion coefficient D, which were plotted versus inverse temperature. The slope of the resulting line was used to calculate the value of the activation energy (Ea) for diffusion into the undamaged bulk LiNbO 3 crystal, which was 0.89 eV (plus or minus about 0.1 eV). The corresponding value of the pre-exponential factor was calculated to be 8.3 x 10 -3, giving the equation for diffusion of H in LiNbO 3 for these conditions of H density of D = 8.3 x 10-3 exp(-0.89 eV/kT) cm2/s. The profiles for the 1.0-MeV implantation energy (Fig. 18) exhibited poorer definition because lH was implanted and diffusion in either direction from the source of H is less well defined, due to the higher background of iH in the samples. Similar calculations for this case were not made. The shoulder on the deep side of the profile for the as-implanted case indicates that the sample probably reached a temperature near 250~ during implantation. Both the energy and flux were high enough that the resulting power density (about 2 W/cm 2) during implantation was sufficient to raise the temperature of the LiNbO 3.

1021E

I

I

I

l

=

1H IN LINbOs

1,0 MeV NO ANNEAL 102o~__1"0 x 1016cm'2 ~ 2 5 0 ~ ~E' I~ 3 0 C I I J300. 350~

kT; oo 1o,.

II- I~~ F" f.J' ~ j250~ .~ 101.~_ ~'~400o, ~ L ~ ~300"400~

1o,, 0

. 5

10 15 OEPTH (pm)

~

.1, 20

!

25

Figure 18. Depth profiles of 1-MeV IH implanted into LiNbO 3 and annealed.

In summary, the expression for diffusion of H in bulk crystalline LiNbO 3 is D = 8.3 x 10"3 exp(-0.89 eV/kT) cm2/s. The diffusion profiles for H deep into the bulk of LiNbO 3 from an implanted source of H (of about 1020 cm -3) exhibited strong "uphill diffusion," probably resulting from paired H atom diffusion in association with the diffusion of defects or defect-H complexes from the implantation damage. Therefore, this diffusion

416

Wide Bandgap Semiconductors

expression may be for defects in the bulk of LiNbO 3. As the LiNbO 3 crystal was annealed to increasing temperatures, most of the H leaves, but some was retained in the profile of the original implantation-caused damage (up to a temperature of 400~ therefore, the damage in the crystal was not removed (annealed) for temperatures up to 400~ 9.2

R a r e Earths in L i N b O 3 and L i T a O 3

The lanthanide rare earth elements, Ce, Pr, Nd, Er, Yb, and Lu, and the actinide rare earth elements, Th and U, were implanted into pieces of a 3-inch commercial wafer of LiNbO 3. Er was also implanted into a piece of similar LiTaO 3. The implantation energy was 300 keV for the lanthanide rare earths, and 600 keV for the actinide rare earths. The fluences were mostly 1.0 x 1014 cm -2 (for the sum of the isotopes of each element), but a few were different, namely, 2 x 1014 for Ce, 5 x 1012 for Lu, and Er was additionally done at 4 x 1013 cm "2. Pieces of some samples were annealed at 600 and 700~ and Er and Yb also at 800~ for 10 min in flowing dry nitrogen. Depth profiles were measured using a PHI 6600 quadrupole SIMS instrument for the lanthanide rare earths, and a CAMECA sector magnet instrument for the actinides, all using an oxygen primary ion beam. Standard SIMS profiling conditions for insulators were used, as described in Ref. 1, for both instrument types. Selected representative depth profiles are shown in Fig. 19 for Er asimplanted, and annealed at 700 and 800~ in Fig. 20 for four isotopes of Yb; and for Th in Fig. 21. The nature of redistribution for lanthanide rare earths in LiNbO 3 is illustrated in Fig. 19. Redistribution begins between 700 and 800~ The rare earth element moves primarily deeper into the LiNbO 3, exhibiting an "uphill diffusion" profile that results from interstitial-vacancy diffusion. A small amount also moves to the surface. SIMS detection sensitivity is sometimes better using the EO + molecule (for O bombardment) than for E +. Most SIMS detection limits for these rare earth elements in L i N b O 3 (and LiTaO3) are near 1 • 1016 cm-3; a few are higher, toward 5 x 10 ]6 cm -3, and Yb is near 1 x 10 ]5 cm -3. SIMS RSFs for the lanthanide rare earth elements in LiNbO 3 and LiTaO 3 are: for the quadrupole, and wrt O, 5 x 1020to 1.2 • 1021 cm -3, and wrt Nb, 2 x 1021 to 3 x 102] cm-3; and for Th using the CAMECA, and wrt both O and Nb, ~3 x 1023 cm -3, and for U, and wrt Nb, --8 • 1022cm -3.

SIMS Analysis 417

1021

1020

-

I

-

Er IN U N b O 3

-

300 keV 1 x 1014 c m "2 3 x 1013 cm "2 THIS ISOTOPE 0 2 SIMS POSITIVE S E C O N O A R Y IONS

-" -

I

I

_

-

"-=

..~

A

~" , , 1- , " 0 1 9

~

-

r Z

-

uJ Q =E

~

800~

110 MIN. A N N E A L

10 TM

.-:_

1017

~

1016 0.0

III 0.2

As IMPLANTED 00~ 110 MIN. A N N E A L

J 0.4 DEPTH (l~m)

I 0.6

0.8

Figure 19. Depth profiles of Er implanted into LiNbO 3 and annealed.

10.0 GROUP

III-NITRIDES

Group m-nitrides are wide bandgap, otten high temperature semiconductors that have a potential application (alone or in conjunction with other III-V materials technologies) for electronics and optoelectronics.[ 17] III-nitrides are otten grown as epitaxial layers (by MOCVD/MOVPE or MBE) on either sapphire or SiC, and at high temperatures, often near 1000~ Unintentional dopants, from either the growth chamber or from impurities in gases used to grow these materials (e.g. NH3), are often introduced during growth. Intentional n- and p-type dopants (otten Si and Mg) are introduced, possibly with impurities. SIMS is often used to measure the concentrations of these intentional and unintentional impurities in III-nitrides. This application of SIMS is discussed here in some detail. Some impurities, especially H, may have effects on the electronic or optical properties of these materials. Their redistribution under the higher temperatures used in the processing of these materials is an issue of some importance. SIMS can be used to determine these effects. SIMS is best quantified in these materials using ion implanted standards.

418

Wide Bandgap Semiconductors

9rjo-oo.o78 Yb'IN LINbO 3 300 keY 1 x 1014 cm "2 TOTAL 0 2 SIMS POSITIVE SECONDARY IONS 174yb (4 x 10 TMcm "2)

10 lg

[ _

72yb (2.5 x 10 TM cm "2) //----~173yb IO TM

~Eu

(2.2 x 1013 cm "2) 71yb (1.3 x 10 TM cm "2)

-= -_

:E

1016

101s

1014

0

0.1

0.2 DEPTH (pm)

0.3

0.4

Figure 20. Depth profiles of Yb implanted into LiNbO3. A fairly extensive study of the elements H, C, O, and Si in many IIInitride materials has been carried out. The implantation of lanthanide rare earth elements in these materials for potential optics and electro-optics applications has also been researched. The following discussion summarizes some of the work in these two areas. Collaborators in some of this work are acknowledged via references, but we wish to acknowledge especially the collaboration of S. J. Pearton and C. R. Abemathy of the University of Florida, J. M. Zavada of the An~y Research Office, and R. N. Schwartz and D. E. Gilder of Hughes research Laboratories. III-nitride materials studied in this work were grown by MBE, ion-assisted MBE, MOMBE, and various MOCVD processes at AT&T Bell Laboratories, Univ. of Florida, Univ. of New Mexico, Univ. of California at Berkeley, North Carolina State Univ., Univ. of California at Santa Barbara, Carnegie Mellon University, and Naval Research Laboratory on substrates of sapphire, SiC, or GaAs. We thank all of our collaborators at those laboratories for growing and sharing these materials.

SIMSAnalysis 419 1020

I

I

I

Th IN LINbO 3 600 keV 1 x 1014 cm "2 o~ S~MS POSITIVE SECONDARY IONS 1019

& 1010

1017

Th

101el 0.0

I 0.1

I 0.2 0.3 DEPTH (pm)

I 0.4

0.5

Figure 21. Depth profile of Th implanted into LiNbO3. SIMS of III-nitrides is especially influenced by the substrate that is used. Sapphire is the most commonly used substrate, and SiC probably the second most common. Sapphire is such a good insulator that difficulty is encountered in trying to compensate entirely for the charging that occurs at the dynamic surface during SIMS analysis when using a sector magnet instrument like a CAMECA. Even good electron flooding techniques are frequently inadequate to compensate for this charging. Charging is less of a problem using a quadrupole instrument, so quadrupole SIMS is often used for samples grown on sapphire substrates. For samples grown on SiC, a CAMECA instrument usually provides good analyses. In particular, mass interferences are significant for the analysis of Cr, Mn, Fe, and Ni in III-nitrides, and high mass resolution is required for these elements. Quadrupole instruments do not provide adequate mass resolution, but sector magnet instruments do. Therefore, CAMECA instruments are used to analyze these elements, which can be introduced from stainless steel components of growth systems. But this analysis is best performed only for samples grown on SiC. An example of such a high mass resolution analysis for GaN grown on SiC is shown in Fig. 22, which shows significant Fe and

420

Wide .Bandgap Semiconductors

some Mn at the beginning of growth of the nitride layer. Growth often begins with a high temperature nitridation step and possibly a high temperature buffer layer growth, during which time stainless steel may be very hot. Detection limits are seen to be about 1016cm-3 for Fe and about 1014cm-3 for Cr and Mn. Depth profiles for Cr, Mn, Fe, and Ni in a sample of GaN grown on sapphire (measured using a quadrupole instrument) are shown in Fig. 23 where detection limits of about 10 ]7 cm -3 are seen.

1020

I GaN ON SIC I I 02 SIMS POSITIVE SECONDARY IONS

101o

I

101s

109

1018

108 _

r,n I-

-

-

z

0 -.: 1 0 7 0 z _0 >,. n.,

u

1017

10 e ZO 0 U .i

0~ 1016

r,n

101s

- 10 s

1014

--_ 104

10131 0

_

_

I 0.5

I 1.0 DEPTH (l~m)

I 1.5

103 2.0

Figure 22. Depth profiles of Cr, Mn, and Fe in GaN (high mass resolution using a sector magnet instrument).

Analysis for H, C, O, and Si (four elements that are well analyzed in a single depth profile using Cs SIMS with either type ofinsmnnent) is routinely performed. Si is a common intentional n-type dopant or may be unintentionally introduced from SiO2 components of the growth environment. H, C, and O are often present in high densities in the growth environment from precursors; N is ot~en introduced using HN 3, and there may be small amounts of water in the HN 3 as well. SIMS quantification standards in nitrides have

SIMS Analysis 421 been developed by implanting H, C, O, and Si, and used in the routine quantification of SIMS analyses ofnitrides. SIMS relative sensitivity factors used in quantification will be published in the future. Many samples ofnitrides from a number of sources have been analyszed, including commercial sources of material and the laboratories mentioned above. There are at least two approaches to impurity analysis. One approach is a depth profile at a moderate sputtering rate to detail the depth distributions of the elements of interest; another is to analyze for one or a few elements at one time using very high sputtering rates. High sputtering rates reduce the contamination on the dynamic surface from the ambient background (gases). Quadrupole instruments typically have better ambient vacuum, but sector magnet instruments typically are capable of higher sputtering rates. Sector magnet instruments may then be run at a high sputtering rate with a possible corresponding reduction in detail in depth resolution, but with improved detection limits. Detection limits in a quadrupole may be superior for 0, under all conditions.

, . 1020 i CrlFe, N i , ~ C u l n G a N I ' 300 keV 3 x 1014 cm "2 0 2 SIMS POSITIVE SECONDARY IONS 101o

I

173o.oo.162 ' ~ 101

52Cr

,~/

108

Ga---~

10r

1,o.

,~..s6F.

-i

_;

,

1017

lo= 101 1016 L 0

,

J 0.2

~

I J 0.4 DEPTH (llm)

= 0.6

m

~100 0.8

Figure 23. Depth profiles of Cr, Mn, Fe, and Ni in GaN using a quadrupole instrument.

422

Wide 8andgap Semiconductors

An example of a SIMS analysis for a relatively "clean" sample of commercial Si-doped GaN grown on sapphire is shown in Fig. 24. Si is high because it is intentionally introduced. Si is possibly 10% electrically active in GaN, so may be introduced at about 102o cm -3 density to achieve an electrical doping of about 1019 cm -3. In this case, H appears to be high, possibly in relation to the Si. However, C and O have unusually low densities, which are the detection limits or less (2 x 1017cm-3 for C and 3 x 1016 cm -3 for O) and are indicated here. An example of a laboratory sample of GaN grown on sapphire but not doped with Si is shown in Fig. 25, in which the Si and H densities are much less, and the C and O are greater. A depth profile that is more typical of most samples of GaN is shown in Fig. 26. The densities of the four elements can be compared with the previous figures.

1022

"r

I

I

COMMERCIAL GaN SI-DOPED

973o-oo-,57107

[I

106

10Ol

~J 105 ~" ~.. 102o

-

w~~'101o Q =f 101e

C 1011

101

0

0.5

1.0 1.5 DEPTH (l~m)

2.0

Figure 24. Depth profiles of H, C, O, and Si in a sample of commercial GaN.

SIMS Analysis 423

10 23 I~

'

I

'

I

'

I

'

I

'

973o-oo-144 ! 10 4

l: GaN/SAPPHIRE Cs SiMS

1022

[_NEGATIVE SECONDARY IONSr._..."

.-= 10 3

1021

u

O

g

i'~

,

-

"_ 1 0 2 ~ < Q

H

.~

1018

1017

Z 0 ill U) 101

1016

1

101S

, 0

I 0.1

,

I 0.2

,

I 0.3

,

I 0.4

,

I 0.5

0o

DEPTH (l~m)

Figure 25. Depth profiles of H, C, O, and Si in a sample of GaN.

Examples of depth profiles of three elements (H, Ca, Se) implanted in samples orGaN and measured using SIMS are shown in Figs. 27-29. While we often use 2H to improve the dynamic range for studies of H in materials, here we used lH (Fig. 27), which would serve well for a SIMS quantification standard. The profile for Ca (Fig. 28) is similar to the one shown as Fig. 4 of Ref. 18, where Ca was studied as an alternative to Mg as a p-type dopant, and where high temperature annealing effects were also studied. Se (Fig. 29) was done as part of a study of all potential n-type dopants from columns 14 and 16 of the periodic table. Impurity densities from the studies of H, C, O, and Si mentioned above as of December 1996 are given in Table 5.

424

Wide Bandgap Semiconductors

~

1023 ~ GaN I~ Cs SIMS I- NEGATIVE SECONDARY IONS 1022

lo4

Io21

102

1o2o O

mzm~101o Q

H

102

Z

< 1018

101

1017

101s

101s

0

0.1

0.2 0.3 DEPTH (jam)

0.4

0.5

Figure 26. Depth profiles of H, C, O, and Si in a sample of GaN.

11.0 A C K N O W L E D G M E N T S The SIMS analyses shown in this work were performed at Charles Evans and Associates during the past 10 years. We thank our colleagues at Hughes Research Laboratories and other organizations who provided the samples that were used as illustrations in this paper, and the staff of SIMS analysts at Charles Evans and Associates. Financial and technical support were received from the Army Research office (Dr. John Zavada) for portions of this work, for which we are grateful, and from DARPA for the commercial CVD diamond plates portion of this work.

SIMS Analysis 425

1021 H IN GaN 50 key 2 x 10is cm.2 Cs SIMS NEGATIVE SECONDARY IONS

1020

-_

_

1019

1018

1017

,

I, 0.2

0

,

I , I 0.4 0.6 DEPTH (j.zm)

,

I 0.8

, 1.0

Figure 27. Depth profile of H implanted into GaN.

1020 I

' I ' I ' I ' ' GaN IMPLANTEO WITH Ca & ANNEALEO AT TEMPERATURES SHOWN

- 107 =10s

1019 F [: ~ A L/ I E ,~o1018

180 keY S X 1014 cm "2

\\

102 n

o2 SIMS

~

POSITIVE SECONDARY IONS

o)

lO4

if) z I1/ a

Z

>. m < z O r uJ r

=[ 1017 <

lO3 Q

NEAL 101s

lO2 _

101s

0

I 0.2

J

I 0.4

~

l 0.6

~

I 0.8

,

_ 101 1.0

DEPTH (pm)

Figure 28. Depth profile of Ca implanted into GaN and annealed at high temperature.

426

Wide Bandgap Semiconductors

Table 5. Impurity Densities in III-Nitrides (cm -3)

Nitride

Growth

GaN

MOCVD

Source

O

Si

5E17 4E18 4E18 2E19 2E19 7E19 1.5E 18 1El8 2E18

1El7 7E16 3E17 2E17 2E17 2E19 8E17 7E17 4E18

IE18 2E18 4E18 5E18 6E18 5E19 1El8 1El8 2.5E1

5E16 2E17 2E16 1El6 3E16 5E17 7E17 7E17

GaN

MOCVD

3E17 5E17 2E17 3E18 2E18

6E15 3E16 3E16 2.5E17 5E16

6E16 2E17 1El7 9E17 2E17

1.5E17 1.5E17 3E16 5E17

GaN

MOCVD

4E18 2E19

1.5E18 3E17

8E18 5E18

8E17 2E17

GaN

MOCVD


1El5 3E17

1El6 5E18

5E15

GaN

MOCVD

1El7

3E15

2E16

2E15

GaN

MOCVD

2E17 5E17

3E16 1El7

1El7 1El7

3E16 1El7

GaN

MBE

4E19 9E18

1.5E18 1.2E18

2E19 2E18

3E17 5E17

GaN

LA

1E20

1E21

1E21

1El9

AIN AIN

MOCVD CVD

1-2E18 3E18 4E20 1El9 4E18

7E17 3E15 1El8 5E18 1.5E18

2E18 5E18 5E19 8E18 8E18

1.5E19 3E16 1El9

A1N AIN

8E17

A1GaN A1GaN

MOCVD MOCVD

2E19 7E18

2E18 8E18

5E18 2E19

3E18 7E17

AIlnN

MOMBE

5E18

3E18

2E19

1E20

InN

MOMBE

8E20

7E19

1E21

2E19

SIMSAnalysis 427

,J/:in oo. m !

1020 I - -

'

I

~

1""

'

I

101e

-,, "

-|

10 e

-ll I l -t- 105

l,o,o

1

1016

1015

~

Se IN GaN 500 keV 1 x 1014 cm 2 Cs SIMS NEGATIVE SECONDARY IONS

10 2

0

|

|

0.2

0.4 DEPTH (pro)

0.6

( -d 01

Figure 29. Depth profile of Se implanted into GaN.

REFERENCES 1. Wilson, R. G., Stevie, F. A., and Magee, C. W., Secondary Ion Mass Spectrometry, Wiley, New York (1989) 2. Wilson, R. G., "Depth Distributions and Range Parameters for Elements Implanted into Single Crystal Diamonds and CVD Polycrystal Diamond Films," Diamond Films 90, Crans Montana, Switzerland (Sept, 1990); Surface and Coatings Technology, 47:559-571 (1991) 3. Roberts, S. G., andPage, T. F.,J. Mat. Sci.,21:457 (1986) 4. McHargue, C. J., Joslin, D. L., and Williams, J. M., Nucl. Instrum. Meth. Phys. Res. B, 46:185 (1990) 5. Burdel, K. K., Suvorov, A. V., and Chechenin, N. G., Sov. Phys. SolidState, 32:975(1990) 6. Avila, R. E., Kopanski, J. J., and Fung, C. D.,J. ,4ppl. Phys., 62:3469 (1987) 7. Marsh, O. J., and Dunlap, H. L., Radiat. Effects, 6:301 (1970) 8. Gardner, J., Rao, M. V., Holland, O. W., Kelner, G., Simons, D. S., Chi, P. H., Andrews, J. M., Kretchmer, J., and Ghezzo, M., J. Electron. Mat., 25:885

(1996) 9. Steckl, A. J., Devrajan, J., Choyke, W. J., Devaty, R. P., Yoganathan, M., and Nova.k, S. W.,s Electron. Mat., 25:869 (1996) 10. Sanford, N. A., and Robinson, W. C., Opt. Lett., 10:190 (9185)

428

Wide Bandgap Semiconductors

11. Wilson, R.G., Betts, D. A., Sadana, D. K., Zavada, J. M., and Hunsperger, R. G., J. Appl. Phys., 57:5006 (1985) 12. Canali, C., Camera, A., Della Mea, G., Mazzoldi, P., A1Shukri, S. M., Nutt, A. C. G., and De La Rue, R. M., J. Appl. Phys. 59:2643 (1986) 13. Landheer, D., Mitchell, D. F., and Sproule, G. I., J. Vac. Sci. Technol. A, 4:1897(1986) 14. Novak, S. W., Matthews, P., Young, W., and Wilson, R. G., Secondary Ion Mass Spectrometry SIMS VIII, (A. Benninghoven, K. T. F. Janssen, J. Tumpner, andH. W. Werner, eds.), Wiley, NY, pp. 471-474 (1992) 15. Novak, S. W., Kanber, H., Bar, S., Sridhar, C, and Wilson, R.G., SPIE, 1186:179(1989) 16. Loni, A., De La Rue, R. M., Zavada, J. M., Wilson, R.G., and Novak, S. W., ECOC Conference (1989) 17. Pearton, S. J., and Kuo, C. K., Materials Research Society Bulletin, pp. 17-57 (Feb., 1997) 18. Zolper, J. C., Wilson, R.G., Pearton, S. J., and Stall, R. A.,Appl. Phys. Lea., 68:1945(1996)

10

H y d r o g e n in W i d e Bandgap Semiconductors Stephen J. Pearton and Jewor W. Lee

1.0

INTRODUCTION

Hydrogen plays an important role in all semiconductors, and its properties have been widely studied in relation to Si, Ge, GaAs and other related materials. These studies have been summarized in past reviews and books on the subject.Ill-[ l~ It is well-established that atomic hydrogen, the much more mobile species relative to molecular hydrogen, can enter the semiconductor during growth or device processing steps and neutralize the electrical activity of both n- and p-type dopants and a variety of defect and impurity states. In GaN for example, the unintentional hydrogen passivation of Mg acceptors prevented observation of p-type doping for a time, delaying the achievement of light-emitting diodes and eventually lasers. This review will focus on the prevalence of H in the growth and processing ambients for wide bandgap semiconductors and the effect of this hydrogen. The different configurations of hydrogen in semiconductors and the techniques for hydrogen incorporation will be discussed, and the diffusion behavior of different forms of hydrogen in diamond, SiC, GaN and II-VI compounds will be detailed.

429

430 2,0

Wide Bandgap Semiconductors HYDROGEN INCORPORATION IN WIDE BANDGAP SEMICONDUCTORS

There are a number of methods available for introducing hydrogen into semiconductors. These may be classed into two groups: methods in which the hydrogen is introduced in an intentional and controlled manner, such as exposure to a plasma or by direct ion implantation; and those methods in which hydrogen is injected into the semiconductor in an uncontrolled, and often unintentional, way. Examples of the latter include crystal growth, simple cleaning, and fabrication processes such as boiling in water, wafer polishing in the presence of hydrogen-containing reagents, heat treatments in molecular hydrogen and avalanche carrier injection in Metal-Oxide-Semiconductors (MOS) structures. It is atomic hydrogen which is the active species for defect and impurity passivation, and, therefore, sufficient concentration needs to be incorporated into the semiconductors if significant changes in the electrically active concentration of these entities are to be observed.

2.1

Hydrogen Plasma Exposure

The most common method for hydrogen insertion is by exposure to a low-power-density H 2 plasma. Basically, any system in which a hydrogen glow discharge or plasma can be maintained, can be used to hydrogenate a semiconductor sample. The hydrogenation systems employed by most workers are relatively simple and generally consist of a quartz tube through which molecular hydrogen is pumped at a reduced pressure (0.10.3 Torr, typical flow rate <103 em3/min). The hydrogen plasma is excited by eapacitively or inductively coupling low-frequency (---30kHz) or radiofrequency (13.56 MHz) power via a high frequency oscillator[ Ill or a copper coil encircling the quartz tube.J12] Schematics of the arrangement used by Pankove and co-workers[ 111 and one of the systems employed in many of the early deep-level passivation studies are shown in Fig. 1. The sample is heated typically to 100-400~ for a period of 0.5-3 h, either downstream from, or directly immersed in, the plasma. This leads to hydrogen incorporation depths of anywhere from a few tenths of a micrometer to as much as 50 ~tm, depending on the material itself, and the doping density within it. The depth of hydrogen permeation can be measured by looking at the carrier profile in the sample, using spreading resistance or capacitance-voltage profiling, or, more directly, by measuring

Hydrogen in Wide Bandgap Semiconductors 431 the atomic distribution of the hydrogen (or deuterium) by Secondary Ion Mass Spectrometry (SIMS). The typical power densities used for the hydrogen plasma are in the range 0.1-1 W/cm 2. Although at higher power densities, the plasma is more ionized and hence is more efficient in providing atomic hydrogen to the sample. The higher ion and electron bombardment energies can cause damage to the near-surface region of the sample.

H2 IN ~ ~"

[ ~ ~P---G

= VACUUM SYSTEM

,,

t

t

Cz

Q

t

C1 SYSTEMA

RF IG E N E R A T O RI -

~OVEN VACUUM SAMPLE J THERMOCOUPLE H 2 IN

=

SYSTEM B Figure 1. Schematic representation of plasma hydrogenation systems. In system A, hydrogen is pumped through the quartz tube Q, and a plasma excited by inductively coupling 13.56 MHz power via a copper coil C2. The sample sits on a graphite block G which is heated by 440 kHz power coupled via another copper coil Cl. The temperature of the graphite block is monitored by a pyrometer P. In system B, a high-frequency oscillator is used to excite the plasma, and the sample is heated via a tube furnace.

Of course, it is possible to use commercially available plasmaassisted deposition or etching reactors in which the electrodes are inside the

432

Wide Bandgap Semiconductors

chamber. Hydrogen is introduced at reduced pressure into the system and ionized by an applied field. The use ofrf, microwave, or multipolar discharges is now routine in semiconductor processing[ 13] and most laboratories contain at least one reactor capable of being used for hydrogenation studies. In general, the energy distribution and composition of most hydrogen plasmas used to insert hydrogen into semiconductors is not well known, particularly for simple flow discharge systems of the type depicted in Fig. 1. What is certain is that the normal hydrogen plasma exposures used for defect neutralization are not usually as benign as might be thought. The plasma is neutral overall, but consists of an agglomeration of positive and negative ions, molecules, neutral atoms, and free radicals.[ 14]The hydrogen gas also exhibits emission from its vibrational levels, giving rise to the familiar light purple color of the discharge. The electron temperature in these plasmas is typically 1-10 eV with an electron density of109-1012 cm-3.[15]The neutral-gas density at 0.1 torr pressure is --3.1015 cm -3, with a mean free path for the atoms of-q). 1 cm. The neutral flux incident upon a sample at 0.1 torr is --1016particles cm-2/s, with an ion flux also of--1016 particles cm-2/s. These energetic ions produced at the plasma edge have energies in the range 100-1000 eV. The threshold energy for atomic displacement in most semiconductors is of the order of 15 eV, and therefore, near-surface damage can be introduced into the sample during the plasma exposure. This can be more significant for low-frequency discharges where the average ion energies are higher than for radio-frequency plasmas. Overall therefore, the sample is subjected to ion and electron bombardment and a high level of UV and visible light exposure unless optical baffles are used, and the plasma is contained upstream from the sample. Such systems have been constructed, but often are not necessary, depending upon the application. Several experiments have demonstrated a gentler technique for the incorporation of atomic hydrogen into Si,[16]-[18] Gel 19] or GaAs[ 2~ using a simple two-electrode electrochemical cell. A variety of aqueous acidic solutions can be used, the choice being determined by whether or not the acid attacks the semiconductor surface. Examples are HC1, H3PO4, and H2SO 4. A schematic of a system for cathodic hydrogenation is shown in Fig. 2. It is generally assumed that this method of hydrogen insertion is damage-flee, and certainly there is not ion bombardment. Similar defect passivation results are obtained using electrolytic hydrogenation compared to plasma exposure. Pure bombardment-related effects can be examined using He plasmas in place of H 2 plasmas. Under high-power density conditions, or when using very pure samples (doping density: < 1014cm-3),damage-related levels can be observed in the near-surface region aider direct exposure to a hydrogen plasma.[ 21]

Hydrogen in Wide Bandgap Semiconductors 433

VOLTAGE SOURCE PT OR C ELECTRODE

SUPPORT ROD

ELECTRICALLY/4 CONDUCTING CONNECTION

\

WIREI

......OHMIC i"r' CON~ I SAMPLE i H CONTAINING i ELECTROLYTE (eg. HCI)

J

ii

BEAKER

Figure 2. Schematic of an electrolytic cell for hydrogenation of semiconductors. The addition of small amounts of H20 or 0 2 (0.1-0.3%) can reduce recombination in hydrogen plasmas, and increase the density of atomic hydrogen by more than an order of magnitude.[15][22] This is useful in passivation reactions involving highly doped p-type Si (_>1018cm-3), where SIMS profiling[ 23] generally reveals that the injected hydrogen or deuterium is totally consumed by the aeeeptor dopants. In any study of diffusion or reaction kinetics, the available hydrogen supply should not become the rate limiting step. The concentrations of the different states of hydrogen that can exist in semiconductors (bound, atomic, or molecular) can be affected by the mode of hydrogen insertion. For example, a high flux of hydrogen ions entering n + Si can be neutralized by electron capture, and then associate to form molecules. An alternative mechanism would be attachment of the hydrogen to Si dangling bonds to form Si-H bonds. These might even be induced by the presence of the hydrogen near a Si atom.

434

2.2

Wide Bandgap Semiconductors

Hydrogen Implantation

The other common method for incorporating hydrogen into semiconductors is to actually implant the hydrogen, usually at low ion energy and high beam current. This is the technique generally used for passivating a-Si for solar-cell application.J7][ 24] The hydrogen is usually obtained from a broad beam, multi-aperture ion source of the type initially developed for space propulsion,[ 25] but now widely used in semiconductor processing (e.g., ion-beam etching, surface oxidation, or thin-film deposition). Commonly called Kaufman ion sources, they have a number of advantages for hydrogenation, including close control of H + ion energy, dose and dose rate, and much shorter exposure times (of the order of several minutes, rather than at least 30 minutes with plasmas). The major disadvantage, of course, is the extensive damage to the near-surface region (<0.1 ~tm from the surface) by the high level of hydrogen bombardment. Within the damaged layer, very large concentrations of hydrogen are incorporated locally, up to ~ at.%.[ 26] Typical ion energies are 400-1500 eV, at a current density of 0.5-1.5 mA.cm 2, for times long enough to achieve total doses of -~1017 H § ions/cm 2. Once again, the sample is held at elevated temperatures (up to 150~ and, in general, will heat up anyway because of beam heating. Conventional-energy (20--400 keV) ion implantation can also be used to incorporate hydrogen. [27]-[29]The range ofH § ions is roughly 1 ~tm per 100 keV energy in GaN and SiC and slightly more in Si. An example of a multiple energy implant into SiC in order to achieve a uniform hydrogen concentration is shown in Fig. 3. There is, of course, lattice-displacement damage associated with such a process. This damage can be annealed, but this can destroy the beneficial effects of the hydrogen in passivating defects. It has been observed, that low-energy hydrogen implants can be selfpassivating (i.e., the hydrogen attaches to the vacancies it has created in collisions with Si atoms). Similar profiles for H § implanted in GaN are shown in Fig. 4. Projected ranges for proton implants in diamond, GaN, and SiC are shown in Figs. 5-7. Whatever method of hydrogen insertion into semiconductors is used, one needs to be aware of the possible pitfalls of that particular technique. High leakage current in hydrogenated diode structures for C-V for DLTS is an indication of the presence of plasma induced surface damage or a contaminating layer deposited during the exposure. Photoluminescence and infrared spectroscopy can usually reveal defect damage-related transitions aider plasma hydrogenation. Of course, isotopic substitution of deuterium for hydrogen, followed by SIMS profiling, is the only sure method to see that the passivating

Hydrogen in Wide Bandgap Semiconductors 435 species actually was incorporated to the depth expected. Such experiments can be invaluable in the interpretation of hydrogenation experiments.

101' .j' ~

l

10 TM

4xlO:Zcm"2, lOk'eV 5xlO:Zcm"2,2$keV 6xl 0'zcm"2, 50keV 8xl 0:2cm"2,80keV lOxlOtZcm"z, lOOkeV

j= s

I-I + ---> GaN

1017

r~

10 TM

101S ,

0

,

:

,

2000

4000

6000

9

,

8000

D E P T H (.~) Figure 3. Hydrogen ion profiles in SiC resulting from multiple energy proton implantation.

'~

1010 t

O

101s

4xlOlZcm"2,lOkeV 5xlO:Zcm"2,25keV 6xlOlZcm"2,$OkeV 8xlOt2cm"2,80keY lOxlOtZcm"2, lOOkeV sum --->SIC ~ .I.

101r O

~

101e

~101s ()

TO00 "10'00"2000"30'0040'00"d00"60'00"7C DEPTH 1.~1

Figure 4. Hydrogen ion profiles in GaN resulting from multiple energy proton implantation.

436

Wide Bandgap Semiconductors

2.0x10 s

A

<9 1.5x10 s

J

~0

10 s

~

5.0x10 4

I

9

0

I

9

200

I

9

400

i

9

600

I

9

800

I

~000 1200 101

Ion energy (keV) Figure 5. Projected range of H + ions in SiC.

3.5x10 s

,

3.0X10 s "<~ 2.5x10 s

~ 2.0x10 s 1"5x10S k

10 s

5.0x104 9

0

I

200

9

I

400

9

I

600

9

I

800

Ion energy (keV) Figure 6. Projected range of H + ions in GaN.

9

l

9

1000 1200

Hydrogen in Wide Bandgap Semiconductors 43 7

1.6x10 s

A

<9

1.2xlo s

~9 8.0x104

I=, 4.0x104

9

)

I

200

.

9 I

4O0

,,

I

600

,

I

800

9

I

1000 1200

Ion energy (keV) Figure 7. Projected range of H + ions in diamond.

2.3 Survey of the Configurations of Hydrogen in Semiconductors A model system (namely Si), for which most understanding exists will be examined. The early high-temperature permeation data on hydrogen in Si showed it was a rapid diffuser (0.48 eV activation energy) and presumably occupied a tetrahedral interstitial site.[ 3~ As experiments on different aspects of hydrogen incorporation into Si were performed over the years, a number of inconsistencies with this picture appeared. For example, the expected high mobility of hydrogen near room temperature was inconsistent with infrared-absorption experiments that indicated implanted hydrogen was chemically bonded to Si atoms at defect sites, as in silane molecules. Early Electron Paramagnetic Resonance (EPR) measurements also failed to observe any unpaired paramagnetic atomic hydrogen in implanted Si. [31] Initial ion channeling experiments on deuterium-implanted Si found that most of the deuterium was located at a unique site displaced 1.6 A from a Si atom along the [ 111 ] direction away from the neighboring atom, the so-called antibonding site.[ 32]-[35] Theoretical calculations of this period were hampered by the use of small clusters only, but were unable to reproduce the experimental result. Extended Huckel Theory (EHT)[ 37]

438

Wide Bandgap Semiconductors

calculations by Singh et al.[36] found that the equilibrium position for atomic hydrogen in a perfect lattice was the tetrahedral interstitial site (Td). They also found that the energy of bound hydrogen would always favor saturating an available dangling bond.[ 34][35]The ground state of hydrogen at the T site was found to be in the valence band, in agreement with the calculation of Wang and Kittel,[ 36] but as the hydrogen was moved away from this position, the resonance level moved into the gap. In other words, hydrogen would be electrically active. The electrical levels associated with bound hydrogen were found to be those of the vacancy perturbed by the presence of one, two, or three hydrogen atoms. There was no electrical activity for fully saturated vacancy defects (VH 4 and V2H6).[38]-[39]These conclusions were basically in agreement with the results of calculations by Mainwood and Stoneham,[ 4~ suggesting the T site for neutral hydrogen, but in disagreement with the antibonding site suggested by Rodriguez et al.[ 42] A schematic of various sites for hydrogen in Si is shown in Fig. 8.

co

~S

,.s~8 "re"

co

Figure 8. Schematicrepresentation of different interstitial sites occupiedby various states of H in Si.

The formation of molecular hydrogen in semiconductors may be expected by analogy with its formation in vacuum, where it is favored over the atomic state. It has been postulated that the open regions around tetrahedral sites in a semiconductor lattice are vacuum-like and may be likely areas for molecule formation.[ 43] The coalescence of hydrogen atoms into molecules was also suggested as a way of explaining some diffusion experiments in which the low-temperature hydrogen permeation was considerably

Hydrogen in Wide Bandgap Semiconductors 439 lower than expected from an extrapolation of high-temperature data.j44][46] This assumption was also consistent with the observation of a high concentration of a relatively immobile, electrically and optically inactive form of hydrogen in the near-surface region of plasma-treated Si and GaAs. Recently H 2 has been detected in both Si and GaAs. [47][48] The hydrogen is usually always bound, either at lattice defects or impurity atoms. There has been some indication of isolated hydrogen in implanted S i as measured by electron spin resonance. On the question of the existence of hydrogen molecules, while there has been no direct experimental verification of their existence, a number of aspects of the diffusion behavior of hydrogen are consistent with their presence. Certainly in SIMS measurements ofdeuterated Si and GaAs, it is often observed that the deuterium is present at concentrations above that of the dopants. This "excess" deuterium is immobile at temperatures below 400~ and is electrically and optically inactive. Most researchers have explicitly assumed that these centers are molecules. Since hydrogen can exist in a number of charge states in semiconductors the reaction: Eq. (1)

H ~ + H ~ --~ H 2

is not necessarily the only one resulting in molecule formation. An analysis of the deuterium profiles in plasma-exposed, reverse-biased n§ junctions indicates that a competing reaction: Eq. (2)

H o + H o _-~ H 2 + h +

may also be important in p-type Si. [49] The microstructure of these molecular species is still unclear. A variety of theoretical calculations have suggested different configurations, including a simple diatomic species located at the T site with the atoms separated by 0.86 A and a binding energy o f - 2 eV,[ 5~ and several metastable diatomic configurations. Deak and Snyder[ 51] found that atomic hydrogen is more energetically favorable in the Si lattice than molecules, provided the atomic hydrogen is at the BC position. At relatively modest temperatures, however, some of the hydrogen can occupy the antibonding sites, and since the energy of the molecule is much lower (-1.6 eV) than that of the isolated atoms at the AB sites, then such molecules can form by the association of the atoms. A diffusion barrier of 0.56 eV was calculated for motion of the molecule through a hexagonal interstitial site. Deak and

440

Wide Bandgap Semiconductors

Snyder determined, by using MINDO/3 calculations, that a least two other di-hydrogen complexes can be stable in Si. These involve two hydrogens at neighboring BC sites, or one hydrogen at a BC site and the other at the nearest AB position. The former is expected to exist only in n-type material, where the T-configuration species would also be expected because of the predominant occupation of AB sites by the atomic hydrogen. Somewhat similar results have been reported by Chang and Chadi using ab initio pseudo-potential calculations.[ 521They proposed a diatomic complex with one hydrogen at a BC site and the other at a T site along the [111] axis, in which the two hydrogens essentially are the H § and Hspecies responsible for acceptor and donor passivation, respectively. This configuration would also be metastable, being higher in energy than an isolated molecule at the T d site. The possible role of this hydrogen in the motion of hydrogen is unclear, but the metastable complex would be expected to be more stable in n-type material. The most reliable direct measurements of the lattice position of hydrogen in silicon are those of Bech Nielsen.[ 53] He implanted 10 keV deuterium ions at a relatively low dose (8.1014 cm -2) into Si at 30 K, followed by channeling analysis using the D(3He,p) 4He nuclear reaction. Immediately after implantation, the D occupies interstitial sites, close to the C and M locations. This is in agreement with some theoretical calculations. The preceding analysis was performed in high-resistivity Si, in which the implanted deuterium concentration was far in excess of the impurity concentration in the material. The case where the deuterium or hydrogen concentration is comparable to the dopant concentration will now be examined (since hydrogen is associated with dopants to passivate their electrical activity). More work has been performed on shallowacceptor passivation in Si, and it has been clearly shown that all of the acceptor species (B, A1, Ga, In, and T1) are neutralized upon association with hydrogen. More than 99% of the acceptors can be de-activated by hydrogenation, and the thermal stability of the effect is a strong function of the acceptor species. In general, the larger the acceptor ion, the more apparently thermally stable is the passivation. Although the bonding configurations suggested to explain this passivation are dealt with in detail later, note that the initial model suggested by Pankove et al.[ 53] and studied theoretically by DeLeo and Fowler[ 54] has gained general acceptance. This involved the formation of a Si-H bond, leaving the acceptor three-fold coordinated with its remaining Si neighbors. The hydrogen atom is located close to a bond-centered position with the acceptor displaced from a substitutional site towards the plane of the three other Si neighbors.

Hydrogen in Wide Bandgap Semiconductors 441 By contrast to the case of hydrogen passivated acceptors, the shallow donors in Si (P, As, and Sb) usually show only a very weak deactivation after exposure to atomic hydrogen. Under many conditions, the change in electrically active donors is negligible.L55]-[ss]The effect is less thermally stable than passivation of the aceeptors and the stability is only weakly dependent upon the donor species.[ 59]The model suggested for donor passivation is formation of a complex in which the hydrogen is located near a tetrahedral interstitial site of a Si atom adjacent to the substitutional donor. In this complex, the H is calculated to behave as an acceptor, with charge transferred from the donor to the hydrogen leaving an overall neutral complex.[ 6~ There have been a number of channeling experiments performed on ptype Si that has been deuterated by plasma exposure and other methods. [611[64] There is agreement that, once again, the majority of the deuterium occupies a near-bond-centered site with the acceptor displaced from a substitutional site. A small portion of deuterium is located in a backbonded position at a near-tetrahedral site. A measured axial scan was found through the (100) and (110) axes, together with a fit to the deuterium scans (assuming 87% at a near bond-centered site and 13% at a near-tetrahedral site). Taken together with channeling curves for the { 111 } plane, it can be concluded that the majority of the deuterium occupies a position displaced 0.2 A from a bond-centered site along a (110) direction perpendicular to the { 111 } direction connecting the two nearby substitutional sites. The remaining deuterium atoms were displaced 0.1 A from a tetrahedral site.[ 631 The near-bond-centered deuterium was consistent with the model proposed earlier for acceptor passivation. A schematic of H occupying a BC site passivating an acceptor dopant is shown in Fig. 9. Some of the first-published results suggesting the bond-centered site as the stable position for neutral hydrogen in Si were included in the context of the properties of hydrogen in diamond and the location of anomalous muonium in Si. [65]-[67] These calculations were later expanded to include hydrogen in Si.[68] The host Si crystal was simulated by a variety of clusters, with hydrogen atoms saturating the dangling bonds. Four or five host atom shells around the various sites investigated were used. First and second nearest neighbor Si atoms were allowed to relax. The calculational techniques employed were the approximate ab-initio method of partial retention of differential diatomic overlap and ab-initio Hartree-Fock scheme with various basis sets. The potential energy surface for neutral interstitial hydrogen had two minima. The most stable was a relaxed BC site, with the Si-Si bond expanding by approximately 34% to accommodate

442

Wide Bandgap Semiconductors

the hydrogen. This resulted in a Si-H bond length of 1.58 A in the bridged SiH-Si bond. The energy level associated with hydrogen at the BC site was found to be deep in the bandgap. The other minimum was found at the tetrahedral interstitial (T) site with the energy level for hydrogen at this position being below the top of the valence band.[ 68]

I I I

C) I

Figure 9. Schematic of H at a BC site, passivating an acceptor (shown shaded).

Similar results were obtained by a group at the State University of New York at Albany, [69][70]using an open shell version of the modified intermediate neglect-of-diatomic-overlap (MINDO/3)[7]] method. In these calculations, the hydrogen atom was placed at all of the major interstitial sites, and the total energy of the hydrogen and of the first two shells of Si atoms were minimized. Two minima, both involving considerable lattice relaxation were found. The BC site was the minimum energy position for H ~ and H § The two Si neighbors of the hydrogen were pushed back symmetrically along the (111 } axis by 0.42 A, with a Si-H bond length of 1.64 A. The energy of the hydrogen atom relative to the lattice was 0.12 eV lower in the BC site than in a H a molecule located at the T site. In other words, molecular formation was not favored over isolated hydrogen, as was indicated from earlier calculations. The diffusion barriers for the neutral hydrogen atom, and for the proton, were 0.84 and 0.70 eV, respectively, at the M point (which is the midpoint between two second-neighbor BC sites). In p-type Si, the hydrogen gave indications of

Hydrogen in Wide Bandgap Semiconductors 443 showing donor-like behavior if it moved off the BC site. An additional local minimum for the atomic hydrogen was found at a distance of 0.42 A from the T site in the (111) direction, the antibonding (AB) site. The activation energy for diffusion between AB sites was ~0.4 eV. For the case of H + at a BC position, the proton seemed to be neutralized by the lattice and exhibited the same diffusion path as for the neutral species. We note that these two minima were consistent with the channeling results in undoped Si. To summarize, the majority of the hydrogen was at the BC site, with a secondary potential energy minimum at the AB side. Molecules would be located at the T site, if formed. First principles, total-energy pseudopotential calculations have also been performed on the stable configurations of hydrogen in n- and p- type Si.[ 60] The lowest energy for hydrogen was in the bond-centered site in Bdoped Si. The hydrogen in the BC position was noted to be neutral, with the HSi hybridization removing the electrical activity of the boron acceptor. In P-doped Si, the stable position of hydrogen was found to be a near-T site adjacent to the substitutional donor. The calculated H-stretching vibrational frequency for H passivating P was not in agreement with the experiment, [72] and indicated inaccuracies in the calculational method. The hydrogen in P-doped Si was found to act as an acceptor, with charge transfer from P to H. The HB pair is more stable than H 2 for hydrogen in either a BC site or an AB site, whereas, in n-type Si, only the T site is more stable than molecule formation. More recent calculations found that, in the most stable P-H configuration, the Si atom to which the hydrogen is attached moves -0.6 A away from the p.[73][75] The calculated vibrational frequencies of this P-H configuration are in much better agreement with experiment results. A schematic ofH passivating a donor is shown in Fig. 10. Chang and Chadi 975]postulated that, in n-type Si, the formulation of a metastable diatomic complex (labeled H2*), with hydrogen atoms occupying BC and T d sites along the (111 ) axis, is relatively stable. The most stable position of hydrogen in n-type Si is at the T d position of the silicon and negatively charged. The overall reaction for donor passivation is then: Eq. (3)

H ~ + P + e- --~ H- + P+ --~ (HP) ~

when phosphorous is the donor. The H 2 molecule was found to be 1.6 eV more stable than two dissociated neutral interstitial atoms. The P-H complex was less stable than B-H, so that donor passivation was less efficient than acceptor passivation. In n-type material, the formation and diffusion of

444

Wide Bandgap Semiconductors

the metastable diatomie complexes led to slower diffusion of hydrogen in n-type Si relative to p-type Si.[75]

I I I

(-') J

I

I I

(

) l'~ I

Figure 10. Schematic of H at a AB site, passivating a donor (shown shaded).

In summary, therefore, theory finds hydrogen predominantly located at a bond-centered site in undoped Si but this has not been observed experimentally except perhaps in the Russian EPR work. In p-type material (where it is in a positive charge state), hydrogen is located at a bondcentered site, passivating the electrical activity of an aceeptor. In n-type material, hydrogen is located at a tetrahedral interstitial site passivating the electrical activity of a donor. It is in a neutral or negative charge state in ntype Si. The possible charge states and configurations include: H +, H ~ H 2, or Si-H in p-type Si; H-, H ~ or H 2 in n-type Si; and H ~ H 2, or Si-H in undoped Si. It is worthwhile at this point to compare the results for hydrogen to those for its lighter analog, muonium. In this ease, the proton is replaced by a positive muon (It+), with an average lifetime 0f2.2 ~s.E761Alarge amount of experimental data are available concerning the lattice location of the two types of muonium, so-called normal muonium and anomalous muonium. Normal muonium actually appears to be a metastable state located at the T D site, and, in some semiconductors, it makes a transition to the more stable state (anomalous muonium) which is located at a BC position near the

Hydrogen in Wide Bandgap Semiconductors 445 center of a covalent bond. Both species appear to be in a neutral charge state. In suggesting the BC site as the location for anomalous muonium in Si, Cox and Symons [77] also pointed out that this was also the logical site for hydrogen. Theoretical calculations by Estreicher [68] and Estle et al., [66] together with experimental data from Westhauser et a1.[78],Patterson et al.,[76] and Kiefl et al. [79], provided strong evidence for anomalous muonium being located at the BC site. A paramagnetic hydrogen center, detected by Electron Paramagnetic Resonance (EPR) in proton implanted Si at low temperature (<150K), and labeled AA9 shows similar characteristics to anomalous muonium.[ 8~ This is evidence for the presence of neutral hydrogen at a BC site in Si.

3.0

H Y D R O G E N IN GaN

Atomic hydrogen plays an important role in GaN, passivating the electrical activity of Mg aeceptors during cool-down after Metal Organic Vapor Deposition growth, and thereby preventing achievement of high p-type doping levels unless an annealing step is performed. Hydrogen is easily incorporated into GaN and related materials during many different process steps, including boiling in water, wet chemical etching in KOH-based solutions, dielectric deposition using Sill 4, or dry etching with C12/CH4H2/Ar Electron Cyclotron Resonance plasmas. In each of these cases, deuterated chemicals were employed and the incorporation of deuterium into GaN using high sensitivity SIMS measurements was detected. Hydrogen was found to have a diffusivity of>10-11em2.s -1 at 170~ and to passivate the electrical activity of Mg and C aceeptors, and also the native shallow donors in InGaN and InA1N. Annealing at 450-500~ restores the electrical activity, but the hydrogen does not physically leave the films until much higher (_>800~ temperatures. The activation energies for dopant reactivation are of the order of 2.5 eV in bulk samples, where hydrogen retrapping is expected to be significant. Hydrogen plays an important role in semiconductors, passivating the electrical activity of shallow and deep level impurities.[ 81] Unintentional incorporation of hydrogen can therefore lead to uncontrolled variations in conductivity of a semiconductor, and this effect has been observed for virtually all group IV, III-V, and II-VI materials. The properties of hydrogen in Group III nitrides and their alloys are recently attracting attention

446

Wide Bandgap Semiconductors

because of the discovery that post-growth annealing is required to attain high conductivity.[ 82] This problem will be especially evident in materials grown using NH 3 as the nitrogen precursor. Past studies in other III-V semiconductors have shown that metalorganics such as trimethylgallium [Ga(CH3)3] can also produce significant hydrogen incorporation.[ 83] It is also well established in conventional III-V materials, that hydrogen can be incorporated during virtually any processing step, since most of the gases and liquids that are used contain at least some hydrogen.[ 84][85] Combined with the high diffusivity of atomic hydrogen in semiconductors, it is then quite common for hydrogen to disrupt the electrical properties of nearsurface regions. For photonic devices such as light-emitting diodes or heterostructure diode lasers, the effect of hydrogen in ternary and quaternary nitrides will also need to be understood, because these will comprise the active or cladding layers. Metal organic Chemical Vapor Deposition (MOCVD) of GaN occurs by the reaction: Eq. (4)

(CH3)Ga + NH 3 --->GaN + CH 4 1' + H 2 1"

While the growth temperature is generally around 1040~ there is a long cool-down under a NH 3 ambient during which atomic hydrogen can easily diffuse into the GaN.[ 81]Undoped GaN is generally n-type due to the presence of native defects or possible impurities, and thus, since this is relatively uncontrolled, the effects of any hydrogen passivation will not normally be noticed. When the GaN is intentionally doped with Mg from the Cp2Mg (cyclopentadienyl magnesium) source, it is invariably found that the material must be post-annealed at >700~ in order to dissociate neutral hydrogen-magnesium pairs that prevent achievement of p-type conductivity:[ 82] Eq. (5)

( M g - H) ~ ~ M g + H +

Similar behavior is observed to a greater or lesser extent in virtually all other semiconductors grown by gas-phase techniques.j86]-[ 1~ However, the high growth temperature (leading to a longer cool-down time than more conventional materials) and the high thermal stability of hydrogen bonding in GaN means that unintentional passivation is likely to be significant. There are many opporttmities for atomic hydrogen to diffuse into semiconductors during processing, since essentially all of the chemicals

Hydrogen in Wide Bandgap Semiconductors 447 that come in contact with the surface contain hydrogen. [102]In Si and III-V semiconductors, such as GaAs and InP, it is well established that hydrogen is among the fastest diffusing impurity,J1~ 1~ and that its solubility at temperatures _<400~ is largely determined by the density of sites to which it can bond.[ 1~ Therefore, in most situations, the concentration of hydrogen in the near-surface region will be related to the doping density in the material. The native defect and impurity concentrations in GaN appear to be relatively high based on photoluminescence and secondary ion mass spectrometry (SIMS) measurements.[ 82] 3.1

M O C V D Growth

A typical growth sequence for MOCVD of GaN on A1203 substrates begins with degreasing of the substrate and cleaning in HC1.[I04]-[106] This was followed by in-situ cleaning in a H 2 flow at 1070~ Next, the GaN buffer layer was grown in order to overcome the large lattice mismatch between the GaN and substrate. A NH 3 flow of 4-20 liter/min and TMG flow of 80-120 ~tmol/min were used to produce a 200-300 A layer of amorphous or polycrystalline GaN at 530~ This thin GaN layer was crystallized by a ramped thermal process. Finally, p-type GaN epitaxial films were grown by the reaction ofNH 3 and TMG at 1040~ with a flow rate of 4-20 liter/min and 150-0 ktmol/min, respectively. The pressure was 30-300 xopp, and the rotating speed was 500-1,000 rpm during the growth process. Biscyclopentadienyl magnesium (CpzMg) was used as the p-type doping precursor. Before the p-type doing, an undoped GaN film growth was achieved with a background carrier concentration ofn~5 x 1016cm-3. A growth rate of 2-3 ~m/h was achieved during the growth cycle. The films were transparent and featureless with a thickness uniformity of--2% across a 1 in. substrate. The growth rate was slightly reduced with an increased flow of CP2Mg. Figure 11 contains a secondary ion mass spectrometry (SIMS) profile of a Mg-doped GaN film. The Mg concentration in this sample was ~5.5 x 1019 atoms/cm 3 uniformly through the film thickness.[ 1~176 The SIMS profile from the annealed sample showed the identical Mg atomic concentration compared to the as-grown sample, indicating that Mg atoms were stable during the thermal annealing process. Since H 2 was used as the reaction carrier gas, the incorporation of H 2 into the deposited Mg-doped GaN films was also studied. The SIMS study showed different concentrations o f H (~1 x 1018cm-3) from the annealed one. This indicated that the

448

Wide Bandgap Semiconductors

passivation of H atoms caused the highly compensated as-grown Mgdoped GaN. The post-thermal annealing broke the bonding between H and Mg. As a result, Mg atoms were activated, while the H atoms were free and evaporated from the films, evidenced by the reduction of H atomic concentration (one order lower). This observation was also consistent with a similar report.[~07]

10 2t o

o

Mg

10 2~

o o!,.~

before annealing

Mg after annealing 9. I 0 Is

H before

"

annealing

.....

9 9 . . . - - , ... ........-,. ..'..',...... ...... . .-..-.,,...,.,,..,..,.,, ... : , 9.... , . ~ , , "

0

"~ I017 s % . .

H after annealing 9

1016

0

AI

"

9

:. --..

"" "

i

.. - ~ .

Depth (~tm)

~'

i.5

"

'

'2

Figure 11. SIMS depth profile of MOCVD grown GaN(Mg) before and after 700~ 60 min anneal in N 2.

As shown in Fig. 12, the hydrogen concentration tracks that of the Mg in as-grown GaN(Mg) layers, suggesting that there is trapping of hydrogen at the acceptors. This phenomenon occurs in virtually all p-type semiconductors grown with gas-phase techniques.[95][ 1~ In both Si and GaAs[l~ 119],injection of minority carriers either by forward biasing of a diode structure or illumination with above-bandgap light, produced dissociation of neutral acceptor-hydrogen or donor-hydrogen complexes at temperatures at which they were normally thermally stable. While the details of the reactivation process are not clearly established, it is expected that for an acceptor A the reactions likely can be described by: Eq. (6)

( A H ) ~ <--->A

H + + e

+ H +

<--->H ~

Hydrogen in Wide Bandgap Semiconductors 449 The neutral hydrogen most likely forms diatomic or larger clusters with other neutral or charged hydrogen species.[ x2~

1020 ,,i, .,,,,

?A E~

m,

9H 0 Mg

r

V

Z O m I-,-

< 1019

It: I...-

Z

~

0

mm

mmm

LM

m

r

m

Z

0

0

1018

1

!

.,J

I

LI

I1~

10

....

I ..

.!

... I..

1 1 111_

100

Cp2Mg CARRIER FLOW RATE (cm3/min) Figure 12. Mg and H concentrations as a function of CP2Mg carrier flow rate for as-grown GaN(Mg)--afier Ref. 107.

Recently, there has been a lot of interest in the stability of hydrogen passivated Mg acceptors in GaN. Amano et al.[121]first demonstrated p-type conductivity in GaN (Mg) after an e-beam irradiation process near room temperature. Later Nakamura et al.[82] showed that simple thermal annealing at --700~ also reactivated the Mg acceptors. It was clear that atomic hydrogen remaining in the GaN, after growth by metal organic chemical vapor deposition (MOCVD) with NH 3 and (CH3)3Ga grown by MOCVD, was annealed under N 2 for 20-60 mins at-700~ to achieve the full level of p-type conductivity.[ 82][1~176 The mechanism for acceptor activation during the e-beam irradiation process has not been studied in detail, to date. To establish that minority carrier enhanced debonding of Mg-H complexes

450

Wide Bandgap Semiconductors

in GaN was responsible for this phenomenon, the effect of forward biasing in hydrogenated p-n junctions was examined. The reactivation of passivated acceptors obeyed second order kinetics and the dissociation of the Mg-H complex was greatly enhanced under minority carrier injection conditions. The samples were grown in c-A1203 by MOCVD using a rotating disk reactor.[ 1~176 After chemical cleaning of the substrate in both acids (H2SO4) and solvents (methanol, acetone), they were baked at 1100~ under H 2. A thin (_<300 A) GaN buffer was grown at 510~ before growth o f ~ l ~tm undoped material, 0.5 ~tm of GaN(Mg) with a carrier density of 5 x 1018 cm-3. Some of the samples were hydrogenated by annealing under NH 3 for 30 mins at 500~ This produced passivation of the Mg aceeptors but had little effect on the Si donors. Mesa p-n junction diodes were processed by patterning 500 ~tm diameter TiA1 ohmic contacts on the n-GaN by lift-off and then performing a self-aligned dry etch with an Electron Cyclotron Resonance BC13/Ar plasma to expose the p-type GaN.[ 911 E-beam evaporated NiAu was pattemed by lift-off to make ohmic contact to the p-type material. The carrier profiles in the p-type layer were obtained from 10 kHz capacitance-voltage measurements at room temperature. Anneals were carried out in the dark at 175~ under two different types of conditions. In the first, the diode was in the open-circuit configuration, while in the second, the junction was forward biased at 9 mA to inject electrons into the p-type GaN. After each of these treatments, the samples were returned to 300 K for re-measurement of the net electrically active acceptor profile in this layer. Figure 13 shows a series of acceptor concentration profiles measured on the same p-n junction sample, after annealing at 175~ under forward bias conditions. After the NH 3 hydrogenation treatment, the electrically active acceptor density decreased form 1.5 x 1017 cm -3 to-~6-7 x 1016 cm -3. If the subsequent annealing was carried in the open-circuit configuration there was no change in the carrier profile for periods up to 20 hr at 175~ By sharp contrast, Fig. 13 shows that for increasing annealing times under minority carrier injection conditions, there is a progressive reactivation of the Mg acceptors with a corresponding increase in the hole concentration. After 1 hr, the majority of these acceptors have been reactivated. Clearly therefore, the injection of electrons has a dramatic influence on the stability of the MgH complexes. The Mg reactivation has a strong dependence on depth into the p-type layer, which may result from the diffusion distance of the injected electrons prior to recombination. Heating of the sample during forward biasing was ruled out as a factor in

Hydrogen in Wide Bandgap Semiconductors

451

the enhanced dissociation of the neutral dopant-hydrogen complexes. The samples were thermally bonded to the stainless steel stage and the junction temperature rise was expected to be minimal (<_10~ Moreover, from separate experiments, reactivation of the Mg did not begin until temperatures above --450~ under zero-bias conditions.

~. ?

15-

E

o

tlD v,-

o ~

M e~ ~ "0

t=O rain. 1=10 min t=20 min t : 3 0 mln t=60 mln

12

9-

L_ L_ L_

0

6

I

0.00

0.05

'1

0.10

I

'

0.15

I

0.20

0.25

depth (pm) Figure 13. Carrier concentration profiles in hydrogenated GaN(Mg), after annealing or various times at 175~ under forward bias condition.

Previous experiments on minority carrier enhanced reactivation of hydrogen passivated dopants in Si[l 12]and G a A s [119] have found that, for long annealing times, the kinetics can be described by a second-order equation:

Eq. (7)

dt

where NA is the uniform Mg acceptor concentration in the non-hydrogenated sample, N(O is the acceptor concentration in the hydrogenated GaN after forward bias annealing for time t and C is a second order annealing parameter. In order to quantitatively analyze the reactivation kinetics of the MgH complexes in GaN, the inactive acceptor concentration NA-N(t) was measured at a depth of 0.1 ~tm in the p-GaN layer. Figure 14 shows that there is a linear relationship between [NA-N(t)]-1 and annealing time (t),

452

Wide Bandgap Semiconductors

confirming that the reactivation process can be described by a secondorder equation with C = 4 x 10-20 cm3s -1. This value was consistent with those obtained in Si and GaAs, where minority carrier enhanced dopant reactivation had also been reported. In that work, the annealing parameter was found to depend on the injected minority carrier density. Moreover, for short annealing times, it was found that the dopant reactivation occurred at a faster rate than predicted by the second-order equation for very short annealing times, and that the annealing process was rate-limited by the formation of stable, electrically inactive, diatomic H species. At this point, there have not been enough studies of the various states of hydrogen in GaN as determined by infra-red spectroscopy, channeling, or secondary ion mass spectrometry to conclude anything about the ultimate fate of the atomic hydrogen once it has dissociated form the Mg-H complex. It is likely that it then reacts with other hydrogen atoms to form diatomic or larger clusters. A strong dependence of reactivation rate on injected minority carrier density would indicate the presence of a charge state for hydrogen and therefore influence the conversion ofH § into the neutral state and then into the final hydrogen complexes.

15" 4.

E '7 O

10.

6.

0 0

1 oo 18'oo 2 oo 3 oo 3e'oo Annealing time

(see)

Figure 14. Plot of inverse net inactive Mg concentration determined from Fig. 13 at a depth of 0.1 ~tm, as a function of forward bias annealing time.

Hydrogen in Wide Bandgap Semiconductors 453 The fact that the MgH complexes are unstable against minority carrier injection has implications for several GaN-based devices. Firstly, in a laser structure, the high level of carrier injection would rapidly dissociate any remaining Mg-H complexes and thus would be forgiving of incomplete removal of hydrogen during the post-growth annealing treatment. In a heterojunetion bipolar transistor, the lower level of injected minority carriers would also reactivate passivated Mg in the base layer, leading to an apparent time-dependent decrease in gain as the device was operated. In summary, hy.drogen passivated Mg aeeeptors in GaN may be reactivated at 175~ by annealing under minority carrier injection conditions. The reactivation follows a second order kinetics process in which the (MgH) ~ complexes are stable >450~ in thin, highly-doped GaN layers. In thicker, more heavily doped layers where retrapping of hydrogen at the Mg aeeeptors is more prevalent, the apparent thermal stability of the passivation is higher and annealing temperatures up to 700~ may be required to achieve full activation of the Mg. Our results suggest the mechanism for Mg activation in e-beam irradiated GaN is minority-earner enhanced debonding of the hydrogen. Neugebauer et al.[ 122,1231have calculated the total energy surface for H and H 2 in GaN. For H-, a tetrahedral Ga interstitial site was found to be energetically most favored, with a large activation energy for migration of 3.4 eV. The neutral charge state H ~ was also most stable at Ga Td site, but with a much lower energy barrier for migration. The H + charge state was most stable at a nitrogen anti-bonding site, rather than the bond-centered position found for H + in Si and GaAs. This difference was ascribed to the smaller lattice constant and more ionic character of GaN compared to the eovalently bonded semiconductors. An activation energy for migration of ---0.7 eV was found for H + in GaN. Figure 15 shows the calculated formation energies for the different charge states of hydrogen in GaN as a function of Fermi level position. These calculations suggest that H + is more soluble in p-GaN than H- in n-GaN, and that H ~ is never a stable charge state. Moreover, H 2 molecules are predicted to be unstable with respect to dissociation into atomic hydrogen. We caution that these are preliminary calculations only, and historically a number of iterations have been necessary to achieve agreement with experiments for similar situations in Si and GaAs. Brandt, Gotz, and co-workers [124]-[126] have reported infrared absorption bands for the Mg-H neutral complexes in GaN. However, the suggested model involves the hydrogen bound to a N atom, rather than to the Mg, as is the case for acceptor-hydrogen complexes in other III-V

454

Wide Bandgap Semiconductors

semiconductors. The reasoning was that in p-GaN, the H + species are more strongly attracted to the electronegative nitrogen than to the more electropositive Mg.[ 122] In these experiments, hydrogen was found to diffuse more readily in p-type GaN relative to n-type material but the hydrogenation temperatures were very high (_>600~ More controlled experiments at lower temperatures are necessary to fully explore the trapping characteristics of hydrogen at n- and p-type dopants in GaN.

2 ~

&

,

1

-3 -4

Mg-H eompl. 0

9

J

1

,

|

2

p(ev)

!

3

Figure 15. Formation of energy for H+, H-, H~ Mg-H and H2 species in GaN, as a function of Fermi level position (after ref. 122).

3.2

Passivation of Other Acceptor Dopants

P-type doping of GaN has been reported using CC14 in Metal Organic Molecular Beam Epitaxy (MOMBE)[127], with maximum hole densities of--3 x 1017cm-3.[128] To prove that C, as well as Mg, acceptors are passivated by hydrogen, GaN(C) was exposed to an ECR H 2 plasma for 0.5 hr at 0~ Figure 16 shows that this treatment reduced the carrier concentration by a factor of 3. Subsequent annealing had little affect up to 350~ where the carrier density again decreased before returning to its original value at -~450~ It was assumed that around 350~ hydrogen motion began and led to additional passivation, while at higher temperatures, the C-H bonds broke, reactivating the acceptors.[ 129] The realization of p-type conductivity in GaN can be difficult because of several factors: the often high background, shallow donor concentration

Hydrogen in Wide Bandgap Semiconductors 455 in some material; the relatively deep ionization levels for acceptors such as Mg (~0.16 eV); and residual hydrogen incorporation from the growth ambient that passivates the electrical activity of Mg.[ 13~ It is possible to reverse the hydrogen passivation phenomenon by either minority carrier injection or thermal annealing.[13~ TM] To date, only (Mg-H) neutral complexes have been examined in any detail. [124][136][137]One would expect, in analogy to acceptor passivation in other p-type semiconductors, that hydrogen interrupts the bond between the acceptor and a neighboring N atom, and occupies a bond-centered position bound predominantly to this nitrogen and having some interaction with the acceptor.[ 138]Preliminary results on p-type GaN(C) have also shown acceptor passivation.[ 139]

3.0

H 2 Plasma 0.5h 2500C

2.5

60 sec anneals

V

" f " 2.0 E U

X

1.0 --0--

0.5

0.0

0

Before Hydrogenation After Hydrogenation I

,

i

I00

200

300

ANNEALING

TEMPERATURE

.

I

400

,

,

500

(oc)

Figure 16. Carrier density versus annealing temperature in hydrogenated GaN(C).

To achieve higher doping levels in p-type GaN, it is desirable to find aeceptor impurities with smaller ionization energies. For example, at room temperature, only a few percent of substitutional Mg acceptors will be ionized because of the Boltzmann factor, e'~a/*r (Ea is the ionization level, k is the Boltzmann constant and T is the absolute sample temperature) that relates total and electrically-active impurity concentration. Theoretical considerations have suggested that Ca might be a shallower acceptor in GaN than Mg.[ 139]Using implantation of Ca § alone (or a co-implantation of Ca + and P+), p-type doping of GaN, followed by rapid thermal annealing at

456

Wide Bandgap Semiconductors

>1100~ has been realized.[ 14~ While the activation efficiency of Ca in both implant schemes was ~100%, temperature-dependent Hall measurements showed that the ionization level of Ca was--168 meV, similar to that of Mg. The Ca atomic profile was thermally stable to temperatures up to 1125~ Since Mg has a substantial memory effect in stainless steel epitaxial reactors (or in gas lines leading to quartz chamber systems), Ca may be a useful alternative p-dopant for an epitaxial grown laser diode or heterojunction bipolar transistor structures in which junction placement, and hence control of dopant profiles, is of critical importance. In considering Ca-doped GaN for device applications, it is also necessary to understand the role of hydrogen. There is always a ready supply of atomic hydrogen available from NH3, the metalorganic group III source [typically (CH3)3Ga ] or from the gaseous dopant source when using chemical vapor deposition techniques. We have also found that Ca acceptors in GaN are also readily passivated by atomic hydrogen at a low temperature (250~ but they can be reactivated by thermal annealing at < 500~ for 1 min in lightly-doped (3 x 1017cm -3) materials. As the carrier density is restored by such annealing treatments, there is a corresponding decrease in hole mobility, indicating that there is a true passivation and not just compensation of the Ca acceptors by the hydrogen. Nominally undoped (n <3 x 1016cm-3) GaN was grown on doubleside polished c-A1203 substrates prepared initially by HC1/HNO3/H20 cleaning and an in-situ H 2 bake at 1070~ [124] A GaN buffer _<300 A thick was grown at-~500~ and crystallized by ramping the temperature to 1040~ where trimethylgallium and ammonia were again used to grow the 2 ~tm thick epitaxial layer. The materials properties were discussed in detail earlier,[ 124-126] but ,in brief, the double crystal x-ray full width at half maxima were--300 arc sec and the total defect density (threading dislocations, stacking faults) apparent in plain view transmission electron microscopy was typically 2-4 x 109 cm -2. The as-gown films are featureless, transparent, and have strong bandedge (3.47eV) luminescence. 4~ ions were implanted at 180 keV and a dose of 5 x 1014cm -2. In some cases, a co-implant of P § to the same dose at an energy of 130 keV was performed to try to enhance the substitutional fraction of Ca upon subsequent annealing (in analogy to the case of Mg implantation in GaN).[ lal] In the case of Ca, it was found that there was little additional activation as a result of the co-implant. After rapid annealing at 1150~ for 15 sec under N 2 in a face-to-face geometry, sheet carrier densities ofp~l.6 x 1012cm-2 with a mobility at 300 K of 6 cm2/Vs were measured. Arrhenius plots of the

Hydrogen in Wide Bandgap Semiconductors 457 hole density showed an ionization level of 169 meV for the Ca in GaN. Samples with alloyed HgIn ohmic contacts were exposed to an Electron Cyclotron Resonance (ECR) H 2 or He plasma (2.45 GHz) with 850 W forward power and a pressure of 10 mxogp. The exposure time was 30 min at 200~ and the temperature was lowered to room temperature with the plasma on. The sheet carrier density and hole mobility at 300 K were obtained from Van der Pauw geometry Hall measurements. Post hydrogenation annealing was performed between 100-500~ for 60 see under flowing N 2 with the ohmic contacts already in place. The initial H 2 plasma exposure caused a reduction in sheet hole density of approximately an order of magnitude, as shown in Fig. 17. No change in electrical properties was observed in the He-plasma treated samples, showing that pure ion bombardment effects were insignificant and the chemical interaction of hydrogen with the Ca aceeptors was responsible for the conductivity changes. Post-hydrogenation annealing had no effect on the hole density up to 300~ while the initial carrier concentration was essentially fully restored at 500~ Assuming the passivation mechanism was formation of neutral Ca-H complexes, then the hole mobility should increase upon hydrogenation. This is indeed the ease, as shown in Fig. 18. Note that the mobility decreases to its initial value with post-hydrogenation annealing. If the carrier reduction was due to introduction of compensating defects or impurities, then the hole mobility would decrease, which was not observed. In other p-type III-V semiconductors, it is generally accepted that atomic hydrogen is predominantly in a positive charge state with the donor level being around midgap. If a similar mechanism exists in GaN then the initial Coulombie attraction between ionized acceptor and hydrogen leads to formation of a neutral close pair, i.e.: Eq. (8)

Ca-+ H § ~ ( C a - H) ~

The existence of the neutral complex should be verified by observation of a vibrational band-(142)-, but to obtain the sensitivity needed for such a measurement will require a relatively thick epitaxial layer of Cadoped GaN. The presently implanted samples do not have a sufficient Ca density-times-thickness product to be suitable for infra-red spectroscopy. If the dissociation of the Ca-H species is a first-order process, then the reactivation energy from the data in Fig. 17 is -~2.2 eV,[1381 assuming a typical attempt frequency of 1014s-1 for bond breaking processes. This is

458

Wide Bandgap Semiconductors

similar to the thermal stability of Mg-H complexes in GaN which were prepared in the same manner (implantation) with similar doping levels. In thicker, more heavily doped samples, the apparent thermal stability of hydrogen passivation was much higher because of the increased probability of retrapping of hydrogen at other acceptor sites.[ 1381This is why for thick, heavily doped (p >1018 r -3) GaN(Mg), a post-growth anneal of at least 700~ for 60 min was employed to ensure complete dehydrogenation of the Mg. True reactivation energies can only be determined in reversebiased diode samples where the strong electric fields present swept the charged hydrogen out of the depletion region and minimized retrapping at the acceptors.[ 143]

,,t-I _]

-'o,,N (Ca)

H~lasma 0.Sh 250"C /

1..,

4

0

100

200

300

400

500

600

annealing temperature (oc) Figure 17. Sheet hole density at 300K in hydrogenated GaN(Ca) as a function of subsequent annealingtemperature.

In conclusion, hydrogen passivation of acceptors in GaN occurs for several different dopant impurities. Post-growth annealing will also be required to achieve full electrical activity in Ca-doped material prepared by gas-phase deposition techniques. The thermal stability of the passivation is similar for Ca-H and Mg-H complexes, with apparent reactivation energies of~2.2 eV in lightly-doped (-~1017cm-3) material.

Hydrogen in Wide Bandgap Semiconductors 459

25

GaN (Ca) Hzplasma 0.Sh 250"C 60sec anneals

20

lO

5

--I!-- before hydrogenation --,k- after hydrogenation 0

!

!

I

I

I

100

200

300

400

500

600

annealing temperature (~ FigurelS. Hole mobility at 300 K in hydrogenated GaN(Ca) as a function of subsequent annealing temperature.

3.3

Diffusion of Hydrogen in the Nitrides

In other compound semiconductors, it is generally observed that annealing to restore the electrical activity of initially passivated acceptors does not actually remove the hydrogen from the material. Rather there is a sequential process, in which, at temperatures around 300-400~ the neutral acceptor-hydrogen complex breaks up to reactivate the dopant, i.e.: Eq. (9)

(AH) ~ --+ A + H +

and the hydrogen remains in the crystal in an electrically inactive, relatively immobile form. A strong possibility is formation of molecules, i.e.: Eq. (10)

H + + H 0 ~ H 2 + h + __~H 2

where h + denotes a hole in the p-type semiconductor. The hydrogen remains in the material until temperatures of~600~ (for GaAs), where it is evolved from the sample. Since the III-N materials, especially GaN and A1N have much wider bandgaps and lower group V vapor pressures than more conventional III-V compound semiconductors, it is of interest to

460

Wide Bandgap Semiconductors

establish the thermal stability of hydrogen in these materials. In this section, measurements of the outdiffusion of deuterium from GaN, A1N, and InN into which the deuterium was incorporated either by plasma exposure or ion implantation, will be discussed. The nitride materials were grown on semi-insulating (100) GaAs substrates by Metalorganir Molecular Beam Epitaxy (MOMBE) in an Intervac Gas Source Gen II system.[ 137] The group III sources were triethylgallium, trimethylindium, or trimethylamine alane, respectively for GaN, InN, and A1N. Atomic nitrogen was supplied from a 200 W, 2.45 GHz electron cyclotron resonance (ECR) source (Wavemat DPDR 610). The InN samples were cubic with smooth morphology, while the GaN and A1N were a mixture of cubic and hexagonal phases. The GaN and A1N were resistive as-grown, while the InN was strongly n-type (-~102-cm3). Layer thicknesses were 1-1.5 ~tm for each material so that interfacial problems were avoided. Typical hydrogen concentrations in as-grown films were <10 is cm -3. The lattice mismatch was accommodated by a high density of stacking faults and dislocations in the single crystal material. Parts of each sample were implanted with 2H+ ions at 40 keV to a dose of 5 x 1015cm -2 at room temperature, while companion sections were exposed to a 2H plasma for 30 min at either 250 or 400~ using the ECR source on the growth chamber of the MOMBE system. A deuterium flow rate of 20 sccm and pressure of 5 x 10-5 ~op9 was employed, with an applied microwave power of 200 W. Furnace annealing was performed (up to 900~ for 20 min for GaN and A1N, and 500~ for InN) in a flowing N 2 ambient. These temperatures were chosen to avoid surface degradation. The GaAs was not degraded under these conditions. SIMS measurements were performed using a Cs § primary ion beam and negative secondary ion detection.[ 144] The raster size was 250 x 250 ~tm2 with counts taken only from the central 75 x 75 ~m2.[145] In the deuterated GaN, two distinct regions in the SIMS profiles were observed. The first region is shown in Fig. 19 for a sample deuterated at 250~ for 30 min. It consists of a thin (<0.3 ~tm), very highly concentrated section, which is slowly evolved from the sample, beginning at 300~ The inset of Fig. 19 shows the areal density of deuterium remaining after 20 min anneals over a temperature range of 300-700~ Similar deuterium profiles may be observed in plasma exposed Si under certain conditions. Transmission electron microscopy reveals the presence of oriented arrays of linear defects comprised of Si-H bonds, the so-called platelet defects. [146] To date, no comparable microscopy studies have been performed in GaN.

Hydrogen in Wide Bandgap Semiconductors 461

,02' f

2H plasma treated GaN at 250~ 2H surface density (cm"2) No anneal 300~ 500~ 700~

1020

2.8 x 1.7 x 1.1 x 5.5 x

1015 10TM 1015 10TM

300~ anneal 500~ anneal ~ 700"C anneal

~'

\ w

1018

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

DEPTH (lUn)

Figure 19. SMS profile of deuterium in the near-surfaceregion ofa GaN sample exposed to a 2Hplasma at 250~ and subsequentlyannealed for 20 rains at temperatures up to 700~

The thermal stability of the deuterium trapped in this near-surface region is much different than that in the second part of the profiles, shown in Fig. 20. This second region extends throughout the GaN epilayer thickness, and the concentration of the plateau increases with increasing plasma exposure temperature.[147] In the work of Brandt et. al., plasma exposure temperatures of 500-~00~ were required to observe measurable passivation in heavily (> 1019 em -3) Mg-doped GaN, since the lower temperatures there would not be of sufficient concentration to affect the doping density. This is a disadvantage of the remote microwave plasmas used in that work, since they typically operate at low pressure (<10 -3 1;099 ) and simply do not produce enough hydrogen flux incident on the sample to incorporate very high concentrations. In the current samples, the plateau of deuterium is most likely due to trapping at impurities or native defects such as nitrogen vacancies.

462

Wide Bandgap Semiconductors

107 :

:oR

1'~ l oe

o __

e~ 10 TM

8

G

rn -- 1018 ~ Z auJ

~ 1017

2H for 250"C ~ I, plasma

101tin

r

1,05oo

~1~

1o3t/! o

~..J103 r/) 102

1

10151, , , , I . . . . I , | I I l I I , ,1101 0 0.5 1.0 1.G 2.0 DEPTH (pro)

1~ I

,nellg00"C

1102

1015 . . . . t . . . . i , , , , j .... 101 0 0.5 1.0 1.5 2.0 DEPTH (pm)

Figure 20. SIMS profiles of deuterium in GaN after exposure to a 2H plasma at either 250 or 400~ (left) and after subsequent annealing at 800~ or 900~ (fight). On these scales the near-surface profiles for the 800 and 900~ samples are indistinguishable.

As shown in the fight-hand side of Fig. 20, no redistribution of deuterium was observed in the deeper, lower concentration region of the profile for annealing temperatures p to 800~ After annealing at 900~ however, essentially all of this deuterium in this region has been evolved from the GaN. The concentration in the near-surface region has been decreased from a peak of~5 x 1020cm -3 to ~6 x 1018cm -3 (i.e. almost 99% has been removed from the sample). Figure 21 shows SIMS profiles of implanted deuterium profiles in GaN for post-implant annealing temperatures up to 700~ Within experimental error there is no redistribution of deuterium for any of these conditions, although once again at 900~ (not shown in the Figure), essentially all of the deuterium had been evolved from the sample. The percentage of deuterium lost from the GaN as a function of annealing temperature is shown in Fig. 22 for both the surface and bulk regions of the profiles obtained in plasma-treated samples, and for the implanted sample. Note that the deuterium from deeper in the GaN, whether it is placed there by direct implantation or by diffusion from the plasma, shows a much different thermal stability than does the deuterium

Hydrogen in Wide Bandgap Semiconductors 463 trapped near the surface of the plasma exposed GaN. The fairly narrow temperature range over which the deuterium in the bulk of the material is released indicates that it is present in a single bonding environment, whereas, the extended temperature range over which the near-surface deuterium is evolved is suggestive of the presence of a variety of different trap sites with different binding energies. Retrapping of deuterium in this region is also likely to be significant and will enhance the apparent thermal stability.

102 1 I

2H in GaN 40 keV 5 x 1015 cm "2 -.--9 No anneal - - - 300"C 500"C ---- 700"C .......

1020

,'; ......

-,9 10 TM "%

0 101a

""L~"'

~"!!LL.,

1017 I

0

I 0.5

.,,

I 1.0 DEPTH (wn)

I 1.5

.I 2.0

Figure 21. SIMS profiles of deuterium implanted into GaN and subsequently annealed up to 700~

A deuterium profile for a 2H plasma exposed (0~ 30 min) A1N sample is shown in Fig. 23. Note that there is a much higher concentration of deuterium incorporated in the A1N than in the thin (--0.1 ~tm) GaN cap layer that was present to prevent possible oxidation of the A1N. The thermal stability of the deuterium was examined by SIMS profiling of annealed samples, and once again no significant loss of deuterium was measured until 900~ where most of it was evolved from the A1N. [147]

464

Wide Bandgap Semiconductors

i'

100

i

,

i

- - 0 - - SURFACE . . . .

o...BuLK

75 0

a~

500 t ~

25 -

0 00

200

400

600

800

lOOO

A N N E A L TEMPERATURE (~ Figure 22. Percentage of deuterium lost from different regions in GaN as a function of annealing temperature. Surface and bulk refer to the two regions observed in the SIMS profiles of deuterium plasma treated samples, while implant refers to samples into which the deuterium was introduced by implantation.

. lOe

1021 ~

2H plasma on AIN at 250~ .

lO2O

10 7

AI (counts) --~

A

lO6 o'~J

10 TM

o

z

(n 1018

z uJ c3

r

=E

z 0

0 ,~ 1017

"

ffl

.

1016

L2 H 10151

0

I 0.5

I 1.0

. .

1.5

DEPTH (l~m)

Figure 23. SIMS profile of deuterium in AIN after exposure to a 2H plasma at 250~

Hydrogen in Wide Bandgap Semiconductors 465 Results for 2H+ implanted A1N and InN are shown in Fig. 24. No measurable redistribution was observed in either material up to 600~ Annealing temperatures above this are not possible for uncapped InN due to prohibitive loss of N from the surface. In the case of A1N, a similar thermal stability for the implanted deuterium, as for the plasma-diffused material, was observed. We cannot measure the stability of any nearsurface deuterium in A1N since this region is absent due to the need for the GaN cap layer. Figure 25 shows a plot of the percentage of deuterium lost from plasma treated A1N as a function of annealing temperature. This result is quantitatively similar to that for outdiffusion of deuterium from the bulk region of GaN.

,o,

2H in AIN 40keY 5xl0t5 cm"2

12H.nN,o,s.,o i,o,

Io2o

1

/

I , No anneal

_~ -

1019

Z w

,on-g

:E 10 TM o

g

,o,,I

,X,

g

1017~ -

I

o

,

J

0.5 1.0 DEPTH (pro)

.

lo2 1.5

/ .... 0

i

j

~1o2

0.5

1.0

1.5

DEPTH (l~m)

Figure 24. SMS profiles of deuterium implanted into either AIN (left) or InN (right) and subsequently annealed at 600~

The fact that temperatures of 900~ for GaN and A1N, and >600~ for InN are required to remove deuterium suggests that the amount of hydrogen in the growth ambient, or during subsequent processing steps, should be minimized. It is clear that the annealing employed in past work on MOCVD grown GaN to reactivate passivated Mg acceptors is not sufficient to actually remove the hydrogen from the material.[ 1371Therefore, subsequent

466

Wide Bandgap Semiconductors

thermal treatments during device processing could lead to re-passivation of the acceptor dopants. Since, in most actual device structures there will be heterostructures involving both binary and ternary III-N materials, there is still a need to investigate the behavior of hydrogen in these more complex layers, and, in particular, the effect of p-n junctions on the outdiffusion. Past results in GaAs suggest that the presence of thin heavily doped layers on top of regions of opposite conductivity type may have a significant effect on the motion o f h y d r o g e n . [ 148,149]

I

z

I

I

I

2H P L A S M A TREATED AIN

< 9

I

100 9

80

.O !

BULK

... !

60

.

i

o~

O ~a

-

l

!

40

!

rl

O

20

0

i o

9.................. 9 .................. 9 ..................o----" I

I

I

100

300

500

,

-

I

I

700

900

A N N E A L TEMPERATURE (~

Figure 25. Percentage of deuterium lost from the bulk of plasma treated AIN, as a function of subsequent annealing temperature.

3.4

Passivation

in Ternary

Nitrides

By analogy with the models for neutral hydrogen-dopant complexes in other III-V semiconductors,[ 81] Fig. 26 shows schematic representations of the likely configurations in GaN. For donor dopants (Fig. 26a and b), hydrogen occupies an anti-bonding position either attached to the dopant, in the case of group IV donors, or attached to the Ga neighbor, in the case of a group IV donor. For aeceptor dopants (Fig. 26c and d), hydrogen is at a bond-centered site, bonded predominantly either to the acceptor or a N neighbor, respectively, depending on whether the acceptor is from column IV or II of the Periodic Table. Upon annealing at >400~ hydrogen is dissociated from the complex, leading to reactivation of the dopant. The hydrogen does not leave the crystal at this temperature but probably

Hydrogen in Wide Bandgap Semiconductors 467 associates with other hydrogen atoms to form molecules and larger clusters. At much higher temperatures (>800~ in GaN), these clusters are evolved from the sample.

(b)

(a)

:

S'

o

e

o

o

o

' Si

'

0 0

(DH

0 Q 0

Mg.

Figure 26. Schematicrepresentationof hydrogen-dopantcomplexesin GaN.

The ternary samples were exposed to an ECR D 2 plasma for 30 min at 250~ Figure 27 shows the fraction of passivated donors remaining in both InA1N and InA1GaN as a function of post-hydrogenation annealing temperature. Both samples displayed a decrease in carrier concentration of approximately an order of magnitude after hydrogen plasma exposure, consistent with previous reports.J15~ On subsequent annealing, the passivated donors begin to reactivate around 400~ and by 500~ 78% of the

468

Wide Bandgap Semiconductors

lost carriers were restored in InA1N and 66% in the InAIGaN. The reactivation o f the donors was fit to the relation:[ 81]

No = l - exp - tv exp - ~ -

Eq. (11)

N

for time t, v is the attempt frequency (assumed to be 1014S"1) and E d is the activation energy for reactivation. The recovery of the donor activity occurred over a slightly broader temperature range than generally observed for other passivated dopants in binary semiconductors, and is consistent with the presence ofa Gaussian distribution of activation energies.[ 811This may be due to nitrogen vacancies with different numbers of specific group III neighbors surrounding them (i.e., 2 In and 2 A1, against 1 In and 3 A1). Assuming a Gaussian distribution of activation energies, values for E d around 2.4 eV were obtained, with a full width at half m a x i m u m of--0.3 eV. The fact that reactivation of these donors occurs around 500~ means that the apparent thermal stability is similar to that ofpassivated M g acceptors in GaN.

I

r,~

o z o

,. I

I

I

I

1.00

0.75 < ;> <

0.50

"-<>-- InGaN InAIGaN

I.z.,

o z o[..., 0.2s I

I

I

I

300

350

400

450

,,,

I

500

ANNEALING TEMPERATURE (*C) Figure 27. Fraction of passivated donors remaining in InAIN r InGaN after deuteration at 250~ and subsequent annealing at different temperatures.

Hydrogen in Wide Bandgap Semiconductors 469 SIMS profiles of deuterium in InAIN for as-hydrogenated plus subsequent 900~ annealing treatments are shown in Fig. 28. The deuterium concentration (---1021 cm-3) throughout the epitaxial layer thickness is well in excess of the doping concentration. The sites to which this deuterium is bonded are at present unclear, but presumably involve the structural defects in the material. Annealing up to 800~ did not measurably alter the deuterium profile, whereas annealing at 900~ reduced the plateau concentration to --5 x 1018cm -3. This is similar to the behavior in the component binaries reported previously.[ 147)

1022

~,

1021

"=

102o

InAIN

t.. i,u

~

1019

j,m ~

L..

~

1018

1017 L

t

!

0.5 1.0 Normalized Depth (~m)

Figure 28. SIMS profiles of deuterium in InAINexposed to D2 plasma for 0.5 h at 250~ and subsequentannealing at different temperatures.

The native donors in InN and InGaN are also passivated by atomic hydrogen. Figure 29 shows the ratios of post-hydrogenation resistance to the as-grown resistance, as a function of the microwave power used to create the ECR hydrogen discharge. In the case of InGaN, the passivation efficiency increases with microwave power, which is consistent with the increased concentration of atomic hydrogen in the plasma. This was

470

Wide Bandgap Semiconductors

confirmed by monitoring the 656 nm emission line of this species using optical emission spectroscopy. For InN however, there was a fall-off in passivation efficiency at higher microwave powers. This coincided with the observation of severe degradation of the InN surface, due to preferential loss of N as NH 3. The resultant surface contained a high density of In droplets. The effects of this surface degradation are two-fold. First, it prevents efficient permeation of further atomic hydrogen into the InN, and secondly, the additional nitrogen vacancy concentration may increase the apparent n-type conductivity of the material. The hydrogen passivation of the shallow donors in both materials can be reversed by subsequent annealing. Figure 30 shows the annealing temperature dependence of the reactivation (1 min anneals). The initial conductivity in both cases is recovered with similar characteristics, with slightly higher stability for the wider bandgap In0.5G%.sN. This would be expected if the native donors had a common origin, either as defects or impurities, whose electrical properties were altered slightly by the different matrixes in which they were incorporated. Similar thermal stabilities have been reported for passivated shallow native donors in InA1N and InGaA1N, with reactivation energies of--2.5 eV. [151]

40sccm H z 1 0 0 W rf 1.5 m T o r r

o

3

--I--

0

I

I

I

I

200

400

600

800

InN InGaN I

1000 1200

Microwave power (W) Figure 29. Ratio of resistance increase (Ro is the initial value) of InN and InGaN layers exposed to ECR H 2 plasmas as a function of microwave power.

Hydrogen in Wide Bandgap Semiconductors 471

6 -----~,

.

.

.

4o

3210 ---~.

I

I

I

i

0

400

450

500

550

350

Anneal temperature (~ Figure 30. Resistance of hydrogenated InN and InGaN layers as a function of subsequent annealing temperature.

3.5

Hydrogen Incorporation During Processing

Atomic hydrogen is found to be unintentionally incorporated into the nitrides during many processing steps, including boiling in water, dry etching, wet etching, chemical vapor deposition of dielectrics, and annealing in H 2 or NH 3. GaN sampled boiled in D20 at ~100~ for 30 min showed a considerable indiffusion of deuterium, as shown in Fig. 31. Similar results were obtained for A1N and InN, and showed that the nitrides behaved the same as Si and GaAs in that even a relatively benign treatment such as boiling in water (or any solvent) can lead to hydrogen incorporation.[ 152] We have found that KOH solutions held at ~85~ can selectively wet etch A1N, but not affect GaN.91531 Figure 31 also shows that exposure of GaN to an etching solution of KOH/D20 for 20 min leads to incorporation of deuterium up to -~0.6 ~tm. This shows that even at quite low processing temperatures, atomic hydrogen can be readily diffused into GaN.

472

Wide Bandgap Semiconductors

lOZ2-

i0 zl

I0 z0

o

Et 1019 I'b

o

m,i,

~o 10TM

2H 1016

2H ...... 0.0. . . . 0.2

0.4

'01.6. . .0.8 . . . . . . . .1.0

Depth (l~m)

1'.2 . . 1.4 ..

Figure 31. SIMS profiles of deuterium in GaN samples after boiling in D20 or etching in a KOH/DEO solution.

PECVD of dielectrics such as SiN x and SiO 2 onto GaN and related alloys will be common for masking and surface passivation. High ion density ECR SiH4-based discharges are known to cause extensive hydrogen passivation in GaAs.[154][]551We deposited ~2,000 A/min of SiN~ onto GaN using an ECR SiD4/N2 discharge at ~I~ The deposition rate was ---250 A/min. Figure 32 shows that this treatment produced extensive deuterium incorporation into the GaN, as might be expected. Further annealing of the SiN~/GaN structure at 500~ for 20 min did not produce further indiffusion of deuterium from the SiN x. The fastest dry etch rates for GaN, InN, and tklN at moderate de biases (_<150V) have been obtained in C12/CH4/H2/Ar or C12/H2 ECR discharges.[ 156] In particular the former plasma chemistry produces the most stoichiometric surfaces for substrate temperatures of~lT0~ which is found to aid desorption of the In in InN, InGaN, and InA1N. Figure 33 shows SIMS profiles of deuterium in In-containing ternaries of various compositions after a 40 see

Hydrogen in Wide Bandgap Semiconductors 473 exposure to an 850 W ECR discharge of C12/CH4/D2/Ar (1 mtorr, 100 W rf). Similar experiments on thin (0.2 ~tm) binary layers showed that deuterium had permeated to the substrate in each case, indicating that the diffusivity of deuterium is >10 -11 cm2/sec at 170~ in these materials, based on a simple (Dr) 1/2 calculation. The C12/CHa/H2/Ar plasma chemistry provides rapid, non-selective dry etching of all of the nitrides, but one should be aware of possible hydrogen passivation effects. [157] The etch rates for the nitrides are between 1000-4000 A/min at moderate de bias, but the hydrogen can diffuse ahead of the etch front and incorporate into the epitaxial layer. 10 7

1021.

102~

SiD,/N2/GaN

~

t06

Gao

10 s O 10 4 o

c 10~8

0

oo 1017[

a_

10 3 ~ 0 0 1,)

0

1016

10 2m 111

l0 ts

0.0

10

0.2

0.4 0.6 Depth (lain)

0.8

1.0

Figure 32. SIMS profile of deuterium in GaN after deposition of SiNxfrom a SiD4/N2ECR plasma. The SiNx was removed by CF4 etching prior to the SIMS measurement.

Both isotropically-substrated mixtures were used to examine possible incorporation of deuterium during this type of etching (for InN see Fig. 34 and for A1N see Fig. 35). The diffusion coefficient of deuterium is >10 11 cm2/s at 170~ in both materials, based on a simple (Dt) 1/2 calculation. Similar results were obtained for GaN. When pure InN is exposed to a H 2rich plasmas preferential loss of N can occur, leading to In droplets on the surface. This does not occur with either GaN or A1N, or any of the ternary compounds.

474

Wide Bandgap Semiconductors

IOZ'~

~o

10 II

inxGal.xN 4OCn,~CHVWAr

I I

~

"

~

"~~ 1017

II~

[nxGat.xN

o~cam.~,,u. 40 sr162

~

1019

x".0 7S

"~ 10I

~-~ I0*6

"<~ iO|

m

I0 t*

lOtS

0"0

10am_ i~

'[~

40 sr162

0.5

1.0

I.S

2.0

|

0.0

0.5

Depth (pm)

1.0

|.5

2.0'"'

2.5

Depth (pm)

Figure 33. SIMS profile of deuterium in InxGal_xN (righO or Ino.36Gao.64N (left) after a 40 sec etch in Cl2/CH4/D2Ar at 170~

10'2o

I0

19

InN I Si ~-- --~

InN

E

0

._~ 1 co~ c

01e

2H

(1)

I:D

E

ld 7 .

0

0.0

I

I

0.2

0.4

0.6

Depth ~m) Figure 34. SIMS profiles of deuterium in InN/Si sample after a short etch in a CI2/CH4/H2/ D2 ECR plasma.

Hydrogen in Wide Bandgap Semiconductors 475

10 21

AIN .,.,~ ~

AIN

lU

Si

~I;:: 0 101~ r 0

E1 d7 0

,

0.0

0.1

,

,

0.2

0.3

0.4

D e p t h t(um) Figure 35. SIMS profile of deuterium in AIN/Si sample after a short etch in a CI2/CH4/HE/D2 ECR plasma.

Hydrogen implantation is used extensively in III-V electronics and photonics for current confinement, or electrical or optical isolation.[ ls81 This technique works by the H § ions creating deep level states in the bandgap of the implanted material. These deep levels trap carriers, removing them from the conduction (valence) band for initially n-type (p-type) material. Therefore, the implanted semiconductor has much higher resistivity than the as-grown material. Binari et al.[ 159]reported on the thermal stability of proton-implanted n-type GaN. Figure 36 shows a comparison of these results with similar data for H+-implanted n § GaAs. The thermal stability of the implant-induced deep levels is surprisingly greater in GaAs, but higher stabilities for GaN can be obtained with heavier ions such as N § O +, o r F+. [160][161 ] The maximum incorporation depth of implanted protons in GaN is --2 ~m for a typical 200 kV implanter. While the deep levels in implanted GaN anneal out at >200~ temperatures above 800~ are needed to cause most of the hydrogen to actually leave the materials.[ 113] A compilation of the processes found to incorporate hydrogen into GaN and related alloys is shown in Table 1, along with the typical depths and maximum concentrations measured. This shows that most of the wet chemical or plasma processes involved in device processing are capable of causing passivation of dopants due to hydrogen incorporation in these

476

Wide Bandgap Semiconductors

materials. Reactivation ofpassivated dopants occurs with activation energies around 2.5 eV in hydrogenated nitrides, corresponding to temperatures of 450--500~

10 8

.

107 e~ ,=,,=

106 . i

105

-

.m r,~ .m r.~

104

H* implant

103

9 l

....~lU 2

"I';__

GaN C.oAs

,

,

100

0

.,.,

,

200

.,

,

,

300

400

Anneal temperature (~

Figure 36. Thermal stability of proton implant isolation in n-GaN or GaAs.

Table 1. Processing Steps in which Hydrogen is Found to be Incorporated into GaN

Process

Temperature

Max. [ H ] (cm"3)

(oc3 H20 Boil

P~S~x Dry Etch Implant Isolation Wet Etch

100 125 170 25 85

1020 3 x 1019 1019-1020 Dose Dependent 2 x 1017

Incorporation Depth 0m0 1.0 0.6 >0.2 2.0 0.6

Hydrogen in Wide Bandgap Semiconductors 477 SIMS profiling of GaN, and related materials exposed to hydrogencontaining gases or chemicals, shows that the hydrogen can diffuse into these layers at temperatures as low as 80~ Dielectric deposition, dry etching, wet etching, boiling in solvents, and, most likely, processes such as sintering of contacts, all lead to indiffusion of hydrogen. The hydrogen can bond to dopants, defects, or impurities in the nitrides, changing their electrical properties. Even though reactivation of dopants can be achieved by relatively low temperature thermal annealing after a given process step, hydrogen may cause similar problems after the subsequent processing so that one should be aware of its effects.

3.6

H in GaN/InGaN Device Structures

The behavior of hydrogen in device structures is likely to be more complicated. For example, light-emitting diodes or laser diodes contain both n- and p-type GaN cladding layers with one or more InGaN active regions. The first laser diode reported by Nakamura et al.[ 162]contained 26 InGaN quantum wells. In other III-V semiconductors, the diffusivity of atomic hydrogen is a strong function of conductivity type and doping level (since trapping by acceptors is usually more thermally stable and more efficient than trapping of hydrogen by donor impurities). Moreover, hydrogen is attracted to any region of strain within multilayer structures and has been shown to pile up at heterointerfaces in the Gams/Si, [163][164]GaAs/I/1P, [163][165] and GaAs/A1As[ 166]materials systems. Therefore, it is of interest to investigate the reactivation of acceptors and trapping of hydrogen in double heterostructure GaN/InGaN samples, since these are the basis for optical emitters. The reactivation of passivated Mg acceptors also depends on the annealing ambient, with an apparent higher stability for annealing under H 2 rather than N 2. Hydrogen is found to redistribute to the regions of highest defect within the structure. The sample was grown by MOCVD in a rotating disk reactor on cplane A120 3. The sapphire substrate was rinsed in H2SO4, methanol, and acetone prior to loading into the growth chamber, where it was first baked at 1100~ under H 2. A low temperature (-~510~ GaN buffer (~300 A thick) was followed by 3 x 3 ~tm ofn+GaN (n = 1018cm -3, Si doped), 0 x 1 ~tm InGaN (undoped), and 0 x 5 ~tm thick p+GaN (p = 3 x 1017 cm -3, Mg doped). The growth temperature was 1040~ for the GaN and-~800~ for the InGaN. Cross-sectional transmission electron microscope (XTEM) analysis was carried out on the MOCVD grown InGaN/GaN double heterostructure.

478

Wide Bandgap Semiconductors

A XTEM bright-field image, obtained using two-beam diffraction conditions with g = (2-1-I0) along the [0-1"I'0]GaNzone axis of the double heterostructure light-emitting diode (DH-LED) structure, is shown in Fig. 37. The interface between the various layers appears to be abrupt with no indication of interfacial phases. Selected-area-diffraction and high resolution electron microscopy revealed that the entire DH-LED structure grew epitaxially on the substrate.

~-~N InGaN

p-GaN

,4

InGaN

,,q

.-GaN -GaN

19

o-__z,p_m

Figure 37. (Left) Bright field XTEM image of GaN/InGaN double heterostructure; (Right) XTEM image of top region of device.

In the immediate vicinity of the n-GaN/A120 3 interface, the defect density was high but was reduced with increasing film thickness. However, after the growth of the active layers (InGaN), the defect density of the threading dislocations increased as shown at the fight of Fig. 37. A possible reason could be the different growth conditions used for growing the active layer and the GaN layers. The growth mechanism for p GaN on InGaN in the DH-LED structure could be similar to that proposed by Hiramatsu.[ 167] During the growth of the subsequent p-GaN layer, the underlying active layer may be undergoing solid-phase epitaxy. Hence, the quality ofp GaN grown on top of the active layer depends on the amount of epitaxy undergone by the active layer. In this structure, the thermal degradation of the InGaN upon raising the growth temperature for the p GaN leads to a higher defect density in this overlayer. XTEM of the DH-LED showed dislocations as dark lines propagating in the direction normal to the substrate. Most of the dislocations appeared to

Hydrogen in Wide Bandgap Semiconductors 479 bend and follow the interface for a short distance before threading out to the surface. The nature of the threading dislocations was studied by conventional XTEM using the g.b = 0 criteria. The dislocation will be invisible when b lies in the reflecting plane. Some of the dislocations were invisible both in g2 = (0002) and g5 = (li01) and, because b was common to both reflections, b was found to be 1/ 3[ 1120]. Assuming that the growth is the same as the translation vector of the dislocation, these defects would be pure edge type in nature. The average threading dislocation density was also found along the plane normal to the growth direction. The dislocation density was found to be---8 x 1010/cm2. The double-heterostructure sample was exposed to an electron cyclotron resonance plasma (500 W of microwave power, 10 mtorr pressure) for 30 min at 200~ The hole concentration in the p-GaN layer was reduced from 3 x 1017 cm -3 to --23 x 1016 cm -3 by this treatment, as measured by capacitance voltage (C - V) at 300 K. Sections from this material were then annealed for 20 min at temperatures from 500-900~ under an ambient of either N 2 or H 2 in a Heatpulse 410T furnace. Figure 38 shows the percentage of passivated Mg remaining after annealing at different temperatures in these two ambients. In the case of N 2 ambients, the Mg-H complexes showed a lower apparent thermal stability (by-~150~ than with H 2 ambients. This was reported previously by Si donors in InGaP and AIlnP, and BE and Zn acceptors in InGaP and AIlnP, respectively,[ 168] and most likely was due to indiffusion of hydrogen from the H 2 ambients, causing a competition between passivation and reactivation. Therefore, an inert atmosphere was clearly preferred for the postgrowth reactivation anneal of p GaN to avoid any ambiguity as to when the acceptors are completely active. Previous experimental results by Brandt et al.[ 168] and total energy calculations by Neugebauer and Van de Walle[169] suggested that considerable diffusion of hydrogen in GaN might be expected at <600~ Other sections of the double-heterostructure material were implanted with 2H+ ions (50 keV, 2 x 1015cm-2) through a SiN x cap in order to replace the peak of the implant distribution within the p+ GaN layer. Some of these samples were annealed at 900~ for 20 min under N 2. As shown in the secondary ion mass spectrometry (SIMS) profiles of Fig. 39, the 2H diffuses out of the p+ GaN layer and piles up in the defective InGaN layer, (seen from the TEM results), and suffers from thermal degradation during growth of the p+ GaN. Note that there is still sufficient 2H in the p+ GaN (--~1019cm3) to passivate all of the acceptors present, but electrical measurements showed that the p-doping level was at its maximum value of---3 x 1017cm-3.These results confirmed that as in other IH-V semiconductors, hydrogen can exist in a

480

Wide Bandgap Semiconductors

number of different states, including being bound at dopant atoms or in an electrically inactive form that is quite thermally stable. It is expected that, after annealing above 700~ all of the Mg-H complexes will have dissociated, and the electrical measurements show that they have not reformed. In other III-Vs, the hydrogen in p-type materials is in a bond centered position forming a strong bond with a nearby N atom, leaving the acceptor threefold coordinated.[ 168] Annealing breaks this bond and allows the hydrogen to make a short-range diffusion away from the acceptor, where it probably meets up with other hydrogen atoms, forming molecules or larger clusters that are relatively immobile and electrically inactive. This seems like a plausible explanation for the results of Figs. 38 and 39, where the Mg electrical activity is restored by 700~ while the hydrogen remains in the layer even at 900~ In the material hydrogenated by implantation, there is almost certainly a contribution to the apparently high thermal stability by hydrogen being trapped at residual implant damage, evident by the fact that the 2H profile retains a Pearson IV type distribution even after 900~ annealing. The other important point from Fig. 39 is that as in other defective crystal systems, hydrogen is attracted to regions of strain, in this case, the InGaN sandwiched between the adjoining GaN layers.

100 P .< ;>

SO

~'~

so

~z ~ ,<:_ .,,

40

20

--e---

~ ~"

0

. . 200

0

300

H z AMBIENT \ Nz AMBIENT ~ .

,

400

.

,

500

.

,

600

.

,

9

,

700 800

,

9

900

ANNEALING TEMPERATURE (~ 38. Fabrication of passivated Mg acceptors remaining in hydrogenated p-GaN after annealing for 20 min at various temperature in either N 2 or H 2 ambients. F i g u r e

Hydrogen in Wide Bandgap Semiconductors 481

I05

1021 ~---SiNx

,~-p-GaN-~InGaN ! n-GaN

oum 104 r~

2H ANNEALED [_/"N 103

O2

Z

10t9

Z 0

H AS-IMPLA 102

l0 ts

Z 0

10

I0 t7 ll

9 A,A .... 0

0.4

AJ,,

.......

0.8

,,

,.,

C

...

,,

1.2

DEPTH(MICRON)

Figure 39. SIMS profiles of 2H in an implanted (50 keV, 2 x 1015cm-2 through a SiN x cap) double heterostructure sample, before and after annealing at 900~ for 20 mins.

In conclusion, the apparent thermal stability ofhydrogen-passivated Mg acceptors in p GaN is dependent on the annealing ambient, as it is in other compound semiconductors. While the acceptors are reactivated at <700~ for annealing under N 2, hydrogen remains in the material until much higher temperatures and can accumulate in defective regions of doubleheterostructure samples grown on A120 3. It will be interesting to compare the redistribution and thermal stability of hydrogen in homoepitaxial defects present in the currently available heteroepitaxial material.

3.7

H in Cubic GaN

Estreicher and Marie [170]have published ab-initio calculations on the behavior of H in cubic GaN. While the AB N site was higher in energy, H § was found to be bonded to N in a BC-like position. The lowest energy configuration was found for H ~ at the BC site, with H bound to Ga and most of the odd electrons localized on N. The diffusion activation energy for H ~

482

Wide Bandgap Semiconductors

was calculated to be <1 eV. The H- state was found to be most stable at the ABGa site, with a corresponding large relaxation of the Ga atom, and an activation energy for diffusion o f - l . 5 eV. Molecules were found to form at the Taa site, and the Mg-H pair had H in a BC position, strongly bound to N. The Mg atom was found to move to the plane of its three N nearest neighbors, and to show negative overlap with H (i.e., a nearly unperturbed N-H bond would form). Other calculations put H at the AB n site of N in Mg-H complexes, with a very high N-H stretching frequency.[ 169]

4.0

H Y D R O G E N IN SiC

There has been a resurgence in interest in SiC in recent times. The main applications are for high temperature/high power electronics, including Metal Oxide Semiconductor Field Effect Transistors, millimeter wave devices and junction field effect transistors. [1711-[177] Blue light-emitting diodes based on SiC are commercially available,J1781 although they face stiff competition from devices based on GaN which has a direct bandgap and therefore higher optical output. [179] Hydrogen is a component of virtually every gas or chemical involved in the growth and processing of SiC. For example, silane (Sill4) and propane (C3H8) diluted with hydrogen can be employed as the sources for SiC epitaxial growth.[ 180] Alternative organosilanes are also gaining attention.E181] Boron doping can be accomplished with diborane (B2H6) diluted with hydrogen.[ 1801The extremely good chemical stability of SiC requires that dry etching be employed for patterning, and hydrogen is generally added to the plasma chemistry to avoid rough surfaces.[182]-[1841 High temperature (1500-1700~ annealing in H2, followed by quenching to room temperature, produced hydrogen passivation (up to---75% decrease in carrier density) in both n-type (N-doped) and ptype (A1, B) SiC. [185) This effect was reversed by subsequent annealing in He at the same temperatures.[ 18s1 Annealing at 1700~ of chemical vapor deposited 6H-SiC epilayers increased the net hole concentration by a factor of approximately three and was accomplished by outdiffusion of hydrogen[180] (consistent with residual passivation ofB acceptors in the asgrown films). Theoretical studies of the properties of hydrogen in SiC have suggested that a tetrahedral interstitial site is the lowest energy site in undoped 3C-SiC, with a barrier to diffusion to the bond centered site of 1.5 eV.[186]-[188]In 6H-SiC, it was suggested that another interstitial position (the R-Site) was the lowest energy site, with the diffusivity being slightly slower than in Si. [186]-[1881

Hydrogen in Wide Bandgap Semiconductors 483 Bulk 6H-SiC substrates from Cree Research were employed in these experiments. The n-type wafer was doped to 1.5 • 1018 cm-3 (N dopant) while the p-type sample was doped to 2.5 x 1018 cm-3 (B dopant). The Si face of each sample was used for the plasma processing and subsequent analysis. The samples were exposed to 2H plasmas in a Plasma Therm SLR 770 reactor, shown schematically in Fig. 40. The plasma is generated in a Wavemat (Model 4400) low profile ECR source (2.45 GHz) at 500 W forward power and a pressure of 10 mtorr. Without any additional biasing of the sample, the average ion energy in the discharge is--eV, with an ion density of~3 x 1011 cm -3. The deuterium neural density is 3 x 1014 cm "3 at this pressure. In some cases, 150 W of power (13.56 MHz) was applied to the sample position to increase the incident ion energy to --175 eV. We have previously found in GaAs that this can dramatically increase the incorporation depth of deuterium.[ 189) Typical plasma exposures were 30 min, with the sample held at 300~ Subsequent annealing at 500 or 700~ for 1 min was carried out under a N 2 ambient in a Heatpulse 410T furnace. The atomic profile was measured in all samples using a Cameca IMS 4f system, using a Cs § ion beam. The 2H concentrations were quantified using ion implanted standards, while the depth scales were established from stylus profilometry of the sputtered analysis craters. Current-voltage (I-V) measurements using stainless steel probes were used to measure the breakdown voltages of the samples for both polarities of voltage. Figure 41 shows SIMS profiles of 2H in n-type SiC after exposure to the 2H plasma, and after subsequent annealing at 700~ The most interesting feature is that the application of rf power to the sample position to increase the ion energy (there were 1D+ and 2D+ species present in the plasma) increased the incorporation depth, but the peak near-surface deuterium concentration was reduced. It is expected that the more energetic ions present when rf power is applied can be implanted into the SiC (projected ranges 20-30 A) which lowers the average deuterium concentration right at the surface. When the rf biasing is not present, there is a greater chance for 2H agglomeration into molecules and large clusters. All the deuterium must enter the SiC fight at the surface, rather than have a certain amount implanted slightly beyond the immediate surface. With less chance of self-trapping in the latter case, the maximum incorporation depth is slightly higher. A qualitatively similar effect in n- and p-type GaAs has been seen.[ 189] Annealing at 700~ for 1 min provided outdiffusion to the surface of more than 50% of the deuterium.

484

Wide Bandgap Semiconductors

MICROWAVE COUPL,INO

AND "n.JN~O

.

r'E711 , ....

,

I

.... I

mm

!1-E71 tt t?

.

HELIUM BACKSIDE '= MAGNET COIL COOLING

"~'

F i g u r e 40. Schematic o f reactor e m p l o y e d for deuteration o f SiC.

Figure 42 shows I-V characteristics from the n-SiC before and after 2H plasma exposure. The ECR-only treatment produced a slight decrease in forward and reverse bias-on. This was related to a slight preferential loss of Si from the surface that lowers the barrier height. This was confirmed by Auger Electron Spectroscopy (AES) measurements of the deuterated surface. The ECR + rf treatment produced a more significant increase in current turn-on, which was believed to be due to the introduction of deep trap states that reduced the shallow donor density in the near-surface region. It was not believed to arise from donor passivation, because similar shifts in turn-on voltage for the He plasma exposures were observed. Figure 43 shows the I-V characteristics before and after annealing at 500~ and 700~ of the ECR + rf exposed SiC. The annealing had no significant effect on the I-V characteristics, suggesting that the plasma-induced defect states were quite stable. This was consistent with past results on Ar-plasma damage in SiC, which persisted even at 1050~ 19~ SIMS profiles of 2H in plasma exposed p-SiC are shown in Fig. 44. Once again, the deuterium penetration is larger, but now the peak concentration is also larger in the samples where rfbiasing of the lower electrode was employed. This is due to the fact that now the deuterium can bond to the B acceptors, forming neutral (B-H) ~ complexes. The microscopic

Hydrogen in Wide Bandgap Semiconductors 485 structure of these complexes has not been investigated closely in SiC, but it is assumed in analogy that the deuterium occupies a bond centered position between a B atom and a neighboring Si or C atom. In Si, the acceptor and its neighboring lattice atom relax away from each other to allow the deuterium into the bonding environment. The acceptor may also trap multiple deuterium atoms (B-Dn) leading to an excess of deuterium relative to the acceptor concentration. Note in Fig. 44 that there is little motion of the deuterium after annealing up to 700~ unlike the case of n-SiC. The incorporation depth of 2H is also much less than in S i ~ i f one takes the depth of incorporation as--0.08 ~tm (where the 2H profile reaches the background sensitivity of the SIMS measurement), then the diffusivity at 200~ can be estimated to be _<3.5 x 10 -14 em2.sec -1. This is roughly two orders of magnitude slower than in p-Si of the same acceptor density.

10

N-SiC 2H PLASMA 200C 10

!

"y 8

I

0

2o0

4oo

6oo

o00

Io#o

DEFTH(ANGSTROMS)

Figure 41. SIMS profiles of 2H in n-SiC exposed to ECR 2H plasmas at 200~ for 30 min at 500 W microwave power, either with or without 150 W or rf power to accelerate the ions. The ECR + rf sample was subsequently annealed at 700~ for 1 min under N 2.

486

Wide Bandgap Semiconductors .

40 .

,

!

'r

"

n-SiC

ECR 500 W RF 150 W 200 ~

20

< 0

D2 9

t--

~

"

L_

tO

t'

-20

.,

-40 -40

-20

~,nneal ~nneal

0

'

210

'

40

Voltage (V) Figure 42. I-V characteristics from n-SiC exposed to ECR 2H plasmas at 200~ for 30 min at 500 W microwave power, either with or without 150 W ofrfpower to accelerate the ions.

!

40

!

!

!

n-SiC ECR 500 W

!

Treatment ~

-J

200 "C D2

A

< ::1. v

0

/ L R F 150 W

;F

i-

t..,. ::3

to -20

-40 9

-40

!

-30

9

-20

|

=

-10

0 ",'o'

!

3o

o

4o

Voltage (V) Figure 43. I-V characteristics from n-SiC exposed to an ECR 2H plasma at 200~ for 30 min at 500 W microwave power and 150 W rf power, and subsequently annealed at 500 or 700~ for 1 min under N 2.

The addition of rf biasing to the sample position was also more effective in producing passivation in the p-SiC, as shown in the I-V characteristics of Fig. 45. The ECR-only plasma exposure cased at ~20% increased in tum-on voltage due to the higher resistivity of the hydrogen passivated SiC. However the ECR + rfplasma exposure increased the ttwn-on

Hydrogen in Wide Bandgap Semiconductors 487 voltage to > + 50 V (the limits of the probe station), a factor of two increased over the unexposed control material This is much larger than can be accounted for by the additional ion-induced damage. Typically, the use of He or Ar plasmas increased the turn-on voltage by ~20-% by introducing deep level traps that reduced the carrier density in the near-surface (-0.1 ~tm) region.

i~ I

'p-SiC 2H PLASMA 200C

10a

CR+rf

!. o

0

400

500

a00

1000

DE~'H(ANGSTROMS)

Figure 44. SIMS profiles of 2H in p-SiC exposed to ECR 2H plasmas at 200~ for 30 min at 500 W microwave power either with or without 150 W of power to accelerate the ions. The ECR + rf sample was subsequently annealed at 700~ for 1 min under N 2.

The I-V characteristics from annealed p-SiC samples that were deuterated with ECR + rfplasmas are shown in Fig. 46. The 500~ anneal had no affect on the resistivity, while the 700~ treatment reduced the turnon voltages to • V. This showed that the reactivation of passivated acceptors was almost complete at this temperature. Most of the residual carder loss was due to the deep levels introduced by the ion bombardment, which does not anneal completely even at 1050~ The strength of the

488

Wide Bandgap

Semiconductors

acceptor-deuterium bonds was comparable to those in GaN(Mg,H) which must also be annealed at-~700~ to reactivate the passivated acceptors. If one assumes the reactivation mechanism is thermal dissociation of the acceptor-deuterium complex, and this is a first order process, [186] with the deuterium moving away from the acceptor and probably attaching to other deuterium atoms to form neutral, immobile D 2 molecules, then the reactivation energy E a for these complexes can be calculated from:

Ea=kzcnl (tv)In/1-~OO31

Eq. (12)

where k is Boltzmann's constant, T is the absolute temperature of the annealing process, v is an attempt frequency for bond breaking (assumed to --1014 S-1) and P / P o the ratio of holes to total acceptors (i.e., the fraction of reactivated acceptors), then E a ~ 3.3 eV for B acceptor passivation in SiC. This is much higher than for acceptors in Si ( E a ~2.1 eV) and attests to the high temperature stability of most processes in SiC. I

40

20A

< ~

No RF

No Treatme~

D2

0

e-

el) L_ t_

L)

i

p-SiC ECR 500 W RF 150 W 200 *C

-20

1

f

RF 150 W

-40 I

-40

-20

i

i

0

I

20

,

40

Voltage (V) Figure 45. I-V characteristicsfromp-SiC exposedto ECR 2H plasmasat 200~ for 30 min with 500 Wmicrowavepower,eitherwithor without 150W ofrfpowerto acceleratethe ions. In summary, acceptor passivation is much more efficient in 6H-SiC than is donor passivation. This is a fairly general phenomenon in semiconductors, and is related to the more stable bond centered position for

Hydrogen in Wide Bandgap Semiconductors 489 hydrogen passivating acceptors compared to the antibonding interstitial position for hydrogen passivating donors. Reactivation ofpassivated acceptors occurs at-~700~ in p-SiC, corresponding to a reactivation energy of -3.3 eV. The diffusivity of 2H is much slower in SiC at 200~ than it is in Si of the same doping level. The incorporation distance can be influenced by the average ion energy in the plasmas used to introduce hydrogen or deuterium, which is probably related to a decrease in self-trapping efficiency at the immediate surface of the SiC. One needs to be aware of acceptor passivation effects that can occur when hydrogen is part of the growth or processing ambient for SiC.

- p-SiC ECR 200 W RF 150 W 20 200 *C

40

<~L

A

i700 *C Post Anneal i

.

D2 0

f

r-

t

As Deposit And

tO -20

500 ~

Post Anneal

-40 -40

-~o

'

o

'

~0

'

40

Voltage (V)

Figure46. I-V characteristics fromp-SiC exposedto an ECR2H plasma at 200~ for 30 min at 500 W microwavepowerand 150 W rfpower, and subsequentlyannealedat 500~or 700~ for I min under N2.

4.1

Thermal Stability of Deuterium in 3C-SiC

Some impressive demonstrations have been reported for high power SiC metal-oxide semiconductor field effect transistors, microwave devices, and high temperature junction field effect transistors. Improved epitaxial growth, ohmic contact ,and processing steps such as dry etching and implant doping/isolation are needed to fully exploit this promising technology. In many of these processes, hydrogen plays an important role as part of the ambient. For example, in chemical vapor deposition of borondoped 6H-SiC, s i n / H E and C3H8/H 2 were employed as precursors, BEH6/H2 as a dopant source and HC1/H2 as an in-situ substrate etchant.[ 18~ Hydrogen

490

Wide Bandgap Semiconductors

concentrations near 1019cm -3 were found in as-grown epitaxial layers with reductions in net hole densities of 3-4 times, due to passivation of the boron. Reactive ion etching processes for SiC have traditionally been plagued by rough, micromasked surfaces, but addition of H 2 to the plasma chemistry generally alleviates these problems.[ 172]Both donors and acceptors in SiC can be passivated during high temperature (1500-1700~ annealing in H 2, followed by quenching to room temperature.[ 185] Muon spin resonance studies in SiC have shown most of the muonium to occupy a preferential interstitial site,[ 187]while extensive calculations of the properties of H have shown that this site is favored over the bond-centered position in 3C-SiC, and is one of the lowest energy sites in 6H-SiC. There have been no systematic studies of the thermal stability of hydrogen in this material, although it has been shown that dopant reactivation is complete at 1700~ in hydrogenated SiC.[ 1801 Nominally undoped 3C-SiC (n--8 x 1017cm-3) substrates cut 3.5 ~ off [0001 ] towards the middle of [ 1010] and [ 1020] were implanted into the polished Si face with 100 keV2H + ions to a dose of 2 x 1015cm-2. Sections from these substrates were annealed under N 2 for 20 min at temperatures up to 950~ in a tube furnace, or for 1 min at temperatures up to 1200~ in a Heatpulse 410T system. Other substrates were exposed to an Electron Cyclotron Resonance 2H plasma (500 W microwave power, 10 mtorr pressure) for 30 min at 300~ The deuterium profiles were measured by Secondary Ion Mass Spectrometry (SIMS) using a Cs 1 ion beam. The concentrations were quantified using the implant standards, while the depth scales were established from stylus profilometry measurements of the analysis craters. Figure 47 shows the 2H atomic profiles in the plasma exposed sample before and after annealing at temperatures of 1000~ for 1 min. The incorporation depth is <_0.1 ~tm, or an effective diffusivity of_<6 x 10"16 cm2/S, much lower than in Si or wide bandgap materials such as GaN and A1N. The fact that the deuterium is present at a high concentration in a shallow region near the surface is indicative that it is present in platelet form or other extended clusters. Transmission electron microscopy has not been performed on the samples to confirm this. In B-doped SiC containing hydrogen, the hydrogen profile follows the B profile because of the presence of neutral (B-H) ~ complexes. In undoped materials, the intrinsic solubility of hydrogen is so low that isolated hydrogen is essentially undetectable and only other states (H 2 and larger clusters, Si-H) are present at significant concentrations. Note that the deuterium concentrations in the plasma-treated material is reduced by --90% by 800~ annealing, and < 1% remains after a 1000~ 1 min anneal.

Hydrogen in Wide Bandgap Semiconductors 491

l

l

l

l

l

j

~

- -

,,,

,

SiC 0.5h -'H plasma treated at 300~ nd processed at temperature shown

ll~/ ~

As treated r s o~176 w'r ~ k 9oo~

<

lo

10[t ....

I

O.O

O.OO

9-

9 9I `~L O,tO

9" I ..... O. tJ

i , ~ .~ I .... O.M

O.aO"

O.M

Depth (/am) Figure 47. SIMS profiles of 2H in SiC exposed to a plasma at 300~ for 30 min, and subsequently annealed at 800, 900 or 1000~ for I min.

Implanted deuterium showed much higher thermal stabilities. Figure 48 shows SIMS profiles of samples before and after 20 min anneals up to 950~ Under these conditions, there is no measurable redistribution of the deuterium. Note that the profiles cannot be described by a single Gaussian, but has a Pearson IV-type distribution. This is fairly typical for implanted hydrogen in semiconductors. The peak of the distribution occurs at-4).75 ~tm. The SIMS profiles of the 2H implanted SiC annealed at higher temperatures are shown in Fig. 49, in both log and linear form. The 2H is lost by outdiffusion to the surface, with the concentration in the region < 0.4 jam increasing with annealing temperature, while the density in the peak region around 0.75 ~tm decreases. It is expected that the deuterium in the implanted material is in isolated interstitial positions, and part of the population is rapped at the residual implant damage, which for 2H overlaps the atomic profile. Therefore, with annealing, there is remnant deuterium that remains in the initial implanted distribution shape, or its peak concentration is

492

Wide Bandgap Semiconductors

reduced by outdiffusion to the surface. The outdiffusion is controlled to some extent by the annealing of the implant damage, which releases trapped deuterium.

. ~

lO1't~1

F 1: lOmL

~,

E

.

!

-

9 .~.~

-

-

~' -

i

. . . .

q.lm'sla looimv Uxlo~Sam '4 &ANNEAL.I~ATIIQ~700tlIOO, Ir/'8,&IIEr~

1

10tin

.,~

10 lm

"~

101T

'~

1011

lol1

11014 I 9 Q.O

,

9 -

t

0.11

. . . . .

t

. . . .

1.0

r

. . . . .

,

IJ

Depth (pm) Figure 48. SIMS profiles of ZHin SiC implantedwith 2H ions (100 keV, 2 x 10~5cm-2)and subsequently annealed at various temperatures for 20 mins at each temperature.

The decrease in 2H concentration in the peak around 0.75 ~tm as a function of the annealing temperature showed little change as measured by the area under the curve up to 1150~ because the deuterium is moving toward the surface but was not lost until 1200~ where 30% was lost. In the range of 1150-1200~ the 2H concentration decreases at the rate o f - 7 x 1017 cm-3: OCl" This is a much higher stability than for implanted deuterium (or hydrogen) in any semiconductor we have examined, including A1N. The difference in apparent thermal stability between the plasma treated and implanted material is the presence of lattice damage in the latter which traps the deuterium. It is well established, that hydrogen is attracted to any regions of stain or disorder in semiconductors, and thus the outdiffusion is retarded. By contrast, in the plasma treated material, there is not an impediment to the deuterium motion once it is released from its initial state. In conclusion, high concentrations (>5 x 1019cm"3) of deuterium are incorporated into undoped 3C-SiC by exposure to plasmas at 300~ but

Hydrogen in Wide B a n d g a p Semiconductors

493

the incorporation depth is shallow (_<0.10 l.tm). Most of this 2H is removed by annealing above 800~ Implanted 2H required higher temperatures (> 1100~ for loss from the crystal. Therefore different methods of incorporation produce different apparent thermal stabilities. Control of the hydrogen concentration in SiC devices requires careful understanding of which growth and processing steps have the potential of introducing it.

I 021 I

.

.

. . 2H in SiC 100 keV 2xl0'Scm 2 annealed at temperature shown

12"Ox1020

1020'

/~/-'/

1o'~

If l 7

~

t

:.So21A! \

I

,

$::

._"''"///~Ii llxlO'~ ~

N lO,,[~~' o <

f~

,,5o~ 1/II

~

II II

~ |

w

I, t

"~

I0'

xlO, ~ ~

'~ O,O

S' O.l

0,4

0.6

O.I

t0.0

1.0

Depth (~tna) Figure 49. SIMS profiles of 2H in SiC implanted with 2H+ ions (100 keV, 2 x 1015cm-2) and subsequently annealed at 1050, l 100, I 150 or 1200~ for I min.

4.2

Diffusion of Hydrogen in 6H-SiC

Linnarsson et al. 9191) made a study of hydrogen diffusion in implanted 6H polytype SiC. In material implanted with 50 keV H+ ions at doses of _<5 x 1015cm-2, hydrogen was found to begin migrating above 700~ and to decorate the defect distribution created by the ion implantation. An activation energy for migration of 3.5 eV was calculated for the ion-damaged

494

Wide Bandgap Semiconductors

region, though the value is likely to be lower in damage-free material where hydrogen trapping was reduced. Roberson and Estreicher[ 192] calculated the properties of hydrogen in 2H, 6H, and 3C Si polytypes based on the muon spin resonance experiments[ 193]that showed no anomalous muonium present in SiC at any temperature. In 3C material, normal muonium and ~t+ were present at low temperature, with ~t+ transforming into Mu as the temperature was increased. In 6H material, three different Mu signals were detected at low temperature.[ TM] The theory [5][188][192] predicts that in 3C-SiC, the lowest energy site for hydrogen is at Tsi position, followed by the BC site. The Tsi-~BC barrier was estimated to be <1.5 eV. In 2H-SiC, the lowest energy position was at the R site with BC being 0.39 eV higher in energy. The barrier for diffusion was < 2 eV. These results were used to predict that the lowest energy site in 6H-SiC would be the R site, followed by Zsi and BC positions.

5.0

DIAMOND

Hydrogenation has been reported to produce significant decreases in resistivity of both natural and synthetic diamond.[195][196]In general, CVD diamond has resistivities on the order of 106~cm, and is expected to contain large quantities of hydrogen, while natural diamond is around 1016~cm. Post-growth annealing of CVD diamond at-~800~ causes H outdiffusion and an increase in film resistivity. The CVD of diamond generally occurs in the presence of large concentrations of molecular hydrogen, [197]-[199]giving rise to a plethora of CH infra-red lines.[ 2~176 Typical nuclear magnetic resonance measurements show H contents up to 1 at .% in CVD diamond films,[197] and that it is incorporated preferentially at grain boundaries and other defects. The thermal conductivity of diamond is also found to generally decrease with hydrogen content. Theoretical calculations find that (B-H) complexes form in diamond, though there is little agreement on the microscopic structure. [201][202] From muon spin resonance and theory, it is well established that the stable configuration for hydrogen in diamond is a relaxed BC site, with the c-c bond relaxing back to incorporate the H. [203][204] There is also agreement that the activation energy for diffusion is very high for H in diamond (between 1.9-5.3 eV in various calculations).[2~ 2~ It is further predicted that a H 2 molecule at a T site will dissociate into the HAB-HBc

Hydrogen in Wide Bandgap Semiconductors 495 complex.[ 2~ Very little quantitative IR data is available for well-characterized diamond samples and thus there is little experimental verification of the theory.

6.0

II-VI COMPOUNDS

ZnSe-based heterostructures are good candidates for blue/green LEDs and lasers, as discussed elsewhere in this book. In gas phase growth of these materials, hydrogen is known to be incorporated from the H2Se precursor.[ 2~ H ~ is theoretically found to diffuse rapidly between Tzn and Tse sites, while the BC site was energetically unfavorable. The common donor in ZnSe, namely C1, is not strongly affected by H, whereas there is efficient passivation of the N acceptors in p-type material.[2~ 21~ The acceptor passivation is quite thermally stable (>450~ 1 hr) and does not show minority-carrier enhanced debonding. IR bands of H-related defects have been found in H § and D § implanted CdS, CdTe, ZnTe and ZnSe, with the H forming stronger bonds with the group VI atom (S, Se, and Te). [211][212]

CONCLUSIONS Hydrogen plays an important role in all the wide bandgap semiconductors, producing strong passivation of acceptors in all the materials, and donor passivation also in some cases. The hydrogen is unintentionally incorporated during growth and processing and remains in the material until quite high temperatures. Reactivation of dopants occurs in the range 400-700~ depending on the material.

ACKNOWLEDGMENTS This work is partially funded by NSF-DMR 9421109 (L D. Hess).

496

Wide Bandgap Semiconductors

REFERENCES 1. Pearton, S. J., Corbett, J. W., and Shi, T. S.,Appl. Phys. A., 43"153 (1987) 2. Myers, S. M., Bases, M. I., Bimbaum, H. K., Corbett, J. W., DeLeo, G. G., Estreicher, S. K., Haller, E. E., Jena, P., Johnson, N. M., Kircheim, R., Pearton, S. J. and Stavola, M., Rev. Mod. Phys., 64:559 (1992) 3. Pankove, J. I., and Johnson, N. M., "Hydrogen in Semiconductors," Vol. 34, Semiconductors and Semimetals, Academic Press, San Diego (1991) 4. Pearton, S. J., Corbett, J. W. and Stavola, M., Hydrogen in Crystalline Semiconductors, Springer-Verlag, Berlin (1992) 5. Estreicher, S. K., Mat. Sci. Eng., R14:319 (1995) 6. Pearton, S. J., Mat. Sci. For., pp. 148-149 (1994) 7. Street, R. J., Hydrogenated Amorphous Silicon, Cambridge University Press, New York (1991) 8. Estreicher, S. K., Hyperfine Interactions, 86:625 (1994) 9. Bambakidis, G., and Bowman, R. C., Jr., Hydrogen in Disordered and Amorphous Solids, NATO ISI Series 136, Plenum, New York (1986) 10. Stutzmann, M., and Chevallier, J., Hydrogen in Semiconductors, North Holland, Amsterdam (1991) 11. Pankove, J. I., Semiconductors and Semimetals, 21D(14):261, Academic Press, New York (1984) 12. Hansen, W. L., Pearton, S. J., and Hailer, E. E., Appl. Phys. Lett., 44:606 (1984) 13. Kohler, K., Cobum, J. W., Home, D. E., Kay, E. and Keller, J. H., J. Appl. Phys., 57:59 (1985) 14. Curran, J. E., Vacuum, 34:343 (1984). 15. Boenig, H. V., Plasma Science and Technology, Ch. 13, p. 36, Comell University Press, Ithaca, New York (1983) 16. Oehrlein, G. S., Lindstr'm, J. L. and Corbett, J. W., Phys. Lett., 81:246 (1981) 17. Pearton, S. J., Hansen, W. L., Haller, E. E., and Kahn, J. M., J. Appl. Phys., 55:1221 (1984) 18. Aucouturier, M., Rallon, O., Mautref, M., and Belouet, C., J. Phys., 43:117 (1982) 19. Pearton, S. J., Kahn, J. M., Hansen, W. L., and Haller, E. E., J. Appl. Phys., 55:1461 (1984) 20. Chevallier, J., Dautremont-Smith, W. C., Tu, C. W., and Pearton, S. J., Appl. Phys. Lett., 47:108 (1985)

Hydrogen in Wide B a n d g a p Semiconductors

497

21. Pearton, S. J., Kahn, J. M., and Hailer, E. E., J. Electron Mater., 12:1003 (1983) 22. Hwang, J. M., Schroder, D. K., and Biter, W. J., d. Appl. Phys., 57:5275 (1985) 23. Chapman, B., Glow Discharge Processes, Wiley, New York (1980); Oeschsner, H., "Thin Film and Depth Profile Analysis," Topics Curr. Phys., Vol. 37 Springer, Berlin, Heidelberg (1984) 24. Muller, J. C., Barhdadi, A., Abadou, Y., and Siffert, P., Re. Phys. Appl., 22:649 (1987) 25. Harper, J. M. E., Cuomo, J. J., and Kaufman, H. R., Ann. Rev. Mater. Sci., 13:413 (1983) 26. Chevallier, J., and Aucouturier, M.,Ann. Rev. Mater. Sci., 18:261 (1988) 27. Ji, C., Shi, T. S. a,nd Wang, P., Nucl. Instr. Meth. Physics Research B, 12: 486 (1985) 28. Bech Nielsen, B., MRS Proc., 59:487 (1986) 29. Yapsir, A. S., Hadizad, P., Lu, T. M., Corelli, J. C., Lanford, W. A., and Bakhru, H., Appl. Phys. Lett., 50:1530 (1987) 30. Van Wieringen, A., and Warmoltz, N., Physica, 22:849 (1956) 31. Kleinhenz, R. L., Lee, Y. H., Singh, V. A., Mooney, P. M., Jaworoski, A., Roth, L. M., Corelli, J. C., and Corbett, J. W., Defects." Rad. Effects in Semicond., (J. H. Albany, ed.) IEE Publishers, Bristol, UK, 1978, pp. 371375 (1978) 32. Picraux, S. T., and Vook, F. L., Phys. Rev. B, 14"1593 (1976) 33. Picraux, S. T., and Vook, F. L., Phys. Rev. B, 18:2066 (1978) 34. Picraux, S. T., Phys. Today p. 22 (June 1978) 35. Hoffman, R., J. Chem. Phys., 39:1397 (1963) 36. Singh, V. A., Weigel, C., Corbett, J. W., and Roth, L. M., Phys. Star. Sol., (b) 81:637 (1977) 37. Singh, V. A., Corbett, J. W., Weigel, C. an,d Roth, L. M., Phys. Lett. A, 65:261 (1978) 38. Wang, J., and Kittel, C. A., Phys. Rev., 87:713 (1973) 39. DeLeo, G. G., Fowler, W. B., and Watkins, G. D., Phys. Rev. B, 29:1819 (1984) 40. Mainwood, A., and Stoneham, A. M., Physica B, 116:101 (1983) 41. Mainwood, A., and Stoneham, A. M.,J. Phys. C, 17:2513 (1984) 42. Rodriquez, C. O., Jaros, M, and Bran&, S., Solid State Commun., 31:43 (1979) 43. Bech Nielsen, B., Phys. Rev. B. 37:6353 (1988)

498

Wide Bandgap Semiconductors

44. Hall, R. N., Metallurgical Soc. AIME, (L. C. Kimerling, and J. M. Parsey, Jr., eds.), Warrendale, PA, 14a:759 (1985) 45. Pearton, S. J., Kahn, J. M., Hansen, W. L., and Haller, E. E., J. Appl. Phys., 55:1461 (1984) 46. Hall, R. N., IEEE Trans., NS-31,320 (1984) 47. Holbec, J. D., Nielsen, B., Jones, R., Sitch, P., and Oberg, S., Phys. Rev. Lett., 71:875 (1993) 48. Vetterhoffer, J., Wagner J., and Weber, J., Phys. Rev. Lett., 77:5409 (1996) 49. Johnson, N. M., and Herring, C., Phys. Rev. B, 38:1581 (1988) 50. Van de Walle, C. G., Denteneer, P. J. H., and Pantelides, S. T., Phys. Rev. B, 39:10791 (1989) 51. Dehk, P., and Snyder, L. C., Rad. Eft. Defects and Solids, 111 & 112:77 (1989) 52. Chang, K. J., and Chadi, D. J., Phys. Rev. Lett., 62:937 (1989) 53. Pankove, J. I., Zanzucchi, P. J., Magee, C. W., and Lucovsky, G., Appl. Phys. Lett., 46:421 (1985) 54. DeLeo, G. G., and Fowler, W. B., Phys. Rev. B 31:6861 (1985); Phys. Rev. Lett., 56:402 (1986) 55. Pankove, J. I., Wance, R. O., and Berkeyheiser, J. E., Appl. Phys. Lett., 45:1100(1984) 56. Hansen, W. L., Pearton, S. J., and Haller, E. E., Appl. Phys. Lett., 44:889 (19984) 57. Sah, C. T., Sun, J. Y. C., and Tzou, J. J.,J. Appl. Phys., 54:5864 (1983) 58. Mikkelsen, J. C., Jr., Appl. Phys. Lett., 46:882 (1985) 59. Bergman, K., Stavola, M., Pearton, S. J., and Lopata, J., MRS Proc., 104:281 (1988) 60. Chang, K. J., and Chadi, D. J., Phys. Rev. Lett., 60:1422 (1988) 61. Marwick, A. D., Oehrlein, G. S., and Johnson, N. M., Phys. Rev. B, 36:4539 (1987) 62. Marwick, A. D., Oehrlein, G. S., and Johnson, N. M., MRS Proc., 104:259 (1988) 63. Bech Nielsen, B., Andersen, J. U., and Pearton, S. J., Phys. Rev. Lett., 60:321 (1988) 64. Wichert, T., Skudlik, H., Deicher, M., Grubel, G., Keller, R., Recknagel, E., and Song, L., Phys. Rev. Lett., 59:2087 (1987) 65. Claxton, T. A., Evans, A., and Symons, M. C. R., J. Chem. Soc. Faraday Trans., 82:2031 (1986) 66. Estle, T. L., Estreicher, S., and Marynick, D. S., Hyperfine Int., 32:637 (1986)

H y d r o g e n in Wide B a n d g a p Semiconductors

499

67. Estle, T. L., Estreicher, S., and Marynick, D. S., Phys. Rev. Lett., 58" 1547 (1987) 68. Estreicher, S., Phys. Rev. B, 36:122 (1987) 69. Dehk, P., Snyder, L. C., Lindstrom, J. L., Corbett, J. W., Pearton, S. J., and A.J. Tavendale, Phys. Lett. A, 126:427 (1988) 70. De~k, P., Snyder, L. C., and Corbett, J. W., Phys. Rev. B, 37:6887 (1988) 71. Bingham, R. C., Dewar, M. J. S., and Lo, D. C.,J. Am. Chem. Soc., 97:1285 (1975) 72. Bergman, K., M.Stavola, Pearton, S. J., and Lopata, J., Phys. Rev. B, 37:2770 (1988) 73. Denteneer, P. J. H., Van de Wall, C. G., and Pantelides, S. T., Phys. Rev. B, 39:10809 (1989); Phys. Rev. B, 41:3885 (1990) 74. Stavola, M., Bergman, K., Pearton, S. J., and Lopata, J., Phys. Rev. Lett., 61:2786 (1988) 75. Chang, K. J., and Chadi, D. J.,Phys. Rev. B, 40:11644 (1989) 76. Patterson, B. D., Rev. Mod. Phys. 60:69 (1988) 77. Cox, S. F. J., and Symons, M. C. R., Chem. Phys. Lett., 126:516 (1986) 78. Westhauser, E., Albert, E., Hamina, M., Recknagel, E., Weidinger, A., and Moser, P., Hyper. Int., 32:589 (1986) 79. Kiefl, R. F., Celio, M., Estle, T., Kreitzman, S. R., Luke, G. M., Riseman, T. M., and Ensaldo, E. J., Phys. Rev. Lett., 60:224 (1988) 80. Gorelkinskii, Y. V., and Nevinnyi, N. N., Soc. Phys. Tech. Lett., 13:45 (1987) 81. Pearton, S. J., Corbett, J. W., and Stavola, M., Hydrogen in Crystalline Semiconductors, Springer-Verlag, Heidelberg (1992) 82. Nakamura, S., Mukai, T., Senoh, M., and Iwasa, N., Jpn.. J. Appl. Phys., 31:L139 (1992) 83. Stavola, M., Mat. Sci. Forum, 148/149:251 (1994) 84. Estreicher, S. K., Mat. Sci. Eng., R 14:319 (1995) 85. Pearton, S. J., Hydrogen in Compound Semiconductors, Trans-Tech Publications, Switzerland (1994) 86. Antell, G. R., Briggs, T. R., Butler, B. R., Kitching, S. A., Stagg, J. P., Chen, A., and Syskis, D. E., Appl. Phys. Lett., 53:700 (1988) 87. Sykes, D. E.,Appl. Phys. Lett., 53:758 (1988) 88. Clerjaud, B., Cote, D., and Naud, C., Phys. Rev. Lett., 58:1755 (1987) 89. Shinar, J., Kanah, A., Cavenett, B. C., Kennedy, T. A., and Wilsey, N., Solid State Comm., 59:653 (1986) 90. Chantre, A., Bouchet, L., and Andre, E., J. Electrochem. Soc., 135:2687 (1988)

500

Wide Bandgap Semiconductors

91. Pearton, S. J., Mat. Sci. For., 148/149:113 (1994) 92. Cole, S., Evans, J. S., Harlow, M. J., Nelson, A. W., and Wong, S., Electron. Lett., 24:929 (1988) 93. Glade, M., Grutzmacher, D., Meyer, R., Welk, E. G., and Balk, P., Appl. Phys. Lett., 54:2411 (1989) 94. Minagawa, S., and Kondow, M.,J. Electron. Meter., 19:597 (1990) 95. Hobson, W. S., Mat. Sci. For., 148/149:27 (1994) 96. Nishikawa, Y., Suzuki, M., Ishikawa, N., Kokubum, Y., and Katakoshi, G., J. Cryst. Growth, 123:181 (1992) 97. Kozuch, D. M., Stavola, M., Pearton, S. J., Pearton, C. R., and Lopta, J., Appl. Phys. Lett., 57:2561 (1990) 98. Woodhouse, K., Newman, R. C., DeLyon, T. J., Woodell, J. M., Scilla, G. J., and Cardone, F., Semicond. Sci. Technol., 6:330 (1991) 99. Stockman, S. A., Hanson, A. W., Jackson, S. K., Baker, J. E., and Stillman, G. E., Appl. Phys. Lett, 62:1248 (1993) 100. Kozuch, D. M., Stavola, M., Pearton, S. J., Abemathy, C. R., and Hobson, W. S.,J. Appl. Phys., 73:3716 (1993) 101. Abemathy, C. R., Mat. Sci. For., 148/149:3 (1994) 102. Pearton, S. J., Corbett, J. W., and Stavola, M., Adv. Mater., 4:332 (1992) 103. Estreicher, S. K., Mat. Res. Soc. Symp. Proc., 240:643 (1992) 104. Yuan, C., Salagaj, T., Gurary, A., Thompson, A. G., Kroll, W., Stall, R. A., Hwang, C. Y., Schurmann, M., Lu, Y., Mayo, W. E., Lu, Y., Krishnankutty, S., Shmagin, I. K., Kolbas, R. M., and Pearton, S. J., J. Vac. Sci. Technol. B, 13:2075 (1995) 105. Yuan, C., Salagaj, T., Thompson, A. G., Stall, R. A., Li, Y., and Pearton, S. J., Proc. Wide Bandgap Semiconductors and Devices Symp., ECS, Pennington, NJ, Vol. 95-21, pp. 10-25 (1995) 106. Yuan, C., Salagaj, T., Gurary, A., P. Zawadzki, C.S. Chem, Kroll, W., Stall, R. A., Li, Y., Schurmann, M., Hwang, C. Y., Mayo, W. E., Lu, Y., Pearton, S. J., Krishnankutty, S., and Kolbas, R. M., J. Electrochem. Soc., 142:2163 (1995) 107. Ohba, Y., Topical Workshop on III-V Nitrides, Nagaya, Japan, (Sept. 21-23, 1995), Solid State Electronics, 47:319 (1997) 108. Estreicher, S. K., Mat. Sci. Eng., Rep 14:319 (1995) 109. Seager, C. H., and Anderson, R. A., Appl. Phys. Lett., 63"1531 (1993) 110. Seager, C. H., and Anderson, R. A., Solid State Comm., 76:285 (1990) 111. Tavendale, A. J., Williams, A. A., Alexiev, D., and Pearton, S. J., Mat. Res. Soc. Symp. Proc., 59:469 (1985) 112. Zundel, T., Weber, J., and Tilly, L., Physica B, 170:361 (1991)

Hydrogen in Wide B a n d g a p S e m i c o n d u c t o r s

501

113. Seager, C. H., and Anderson, R. A., Appl. Phys. Lett., 59:585 (1991) 114. Kamiura, Y., Yoneta, M., Nishiyama, Y., and Haskimoto, F., J. Appl. Phys., 72:3394 (1992) 115. Yoneta, M., Kamiura, Y., and Haskimoto, F., J. Appl. Phys., 70:1295 (1991) 116. Szafranek, I., Bose, S. S., and Stillman, G. E., AppL Phys. Lett., 55" 1205 (1989) 117. Szafranek, I., and Stillman, G. E., J. Appl. Phys., 68:3554 (1990) 118. Tavendale, A. J., Pearton, S. J., Williams, A. A., and Alexiev, D., Appl. Phys. Lett., 56:1457 (1990) 119. Leitch, A. W. R., Prescha, T., and Weber, J., Phys. Rev. B, 44:5912 (1991) 120. Pearton, S. J., Hobson, W. S., and Abemathy, C. R., Appl. Phys. Lett., 61" 14 (1992) 121. Amano, H., Kito, M., Hiramatsu, K., and Akasaki, I., Jpn.. J. Appl. Phys., 28:L112 (1989) 122. Neugebauer, J., and Van de WaUe, C. G., Phys. Rev. Lett., 74:4452 (1995) 123. Neugebauer, J., and Van de Walle, C. G., Phys. Rev. B, 50, 8069 91994; Mat. Res. Soc. Symp. Proc., 378:503 (1995) 124. Bran&, M. S., Ager, J. W., Gotz, W., Johnson, N. M., Harris, J. S., Molnar, R. J., and Moustakas, T. D., Phys. Rev. B., 49:1478 (1994) 125. Brandt, M. S., Johnson, N. M., Molnar, R. J., Singh, R., and Moustakas, T. D., Appl. Phys. Lett., 64:2264 (1994) 126. Gotz, W., Johnson, N. M., Walker, J., Bour, D. P., Amano, J., and Akasaki, I., Tech. Dig. Int. Conf. SiC, Kyoto, Jpn.an, paper WEB-11-2 (Sept. 1995) 127. Abemathy, C. R., J. Vac. Sci. Technol. A, 11:869 (1993) 128. Abemathy, C. R., MacKenzie, J. D., Pearton, S. J., and Hobson, W. S., Appl. Phys. Lett., 66:1969 (1995) 129. Pearton, S. J., Abemathy, C. R., and Ren, F., Electronics Lett., 30:527 (1994) 130. Akasaki, I., Amano, H., Koide, N., Kotaki, M., and Manate, K., Physica C, 185:428 (1993) 131. Nakamura, S., Mukai, T., and Senoh, M., Appl. Phys. Lett., 64:1687 (1994) 132. Perlin, P., Gorczyca, I., Christensen, N. E., Grzegory, I., Teisseyen, H., and Suski, T., Phys. Rev. B, 45:13307 (1992) 133. Hwang, S. J., Shah, W., Hauenstein, R. J., Song, J. J., Lin, M. E., Strite, S., Sverdlov, B. N., and Morkoc, H., Appl. Phys. Lett., 64:2928 (1994) 134. Hong, C. H., Pavilidis, D., Brown, S. W. Q., and Rand, S. C., J. AppL Phys., 77:1705 (1995) 135. Davis, R. F., Physica, B, 185:1 (1993)

502

Wide Bandgap Semiconductors

136. Van Vechten, J. A., Zook, J. D., Homing, R. D., and Goldenberg, B., Jpn.. J. Appl. Phys., 31:3662 (1992) 137. Nakamura, S., Iwasa, N., Senoh, M., and Mukai, T., Jpn.. J. Appl Phys., 31:1258 (1992) 138. Pearton, S. J., Mat. Sci. For., 148/149:461 (1994) 139. Strite, S., Jpn.. J. App. Phys., 33:L699 (1994) 140. Zolper, J. C., Wilson, R. G., Pearton, S. J., and Stall, R. A., Appl. Phys. Lett., 68:1945 (1996) 141. Pearton, S. J., Vartuli, C. B., Zolper, J. C., Yuan, C., and Stall, R. A., Appl. Phys. Lett., 67:1435 (1995) 142. Stavola, M., and Pearton, S. J., Hydrogen in Semiconductors, (J. Pankove, and N. M. Johnson, eds.), Academic Press, San Diego (1991) 143. Zundel, T., and Weber, J., Phys. Rev., B 39:135449 (1989) 144. Wilson, R. G., and S.W. Novak, J. Appl. Phys., 69:466 (1991) 145. SIMS measurements performed at Charles Evans and Associates, Redwood City, CA or Evans East, Plainsboro, NJ. 146. Johnson, N. M., Ponce, F., Street, R. A., and Nemanich, R. J., Phys. Rev. B, 35:4166(1987) 147. Zavada, J. M., Wilson, R. G., Abemathy, C. R., and Pearton, S. J., Appl. Phys. Lett., 64:2724 (1994) 148. Pearton, S. J., Dautremont-Smith, W. C., Lopata, J., Tu, C. W., and Abemathy, C. R., Phys. Rev. B, 36:4620 (1987) 149. Stockman, S. A., and Stillman, G. E., Hydrogen in Compound Semiconductors, (S. J. Pearton, ed.), Trans Tech, Switzerland, pp. 501-536 (1994) 150. Pearton, S. J., Abemathy, C. R., Wisk, P., Hobson, W. S., and Ren, F., Appl. Phys. Lett., 63:1143 (1993) 151. Pearton, S. J., Abemathy, C. R., MacKenzie, J. D., Wilson, R. G., Ren, F., and Zavada, J. M., Electron. Lett., 31:327 (1995) 152. Tavendale, A. J., Williams, A. A., and Pearton, S. J., Appl. Phys. Lett., 48:590 (1986) 153. Mileham, J. R., Pearton, S. J., Abemathy, C. R., MacKenzie, J. D., Shul, R. J., and Kilcoyne, S. D., Appl. Phys. Lett., 67:1119 (1995) 154. Seaward, K. L., Appl. Phys. Lett., 61:3002 (1992) 155. DeSouza, J. P., Sadana, D., Baratte, H., and Cardone, F., Appl. Phys. Lett., 57:1129 (1990) 156. Shul, R. J., Kilcoyne, S. P., Hagerott-Crawford, M., Parmeter, J. E., Vartuli, C. B., Abemathy, C. R., and Pearton, S. J., Appl. Phys. Lett., 66:1761 (1995)

H y d r o g e n in Wide B a n d g a p S e m i c o n d u c t o r s

503

157. Pearton, S. J., Abemathy, C. R., Vartuli, C. B., MacKenzie, J. D., Shul, R. J., Wilson, R. G., and Zavada, J. M., Electron. Lett., 31:836 (1995) 158. Pearton, S. J.,Mat. Sci. Rep., 4:313 (1990) 159. Binari, S., Dietrich, H. B., Keiner, G., Rowland, L. B., Doverspike, K., and Wickenden, D. K., J. Appl. Phys., 78:3008 (1995) 160. Zolper, J. C., Vartuli, C. B., Abemathy, C. R., and Pearton, S. J., Appl. Phys. Lett., 6:3042 (1995) 161. Pearton, S. J., Vartuli, C. B., Zolper, J. C., Yuan, C., and Stall, R., Appl. Phys. Lett., 67:1527 (1995) 162. Nakamura, S., Iwasa, N., Senoh, M., and Mukai, T., Jpn.. J. Appl. Phys., 35:L74 (1996) 163. Chakrabarti, U. K., Pearton, S. J., Hobson, W. S., Lopata, J., and Swaminathan, V.,Appl. Phys Lett., 57:887 (1990) 164. Pearton, S. J., Wu, C. S., Stavola, M., Ren, F., Lopata, J., and DautremontSmith, W. C., Appl. Phys. Lett., 51:496 (1987) 165. Chatterjee, B., Ringel, S. A., Sieg, R., Hoffman, R. and Weinberg, I., Appl. Phys. Lett., 65:58 (1994) 166. Zavada, Z. M., and Wilson, R. G., Mat. Sci. For., 1448/149:189 (1994) 167. Hiramatsu, K., Itoh, S., Amano, H., Akasaki, I., Kumano, N., Shiraishi, T., and Oki, K., J. Cryst. Growth, 115:628 (1991) 168. Brandt, M. S., Ager, J. W., Gotz, W., Johnson, N. M., Harris, J. S.~ Molnar, R. J., and Moustakas, T. D., Phys. Rev. B, 49:14658 (1994) 169. Neugebauer, J., and Van de Walle, C. G., Phys. Rev. Lett., 75:4452 (1995) 170. Estreicher, S. K., and Marie, D. M., Mat. Res. Soc. Symp. Proc., 423:613 (1996) 171. Palmour, J. W., Carter, C. H., Jr., Weitzel, C. E., and Nordquist, K. J., Mat. Res. Soc. Symp. Proc., 339:133 (1994) 172. Flemish, J. R., Xie, K., and McLane, G. F., Mat. Rec. Soc. Symp. Proc., 428:106 (1996) 173. Bhatnagar, M., and Baliga, B. J., IEEE Trans. Electron. Dev., 40:645 (1993) 174. Trew, R. J., Yan, J. B., and Mock, P. M., Proc. IEEE 79:598 (1991) 175. McGarrity, J. M., McLean, F. B., DeLancey, W. M., Palmour, J. W., Carter, C. H., Edmond, J. A., and Oakley, R. E., IEEE Trans. Nucl. Sci., 39:1974 (1992) 176. Casady, J. A., Cressler, J. D., Dillard, W. C., Johnson, R. W., Agarwal, A. K., and Seirgeij, R. R., Solid State Electron., 39:1208 (1996) 177. Cooper, J. A., Jr., 3rdIntl. Conf. Expert Evaluation and Control in Compound Semicond., Freiburg, Germany, Mat. Sci. Eng. B, 72:269 (1997)

504

Wide Bandgap Semiconductors

178. Edmond, J. A., Kong, H.-S., Leonard, M., Dimitdev, V., Irvine, K., and Bulman, G., Topical Workshop on III-Nitrides, Nagoya, Japan, to be published in Solid State Electron (Sept. 1995) 179. Nakamura, S., Mukai, T., Senoh, M., Nagahama, S., and Iwasa, N., J. Appl. Phys., 74:3911 (1993) 180. Larkin, D. J., Sridhara, S. G., Devanty, R. P., and Choyke, W. P., J. Electron. Mat., 24:289 (1995) 181. Grow, J. M., Proc. SOTAPOCSXXIV, (F. Ren, S. J. Pearton, S.N.G. Chu, R. J. Shul, W. Pletschen, and T. Kamijoh, eds.), J. Electrochem. Soc., Pennington, NJ, 96-2:60 (1996) 182. Steckl, A. J., and Yih, P. H., Appl. Phys. Lett., 60:1966-1969 (1992) 183. Yih, P. H., and Steckl, A. J., J. Electrochem. Soc., 140:1813 (1993); 142:312 (1995) 184. Flemish, J. R., Wide Bandgap Semiconductors and Devices, (F. Ren, ed.) Electrochem. Soc., Pennington, NJ, 95-2:231 (1995) 185. Gendron, F., Porter, L. M., Poole, C., and Bringuier, E., Appl. Phys. Lett., 67:1253 (1995) 186. Estreicher, S. K., Mat. Sci. Eng., R14:319 (1995) 187. Estreicher, S. K., Wide Bandgap Semiconductors and Devices, (F. Ren, ed.) Electrochem. Soc., Permington, NJ, 95-21:78 (1995) 188. Robertson, M. A., and Estreicher, S. K., Phys. Rev. B, 44"10578 (1991) 189. Pearton, S. J., Abemathy, C. R., Wilson, R. G., Ren, F., and Zavada, J. M., Electron. Lett., 31:496 (1995) 190. Pearton, S. J., Lee, J. W., Grown, J. M., Bhaskaran, M., and Ren, F., Appl. Phys. Lett., 68:2987 (1996) 191. Linnarsson, M. K., Doyle, J. P., and Svensson, B. G., Mat. Res. Soc. Symp. Proc., 423:625 (1996) 192. Roberson, M. A., and Estreicher, S. K., Mat. Res. Soc. Symp. Proc., 242:355 (1992) 193. Schneider, J. W., Baumler, H., Keller, H., Kiefl, R., Kundig, W., Odermalt, W., Patterson, B. D., Estle, T., Rudaz, W., Blazey, K., and Schwab, C., Hyp. Int., 32:607 (1986) 194. Patterson, B. D., Baumler, H., Keller, H., Kiefl, R., Kundig, W., Odermalt, W., Schneider, J., Charpe, W., Estle, T., Spencer, D., Blazey, K., and Savic, I., Hyp. Int., 32:625 (1986) 195. Landstrass, M. I., and Ravi, K. V., AppL Phys. Lett., 55:975 (1989); Appl. Phys. Lett., 55:1391 (1989) 196. Albin, S., and Watkins, L., Appl. Phys. Lett., 56:1454 (1990) 197. Rutledge, K. M., and Gleason, K. K., Chem. Vap., Dept. 2:37 (1996)

H y d r o g e n in Wide B a n d g a p S e m i c o n d u c t o r s

505

198. McNamara, K. M., Williams, B. E., Gleason, K. K., and Scruggs, B. E., J. Appl. Phys., 76:2466 (1994) 199. McNamara, K. M., Scruggs, B. E., and Gleason, K. K., J. Appl. Phys., 77:1459 (1995) 200. Dischler, B., Wild, C., Muller-Serbet, W., and Koidl, P., Physica B, 185:217 (1993) 201. Mehandru, S. P., and Anderson, A. B.,J. Mater. Res., 9:383 (1994) 202. Brevet, S. J., and Briddon, P. R.,Phys. Rev. B, 49"10332 (1994) 203. Claxton, T. A., Evans, A., and Symons, M. C. R., J. Chem. Soc. Faraday Trans., 82:2031 (1986) 204. Estle, T. L., Estreicher, S. K., and Marynick, D. S., Hyp. Int., 32:637 (1986) 205. Mehandru, S. P., Anderson, A. B., and Angus, A. C., J. Mater. Res., 7:68 (1992) 206. Estle, T. L., Estreicher, S. K., and Marynick, D. S., Phys. Rev. Lett., 58:1547 (1987) 207. Briddon, P., Jones, R., and Lisher, G. M. S., J. Phys. C., 21 :L 1207 (1988) 208. Fisher, P. A., Ho, E., House, J. L., Petrich, G. S., and Kolodziejski, L. A., Mat. Res. Soc. Syrup. Proc., 340:451 (1994) 209. Coronado, C. A., Ho, E., Fisher, P. A., House, J. L., Lu, K., Petrich, G. S., and Kolodziejski, L. A., ./. Electron. Mater., 23:269 (1994) 210. Ho, E., Fisher, P. A., House, J. L., Petrich, G. S., Kolodziejski, L. A., Walker, J., and Johnson, N. M., Appl. Phys. Lett., 66:1062 (1995) 211. Tatarkiewicz, J., Witowski, A., and Johne, E., Phys. Stat. Sol., (b) 142:K101 (1987) 212. Tatarkiewicz, J., Breitschwerdt, A., and Witowski, A., Phys. Rev. B, 39:3889 (1989)

11 Diamond Deposition Ch ar a cte riz atio n

and

Donald R. Gilbert and Rajiv K. Singh

1.0

INTRODUCTION

In recent years, diamond has experienced a surge in interest from the industrial and scientific communities, based largely on the promise of developing chemical vapor deposition (CVD) processes for a wide range of previously unattainable applications. As a result, several books, numerous journal articles and many symposia proceedings have been dedicated to the examination of diamond properties, fabrication, and applications. Therefore, this chapter will not attempt to provide either a detailed history of diamond fabrication or an in-depth scientific analysis and theoretical development of diamond formation, but rather a practical overview on the properties, fabrication, and application of diamond in its role as a wide bandgap material.

2.0

PROPERTIES

Diamond is a carbon allotrope, bonded tetrahedrally in the diamondcubic lattice structure (two interpenetrating face-centered cubic lattices with a displacement of one quarter body diagonal) as the result ofsp 3 orbital 506

Diamond Deposition and Characterization 507 hybridization. Figure 1 shows the diamond cubic lattice unit cell. This bond structure, in combination with carbon's low atomic number, gives diamond the highest atom density of any material, which in turn is responsible for many of diamond's superlative properties. Table 1 provides a brief listing of physical properties for diamond, along with corresponding silicon values for comparison. The clean, unreconstructed diamond surface exhibits very high surface energy values. However, diamond is typically terminated by impurity species at its surface. This surface termination, commonly hydrogen and oxygen, acts to reduce the surface energy.

(~ CornerSite C) Face-Center Site

~i In-CellSite

Figure 1. The diamond cubic lattice unit cell.

Diamond is a semi-insulating material possessing a bandgap with a minimum value of 5.45 eV. Like silicon, it is an indirect bandgap material with the minimum of the conduction band offset from the maximum of the valence band. In its intrinsic state, diamond exhibits extremely high resistivity, with measured values of greater than 1015~')-r As a result of its wide bandgap nature, the diamond surface shows negative electron affinity (NEA) characteristics. This means that the conduction band minimum energy lies above the vacuum level, providing no potential barrier to electron emission from the diamond surface. A schematic of an ideal NEA energy band configuration is shown in Fig. 2. The exact nature of the electron affinity characteristics of diamond is sensitive to both crystallographic orientation and surface impurity termination. Diamond typically shows a hydrogen termination layer at its surface, and may also form a high degree of oxygen termination, depending on the process environment. A

508

Wide Bandgap Semiconductors

hydrogen terminated diamond surface will exhibit NEA, whereas oxygen termination reduces or eliminates N E A behavior.

T a b l e 1. Physical Properties of Diamond and Silicon

Property Lattice Constant (A) Density (g.cm-3) Atomic Mass (g.mol-l) Hardness (kg.mm-2) Young's Modulus (GPa) Sound Velocity (m/s) Poisson's Ratio Bandgap (eV) Resistivity (~ cm) Electron Mobility (cm2.V-l.s-I) Hole Mobility (cm2.V-l.s-I) Dielectric Constant Thermal Conductivity (W.cm-l.K-1) Molar Heat Capacity, cp, 298K(J'm~ Debye Temperature, ~ i) (K)

Diamond

Silicon

3.567 3.515 12.011 .~104 10.5 x 103 18000 0.07* 5.45 >lO ts

5.430 2.328 28.0855 103 1.13 x 103 7500 0.223 1.1

2200 1600 5.7 20.0 6.19 1860•

103

1500 480 11.8 1.47 19.85 650

*Isotropic aggregate value;v2m= 0.104

Optically, diamond is transmitting over a very wide spectral range. It exhibits a near-ultraviolet (bandgap) cut-off at 226.5 nm and transmits into the far infrared. Depending on the impurity characteristics of the material, there may be several near-IR absorption peaks in the spectral region of 2-12 ~tm. There is a characteristic diamond two-phonon absorption peak in the region from 1333-2666 cm -1. Diamond has an Abbe number [(n589.29-1)/(na86.13n654.28)] of 55.3 (n589.29= 2.4173, n486.13= 2.4355, and n654.28= 2.4099). The Abbe number represents a measure of the dispersion characteristics of a material over the visible spectral range. It is used in the evaluation of materials in lens systems, where it has direct bearing on the amount of chromatic aberration observed. For comparison, fused silica and BK-7

Diamond Deposition and Characterization 509 (borosilicate) glass have Abbe numbers of 67.8 and 64.2, respectively, while high refractive index plastics have Abbe numbers between 30 and 40. Higher Abbe values indicate less distortion in lens applications.

Diamond

Vacuum

oa

Conduction Band ~n .

EB

BandGap

.

.

.

.

. . . .

I I I I ! ! ! ! !

, I I I I I I I I I

Ev

.

Ev = Valence Band M a x i m u m Ec = Conduction Band M i n i m u m E . = Ec - Ev = Band Gap M i n i m u m ~v = V a c u u m Level X= ~

- Ec = Electron Affinity

Figure 2. Schematic of an ideal negative electron affinity energy band configuration.

3.0

FABRICATION

3.1

High Temperature, High Pressure

The desire to manufacture diamond has existed since it was first determined to be a form of carbon around the turn of the 19th century. Repeatable high pressure, high temperature (HPHT) production of diamond was first demonstrated in 1955 by researchers at General Electric.[ 1] High pressure techniques have been used for the production of synthetic diamonds and diamond grit ever since. Formation of diamond at standard thermodynamic equilibrium occurs at pressures greater than 100,000 atmospheres and temperatures of 1000~ or more. Molten metal catalysts are generally used to reduce these required temperatures and pressures. Diamond

510

Wide Bandgap Semiconductors

grit so produced is commonly used in grinding, polishing, and cutting applications. These applications typically involve incorporation of the diamond particles in a metallic binder to produce the desired shape. Such materials are called polycrystalline diamond or PCD.

3.2

Chemical Vapor Deposition

Chemical vapor deposition of diamond represents a radical departure from the standard HPHT process. Diamond CVD covers a broad range of possible system pressures, from the millitorr range up to atmospheric pressure, and substrate temperatures, reportedly as low as 200~ to greater than 1000~ 31 However, there are certain factors that are common to all diamond CVD processes. First and foremost is the requirement for a gaseous carbon source, typically in the form of a monomer such as methane (CH4), acetylene (C2H2), carbon monoxide (CO), or methyl alcohol (CH3OH). Also critical to diamond CVD is the presence of a diluent gas, usually hydrogen. There is also the necessity of an energetic activation source, such as a hot filament or an electric field, to dissociate some of the hydrogen and to produce appropriate growth species, typically the methyl radical (CH3). Finally, some material must be present upon which diamond may form under the given process conditions. Substrate temperature primarily affects the structural quality of the deposited material and the rate of deposition. If a ternary compositional diagram of the process gas is drawn consisting of carbon, oxygen, and hydrogen (the primary elemental components in most diamond CVD systems), the possible atomic fraction compositions for depositing diamond fall in a roughly triangular region about the CO line. This was first determined in a survey conducted of the reported compositions of successful diamond formation from a wide range of researchers, and became known as the Bachmann triangle.[ 31 Figure 3 shows a schematic approximation of this region. Vital to the interpretation of this figure is the understanding that gas activation must be sufficient to neglect the original bonding character of the source gases. Source monomers must be sufficiently dissociated to form active growth species, some amount of atomic hydrogen must be produced, and, if present, molecular oxygen must dissociate and react sufficiently with the other species to play a role in the growth process effectively. Implicit in this statement is the effect of gas pressure on the activation process: changes in gaseous diffusion lengths, interaction cross sections and transient species lifetimes

Diamond Deposition and Characterization 511 will accordingly affect the outcome of the deposition process. Also, the relative structural quality of the resulting deposition, in terms of defects and non-diamond inclusions, may vary greatly within this compositional region of "possible" diamond deposition.HI

C

No Growth

H

0

O/(O+H) Figure 3. Schematic approximation of the region of viable diamond deposition gas compositions in the C/H/O system.

A great deal of effort has been put into developing theories and models for the mechanisms by which diamond CVD o c c u r s . [3][5]-[9] Diamond deposition can proceed from many different monomers. The methyl radical, CH3, is the most efficient growth monomer for most systems.[ 1~ Hydrogen plays several important roles in the deposition process.[ 12]-[13]As carbon is deposited on the pre-existing material, the diamond crystal structure is stabilized through hydrogen termination of the surface. Graphitic

512

Wide Bandgap Semiconductors

phase or highly defective sp 3 material is etched by atomic hydrogen at a higher rate than diamond phase carbon. Abstraction of surface terminating hydrogen occurs through interaction with impinging atomic hydrogen, which subsequently leaves active sites available for carbon addition. The role of oxygen in diamond deposition is similar in many ways to that of hydrogen.[ 14] Oxygen is a more efficient carbon etchant than hydrogen, and also has a higher etch rate for the graphitic phase than for diamond. Like hydrogen, oxygen may adsorb onto and partially terminate the diamond surface. Another role of oxygen in the growth process is to combine with gas phase carbon as carbon monoxide, which is extremely stable. This may effectively remove "excess" hydrocarbons from the deposition reaction.[ 15] Nucleation and Growth. Two critical facets that must be considered in diamond CVD, as with many thin-film CVD processes, are the processes of nucleation and growth. Both aspects play critical roles in the resulting characteristics of the deposited film. The density of nuclei, which is primarily determined by nucleation rate, has significant influence on initial lateral grain size and can substantially affect as-grown surface roughness. Growth rate, particularly as it pertains to crystallographic direction, can also play a critical role in determining resulting film properties such as surface roughness and defect density.[ 4] Both nucleation rates and growth rates are influenced by temperature, gas chemistry, and gas activation. Nucleation may be particularly sensitive to the choice of substrate material, as well.[16]-[ 19] Regarding diamond CVD growth rates, higher gas phase temperatures generally correlate with faster deposition rates. This is observed in the high deposition rates of torch CVD systems, to be discussed later. In plasma-based systems, higher growth rates are observed at higher plasma densities. DC arc jets have achieved some of the highest diamond deposition rates to date. Nucleation Enhancement. Homogeneous nucleation of diamond on untreated, non-diamond materials is an energetically unfavorable process. Initial nucleation on untreated surfaces tends to occur at defect sites and can result in discontinuous coverage and non-uniform film characteristics. This process can be enhanced by abrading the substrate surface with diamond grit, usually by means of exposure to a diamond colloid (particle sizes from 0.25-100 ~tm, typically) in an ultrasonic bath.[2~ 21] This provides two possible enhancement mechanisms: first, the surface is thoroughly scratched, providing a high density of surface defect sites; second, microcleavage of the impinging particles leaves nanometer sized diamond

Diamond Deposition and Characterization 513 remnants imbedded in the surface that may act as nucleation "seeds" for diamond growth.[ 22] Such abrasion has been shown to provide nucleation densities as high as 101~ cm -2 in a microwave plasma CVD system. Observed nucleation densities following diamond abrasion are inherently influenced by the deposition process conditions.[21][ 23] Low temperature deposition will show inherently lower nucleation rates than deposition at higher substrate temperatures. Nucleation rate may also change as a function of surface coverage by previously formed diamond nuclei during the early stages of CVD.[ TM One drawback of the abrasion method for enhancing diamond nucleation is that the substrate surface is being damaged, which is undesirable for applications such as optical coatings. A common method of enhancing diamond nucleation in such cases is through the controlled seeding of the growth surface with diamond particles.[25H28] In fact, this functionally avoids the nucleation process because there is no real barrier to growth of the seed particles once the deposition environment is established. Particle seeding is also useful for low temperature deposition processes where abrasion would be of limited benefit.[29]Diamond particles may be delivered in a colloid to the substrate surface by several means, depending on the suspending medium and the substrate surface characteristics. Various delivery methods for colloid-based seeding include spraying, spin-coating, brush painting, direct writing, and dipping.j26]-[2s] More reliable seed coatings with better particle adhesion characteristics can be obtained using an intermediate monolayer coating of polymer that may provide an electrostatically favorable surface relative to the colloidal particle surface charge.[27]Eleetrophoretic deposition, which involves the application of a bias potential to the substrate of opposite polarity to the intrinsic surface charge of the colloidal diamond particles, has also been used to seed conducting and semieonducting substmtes.[25][3~ If seed layers can be imparted at predetermined areal densities, they can be used to control the microstructural characteristics of the diamond film, primarily in regard to lateral grain size upon film coalescence. The use of nanometer-sized particles can allow very high nucleation densities (< 1011 cm -2) to be achieved.[2s][29] Such high nucleation densities allow the formation of very thin (<0.1 ~tm) continuous films with concomitantly low surface roughness.j26][2s] Also, if seeding can be controlled lithographically, where specific pattems can either be covered or be leR bare at the micron scale, then microelectronic and micromechanical applications are facilitated.j25][311 Figure 4 shows examples of lithographically defined features formed using a controlled seeding process and subsequent deposition.

514

Wide Bandgap Semiconductors

Figure 4. Micrographshowinglithographicallydefinedregions of diamond deposition. Perhaps one of the most encouraging developments in diamond CVD is the bias-enhanced nucleation (BEN) process.[321-[Sl In this procedure, a negative bias voltage is applied to the substrate during the nucleation stage of deposition. Resulting bombardment of ionic species causes the formation of a thin carbide layer (epitaxial f3-SiC for silicon under the proper conditions) on the substrate surface.[331-[351 In addition to carbide formation, biasing also enhances the diamond nucleation rate, in part through the rapid formation of amorphous carbon on the substrate surface, thus facilitating continuous film coverage. Texturing and Growth Direction. If proper process conditions are used, the resultant nuclei that form during bias-enhanced nucleation on single crystal substrates are highly oriented with respect to the substrate and hence to each other, thus forming a near-epitaxial film, typically referred to as highly oriented diamond (HOD). [35][36]These films have been characterized by low angle grain boundaries with misorientations of--4 ~ Electrical and thermal characteristics of such films can be expected to be substantially

Diamond Deposition and Characterization 515 improved over those of otherwise comparable (i.e., grain size, film thickness, impurity/defect structure, etc.) non-oriented polycrystalline films.[ 38] Alternatively, HOD films have also been produced without the use of electrical biasing. Imparting diamond particles onto nickel substrates with subsequent heat treating can produce highly oriented nuclei.[ 39] Successive deposition leads to HOD formation, with corresponding low angle grain boundaries and improved thermal and electrical characteristics. Another method of producing preferentially oriented nuclei has been demonstrated through electrophoretic seeding.[ 3~ Using a colloid of 5 pm average sized diamond particles, the electrophoretic process can produce a seed layer with a statistically preferred (>80%) orientation toward the (111) direction. This phenomenon has been observed for seeding at or below one monolayer coverage of the substrate surface. This leads to another important topic of interest, namely texture formation in typical polycrystalline diamond films. As mentioned above, the BEN process can produce HOD films that form in a nearly epitaxial (highly textured) state. However, any polycrystalline diamond film exhibiting columnar grain growth will develop threading texture characteristics under constant growth conditions if deposited to sufficiently large thicknesses. This is the result of the competition between grains having different crystallographic orientation with respect to the growth surface, based on the direction of fastest growth for the deposited diamond. Van der Drift first described this process, known as evolutionary selection, in which grains whose fastest growth direction is best aligned with the film surface-normal overgrow adjacent grains that are not so aligned.[4~ As processing conditions of substrate temperature and gas composition are varied, different crystallographic directions will exhibit higher growth rates. Typically, this is characterized by a term known as the growth parameter (c~) which is proportional to the ratio of the growth velocity in the (100) direction to the growth velocity in the (111) direction. [41]-[43] As a result, the growth parameter for regular crystals covers a range 1 < c~ < 3, with the former value signifying a cubic grain structure and the latter an octahedral shape. Figure 5 shows several examples of ideal regular crystal forms in the evolution from cubic to octahedral crystal forms along with corresponding approximate growth parameter values. Substrate Materials. Thus far, we have discussed fundamental issues of nucleation and growth without much consideration of the substrate material. However, this is another prime concern in diamond CVD. While diamond may be formed on a fairly wide range of materials, it will not necessarily remain intact upon removal from the growth system. Due to

516

Wide Bandgap Semiconductors

diamond's low thermal expansion coefficient and high modulus, as well as the rigors of the deposition environment, the range of suitable substrate materials is somewhat limited. Film adhesion is a critical consideration in diamond growth. Poor matching of thermal expansion characteristics over the process temperature range can result in excessively large thermallyinduced residual stresses that can cause structural failure. Formation of a thin carbide layer at the substrate surface can greatly aid in film adhesion strength via chemical bonding. Also, adhesion may be improved through controlled roughening of the substrate surface for applications that will allow such an interface.[ 441 Microroughened surfaces provide increased interfacial area and may provide an element of mechanical interlocking.

a=3

a=2

a=1.2

a=1.5

a=l

Figure 5. Ideal regular crystal forms and their approximate growth parameter values.

The majority of work done on CVD diamond has involved polycrystalline films deposited on non-diamond substrates. This is primarily due to the lack of suitable, readily available large-area substrates for epitaxial

Diamond Deposition and Characterization 517 growth. Available diamond single crystals, both natural and synthetic, are relatively small and of variable quality, as well as being relatively expensive. Cubic boron nitride (c-BN), which is the most similar material to diamond and hence the most likely candidate material for heteroepitaxial diamond deposition, also shares diamond's limitations of problematic fabrication and is not available in large single crystals. Other materials that may provide reasonable low-index-plane lattice matching, such as copper or nickel, have other materials problems of large thermal-expansion mismatch, high carbon solubility/diffusivity and lack of stable carbide formation. As a result, epitaxial growth investigation has been more limited in scope than polycrystalline growth. Candidate materials for diamond deposition must withstand the deposition environment with minimal effects. This implies insensitivity to atomic hydrogen exposure at growth temperature, phase/structure stability at temperatures of at least 500~ (more typically up to -~700-800~ or higher), and low vapor pressure at growth temperature. Materials such as ZnS and GaAs will generally not meet these requirements without some kind of encapsulating interlayer. As well as standing up to the growth environment, the substrate should not exhibit too high a solubility/diffusivity for carbon, as happens with iron. This can prevent effective nucleation due to the rapid absorption of diamond nuclei. Such materials will require a lowdiffusivity interlayer for effective film formation. All of the above considerations have resulted in silicon being the most widely used substrate in diamond CVD research (i.e., low thermal expansion mismatch, stable carbide formation, low vapor pressure, etc.). Refractory metals such as tungsten and molybdenum have also been used extensively as substrates, although they are less ideal materials due to their higher thermal expansion values and carbon diffusivity. Deposition Systems. As previously mentioned, there are many systems that can be used for the deposition of diamond. These systems cover a wide range of process pressures and temperatures, with corresponding variations in deposition rates and material qualities. Hat-Filament. The earliest work in diamond chemical vapor deposition was done using a thermal gas-activation process. [45][46] This basic type of system, utilizing a hot filament (HF-CVD), has been extensively used and explored over the years, establishing it as the most fundamental of all diamond CVD systems. A refractory metal filament (such as W or Ta) is resistively heated to temperatures of 2000-2200~ sufficient to thermally dissociate hydrogen, in a vacuum chamber at typical operating pressures of

518

Wide Bandgap Semiconductors

20-100 torr. Substrates are typically placed within two centimeters of the filament, where they may be independently heated/cooled or they may be heated by the filament. Growth rates may run up to several microns per hour, depending on the conditions used. While the potential deposition area for an individual element is limited to a few centimeters, properly configured arrays of filaments may allow deposition over very large areas. Carbon incorporation, or carburization, of the filament under these conditions is inevitable. In tungsten, this carburization results in significant embrittlement of the filament, requiring care in handling to maximize the lifetime of the unit. Rhenium filaments experience less embrittlement than tungsten, but do exhibit swelling because of carbon incorporation. Also, there is a significant change in the resistivity of the filament during the carburization process. Common practice when a new filament is installed in a reactor is to allow the filament to fully carburize before attempting film deposition, in order to maintain consistent conditions. This may be determined through simple observation of the change in resistance of the filament over time, where "saturation" corresponds to steady state resistance. The time required for full carburization depends on the radius of the filament and the temperature of operation. HF-CVD diamond typically exhibits metal impurities originating from the filament. Oxygen addition to the process gas in the hot-filament system can cause degradation of the filament due to oxidation. The hot-filament CVD system has the advantage of simple basic design and concomitantly low cost. It requires only a single stage pumping system and modest chamber materials requirements. A quartz enclosure (either a belljar or a capped cylinder) is common for small systems, facilitating visual process observation. Filament temperature is commonly monitored with an optical pyrometer. A diagram of a simple hot-filament system is shown in Fig. 6. Plasma Enhancement. Plasma-enhanced (PE-CVD) deposition of diamond has been demonstrated using a wide range of excitation sources, including DC, RF, and microwave.J3][47]-[6~ These systems provide various possible advantages over the thermal deposition process, from lower possible deposition temperatures and reduced metal impurity incorporation (especially for electrodeless systems) to greatly enhanced growth rates and novel gas chemistries. As previously mentioned, diamond deposition rates for plasma-based systems generally correlate directly with the plasma density. DC plasma jet/ torch systems have reported high deposition rates approaching 1 mm/hr, but

Diamond Deposition and Characterization 519 are more commonly operated at growth rate conditions of a few microns per hour.[ 49] However, DC plasma torch systems generally offer relatively small deposition areas with large growth non-uniformity.

Backside Mounted Thermocouple ~.

Support~

I II

Substrate .,

~

r~

I

I

F/J

~

Filament

Electrical Feedthrough Vacuum Flange Quartz Bell Jar

Vacuum Pump H2+CH 4 Gas

Figure 6. Schematic diagram of a hot-filament deposition system.

Microwave plasma based systems may be the most common type of diamond deposition system next to the hot-filament type. [53]-[55] Systems designed specifically for diamond deposition are commercially available and typically operate over the same pressure range as hot filament systems. Deposition rates in these systems are typically several microns per hour using common hydrogen-based compositions. Work has also been

520

Wide Bandgap Semiconductors

done using microwave PE-CVD with an argon ambient and fullerenes (C60) as the source chemistry for diamond deposition.[ 561This compositional system exploits the role of the carbon dimer (C2), which is formed through fragmentation of the C60 structure, as a diamond growth precursor, in contrast to the methyl radical mentioned previously. By avoiding the presence of large amounts of atomic hydrogen in the system, it may be possible to expand the list of viable substrate materials as well as reducing the amount of incorporated hydrogen in the deposited material. Electron cyclotron resonance (ECR) enhanced microwave plasma systems operating at pressures of less than 1.0 torr have reportedly deposited diamond at substrate temperatures approaching 200~ or less.[ 2][5711581ECR systems use magnetic fields to enhance the interaction between the oscillating electric fields of microwaves and free electrons in the plasma region. Free electrons follow helical trajectories in the magnetic field with a characteristic cyclotron frequency determined by the strength of the field. Matching this cyclotron frequency to the microwave frequency enables resonant coupling to occur, providing efficient energy transfer and producing very high "temperature" (>10000 K possible) electrons that interact with atomic species through impact ionization and dissociation to maintain the plasma. Such enhancement is most effective at pressures where the electron mean free path allows several cycles to occur between collisions with other species (generally <10 millitorr), but will provide measurable improvement to the plasma generation process at pressures approaching 10 torr. [59][60] A schematic illustration of an ECR plasmaenhanced CVD system is shown in Fig. 7. Torch. Possibly the simplest of all diamond deposition systems is the oxy-acetylene torch. This system operates at atmospheric pressure with very high growth rates (up to hundreds of microns per hour). Samples must be cooled during deposition to prevent substrate and film degradation. However, despite the high deposition rate, the torch system is very inefficient, consuming large amounts of fuel during deposition and having a relatively small effective deposition area. Also, the quality of the deposited material tends to be low at such high growth rates, with significant graphitic/non-diamond inclusion. Other flame-based systems have been developed to improve efficiency and increase the deposition area and diamond quality. These systems incorporate novel nozzle designs for more tmiform, larger area flames and may operate at sub-atmospheric pressures. The price paid for improved quality is reduced deposition rates, as in essentially all systems discussed here.

Diamond Deposition and Characterization 521

Plasma~ S a m p l e

MagnetsA

~

Stage

Substrate

Top viewof ECRmoduleshowingmagnetring configuration.

2.45GHzMicrowaves MethanolVapor

QuartzWindow

'it

Fe-Nd-B Permanent Magnets

CoolantLines Gas Diffuser Ring HeatedSample Stage AccessPort Adjustable Mounting Post

PumpSystem. and Exhaust Hydrogen

Figure 7. Schematicdiagramof an electron cyclotronresonance enhancedplasma deposition system.

4.0

MODIFICATION

Depending on the intended application, the as-grown diamond material may be suitable for use immediately upon removal from the deposition environment. However, many applications require modification of the asdeposited film. This modification will generally take the form of structural and/or electrical alteration.

522

Wide Bandgap Semiconductors

4.1

Structural Modification

One of the most fundamental of all possible film modifications is polishing for the reduction of surface roughness and etching for the production of specific structures. Polycrystalline diamond films tend to produce surfaces with high roughness that is unsuitable for many applications. Because of diamond's extreme hardness, conventional mechanical polishing can be very slow. Standard chemical etching of diamond is also problematic due to diamond's relatively high chemical inertness. Techniques have been developed which specifically address the etching/polishing of diamond. Diffusive Polishing. Carbon's high diffusivity and solubility in iron has been exploited in the polishing of diamond for many years. Contact of a rough diamond surface to a polished piece of iron at elevated temperatures (>~ 1000~ in an oxygen free atmosphere (to avoid excess graphitization and etching of the diamond) results in reduced roughness through surface graphitization and carbon absorption by the iron.[ 61] Diamond polishing is commonly accomplished using diamond grit on an iron grinding wheel. The application of an iron wheel and olive oil with diamond grit can produce a very good polished diamond surface, suitable for surface science studies.[ 62] Molten rare earth metals have also been used for the rapid etching and thinning of diamond films through diffusive removal of graphitized surface carbon.j63][64] Figure 8 shows diamond film surfaces before and after prolonged contact to a hot (1200~ steel plate. Beam Planarization. Ion beam sources have been investigated for the polishing and etching of diamond as well. [65]-[69]Again, strong bonding and high atomic density reduces the effectiveness of the ionic sputtering process for inert species such as argon. Fastest results can be expected for species including fluorine or oxygen ions, due to the reactive formation of volatile carbon compounds. Also, oxygen enhances the process of surface graphitization at elevated temperatures. Graphitic phase carbon is more rapidly etched than diamond. Through lithographic masking techniques, ion beams and low pressure plasmas can be used to produce microelectronic and micromechanical structures. Etching and planarization using lasers has been another area of interest.[7~ 72] Despite diamond's wide spectral transparency, lasers can effectively be used both to etch away desired pattems and to reduce surface roughness. The optical properties of the diamond film affect the characteristics of the laser etching process. Crystal size, impurities and film quality (i.e., crystalline perfection) play a major role in the etch characteristics. Surface polishing generally requires a glancing incidence angle for the

Diamond Deposition and Characterization

523

laser relative to the substrate surface. Also important to the laser polishing process is the ambient atmosphere in which it is performed. Again, the presence of oxygen will enhance graphitization at the surface and subsequent removal of carbon through oxide formation. Graphitization of the surface also increases the effective absorption of the laser radiation at the surface. Figure 9 shows resultant polycrystalline diamond surfaces after processing with laser or ion beams.

(b)

-,--,-,-----,-

lOpm

Figure 8. Micrographs of a CVD diamond film surface (a) before and (b) after prolonged contact to a 1200~ steel plate.

524

Wide Bandgap Semiconductors

.:11 Figure 9. Micrographs of (a) ion beam and (b) laser polished polycrystalline films.

Marchywka Effect. It has been discovered that highly defective, amorphous carbon inclusion in diamond is subject to etching when placed in a deionized water electrochemical,cell between platinum electrodes. [73][74]

Diamond Deposition and Characterization 525 This process is known as the Marchywka effect and may be exploited for reduction of surface roughness or for the removal of thin diamond membranes from a thick diamond source. Ifa diamond film is sufficiently implanted with carbon ions such that an amorphous layer is formed, the Marchywka effect can be used to remove the layer, leaving a low roughness surface.

4.2

Electrical Modification

Diamond's promise as a wide bandgap electronic material necessarily requires the tailoring of its electronic properties through the use of impurity doping. Boron, which is a substitutional impurity in diamond, has shown the most effective doping characteristics, producing p-type material. It is frequently incorporated during growth through the addition of a gas such as diborane (B2H6) to the deposition ambient.[75]Boron may also be incorporated through ion implantation and annealing.[76]A notable characteristic of boron doped diamond is that at room temperature the dopant impurities are only partially ionized due to a relatively large activation energy (--0.37 eV). [77]This may significantly reduce the effectiveness of the p-type material unless it is used at high temperatures. Production of n-type diamond has proven more problematic than p-type material.[78] Although impurities such as nitrogen, lithium, sodium, and phosphorus may be Used to produce n-type behavior in diamond, this material has not shown standard extrinsic behavior. Nitrogen, which may also be incorporated during CVD, forms a deep donor level (--1.7 eV) in diamond. Lithium and sodium are interstitial dopants in diamond. Excessive diffusivity of lithium in diamond may prevent its effective use. Also, sodium, lithium, and phosphorus are not readily incorporated into diamond during deposition. This indicates the need for ion implantation. However, due to its metastable nature, diamond is not a good candidate for ion implantation because of the necessary post-implant annealing required to repair damage and activate dopants.

5.0

CHARACTERIZATION

Characterization of diamond can be complicated by the very properties that make it remarkable. As a carbon allotrope, accurate determination of the relative amounts of diamond and non-diamond phases that might be

526

Wide Bandgap Semiconductors

present in deposited material is non-trivial. Because of its low atomic mass, diamond has a relatively low scattering cross section for x-ray diffraction analysis. High resistivity can cause difficulties in electron beam-based analyses due to charge build-up effects. Extreme hardness and high modulus make mechanical testing difficult to implement effectively. For any material, the choice of characterization method is dictated by the information being sought. Material characteristics such as conductivity, atomic mass, optical transparency, and others will prescribe which specific techniques will be appropriate for the desired analysis. Table 2 lists some more common analytical techniques and their relevance to diamond characterization.

Table 2. Common Material Characterization Techniques Technique

Properties Observed

Comments

Scanning Electron Microscopy, Atomic Force Microscopy

surface morphology, thickness/roughness, growth parameter

smooth surfaces desirable for most applications

Raman spectroscopy

phase purity, crystallinity, stress

sensitive to film optical properties, quantitative observationsdifficult

X-ray diffraction

texturing, crystallinity, stress

small scattering cross section

Photoluminescence, Cathodoluminescence

impurities and defects

FTIR

impurities, IR optical transmittance

Secondary Ion Mass Spectroscopy, Auger electron spectroscopy

impurities, depth profiling

Transmission Electron Microscopy

morphology, defects, crystal structure

sensitive to surface scattering effects

tedious sample preparation, novel methods

Diamond Deposition and Characterization 527 5.1

Microscopy

Surface specific microscopies, such as scanning electron microscopy (SEM) and atomic force microscopy (AFM), provide very basic information regarding resultant film morphology and roughness. SEM can be complicated by the highly insulating nature of the diamond (especially for thick films) that can cause sample charging and image distortion. When this is encountered, a thin conductive film must be deposited on the diamond surface to allow undistorted observation. AFM has the advantage of being insensitive to the electrical characteristics of the film, as well as providing directly quantitative data regarding surface roughness. However, thick films with very high surface roughnesses may exceed the capabilities of the AFM. Observation of film morphology provides information about the growth characteristics of the deposited diamond.[ 79]Excessive twinning and renucleation produce films with an irregular "cauliflower" morphology, lacking clear faceting. Individual crystal shape provides evidence of which crystallographic direction (i.e., growth parameter) has the fastest growth rate for the deposition conditions being used. Figure 10 shows examples of films exhibiting regular faceting and irregular morphology.

Figure 10. Micrographsof two CVD filmsshowingregularfacetedcrystallinefeaturesand nodular"cauliflower"morphologies.

528

Wide Bandgap Semiconductors

Transmission electron microscopy (TEM) is a very powerful tool in the examination of material microstructure. However, TEM requires significant sample preparation to provide samples that are thin enough for electron transmission. Diamond is non-ideal for such preparation because of its extreme hardness, typically requiting much more time and effort on the part of the analyst. One trick that has been used is the direct deposition of diamond onto high temperature TEM sample grids (thus avoiding the need to remove the diamond from a substrate or thinning of the substrate in preparation for analysis).[ 8~ TEM analysis has been used to elucidate the role oftwinning in the morphological development of diamond thin films, as well as examining the nature of the as-deposited CVD diamond surface.[8~ 85]

5.2

Spectroscopy Raman. Raman spectroscopy is a popular and powerful tool in the

analysis of diamond films. [86]-[89]It provides a relatively simple means of determining the presence of the diamond phase in a carbon film. Analysis is nondestructive and, for standard observation, there is essentially no sample preparation required. Inelastic Raman scattering of probe illumination (typically the 514.5 nm line of an argon ion laser) produces a characteristic peak in the scattered spectrum at a shift of 1332 cm -1. Polycrystalline graphitic (sp 2 bonded) carbon produces broad Raman spectral peaks around 1580 (G-band [graphitic]) and 1360 (D-band [disordered]) cm -1 and has a Raman scattering efficiency approximately 50 times greater than diamond, making it a sensitive probe of film phase purity.[ 86] However, quantitative analysis of composition via Raman is difficult due to variations in optical properties in films. The relative sensitivity of Raman scattering to sp 2 bonded carbon is dependent on the wavelength of the probe illumination. Longer wavelength radiation is more sensitive to graphitic constituents, while shorter wavelength illumination becomes relatively insensitive to graphitic content.[ 87] Highly defective material with small crystalline domains causes broadening of the diamond characteristic peak. Table 3 lists some common Raman signatures and their peak shift positions in the scattered spectrum. Figure 11 shows representative spectra of a singlecrystal diamond chip and a freestanding polycrystalline film for comparison. Crystalline defects and intrinsic stress in the polycrystalline film cause both broadening and shifting of the diamond peak relative to that of the single crystal diamond.

Diamond Deposition and Characterization 529 Table 3. Characteristic Carbon Raman Spectral Features Peak Position

Material

(cm -1)

a-Diamond (sp 3 bonded a-C)

1140

Diamond (sp a bonded C)

1332

Graphite (sp 2 bonded C)

-1360 (D-band), 1580 (G-band)

Polyacetylene (a-C:H) Broad a-C band

,-1450 - 1200-1600

C60

1468

'

'

'

I

'

'

'

I

I

. . . .

'

'1

'

'

~

wA ,--K

'

I

'"

i

'"'

!''

'

""

i

f) i

1300

1350

1400 1450 1500 R a m a n S h i f t ( c m "1)

1550

1600

Figure 1 l. Raman spectra comparing the characteristic diamond signal from a single crystal and a polycrystalline film.

Developments in spectroscopic design and technology have made Raman a fast analysis technique, at least in its basic form. Use of chargecoupled device (CCD) arrays and multicharmel analyzers with fixedgeometry diffraction gratings (instead of scintillation detectors used with scanned gratings) allow concurrent broadband data acquisition. This can greatly reduce the necessary data acquisition time from mechanical scans,

530

Wide Bandgap Semiconductors

which may require from 10-60 minutes (depending on desired scan range and sample properties), to spectral array observations of 10-20 seconds. Raman analysis consists of two primary configurations: standard Raman spectroscopy, where the probing laser is focused using standard optics to spot sizes on the order of a millimeter; and micro-Raman, where an optical microscope is used to produce spot sizes down to the micron scale. Micro-Raman analysis, which utilizes a back scattering geometry, accommodates very localized analysis for the examination of individual grains or grain boundary regions in polycrystalline samples, and also allowing depth profiling in thick films. Standard Raman, which is typically done with a low-angle incident-beam geometry for film samples, provides bulk average scattering properties for the illuminated area. Along with its value in structural and compositional characterization, Raman may be used for analysis of residual stress in diamond.J9~ 91] Strain in the crystalline lattice modifies the elastic constants acting between the carbon atoms, causing a shift in the characteristic diamond scattering peak from its unstressed position. Compressive stress results in shifts to higher peak values, while tensile stress produces lower peak values. The magnitude of the stress can be estimated from the amount of the shift from the unstressed position. If the stress is sufficiently large and non-hydrostatic (e.g., biaxial), a resolvable splitting of the previously degenerate phonon modes will occur, producing multiple characteristic peaks: one for the nondegenerate mode and one for the remaining degenerate modes. Other specialized variants of Raman analysis have been used in diamond analysis. Surface enhanced Raman spectroscopy (SERS), wherein a thin (-~5-15 nm) metal coating (e.g., silver) is deposited on the film surface to enhance Raman scattering, has been used in the examination of diamond films to better investigate non-diamond inclusions, such as polyacetylene. [92][93 ] Polarized micro-Raman inve stigati on 0 fpolycrystalline diamond, in which individual crystals have been investigated, has shown significant dependance of measured film purity on incident laser polarization.[94]This effect arises as a result of the difference in polarization characteristics in light scattering from diamond phase components (having strong polarization) versus non-diamond components (largely depolarized). XRD. X-ray diffraction is commonly used for the determination of film texturing and possibly for stress analysis in diamond films.[ 95]-[98] Standard diffractometry can provide basic indications of texturing based on relative diffraction peak intensities as compared with a standard powder diffraction spectrum. XRD of sufficiently thick films can be used to

Diamond Deposition and Characterization 531 determine the growth parameter of the film, based on the developed threading texture.J95]Using the sin2tg method, as described by Cullity, [99]the full stress tensor ofpolycrystaUine diamond films can be determined. Luminescence. Luminescence spectroscopies [both photoluminescence (PL) and r (CL)] are valuable characterization techniques that provide evidence of optically active (light emitting) defects such as vacancies and impurities in deposited films.[89][l~176176Photoluminescence studies of diamond films commonly exhibit a broadband feature centered about 2 eV, which has been attributed to sp 2 bonded amorphous carbon.[ 89]Careful analysis of PL line-shapes can also provide information on stress sources in diamond.[ 89] Nitrogen and hydrogen are primary impurities involved in optically active defect structures in diamond. Silicon impurities in diamond are also readily detectable using both PL and CL. Cathodoluminescenee is particularly well suited for the optical characterization of diamond because of diamond's wide bandgap, which requires a relatively high-energy excitation source. FTIR. Fourier transform infrared spectroscopy (FTIR) is frequently used for the infrared spectral examination of diamond due to diamond's attractiveness as an IR optical material.[1~ TM]FTIR has also been used in the examination of hydrogen inclusion in deposited diamond films.[l~ 1~ The presence of bound hydrogen in the material produces a characteristic absorption peak in the region 2750-3300 cm -1 in the resulting spectrum. Because this peak is proportional to the amount of bound hydrogen, it may be used to estimate the relative amount of hydrogen incorporated in the diamond. This technique will not provide information on non-bonded hydrogen that may be incorporated, however. Incorporated bonded oxygen may also be detected using FTIR. SIMS/Auger. Quantitative evaluation of incorporated impurities may be most readily accomplished using secondary ion mass spectroscopy (SIMS). As is standard for SIMS, depth profiling may be done to look at variations in film composition and also to observe the diffusive characteristics of the substrate (i.e., how readily carbon diffuses into the substrate surface). Investigation of species other than oxygen allows the use of an oxygen ion primary beam, which can improve sputter rate for depth profiling. Auger electron spectroscopy (AES) can be similarly useful, although diamond's dielectric character can once again cause difficulties in the analysis of thick samples. NMR. Nuclear magnetic resonance (NMR) is another possible analysis tool in the investigation of diamond. NMR may be used for the

532

Wide Bandgap Semiconductors

determination of incorporated hydrogen in diamond.[ 1~ This technique has the advantage of detecting both bonded and non-bonded hydrogen, thus giving a more thorough determination of incorporation.

6.0

APPLICATIONS

The potential for application of diamond in engineering and industry is extremely broad. Unique and extreme physical properties offer the possibility of enhancement to products as common as drill bits and machining inserts to microelectronic sensors and fiat-panel displays. Here we will conclude by mentioning just a few applications that specifically take advantage of diamond as a wide bandgap material.

6.1

Thermal Management

At the current state of technology, diamond's thermal and electrical properties may most effectively be exploited in passive microelectronic applications. It has great potential as a heat spreading material due to its high thermal conductivity.[ 11~When combined with its high dielectric strength, low dielectric constant, and chemical inertness, diamond is seen as an attractive packaging material for microelectronic devices. Devices such as laser diodes and high-power field effect transistors (FET's) may significantly benefit from the use of diamond for thermal management, improving both performance and reliability. Heat spreaders for such applications are currently commercially available. Thermal conductivity in polycrystalline films is significantly influenced by microstructure. Figure 12 shows a plot of thermal conductivity values in a thick film for heat flow parallel and perpendicular to the substrate surface as a function of distance along the growth direction.[1111 Included in this figure is an illustrative schematic representation of the variation in grain structure with distance from the substrate surface. Conductivity shows improvement both parallel and perpendicular to the substrate surface as the grain size increases with thickness, indicating the disruptive effects of the grain boundary structure.

Diamond Deposition and Characterization 533

G R A I N S I Z E (t=m) 30

'

'

'

10 " ~1

''

'

' '

'"'

'"

20 I

,

" "r'

10

0 0

,l

,,,J.

100

200

DISTANCE

1 300

FROM BOTTOM

. . . .

400

(vm)

K .1.IOOal

KII I~ Figure 12. Plot of thermal conductivity values for heat flow parallel and perpendicular to substrate surface as a function of distance along the growth direction. A schematic representation showing the variation in the columnar grain structure with depth is included for illustration.[1 ~~]

6.2

Optics

The high cost of fabrication still prevents diamond from being a commercially viable material for most free standing optical devices. However, diamond has some promise as a wear resistant coating for softer optical materials. High quality (so-called "white") polycrystalline films

534

Wide Bandgap Semiconductors

provide sufficient transmittance for many optical applications extending into the ultraviolet spectral range. Diamond is especially attractive for infrared optical uses because of its broad spectral transmittance at long wavelengths. However, surface roughness and grain boundaries can adversely affect transmittance due to scattering.[Z7][1~ In its role as a protective optical coating, the interfaeial bonding characteristics of diamond are critical. Excessive thermal residual stresses resulting in delamination or cracking of the film may render a directly deposited diamond coating ineffective. This may be overcome by depositing the diamond film on a dummy substrate and subsequently removing the coating and bonding it onto the optical piece.

6.3

Electron Emission

As a result of its negative electron affinity characteristics, diamond is an attractive candidate material for cold-cathode electron emission applications, including emitter arrays used for flat panel displays.[ 112]-[116]Because of a lack of available charge carriers (i.e., electrons) in diamond (specifically due to the lack of effective n-type doping), there has been more focus on the use of diamond to functionally modify the effective work function of other materials such as silicon or molybdenum.[ 117][1181These other materials may readily be formed into emitter arrays consisting of sharp tips upon which diamond may be deposited by various means, including CVD and dieleetrophoresis. Emission may then occur through tunneling of carriers from the host material into the diamond conduction band and subsequent transport to the diamond surface under the impetus of the emitter bias voltage. The mierostrueture of the deposited diamond plays a critical role in its effectiveness for field emission, as has been shown in studies involving nanoerystalline films deposited using argon/fullerene growth chemistry.[ 119]

6.4

Microelectronics

Finally, if the problems of dopant control and defect density can be overcome, diamond may be used for enhanced and specialized mieroeleetronir applications. Theoretically, diamond has very high figures of merit for electronics as a result of its extreme physical properties. In particular, diamond could be well suited for high temperature and high power devices

Diamond Deposition and Characterization 535 for specialized applications. However, present problems with partially ionized charge carriers in diamond inhibit its use with standard microelectronic device structures. Also, problems inherent in polycrystalline film structures are a major obstacle to effective microelectronic application. This is particularly important as a result of the currently limited availability of standard substrates for epitaxial diamond deposition. The development of HOD film deposition, with its improved materials characteristics, may partially offset these problems, although the ultimate realization of diamondbased microelectronics will most likely rely on the development ofepitaxial diamond material. One method investigated to improve diamond performance and avoid the dopant ionization problem is electron beam induced conductivity.[ 12~ Bombardment of a diamond target using a modulated electron beam provides a means to generate current (via impact generation of electron-hole pairs) in the diamond in a controlled fashion.[121] Figure 13 shows a simple schematic of an electron beam activated device. Such a device exhibits current/voltage behavior similar to a bipolar transistor, as is shown in Fig. 14.[121]

Electron Beam

High Voltage Input Gold Electrode

/

Diamond

Output Figure 13. Schematicof a simple electron beam activateddiamond device.[1211

536

Wide B a n d g a p S e m i c o n d u c t o r s

5

0

"~

a5

....

-

~

-

- ~-'~

-.D.-9

1.0 m A

-'-0-"

2.0 m A

. -9o . - -

3.0 m A

~

4.0 m A

-.B--

4.9mA

--4l---

5.9mA

v

~ 0

~oo..r ~a~.,o. o - o 25

50

75

100

DIAMOND VOLTAGE DROP (VOLTS) Figure 14. Current-voltage plot for an electron bombarded diamond target. The different curves represent different bombardment currents (all with an accelerating voltage of 50 kV). [12q

REFERENCES 1. Nassau, K., Diamond Films and Coatings, (R. F. Davis, ed.), Noyes Publications, NJ, pp. 40-55, (1993) 2. Muranaka, Y., Yamashita, H., and Miyadera, H., Diam. Rel. Mater., 3:313 (1994) 3. Bachmann, P., Leers, D., and Lydtin, H., Diam. Rel. Mater., 1:1 (1991) 4. Angus, J. C., Cassidy, W. D., Wang, L., Wang, Y., Evans, E., Kovach, C. S., and Tamor, M. A., Mat. Res. Soc. Symp. Proc., 383:45 (1995) 5. Bar-Yam, Y.,Moustakas, T.D.,Nature, 342:786 (1989) 6. Frenklach, M., Diamond and Diamond-Like Films and Coatings, (R. E. Clausing, ed.), Plenum Press, NY, 499 (1991) 7. Chu, C. J., D'Evelyn, M. P., Hauge, R. H., and Margrave, J. L., J. Mater Res., 5:2405(1990) 8. Piekarczyk, W.,Roy, R., and Messier, R.,J. Crystal Growth, 98:765 (1989)

Diamond Deposition and Characterization .

53 7

Mehandru, S. P., and Anderson, A. B., J. Mater. Res., 5:2286 (1990)

10. Harris, S. J., and Weiner, A. M.,J. Appl. Phys., 75:5026 (1994)

11. D'Evelyn, M. P., Chu, C. J., Hauge, R. H., and Margrave, J. L., J. Appl. Phys., 71:1528(1992) 12. Frenklach, M., J. Appl. Phys., 65:5142 (1989) 13. Muranaka, Y., Yamashita, H., Sato, K., and Miyadera, H., J. Appl. Phys., 67:6247(1990) 14. Belton, D. N., and Schmieg, S. J.,J. Appl. Phys., 69:3032 (1991) 15. Pinter, I., Tebano, A., Paoletti, A., and Paroli, P., Diam. Rel. Mater., 3:126 (1993) 16. Weiner, P. S., Prawer, S., Hoffman, A., Manory, R. R., Paterson, P. J. K., and Stuart, S.,J. Appl. Phys., 72:4643 (1992) 17. Park, S., and Lee, J., J. Appl. Phys,. 69:2618 (1991) 18. Narayan, J., Godbole, V. P., and Matera, G., J. Appl. Phys., 71:966 (1992) 19. XiLing, P., and ZhaoPing, G., Thin Solid Fdms, 239:47 (1994) 20. Ramesham, R., and Roppel, T.,J. Mater. Res., 5:1144 (1992) 21. Fayette, L., Mermoux, M., and Marcus, B., Diam. Rel. Mater., 3:480 (1994) 22. Yang, W. S., and Je, J. H., J. Mater. Res., 11:1787 (1996) 23. Kim, J. W., Baik, Y., Eun, K. Y., and Yoon, D. N., Thin Solid Films, 212:104 (1992) 4. Stiegler, J., Von Kaenel, Y., Cans, M., and Blank, E.,J. Mater. Res., 11:716

(19963 25. Valdes, J. L., Mitchel, J. W., Mucha, J. A., Seibles, L., and Huggins, H., d. Electrochem. Soc., 138:635 (1991) 26. Yang, G. S., Aslam, M., Kuo, K. P., Reinhard, D. K., and Asmussen, J., J. Vac. Sci. Technol. B, 13:1030 (1995) 70 Carasso, M. L., Adair, J. H., Demkowicz, P. A., Gilbert, D. R., and Singh, R. K., SPIE Proceedings, 3060:203-12 (1997) 28. Makita, H., Nishimura, K., Jiang, N., Hatta, A., Ito, T., and Hiraki, A., Thin Solid Films, 281:279 (1996) 29. Yara, T., Makita, H., Hatta, A., Ito, T., and Hiraki, A., Jpn. J. Appl. Phys., 34:L312 (1995) 30. Lee, D. G., and Singh, R. K., Appl. Phys. Lett., 70:1542 (1997) 31. Masood, A., Aslam, M., Tamor, M. A., and Potter, T. J., J. Electrochem. Soc., 138:L67(1991)

538

Wide Bandgap Semiconductors

32. Yugo, S., Kanai, T., Kimura, T., and T. Muto, T., Appl. Phys. Lett., 58:1036 (1991) 33. Gerber, J., Weiler, M., Sohr, O., Jung, K., and Ehrardt, H., Diam. Rel. Mater., 3:506(1994) M~ Chen, C. J., Chang, L., Lin, T. S., and Chen, F. R., J. Mater Res., 10:3041 (1995) 35. Reinke, P., Kania, P., and Oelhafen, P., Thin Solid Films, 270:124 (1995) 36. Wild, C., Koidl, P., Miiller-Sebert, W., Walcher, H., Kohl, R., Herres, N., Locher, R., Samlenski, R., and Brenn, R., Diam. Rel. Mater., 2:158 (1993) 37. Maeda, H., Ohtsubo, K., Irie, M., Ohya, N., Kusakabe, K., and Morooka, S., J. Mater. Res., 10:3115 (1995) 38. Stoner, B. R., Kao, C., Malta, K. M., and Glass, R. C., Appl. Phys. Lett., 62:2347 (1993) 39. Yang, P. C., Zhu, W., and Glass, J. T., or. Mater. Res., 9:1063 (1994) 40. Van der Drift, A., Philips Res. Reports, 22:267 (1967) 41. Barrat, S., Dieguez, I., Michel, H., and Bauer-Grosse, E., Diam. Rel. Mater., 3:520(1994) 2, Wild, C., Koidl, P., Muller-Sebert, W., Walcher, H., Kohl, R., Herres, N., Locher, R., Samlenski, R., and Brenn, R., Diam. Rel. Mater., 2:158 (1993) 3, Baik, Y., and Eun, K. Y., Thin Solid Films, 214:123 (1992) 44. Singh, R. K., Gilbert, D. R., Fitz-Gerald, J., Harkness, S., and Lee, D. G., Science, 272:396 (1996) 5. Eversole, W. G., U.S. Patent No. 3,030,187 (1962) 46. Angus, J. C., Will, H. A., and Stanko, W. S.,J. Appl. Phys., 39:2915 (1968) 47. Nakao, S., and Maruno, S.,J. Crystal Growth, 99:1215 (1990) 48. Reeve, S. W., Weimer, W. A., and Cerio, F. M., J. Appl. Phys., 74:7521 (1993) 49. Ohtake, N., and Yoshikawa, M.,J. Electrochem. Soc., 137:717 (1990) 50. Watanabe, I., Matsushita, T., and Sasahara, K., Jpn. J. Appl. Phys., 31" 1428 (1992) 51. Shimada, Y., Mutsukura, N., and Machi, Y., Jpn. J. Appl. Phys., 31" 1958 (1992) 52. Rudder, R. A., Hudson, G. C., Posthill, J. B., Thomas, R. E., Hendry, R. C., Malta, D. P., Markunas, R. J., Humphreys, T. P., and Nemanich, R. J., Appl. Phys. Lett., 60:329 (1992) 53. Watanabe, I., and Sugata, K.,Jpn. J. Appl. Phys., 27:1397 (1988) 54. Amaratunga, G., Putnis, A., Clay, K., and Milne, W., AppL Phys. Lett., 55:634 (1989)

D i a m o n d Deposition a n d Characterization

539

55. Chang, C. P., Flamm, D. L., Ibbotson, D. E., and Mucha, J. A., J. Appl. Phys., 63:1744(1988) 56. Gruen, D. M., Liu, S., Krauss, A. R., and Pan, X., J. Appl. Phys., 75" 1758 (1994) 57. Eddy, C. R., Youchison, D. L., and Sartwell, B. D., Diam. Rel. Mater., 3:105 (1993) 58. Singh, R. K., Gilbert, D., Tellshow, R., Holloway, P. H., Ochoa, R., Simmons, J. H., and Koba, R.,Appl. Phys. Lett., 61:2863 (1992) 59. Lax, B., Allis, W. P., and Brown, S. C., J. Appl. Phys., 21" 1297 (1950) 60. Kawarada, H., Mar, K. S., and Hiraki, A., Jpn. J. Appl. Phys., 26:L 1032 (1987) 61. Tokura, H., and Yoshikawa, M., Applications of Diamond Films and Related Materials, (Y. Tzeng, M. Yoshikawa, M. Murakawa, and A. Feldman, eds.), Elsevier Science Publishers, 241 (1991) 2. Pate, B. B., Diamond." Electronic Properties and Applications, (L. S. Pan, and D. R. Kania, eds.), Kluwer Academic Publishers, Boston, 41 (1995) 3. Jin, S., Graebner, J. E., McCormack, M., Tiefel, T. H., Katz, A., and DaulremontSmith, W. C.,Nature, 362:822 (1993) 4. Tokura, H., Yang, C. F., and Yoshikawa, M., Thin Solid Films, 212:49 (1992) 65. Hirata, A., Tokura, H., and Yoshikawa, M., Thin Solid Films, 212:43 (1992) 66. Zhao, T., Grogan, D. F., Bovard, B. G., and Macleod, H. A., ,4ppl. Opt., 31:1483(1992) 7. Lee, D. G., and Singh, R. K., Mater. Res. Soc. Symp. Proc., 354:699 (1995) 68. Pearton, S. J., Katz, A., Ren, F., and Lothian, J. R., Electron. Lett., 28:822 (1992) 9. Vivensang, C., Turban, G., Anger, E., and Gicquel, A., Diam. Rel. Mater., 3:645(1994) 0. Tezuka, S., and Yosikawa, M.,J. Jpn. Soc. Precision Eng., 56:2255 (1990) 71. Pemenov, S. M., Smolin, A. A., Ralchenko, V. G., Likhanski, S. V., Veselovski, I. A., Sokolina, G. A., Bantsekov, S. V., and Spitsyn, B. V., Diamond & Rel. Mater.,2:291 (1993) 2. Singh, R. K., and Lee, D. G.,J. Electron. Mater., 25(1): 137 (1995) 73. Marchywka, M., Pehrsson, P. E., Vestyck, D. J., and Moses, D., Appl. Phys. Lea., 63:3521 (1993) 4. Pehrsson, P. E., Marchywka, M., Long, J. P., and Butler, J. E., Applications of Diamond Films and Related Materials." Third International Conference, (A. Feldman, Y. Tzeng, W. A. Yarbrough, M. Yoshikawa, and M. Murakawa, eds.) NIST SP 885:267 (1995) 5. Shiomi, H., Nakahata, H., Imai, T., Nishibayashi, Y., and Fujimori, N., Jpn. J. Appl. Phys., 28:758 (1989)

540

Wide Bandgap Semiconductors

6. Prins, J. F., The Properties of Natural and Synthetic Diamond, (J. E. Field, 7.

78. 9~ 80.

81. 2.

83. 4.

85. 6.

87. 88. 89. 0~

91. 2, 30 4,

95. 6,

97.

ed.), Academic Press, London, 301 (1992) Glover, G. H., SolidState Electronics, 16:973 (1973) Okana, K., Diamond: Electronic Properties and Applications, (L. S. Pan, and D. R. Kania, eds.), Kluwer Academic Publishers, Boston, 139 (1995) Saparin, G. V., Diam. Rel. Mater., 3:1337 (1994) Jiang, N., Hatta, A., Won, J., Mori, Y., Ito, T., Sasaki, T., and Hiraki, A., Jpn. J. Appl. Phys., 34:L782 (1995) Silva, S. R. P., Knowles, K. M., Amaratunga, G. A. J., and Putnis, A., Diam. Rel. Mater., 3:1048 (1994) Wang, Z. L., Bentley, J., Clausing, R. E., Heatherly, L., and Horton, L. L., J. Mater. Res., 9:1552 (1994) Zhang, Y., Ichinose, H., Ishida, Y., Ito, K., and Nakanose, M., Mater. Res. Soc. Syrup. Proc., 416:355 (1996) Joksch, M., Wurzinger, P., Pongratz, P., Haubner, R., and Lux, B., Diam. Rel. Mater., 3:681 (1994) Ohsawa, Y., Tamou, Y., Kikuchi, N., Hiraga, K., and Oku, T., J. Mater. Science, 26:3748 (1991) Shroder, R. E., Nemanich, R. J., and Glass, J. T.,Phys. Rev. B, 41:3738 (1990) Bou, P., and Vandenbulcke, L.,J. Electrochem. Soc., 138:2991 (1991) Ager, J. W., Veirs, D. K., and Rosenblatt, G. M., Phys. Rev. B, 43:6491 ( 1991) Bergman, L., and Nemanich, R. J., Handbook of Optical Properties Vol. II, (R. E. Hummel, and P. Wibmann, eds.), CRC Press, Boca Raton, 331 (1997) Ager, J. W., andDrory, M.D.,Phys. Rev. B, 48:2601 (1993) Rats, D., Bimbault, L., Vandenbuleke, L., Herbin, R., and Badawi, K. F.,J. Appl. Phys., 78:4994 (1995) Lopez-Rios, T., Sandre, E., Leclercq, S., and Sauvain, E., Phys. Rev. Lett., 76:4935(1996) Knight, D. S., Weimer, R., Pilione, L, and White, W. B., Appl. Phys. Lett., 56:1320(1990) Prawer, S., Nugent, K. W., and Weiser, P. S., Appl. Phys. Lett., 65:2248 (1994) Specht, E. D., Clausing, R. E., and Heatherly, L.,J. Crystal Growth, 114:38 (1991) Windischmann, H., and Epps, G. F., J. Appl. Phys., 69:2231 (1991) Chalker, P. R., Jones, A. M., Johnston, C., and Buckley-Golder, I. M., Surf. Coat. Tech., 47:365 (1991)

Diamond Deposition a n d Characterization

541

8. Schreck, M., Hessmer, R., Geier, S., Rauschenbach, B., and Stritzker, B.,

Diam. Rel. Mater., 3:510 (1994) 9. Cullity, B. D., Elements of X-ray Diffraction, 2nd Ed., Addison Wesley,

100. 101. 102. 103. 104. 105. 106.

Reading, MA, 447 (1978) Ruan, J., and Choyke, W. J.,,4mer. Ceram. Soc. Bull., 72:139 (1993) Kawarada, H., Yokota, Y., Moil, Y., Nishimura, K., and Hiraki, A., Jr. Appl. Phys., 67:983 (1990) Shing, Y. H., Rich, D. H., and Pool, F. S.,J. Appl. Phys., 71:6063 (1992) Graham, R. J., Moustakas, T. D., and Disko, M. M.,J. dppl. Phys., 69:3212 (1991) Yacobi, B. G., Badzian, A. R., and Badzian, T., or. Appl. Phys., 69:1643 ( 1991) Chen, C., Chen, S., Hong, T., and Leu, J., Thin Solid Films, 253:162 (1994) Singh, R. K., Gilbert, D. R., and Laveigne, J., Applied Physics Letters, 69:371

(1996) 107. McNamara, K. M., Scruggs, B. E., and Gleason, K. K., Thin Solid Films, 253:157(1994) 108. Golshai, F., Miller, W. H., Prelas, M. A., Sung, T., Popovici, Manning, G., Loyalka, S. K., Shahedipour, F., White, H. W., Brown, W. D., Malshe, A. P., and Naseem, H. A., Mat. Res. Soc. Syrup. Proc., 416:361 (1996) 109. McNamara, K. M., and Gleason, K. K., Diamond and Diamond-Like Films and Coatings, (R. E. Clausing et al., eds.), Plenum Press, New York, 541

(1991) 110. Hoover, D. S., Lynn, S. Y., and Garg, D., Solid State Technol., 34:89 (1991) 111. Graebner, J. E., Jin, S., Kammlott, G. W., Herb, J. A., and Gardinier, C. F., Nature, 359:401 (1992) 112. Geiss, M. W., Twichell, J. C., and Lyszczarz, T. M., J. Vac. Sci. Technol. B,

14:2060(1996) 113. Okano, K., Koizumi, S., Ravi, S., P. Silva, P., and Amaratunga, G. A. J., Nature, 381:140 (1996) 114. Zhimov, V. V., Wojak, G. J., Choi, W. B., Cuomo, J. J., and Hren, J. J., 3:. Vac. Sci. Technol. A, 15:1733 (1997) 115. Kang, W. P., Davidson, J. L., George, M. A., Milosavljevic, I., Li, Q., Xu, J. F., and Kerns, D. V., d. Vac. Sci. Technot. B, 15:460 (1997) 116. Givargizov, E. I., Zhimov, V. V., Chubun, N. N., and Voronin, A. B., J. Vac. Sci. Technol. B, 15:442 (1997) 117. Choi, W. B., J. Liu, M. T. McClure, A. F. Myers, Zhimov, V. V., Cuomo, J. J., and Hren, J. J., J. Vac. Sci. Technol. B, 14:2050 (1996)

542

Wide Bandgap Semiconductors

118. Givargizov, E. I., Zhimov, V. V., Chubun, N. N., and Stepanova, A. N., 3. Vac. Sci. Technol. B, 15:450(1997) 119. Zhou, D., Krauss, A. R., Corrigan, T. D., McCauley, T. G., Chang, R. P. H., and Gruen, D. M., d. Electrochem. Soc., 144:L224 (1997) 120. Schoenbach, K. H., Kennedy, M. K., Joshi, R. P., Brinkmarm, R. P., and Ho, P. T., SPIE Proceedings, 1632:203 (1992) 121. Joshi, R. P., Kennedy, M. K., and Schoenbach, K. H., J. Appl. Phys,. 72:4781 (1992)

Index

Ab initio 5 calculation 6, 54, 481 Hartree-Fock scheme 441 Abbe number 508 Abrasion 513 Abrupt junction 341 Absorption 57 carbon 522 peak 508, 531 Accelerator high energy 301 Accelerometer 231 Acceptor 21, 97, 120, 194, 322, 346, 410, 444, 480, 485, 495 activate 302, 449 activity 321 boron 443 concentration profile 450 density 485 dopant 466 energy level 189 extrinsic 6 hydrogen passivated 441 impurity 455 ion 440 ionization energy 311 level 189 nitrogen 56

passivate 455, 487, 488 reactivation of 477 species 317 substitutional 323 Acheson technique 180 Acoustic phonon scattering 223 Actinide 416 Activation 456 annealing process 346 diffusion energy 494 electrical 410 energetic source 510 energy 34, 36, 45, 46, 59, 194, 307, 315, 443, 445, 453, 468, 525 Gaussian distribution of energy 468 implantation 315 optical 410 process 510 temperature 318 Active zone 3 Adatoms 93 isolated 95 Additive gases 292 Adduct 45, 59, 60, 69 compound 47, 52 Adsorbates 155

543

544 Wide Bandgap Semiconductors Adsorption 54 site 155 Aerospace application 178 AES 259, 484, 531 near-surface 270 AFM 275, 282, 527 AIN amorphous 252 AIGaAs 355 AIGaN 15 Alkane 56 Alkene 46 production gain 46 radical 56 Alloy broadening 67 Alloying 100 AIN polycrystalline 252 single crystal 252 Ambient 429 vacuum 396, 421 Amine 52, 56 Ammonia 456 Amorphization 127, 333 level 335 Amorphous 127 regime 335 Amphoteric 99 character 97 Amphoteric Defect Model 96 Amplifier operational 229 Analogous calculation 26 Anion 55, 95 Anionic component 65 mixed systems 65 Anisotropic 159, 164, 172, 251, 263 Anisotropy 169, 250, 255, 262, 274, 281 mobility 3 Anneal 108, 127, 307 activation 321, 325 furnace 319 post oxidation 209 postgrowth reactivation 479

rapid thermal 321 temperature 321 Annealing 20, 82, 110, 123, 126, 312, 313, 321, 405, 471, 525 ambient 477, 481 condition 367 dynamic 333 furnace 460 high temperature 423 in-situ 319, 335 isothermal 105 post hydrogenation 457 post-growth 446, 453, 458 post-implant 462, 525 post-thermal 448 rapid thermal 323 temperature 314, 316, 334, 462, 465 thermal 102, 301, 447, 449, 455, 456 vacuum 108 Application electronic 359 high-temperature 232 optoelectronic 359 power-switching 151 Applied bias 87 Aqueous acidic solution 432 Arc-jet 402 Argon 269 Arrenhius plot 252 Arrhenius plot 302, 314, 413, 456 Arsenides 112 Atmospheric pressure growth 66 Atom rare earth 354, 355 RE 362 Atomic hydrogen 429, 471 mass 526 nitrogen 19, 20 nitrogen flux 20 Atomic force microscopy 202, 275, 527 Attempt frequency 468, 488 AuGeNi 82

Index Auger electron spectra 167, 277, 283 spectroscopy 160, 259, 484, 531

Bachmann triangle 510 Back scattering geometry 530 Background 396, 415 Backside heating 276 Band alignment 15 bending 84, 289 conduction density 186 edge 8, 22 relative edge position 9 structure 368 valence density 186 Bandgap 92, 94, 136, 138, 180, 183, 188, 209, 303, 305, 306, 356, 442, 459 energy 187, 362 narrower 132 photons 54 state 136 values 187 Bandpass 398 Barrel reactor 156 Barrier 36, 494 contact 136 height 34, 81, 88, 89, 91, 95, 98, 130, 217 layer 123 rectifying 82 width 81 Beam current 260 BEN 514, 515 Bias 168, 223 reverse 383, 387 self-induced 204 Bias-on forward 484 reverse 484 Biasing 514 electrical 515 Bimetallization 120 Binary alloys 130 phase 118, 128

Bipolar Junction Transistor 180, 218, 220 Biscyclopentadienyl magnesium Blocking layer 226 voltage 217, 222 Blue diodes 42 Blue emission 343 Blue laser 250 Blue laser diode 359 Blue LED 180, 289 BN emission line 267 Boiling 445, 471, 477 Boltzmann constant 85, 488 factor 455 statistic 311 Bombardment 432 Bond breaking 260, 271, 275, 276, 292, 457 Bond centered position 480 site 482 Bond energy 261 Bond length 182 Bonding 102, 522 configuration 440 energy 286 ionic 134 tetrahedral 182 wurzite 183 zinc blend 183 Boron carbide 214 Borosilicate 509 Branch point 94 Breakdown field 300 strength 232 Brewster angle reflection 66 Brightness mode 19 Buffer 477 layer 67, 420, 447 Bulk content 24 crystal 413 crystal growth 99 material 2 material growth 189 wafer form 183 Burgers vectors 48, 190

545

447

546 Wide Bandgap Semiconductors C-BN 517 CAIBE 251, 259 Calculation ab initio pseudo-potential 440 approximate ab-initio 441 CAMECA 394, 395, 413, 419 Cap layer 30 Capacitance, characteristic 341 Capillarity 103 force 132 Capping procedure 137 Carbon adventitious 283 allotrope 506, 525 amorphous 514 etchant 512 incorporation 518 removal 202 source 510 Carburization 518 Carrier 24 charge 355 concentration 15, 24, 26, 89, 99, 454, 457, 467 density 450, 456, 487 freeze-out 189, 230 injection 453 ionization energy 311 lifetime 368 majority 83, 383 type 322 Cathodoluminescence 531 Cation 55, 62 Cationic component 65 mixed systems 64 CBM 94 CCD 529 Cd diffusion 69 CD-ROM 1 Channel resistance 222 Channeling 443 analysis 440 experiment 441 ion 437 Charge compensate 412 distribution 6 state 444

transfer 83 Charge-coupled device 529 Charge-Neutrality Level 94 Charging 393, 419 Chemical activity 63 behavior 251 deuterated 445 potential 6 potential difference 62 Chemical Reaction Models 95 Chemical vapor deposition 370, 456, 489, 506. See also CVD plasma enhanced 173 Chlorine gas mixtures 156 Chromatic aberration 508 CIRA 322, 343 Circuit digital 229 integrated 284 optoelectronic 289 photonic 289 CL 531 Cladding 3 layer 27, 71 material 17 Clamping technique 276 Cleaning 208, 430 Cluster, diatomic 449 Co-implantation 317, 322, 323, 328 Co-pyrolysis 46 Coalescence 438 Coefficient, inter-diffusion 316 Coil, inductive 258 Cold-implantation-rapid annealing 322 Collisional frequency 273 Collisional recombination 273 Colloid 513 Color center 328 Communication, fiber optic 355 Compensate 313, 315 Compensating center 10, 23 Compensating defect 8, 22

Index Compensation 20, 305, 307, 456 chemical 302, 306 damage 301 implant-damage 302 mechanism 19, 22 phenomena 5 picture 22 thermal characteristics 302 Complex di-hydrogen 440 metastable diatomic 443 Composition 67 chemical 280 profile 132 Compound beryllium 1 Mg 1 Mn 1 quarternary 42 ternary 42 Concentration fluctuation 72 free electron 303, 311 hole 311 sheet electron 314 Concentric ring geometry 344 Conductance 209 Conduction 83 activation energy of 323 band 9, 11, 12, 13, 100, 310, 362, 475, 507 band edge 15, 37, 306 band offset 15 temperature activated 306 trap assisted 341 Conduction Band Minimum 94 Conductivity 102, 133, 318, 322, 454, 466, 526, 532 electron beam induced 535 thermal 3, 300 Conductor 394 Conductor trace 229 Confinement effect 25 Contact degradation 36 delamination 215 growth temperature 30 low resistance 100

547

material 36 metal 280 multi-quantum well 138 multielement/GaAs 109 n-type 215 nonalloyed 99 ohmic 80, 81, 82, 83, 87, 89, 98, 99, 102, 105, 108, 109, 110, 111, 112, 113, 114, 118, 120, 123, 125, 128, 129, 130, 131, 136, 137, 138, 152. See also Ohmic: contact p-type 214 rectifying 109, 111, 136 resistance 29, 30, 88, 89, 99, 112, 114, 123, 127, 128, 129, 130, 132, 134, 212, 344 thermally stable 108 Contaminant 156 halogen 56 layer 136 Cooling 179 Coulombic attraction 457 Covalent bonding 81 Cp2Mg 447 Critical dimension 256 Critical dose 338 Critical electric field 215 Cross-sectional transmission electron microscope 477 Crucible tantalum coated 194 Crystal damage 363 defect 301 face reactivity 162 grow 23, 430 lattice 333 quality 389 structure 99, 106 symmetry 182 Crystalline defect 528 perfection 9 522 Si 355 Crystallinity 337

548 Wide Bandgap Semiconductors Crystallites 59 Crystallographic category 182 Crystallographic orientation 507, 515 Cubic boron nitride 517 Cubic phase 61 Cubic ZnTe growth 48 Current confinement 475 density 31, 87, 387 generation 186 leakage 342 pn junction leakage 186 reverse bias density 216 Current-voltage characteristics 87 data 133 Cutoff frequency 340 CV profiling 24 CVD 384, 506, 510. See also Chemical vapor deposition bias-enhanced nucleation process 514 diamond growth rate 512 growth 512 hot filament system 517 microwave plasma system 513 plasma-based system 512 plasma-enhanced 518 torch system 512 vertical low pressure system 195 Cyclotron frequency 520

Damage 168 accumulation 320 density 307 etch induced 167 implantation induced 333, 334 ion induced 168, 487 kinetic 257 lattice-displacement 434 level 304 near-surface 432 profile 307 residual implant 480

Dangling bond 159, 285, 433, 441 at vacancies 8 density 162 DCarcjet 512 DC biases 162 DC current gain 220 DCplasma 17, 20 nitrogen 37 DC self-bias 163 DC-bias 258, 271, 273, 276, 286 Deactivation 441 Decay 375 transient 375 Decomposition 54, 60, 107, 118, 119, 127 divorced eutectic 103 temperature 68, 366 thermal 52 Deep donor complex 10 level 525 Deep trap state 484 Defect 98, 220, 232, 304, 413 bulk 190 charged native 8 compensating 457 configuration 7 density 302, 368, 456, 478 distribution 323 edge type 479 extrinsic 95 formation energy 97 generation 346 homoepitaxial 481 implantation-induced 303 level 302, 306 linear 460 micropipe 190 native 95, 446 native donor 302 site 512 structural 469 structure 343, 515 triangular morphological 196 Degeneracy 13, 55, 311 Degenerate carrier density 58 Degradation 52, 470 Degreased 51

Index Dehydrogenation 458 Densification 209 Density 400, 422 areal 460, 513 atom 507, 522 dislocation 48 impurity 423 interface state 209 nonuniform 191 nucleus 512 oxide trap 206 threshold current 3 Deoxidation 47 Depletion distance 88 length 83, 84 Deposition 98, 102, 127, 285, 339 gas-phase 458 polymeric 162 process 152 rate 59 Depth 395 diffusion 415 distribution 332, 413 profile 286, 413, 416, 423 resolution 395 scale 396 Desorption 260, 261, 277 rate 276 Detection limit 420, 422 sensitivity 413 Detector 300 Deuterate 441 Deuterium 462 concentration 483 distribution 491 flow rate 460 incorporation 471 near-bond-centered 441 neural density 483 penetration 484 SIMS profiles of 469 thermal stability of 489 Device bipolar 152 demonstration 338

549

electronic 250 fabrication 152, 254, 393 GaN-based 339 heterojunction 2 heterostructure 183 high voltage 174 high-speed 126 isolation 339 microwave 126 passivation layer 180 patterning 251 photonic 250, 446 trench-MOS 174 Diamond 307, 322, 332, 338, 343, 397, 405, 494, 506 abrasion 513 application of 532 atomic fraction composition 510 bandgap 507 characterization of 525 chemical vapor deposition 510 colloid 512 cubic lattice unit cell 507 deposition 511, 517, 535 deposition rate 518 diode 343 electrical properties 532 electron affinity 507 elemental component 510 fabrication 509 film 397 grit 509 HF-CVD 518 impurity 531 nucleation rate 514 optical properties 508 optics 533 particle 513 plates 404 polishing 522 polycrystalline 510 properties 506 protective optical coating 534 resistivity 507 single crystal 517 synthetic 397, 402 thermal properties 532

550 Wide Bandgap Semiconductors Diamond application electron emission 534 microelectronic 534 passive microelectronic 532 Diatomic configuration 439 overlap 441 Dielectric 223, 362, 472 chemical vapor deposition of 471 constant 83 crystals 366 deposition 445, 477 interfacial layer 135 Diffraction analysis 106 pattern 201 Diffractometry 530 Diffuse 400 Diffuser 328 Diffusion 20, 30, 31, 33, 34, 35, 37, 68, 111, 112, 252, 301, 331, 362, 366, 405, 433, 439, 450 activation energy 481 barrier 442 behavior 429 coefficient 123, 413, 473 constant 69 effects 36 external source 329, 332 growth 61 interstitial-vacancy 416 length 34 limited growth 65 path 443 profile 114 short-range 480 substitutional 315 uphill 413, 415 Diffusivity 326, 329, 411, 446, 473, 477, 489, 490, 517, 522, 525 Digital TV 227 Diluent gas 510 Dilution 123 Dimension control 250 Dimers 54 DIMOS 222

Diode 30, 32, 102, 196, 341 fabrication 341 InGaN 359 light emitting 1, 300, 359, 386, 446, 477 pn-junction 218 reverse-biased 458 Diode laser, heterostructure 446 Dislocation 35, 47, 52, 61, 193, 478, 479 density 3, 36 Dispersion 508 Dissociation 110, 114, 124, 133, 448, 451, 457, 520 Dissolution, solid-state 103 Distributed flight control advantages 179 Distribution 406 Pearson IV type 480, 491 DMOS 222. See also DIMOS Dominant resistance 99 Donor 21, 194, 410 complex 97 concentration 84, 333 dopant 466 level 457 native 470 native shallow 445 nitrogen 386 passivated 467 passivation 443, 484, 488, 495 passivation reaction 443 species 319 substitutional 443 Dopability 2, 8, 26, 27 maximum 10 Dopant 23, 49, 56, 110, 133, 189, 363 acceptor 433 activated implanted 312 activation 319 amphoteric 116 concentration 440 flux 12, 16 implanted 334 incorporation of 82, 99, 137 intentional 397 level 195 n-type 420

Index p-type 423 phases 23 profile 386 substitutional 97, 405 unintentional 417 Dopant reactivation minority carrier enhanced 452 Dope 2 Doping 2, 98, 131, 405 aluminum 196 amphoteric model 120 behavior 6, 17 boron 323 concentration 88, 301 density 184, 430, 447 diffusion model 112, 114 drain 342 electrical 422 epitaxially controlled 194, 201 experiment 58 external 346 implantation 338 ion implantation 301, 346 level 12, 16, 22, 346 limits 22 n-type 10 limitations 15 nitrogen 5, 196 procedures 56 profile 301 RE 379 selective-area 152 unintentional 15 Dose dependence 307, 322 implanted 307 saturation level 320 Double implanted MOS 222 Drain-drift layer 222 region 222 Dry etch 174, 445, 471, 477 rates 472 self-aligned 450 DTBSe 50 Ductility 102 Durability 173 Dynamic surface 393

ECR

551

152, 165, 175, 201, 251, 263, 266, 457, 460, 520 advantages 165 discharge 289 microwave power 283 power source 287 source 169 EDFA 355 Edge termination 174, 217, 224, 226 EDX 118 EF, unpinned 134 Effective mass 188 Effective Work Function Model, (EWF) 95 Effusion cell temperature 365 EL device 388 emission 386, 387 Electric dipole transition 356 Electric field 84, 226 breakdown strength 183 enhancement 168 strength 222 Electrical characteristic 105 excitation 383, 384 measurement 480 Electrochemical processing 360 Electroluminescence 58, 388 Electrolysis 362 Electron affinity 83, 85, 89, 95, 130, 132, 135, 136, 138, 209, 212, 397 barrier 209 concentration 12, 320 confinement 69 donors 54 flooding 394, 412, 419 hole pairs 54 mobility 50, 340 thermal velocity 87 tunneling of 87 velocity 184

552 Wide Bandgap Semiconductors Electron Cyclotron Resonance 1 5 2 , 175, 201, 251, 450, 457, 46O, 490, 52O plasmas 165 Electron diffraction analysis 118 Electron mean free path 520 Electron Paramagnetic Resonance 437, 445 Electron spin resonance 439 Electron velocities peak 126 saturation 126 Electron wall loss 258 Electron-hole pair 368 Electron-hole recombination 388 Electronegative nitrogen 454 Electronegativity 94, 95 difference 81 Electronic 1, 394 device 290 high power 2, 300 high-temperature 300 Electrophoretic deposition 513 Electrostatic potential 83 Ellipsometry 73 Emission 356 decay 368 efficiency 388 field 88 intensity 267, 386 spectra 364, 373 spectrum 3, 20 thermionic 34, 88 Emitter arrays 534 Energetic ion acceleration 256 Energetic position 8 Energy activation 316 backtransfer 362 carrier activation 315 configuration 481 dispersive x-ray analysis 118 distribution 432 donor activation 189 gap 3, 6, 10, 15, 17, 28 high ion 256 implantation 416 incident ion 257

nitrogen activation 189 recombination 383 site 494 surface 453 Epilayer 89, 133, 137, 363, 371, 386, 405 growth 102 thickness 461 Epitaxial film 447 growth 362, 363, 365 layer 59, 152, 225, 417, 473 method 72 orientation 106 process 73 regrowth 109, 124, 125, 137 relation 113 Epitaxy 36, 42, 108, 114, 306, 363 metal organic vapor phase 42 solid-phase 478 EPR 437, 445 Equilibration 395 Equilibrium constant 63 Fermi energy level 83 phase 107, 123 position 438 thermodynamic 62 Erbium doped fiber amplifier 355 Etch anisotropity 204 characteristics 273 chemistry 202 crystallographic 252 damage 152 dopant-selective stops 205 dry 201 front 473 high pressure 204 ion-beam 434 isotropic 252 mechanism 255 morphology 263 nitrogen-based product 261 process 291 profile 251, 293 rate 202, 203 reaction-limited 252

Index reactive ion 490 recipe 203 removal 323 residue-free 204 results 260 selectivity 263 sidewall 154, 168 wet chemical 445 Etch rate 155, 157, 158, 159, 166, 173, 255, 2 5 6 , 258, 260, 261, 265 GaN 262, 265, 266, 267, 269, 270 isotropic 253 maxima 162 MIE 163 monotonic decrease 277 peak 267 polycrystalline material 159 RIE 159 trend 273 Etched surface morphology 168 Etching 102, 522 anisotropic 154, 231, 257 anodic 379 behavior 159, 162 channel 152 characteristics 155 chemically assisted ion beam 251, 259 deep trench 205 dry 151, 152 ECR 166, 258 GaN 255 gate recess 152 high-density plasma 257 ICP 258 isotropy 204 lateral 257, 273 LE4 260 low energy electron enhanced 260 magnetron ion 163 photoassisted dry 260 photoenhanced wet chemical 253 plasma 156, 158, 250 preferential 283, 338 process 152, 172

553

reactive ion 259, 339 reactive sputter 158 SiC 151 via 205 wet 151, 201, 250, 251, 254 Ethylazide 57 Eutectic 109, 112, 123 phase 109 temperature 109 Evaporation 128 peak 127 Evolutionary selection 515 Ex situ contact schemes 98, 102 Excitation electrical 383 electronic 260 optical 369, 375 process 370 vibrational 260 Excitation impact 383, 388 Exciton 52, 367 neutral-donor-bound 286 Excitonic region 57 structure 66 Extended HuckelTheory, (EHT) 437 Extraction voltage secondary ion 394 Extrinsic effects 93, 98

FA 319 Fabrication 430 laser facet 250 Faceting 196 Fall time 224 Fast-decay component 375 FE 82, 85 Fermi level 8, 37, 56, 81, 90, 130, 135, 311 model 15, 17, 22, 24 pinning 29, 81, 82, 89, 90, 92, 95, 96, 98, 99, 130 position 13, 97, 453 stabilization energy 97 Field effect transistor (FET) 532

554 Wide Bandgap Semiconductors Field emission 82, 85, 136, 217 Filament, rhenium 518 Film adhesion 516 amorphous 160 coalescence 513 continuous 513 crystalline 160 ITO masking 168 modification 521 nanocrystalline 534 near-epitaxial 514 phase purity 528 polycrystalline 516, 533, 535 polycrystalline diamond 515 purity 530 resistivity 494 thickness 59, 447, 515 Finger length 225 Flaking 162 Flight control system 179 Flow dynamic 252 Fluctuation 24 Fluence 370, 406, 416 Flux ion 257, 274, 286 neutral reactant 257 plasma 275, 276, 293 Foot 263 Formation 271 enthalpy 6, 8, 9 polymer 271, 273 Forward bias 383, 386, 388, 448, 450 Forward current 224, 343 Fowler-Nordheim injection 222 Fraction constituent 65 Free carrier concentration 7, 8, 13,20, 21 Free carrier ionization 309 Free electron concentration 13, 16 Free electron density 310 Free energy of formation 64 Free exciton resonance 286 Free hole concentration 9, 18, 19, 20, 21, 23, 24, 35, 37, 136 density 132 FTIR 531

Furnace Heatpulse 410T 479, 483 /f-induction 322

GaAs 15, 24, 82, 98 Gain bandwidth 229 GaN 130, 302, 325, 333 bond 256 buffer 456 wafer 286 Gap energy 61, 71 Gas additive 155 phase 62 phase temperature 512 pressure 510 reactivity 162 Gas mass spectrometry 129 Gas-phase, technique 446 Gaseous diffusion length 510 Gate bias 222, 339 breakdown 154 capacitance 339 dielectrics 180 maximum voltage 220 oxide 222 oxide failure 222 pulse 224 ring 344 Gate Turn-Offthyristors, (GTO) Gate-to-drain breakdown 221 Geothermal wells 178 Germanides 110 Gettered 112 Gibb's free energy 63 Glide planes 48 Glow discharge 154 Gold conductor trace 230 Grain boundary 160 Grain size 59, 515 lateral 512, 513 Graphite phase 511 region 338 sheet 162

218

Index Ground state 438 Group III methyl etch product 277 nitrides 302 Growth active species 510 ambient 465 argon/fullerene chemistry 534 boule 151 bulk 192 bulk method 362 characteristic 527 columnar 201 columnar grain 515 condition 51, 53, 57, 62 direction 514 environment 517 epilayer 151 epitaxial 211, 516 epitaxial method 362 experiment 59 gas-phase technique 448 gate oxide 211 heteroepitaxial 183 homoepitaxial 195 optimization 44 parameter 15, 515, 527, 531 photoassisted 53, 57 rate 44 regime 44 sequence 447 side 404 step flow mode 196 technique 279 temperature 50, 72, 446, 477 velocity 515 Gunn diodes 125

Hall characterization 323 data 67 measurement 49, 456 Halogen 270 etch product 261 Hardness 526

555

HBT 218 Heat spreader 532 Heat treatment 98, 131, 138 Heterointerfaces 477 Heterojunction 10, 98, 99, 111, 120 bipolar transistor 218, 456 device 80 formation 99 Heterojunction FET, (HFET) 218 Heterostructures 495 Hexagonal phase 71 pits 190 HF-CVD 517 High density discharge 165 High ion energy 282 High mass resolution 397 High resolution electron microscopy 478 High resolution transmission electron microscopy 48 High-bias step 169 Highly oriented diamond 514 Hillocks 190, 196 HOD 514, 535 Hole concentration 27, 323, 450, 479 conduction 322 density 454 mobility 457 traps 209 Homogeneity 106, 108 Hopping 24 conduction 302 process 315 Host lattice 23 semiconductor 361 HPHT 509 HRTEM 48 Hybrid contact metallizations 120 Hydrazines 57 Hydrocarbons 72, 270 Hydrogen 263, 413, 429 atomic 433, 439, 445, 452, 520 bombardment 434

556 Wide Bandgap Semiconductors bound 433 configurations of 429 diffusion 493 flux incident 461 implantation 475 incorporation 429, 430, 437, 471 incorporation process 475 insertion 430, 433, 434 molecular 433 molecule 439 neutral interstitial 441 paramagnetic center 445 passivation 429, 470, 473 passivation acceptor 458 permeation 430 retrapping 445 termination 511 Hydrogenate 450 Hydrogenation 430, 440, 450, 457, 494 cathodic 432 electrolytic 432 temperature 454

ICP 204, 251, 263, 266 Idealityfactor 204, 217, 218 IGBT 154 II-VI-compounds 1 III-nitrides 417 SIMS of 419 Illumination 253 Image distortion 527 IMOS. See DMOS Implant 197, 434, 475, 480 damage 301, 492 distribution 413 energy 335 He-damage 323 isolation 301, 302, 303, 304, 306, 339 MeV 200 multiple energy 434 nitrogen 197 proton 434 self-aligning 340 single energy 197

temperature 198, 319 Implantation 211, 301, 302, 306, 338, 346, 413 activation temperature 312 boron 200, 217 damage 313, 318, 319, 333, 415 doping 309, 319 elevated temperature 319 energy 415 hot ion 194 hydrogen 200 induced defect 346 ion 197, 201, 301, 313, 316, 322, 329, 339, 405 isolation 301, 307, 346 multiple energy 307 room temperature 336 SiC 341, 346 Implanted depth distribution 413 Implanted etch ion 285 Impurity 189, 325, 332, 393, 412, 417, 515 activation 336 analyses 398 atom 439 compensating 301 concentration 440, 455 deep level 445 doping 301, 525 film-grown 400 implanted 400 natural 400 RE 360 shallow 445 species 396, 507 In Situ Contact Scheme 98 In-situ cleaning 447 InAIN 305 Incident ion beam 394 energetic 395 Inclusion 511 hydrogen 531 non-diamond 530 Incorporation depth 483, 485, 490, 493 distance 489 unintentional 445

Index Indiffusion 109, 123, 471, 477, 479 deuterium 472 Indium-tin oxide 163 Inductively Coupled Plasma 152, 204 Inelastic Raman scattering 528 Infra-red absorption 437 bands 453 Infrared spectra 376 InGaAIP 355 InGaN 304 Inhomogeneity 24, 68, 72, 103 Injection efficiency 225 lasers 42 InP 82 Input partial pressure 62, 65 Instrument quadrupole 419 sector magnet 419, 421 Instrumentation 178 Insulator 210, 222, 232, 393, 419 Integrated circuit 363 Interaction electronic 333 nuclear 333 Interdiffusion 27, 29, 33, 34, 56, 82, 99, 122, 316 Interface 95, 118, 119, 120, 126, 206 broadening 395 morphology 123, 126 phase 136 quality 222 reaction 130 region 82 state 220 trap 223 trap density 209, 232 zone 48 Interracial barrier 111, 132, 137 contamination 137 morphology 99, 108, 123, 128, 129 phase 116, 131, 132, 478 reaction 82, 95, 102, 105, 112, 131, 132, 133, 135

557

Interference 406 Interlayer, Iow-diffusivity 517 Interlevel dielectric film 280 Intermetallic compound 110 Interstitial 323 dopant 525 location 367 position 491 Intervac Gas Source Gen II system 460 Intrinsic carrier concentration 186 defect 6, 7 layer 341 property 97 solubility 490 state 91 Intrinsic effects 98 Ion beam source 522 bombarding 156 bombardment 152, 155, 156, 169, 283, 457 cesium primary bombardment 395 collisional scattering 255 density 257, 283, 483 energy 163, 259, 273, 275, 279, 286, 483 flux 432 high-temperature implantation 197 impinging 255 implant 223 implantation 362, 430, 434, 525 incident primary beam 393 Kaufman source 434 milling 255 oxygen primary bombardment 395 primary energy 395 rare earth 354 RE 357 scattering 257 technique 259 yield 412 Ionicity 92 Ionization 194 cross section 163 efficiency 258 energy 309, 311 level 317 potential 397

558 Wide Bandgap Semiconductors Ionization impact 520 Irradiation, e-beam process Isolation 201, 301, 346 electrical 475 hydrogen implant 346 optical 475 species 302 thermally stable 346 Isotropic 156, 273 Issue, materials 362 ITO 163 mask 166 maskedge 169 IV characteristic 34

449

Joule heating 81 Junction abrupt 98 gradual 98 M/S 87 Junction Field Effect Transistor (JFET) 189,218, 290, 339, 482

Kelvin Cross Bridge Resistor 89 Kinetics gas-surface reaction 277 reaction 277 Knee voltage 342

Lanthanide 356, 416, 418 Laser 3, 300, 453, 522 ablation 362 Ar + 370 argon 376 blue-green 27 blue-violet 1 current channeling 301 diode 2, 52, 289, 384, 456, 477, 532 diode structure 2, 289 etching 522 excitation energy 368 facet 251 facet, etched 293

incidence angle 522 microdisk GaN-based 289 Nd:YAG 373 pump 368 radiation 368 semiconductor 355 Lattice 71, 133, 303, 405 atom 485 constant 29, 61, 453 crystalline 530 damage 197, 492 defect 285, 439 dislocations 285 location 444 match 99, 114 misfit 61 mismatch 180, 447 next neighbors 187 occupation 311 parameters 107 point 20 position 440 relaxation 442 site 34, 54, 322, 367 Lattice constant 1 Layer blocking 195 epitaxial 137 hexagonal 60 homoepitaxy 194 non-epitaxial 137 LE4 260 Leakage current 434 path 307 LED 1, 250, 300, 359 blue 359 green 360 Lely platelet 190, 194 Liftoff 344 Ligands 362 Light blue/green emitter 1 emission 363, 367, 368 guiding 3 scattering 289 short wavelength emitter 300 spectrum 2

Index LiNbO 3 412 bulk crystalline 415 Liquid phase epitaxy (LPE) 363 Lithographic masking technique 522 Low bias exposure 289 Low-Damage-Drive-In Implantation, (LODDI) 323 Luminescence 45, 52, 343, 354, 360, 365, 367, 370, 371, 388, 531 energy 12 intensity 368 room temperature 378 signal 11

Magnetic confinement 257 Magnetic field-enhanced RIE 163 Magnetic fields 163 Magnetron enhanced RIE 175 Manifolds 368 Marchywka effect 524 Mask 204 aluminum 203 ITO 204 Mask-edge erosion 263, 281 Masking material 153, 157 Mason's unilateral gain 221 Mass electron effective 186 interference 419 resolution 405 spectroscopy 103 Material analysis 413 double-heterostructure 479 metastable 1 wide gap 6 Matrix effect 412 element 397 Maximum electric field 217 MBE 99, 136, 363, 417 MDD 190 Mean free path 273, 276, 432 Measured axial scan 441 Measurement nuclear magnetic resonance 494 stylus profilometry 490

559

Mechanism chemical 301 excitation 375, 388 microscopic 5 sputter desorption 256 Melt-growth technique 190 MERLE 163 Mesa edge 307 etch isolation 339 isolation 152 p-n junction diode 450 MESFET 154, 215, 218, 342 Metal induced gap states 94 Metal Organic Chemical Vapor Deposition 2, 263, 339, 363, 446, 449. See also MOCVD Metal Organic Molecular Beam Epitaxy, (MOMBE) 263, 363, 371, 373, 374, 375, 376, 454, 460 Metal Organic Vapor Deposition 445 Metal organics 45 Metal Oxide Semiconductor 430 Metal Oxide Semiconductor Field Effect Transistor 215, 218, 482 Metal step coverage 280 Metallization 82, 98, 102, 116 Au/GaAs 102 Au/Ge/GaAs 109 Au/Ge/Ni/InP 129 Au/InP 126 Au/Ni/InP 129 bielement 110 Ge/GaAs 108 in situ 99 multicomponent 131 multielement/InP 128 multiple element 82 Ni/GaAs 105 Ni/Ge 110 Ni/InP 127 Pd/GaAs 108 Pd/Ge 110 Pd/InP 127 Pd/Si/GaAs 110 refractory metals/GaAs 108

560 Wide Bandgap Semiconductors schemes 136 single component 130 single element/InP 126 Metallurgical reaction 82 Metalorganic Molecular Beam Epitaxy 460 Metalorganics 446 Metastable 22, 107, 440 diatomic complex 444 Methyl radical 511 Methylmercaptane 57, 71 Mg-profile 328 Micro-cluster 160 Micro-masking 157, 163, 167, 263, 282 effect 156, 160 residue 174 Micro-Raman analysis 530 Micro-roughness 160 Microcleavage 512 Microcluster 69, 95 Microdisk laser, whispering-gallerymode 289 Microelectronics 301 diamond-based 535 Micropipe 193, 232, 331 defect density 190 Microscopy 527 Microtwins 110 Microwave 112 device 359, 489 frequency 520 PE-CVD 520 performance 184 power 469 source 257 Mid-bandgap 94 Midgap 302, 304, 457 MIE 163, 175, 251, 258 technique 163 MIG 94, 95 Migrating 493 Migration 453 Millimeter wave device 482 MINDO/3 442 Miniaturization 180 Miniband calculation 29 conduction 25

position 26 valence 25 Minima 310 Minority carrier 383, 448 density 452 enhanced debonding 495 injection 450, 455 Misfit dislocation 48, 67, 69 Misorientations 514 Missile plume detection 231 Mixed crystal 25 Mixed-mode operation 228 Mixing thickness 395 Mixture, isotropically-substrated 473 Mobile species 107 Mobility 183 anisotropy 220 bulk 220 carrier 188 channel 206, 220, 222 electron 184 extracted 220 hole 184 value 220 MOCVD 263, 281, 339, 362, 363, 370, 417, 446, 449. See also Metal Organic Chemical Vapor Deposition Model 9 fermi level pinning 10 microscopic 10, 20 phenomenological 5 Modification electrical 525 structural 522 Modulation characteristic 339 Modulus 516, 526 Molecular hydrogen 429 species 439 Molecular Beam Epitaxy 2, 99, 136, 363 Molten metal catalyst 509 Monolithic integration 80 Monomer 510 growth 511 More electric aircraft concept 178 benefits 178

Index Morphology 57, 82, 99, 103, 109, 110, 157, 250, 460 changes 103 control 126 etched surface 275 interfacial 107 reaction 110 surface 112, 280 tetrapodlike 59 MOS 430 MOSFET 168, 206, 218, 220 power 222 MOVPE 15, 42, 52, 59, 62, 381, 387, 417 Multi-quantum well 30, 32 Multichannel analyzer 529 Muon spin resonance 494 resonance experiment 494 theory 494 Muonium 444 anomalous 444, 494 normal 444

N-dopability 7, 28 N-doping 22, 55, 66, 67 limitations 17 N2-plasma treatment 57 Native oxide 289 NEA 507 Negative electron affinity 507 Net carrier concentration 387 Net hole concentration 482 Neutral charge state 445 complex 455, 457 flux incident 432 hydrogen-dopant complex 466 Neutralization 432 Nickel silicide 215 Nitride 255, 356, 360 material 250 Nitrogen 267 complexes 20 diffusion 36 doping 7 incorporation 71

561

pairs 20 sublattice 20 Non-Volatile Random Access Memory (NVRAM) 231 Nonlattice sites 126 Nuclear magnetic resonance (NMR) 531 Nuclear stopping 338 Nucleation 512 barrier 107 density 513 enhancement 512

Occupation 94 Ohmic alloy 339 behavior 89, 102, 110, 112, 119, 135 contact 22, 29, 30, 152, 218, 344, 450. See also Contact: ohmic formation 105, 125, 129 p-type contact 212 On-resistance 184, 206, 222 Op-amp monolithic 229 Open-circuit configuration 450 OPO 373 Optical activity 367 amplifier 354, 383 characterization 373 coating 513 device 289 device fabrication 180 dielectric constant 95 emission 362, 367 excitation 383 fiber amplifier 357 fiber communication 383 parametric oscillator 373 property 57, 60, 358, 388 pumping 374 quality 66 radiation 374 source 383 storage 1

562 Wide Bandgap Semiconductors transition 355 transmitter 355 transparency 526 Optical emission spectroscopy, (OES) 155, 159, 162, 267, 470 Optoelectronic 359, 394 applications 80 device 284, 363 material 354 source 355 Orbital hybridization 506 Organo-silanes 482 Orientation, crystallographic 103 Oscillating electric field 520 Oscillation frequency 340 Ostwald ripening 105 Outdiffusion 460,465, 482, 483, 491,494 Outputgain 184 Overgrowth 123 hexagonal 61 Oxidation 95, 463 aluminum 213 procedure 208 rate 210 sacrificial 199 Oxide growth 206 native 105, 283 quality 206 Oxidizing agent 59 Oxygen carrier 60 primary ion beam 416 vacancy 60 Oxynitride 211

P-dopant 2 P-doping 7, 22, 49, 56 levels 17 P-implantation 323 P-ZnSe 135 PAE 227 Parasitics 3 Parity conservation 357 Partial pressure 68 ratio 67

Passivation 201, 261, 280, 446, 448, 450, 454, 456, 461, 475, 486, 495 acceptor 440, 441, 443 boron 490 defect 432 donor 440, 441 efficiency 469 hydrogen 314, 446, 458 mechanism 457 reaction 433 residual 482 shallow-acceptor 440 sulfur 102 Pattern 251, 482 transfer 292 PCD 510 PE-CVD 518 Peaks 376 PEC 205 Periodicity 183 Periphery device 227 Permeation 438 high-temperature 437 Phase binary 116 diamond/graphite 346 formation 95, 130 instabilities 122 ternary 115 Phonon mode 530 Photodissociation 58 Photoelectrochemical etching 205 Photoemission 405 Photoemission spectroscopy 216 Photogeneration 253 Photoirradiation 71 Photoluminescence 10, 286, 313, 361, 434, 447, 531 excitation spectroscopy 367 excitation spectra 374 experiment 10 spectra 356 spectroscopy 367 studies 187 Photoresist 173, 263 developer solution 252 mask 282

Index Physical vapor deposition (PVD) 190, 194 Pinch-off 339, 344 Pinning 8 level 22 mechanism 23 model 23 position 18, 22 Pitting 282 PL 286, 531 band edge 289 decay 375 emission 286, 288 intensity 286, 287 spectrum 52, 370, 376, 381, 383 Planar device electronic 306 topology 339 Planarization beam 522 laser 522 Planck's constant 85 Plasma 154 BCI3 265 chemistry 155, 260, 269, 271 CI2-based 292 composition 19 condition 273 density 163, 273, 276, 518 electron cyclotron resonance 479 emission spectra 19 enhancement 518 exposure 483 fluorinated 155 halogen-based 269 halogen-containing 270 high-density 251, 257, 282, 293 hot filament 402 hydrogen 430 inductively coupled 251 microwave 402 microwave, remote 158, 461 processing 483 species 283 Plasma etch application 289

563

condition 269 electron cyclotron resonant 339 mechanisms 254 process 289 Plasma-induced-damage 254, 259, 285, 286, 293 PlasmaTherm ICP 204 Plateau concentration 469 Platelet 490 defect 460 Lely 190, 194 PLE 367, 374 Point defect 315, 323 concentration 7 Poisson's equation 83 Polarized micro-Raman investigation 530 Polishing 522 diffusive 522 Polymer 513 Polytype 3, 151 control 195 electrical property 183 impurity center 183 Polytypism 182 Porous Si 360, 378 Position interstitial 6 radial 286 substitutional 6 Potential 84 Power density 157, 415 electronics 2 Power added efficiency 227 Precipitates 103, 110 Precursor 53, 56, 71, 72, 420 Prefactor 45 Preferential loss 473 Prereaction 45, 59, 72 Pressure 273 Process dry etch 154, 172 non-equilibrium 363 parameters 202 photocarrier mediated 374 ramped thermal 447

564 Wide Bandgap Semiconductors Processing issues 189 post-growth 371 thermal budget 301 Production high pressure, high temperature 509 Profile 168, 328, 396 angle 173 anisotropic 256, 258, 260, 261 capacitance-voltage 430 carrier 430 depth 394, 405, 421, 422, 531 deuterium 463, 490 diffused 413 doping 329 etch 159, 169, 259, 280 etch, GaN 263 etch, vertical 256 overcut 255, 261 Propane 482 Property electrical 108, 393 electronic 188 optical 393 thermodynamic 96 Proximity 169 Pumping equipment contamination 162 Physical Vapor Transport (PVT) 194 Pyramidal pits 103, 110 Pyrolysis 51 experiment 47

Quadratic dependence 62 Quadrupole instrument 394, 395 Quadrupole Mass Spectrometry (QMS) 263 Quantification 397 Quantum well 10, 27, 30, 68 heterostructure 80 InGaN 359 Quartz enclosure 518 Quaternary compound 61, 80 system 71, 72

Quenching 490

R-Site 482 Radiation resistance 180 Radiation-hard 3 Radicals 47 Radio frequency 37 Raman analysis 530 signature 528 Range parameter 332 Rapid Thermal Annealing (RTA) 320 Rare earth element 356, 416 ion 356 Ratio partial pressure 62 RBS 197, 320, 366. See also Rutherford backscattering Re-passivation 466 Reactand 53 Reactant 155 Reaction diffusion-limited 252 interfacial 105 kinetics 433 metallurgical 103 pits 126 solid-state 106 Reactivation 450, 453, 468, 476, 487 dopant 459, 495 energy 457 kinetics 451 process 448 rate 452 Reactive gas flow rate 260 gases 286 ion etch 174 species 258, 274, 276 Reactive Ion Etching 201 Reactor axis 44 cells 43 etch 431 etch, reactive ion 151 growth, clean 51

Index growth, epitaxial 195 hot-wall CVD 195 hydrodynamic cell 43 ICP 265 parallel plate 157, 158 Plasma Therm SLR 770 483 plasma-assisted deposition 431 rotating disk 450, 477 stainless steel epitaxial 456 Recombination 343, 433 Recombination center 362 Reconstruction 54 Recrystallization 200 Redeposition 271, 273 Redistribution 325, 329, 331, 400, 411, 413, 416, 4 1 7 , 462, 465, 491 Redox potential 54 Reflections 479 Reflective coating 289 Reflectivity 57 Refractory metal 109, 517 filament 517 Regime adsorption limited 274 reactant limited 273 Region charge depleted 83 diffusion limited 54 Regrowth 110, 111, 118, 119, 280 guidelines 137 model 116 Relative change 10 Relative sensitivity factor (RSF) 3 9 4 , 397, 402, 406, 421 Relaxation 48, 368, 482 Reliability 102, 215 data 218 Renucleation 527 Replot software 396 Reproducibility 99, 213, 250 Reservoir 31 Residue 203 formation 162, 167 Resistance 302, 307, 340 as-grown 469 post-hydrogenation 469

565

Resistivity 71, 194, 301, 302, 307, 323, 475, 486, 487, 494, 518, 526 region 342 Resistor thickfilm 229, 230 Resolution photolithographic 201 Resonance level 438 near band-edge 286 Resonant coupling 520 tunneling 30 tunneling variant 35 Retrapping 453, 458, 463 Reverse leakage current 217 Rfbias 257, 258, 483, 484 Rfplasma 17, 20, 258 doping 22 Rf power 2 Rf-cathode-power 270, 271, 275, 276, 279, 282, 286, 287 RHEED 52 RIBE 259 Richardson constant 85, 217 RIE 151, 158, 174, 201, 251, 255, 339 residue-free 202 Rise time 224 Rms 275 roughness 280 Roughness 522 Rutherford backscattering 197, 320. See also RBS Rutherford backscattering analysis 366 Rutherford backscattering spectroscopy 218

Sample charging 527 double-heterostructure Sapphire 417, 419 Satellites 178 Saturation 12, 48, 344

479

566 Wide Bandgap Semiconductors current 86, 87, 217 velocity overshoot phenomenon 184 SBH 85, 216. See also Schottky Barrier Height Scalar product 47 Scanning electron microscopy (SEM) 527 micrograph 281 Scattering loss 251 Schottky behavior 90 contact 154, 215, 218 diode 168, 204, 215, 217, 218 rectifier 215 Schottky barrier 29, 88, 95, 204, 212, 302 contacts 135 height 85, 94, 97, 102, 108, 130, 132, 215 Schottky-Mott rule 85 Schrodinger equation 87 Screw dislocations 190 Secondary ion 397 intensity 397 Secondary Ion Mass Spectrometry, (SIMS) 326, 363, 365, 393, 394, 413, 431, 447, 479, 490, 531 analysis 371, 403 detection limit 416 detection sensitivity 416 H2 plasma profile 484 issues 395 profile 460, 462 quantification standard 420 Secondary Ion Mass Spectroscopy 326 Seed coating 513 Seed crystal 193 Seeding 513 colloid-based delivery 513 electrophoretic 515 Segregation 24, 105 Selected-area-diffraction 478 Selectivity 172, 262 etch 173 Self passivating 434 Self-bias 156

dc 158 Self-trapping 483 efficiency 489 Semi-insulator 507 Semiconductor covalently bonded 81 II-VI group 300 ionically bonded 81 lattice 366 low gap 388 wide bandgap 1 wide gap 359, 388 Sensitivity 528 Sensor 229 flame-indication 231 gas 231 MOS oxygen 231 temperature 229 Shallow donor 470 density 484 Sharp columns 158 Sharp emission spectra 356 Sheath potential 156 Sheet carrier density 456 Sheet hole density 457 Sheet resistance 197, 302, 304, 305, 314, 316, 318, 321, 342 Shell, host atom 441 Si-doping, epitaxial 333 Si-dropletformation 196 Si-redistribution 328 SiC 306, 319, 331, 334, 405, 482 advantages 180 amplifier 152 circuit 229 cubic 182 devices 218 dopant ion implantation 197 electronic properties 231 growth, boule 190 growth, crystal 180 growth, epitaxial 194 hexagonal 183 low power devices 220 MESFET 221 ohmic contact 212 oxidation 206

Index passivation 206 polytypes 182 power devices 222 power switch 152 processing 180 rhombohedral 183 sensor 229 thermistor 230 UV photodiode 230 vapor phase 190 vendor 180 Sidewall 172 angle 172, 173 morphology 293 polymer 263, 281 Silane 482 Silica furnace 413 Silicon values 507 Silicon-on-Insulator technology 3 Simulation, Monte Carlo 20 Single metal/GaAs contacts 102 Single metals 102 Sintering 477 Static Induction Transistor (SIT) 218,227 Site antibonding 437, 439 bond-centered 441, 444 density 447 interstitial 366, 440, 442 interstitial, hexagonal 439 interstitial, tetrahedral 438, 441, 444, 482 near bond-centered 441 near-T 443 near-tetrahedral 441 substitutional 366, 440 tetrahedral 438 trap 463 Site-competition effect 196 Slope parameter 95 Smoothness 280 Sn 338 Solid phase epitaxial regrowth 116 Solid phase regrowth 114, 116 model 114 Solid state laser 357 Solid-vapor distribution function 64, 65

567

Solubility 6, 29, 363, 366, 517, 522 limit 120, 126 Solutions, solid 103 Source gas 194 Source power 274 Space exploration 178 Spatial distribution 83 Spatial resolution 106 Species damage implant 323 ionized 255 reactive 255 Specific contact resistance 89, 122, 131, 135, 214, 222 Spectral array observation 530 Spectral transparency 522 Spectroscopic results 155 Spectroscopy 73, 528 Fourier Transform Infrared 531 Raman 528 Spectrum 369 Spikes 158 Spreading resistance 430 Sputter deposition 157, 339 Sputterdesorption 254, 261, 271, 273, 275, 276 efficiency 269 Sputter rate 287 Sputtering 283, 362, 394 ionic 522 preferential 282 rate 395, 396, 421 Square mesa feature 169 Stability 7, 123, 222, 492 Stabilization 395 Stacking fault 3, 110, 456 Standard, ion implanted 417 Static Induction Transistor 218 Step height 335 Step-graded layer 32 Sticking coefficient 23 Stoichiometric 60, 199 Stoichiometry 6, 108, 118, 255, 258, 277, 280, 283 crystal 322 near-surface 283 surface 105, 321

568 Wide Bandgap Semiconductors Strain compressive 48 tensile 48 Stress biaxial 530 compressive 530 intrinsic 528 residual 516, 530 tensile 530 tensor 531 Striation, vertical 282 Structure complementary logic 201 crystallographic 183 failure 516 ideal 89 multi-quantum well 136 Stylus profilometry 483 Sublimation growth technique 190 seeded 190, 192 Submonolayer 95 Subshell, 4f 356 Substitution fraction 456 isotopic 434 Substitutional 311, 316 incorporation 23 Substrate 3, 111 material 515 micropipe 3 off-axis 195 preparation 47, 51 sapphire 250, 282, 368 semi-insulating 194 side 404 temperature 30, 32, 260 Sulfur hexafluoride 265 Superlattice 25, 27 1-D structure 182 short period 29 staggered type 28 Superposition 6 Surface barrier 132 damage 204, 285 dechanneling 333 degradation 346, 460, 470

energy 507 graphitization 522 hardness 405 hole-catalyzed dissolution 205 impurity termination 507 kinetics 99 mobility 277 morphology 112, 122, 126, 168, 195 non-stoichiometric 255 oxidation 434 partially pinned level 130 passivation 102 pinning 102 roughness 275, 281, 282, 512 segregation 24 states 89, 90, 91, 94 stoichiometric 472 Surface enhanced Raman spectroscopy (SERS) 530 Surface valence band edge 24 Susceptor 44, 314 Switching speed 3 Symmetry rhombohedral 183 System DC plasma torch 519 deposition 517 flame-based 520 flow discharge 432 hot-filament 519 magnetron reactive ion etch 251 microwave plasma based 519 reactive ion etch 251 torch 520

Tailing 329 Target 394 Tellurium (TE) 85 Technical support 424 Technique lateral overgrowth 2 Technology automatic tuning 258 material growth 250

Index Tellurides 7 Temperature 181 ambient material 361 dependence 88, 277 high growth 446 melting 312 nitridation 420 plasma exposure 461 quenching 379 substrate 277 TEN 52 Termination, power device 201 Ternary 304, 472 compound 61, 80 decomposition 130 film 118 phase 106, 118, 125, 127, 128, 130, 134 sample 467 system 72 Tertiary-butyl-mercaptane 57 Test structure, transmission line method 340 Texture 514 formation 515 TFE 82, 85, 105 Thermal conductivity 180, 183, 494 degradation 478, 479 energy 189 expansion coefficient 99, 516 management 3, 402 oxide 180 quenching 361, 362, 387, 388 stability 82, 99, 110, 122, 123, 132, 168, 401, 440, 446, 453, 458, 460, 461, 462, 465, 470, 475, 479, 481, 492, 493, 495 Thermionic current coefficient 85 emission 85, 87, 216 field 136 field emission 82, 85, 217 Thermodynamic 62 arguments 155 equilibrium 99, 102, 127, 509 Thickness uniformity 447

569

Thin film deposition 434 reaction 107 system 107 Threading, dislocation 456, 478, 479 Threading out 479 Threshold 251 current 3, 289 energy 432 voltage 220, 289 Thyristors 152, 218, 224 Time 395 TLM 89 Topography 132, 384 Torch, oxy-acetylene 520 Transconductance 184, 220, 227, 339, 344 Transistor 250, 300, 359 action 343 bipolar 535 bipolar diamond 343 bipolar fabrication 203 circuit 301 field effect 189 field effect, diamond 344 field-effect, GaN junction 290 heterojunction bipolar 453 high temperature junction field effect 489 insulated gate bipolar 154 metal-oxide semiconductor field effect 489 static induction 227, 228, 232 vertical static induction 154 Transistor curves 344 Transition 370 damage-related 434 Translation vector 479 Transmission electron microscopy (TEM) 113,460, 490, 528 Transmission electron microscopy 456 Transmission Line Method 89 Transport 87, 300 mechanism 85, 87 physical vapor system 193

570 Wide Bandgap Semiconductors property 334 solid state 103 vertical 27 Trapping 448, 454, 477 Trench 169, 281 bottom 154 Triangular barrier 34 Triethylgallium 460 Trimethylamine alane 460 Trimethylgallium 446, 456 Trimethylindium 460 Trivalent praesodymium ion 356 Tunneling 37, 88, 100 contact 88 current 87 Fowler-Nordheim 209 probability 36, 87, 217 Twinning 106, 527

UMOSFET 154, 206 Unified defect model, (UDM) 92, 93, 95 Unintentionally incorporate 495 Unipolar device 152, 154 Unit cell 107 Universal gas constant 62 Unpin 102 UV dosimetry 231

Vacancy

9, 97, 209, 302, 318, 322, 438 concentration 7 defect 438 diffusion model 114 flux 112, 125 nitrogen 461, 468 Vacuum level 8, 37, 507 Valence band 11, 17, 29, 83, 507 edge 9, 23, 29, 37 maximum 94 offset 29 Van der Pauw geometry hall

measurement 457 Vanadium 306 Vapor pressure 64, 459 VBM 94, 99 Velocity, surface recombination Vessel, dielectric 258 Vibrational band 457 Vibrational frequency 443 H-stretching 443 ViGS 94 Virgin channeling yield 337 Virtual Gap States 94 VMOSFET 206 Voids 132, 190 Volatility 156, 261, 274, 276, 277 Voltage 35, 88 blocking 226 blocking, forward 224 breakdown 224 breakdown, reverse 342 characteristic 341 forward drop 224 offset 398 threshold 3 turn-on 344, 487

Wafer 123, 190 single wafer 110 single-crystal 151 Wavefunction tails 94 Waveguides 413 Wavelength illumination 528 lasing 355 radiation 528 Wet anodization 378 Wet etching 47, 471, 477 Wetting 112 Wirebonding 215 Work function 81, 83, 89, 95, 131, 132, 135, 136, 138, 212

289

Index X-ray diffraction 252, 526, 530 X-ray photoelectron spectroscopy 199 XPS 199 XRD 530 XTEM 477 DH-LED 478

Zincblende 1 ZnCdS 71 ZnMgSeS 71 ZnO 59 ZnSe 50, 130, 410 ZnSSe 66 ZnTe 45 ZnTeSe 65 ZnxCdl.xSe 68

571

This page intentionally left blank