Effect of Disorder and Defects in Ion-Implanted Semiconductors: Optical and Photothermal Characterization, Volume 46 (Semiconductors and Semimetals)

Effect of Disorder and Defects in Ion-Implanted Semiconductors: Optical and Photothermal Characterization SEMICONDUCTORS AND SEMIMETALS Volume 46

Semiconductors and Semimetals A Treatise

Edited by R. K . Willardson

Eicke R. Weber

DEPARTMENT OF MATERIALS SCIENCE SPOKANE, WASHINGTON AND MINERAL ENGINEERING UNIVERSITY OF CALIFORNIA AT BERKELEY CONSULTING PHYSICIST

In memory of Dr. Albert C. Beer, Founding Co-Editor in 1966 and Editor Emeritus of Semiconductors and Semimetals. Died January 19, 1997, Columbus, OH.

Effect of Disorder and Defects in Ion-Implanted Semiconductors: Optical and Photothermal Characterization SEMICONDUCTORS AND SEMIMETALS Volume 46 Volume Editors

CONSTANTINOS CHRISTOFIDES DEPARTMENT OF NATURAL SCIENCES UNIVERSITY O F CYPRUS NICOSIA, CYPRUS

GERARD GHIBAUDO LABORATOIRE DE PHYSIQUE DES COMPOSANTS A SEMICONDUCTEURS ENSERG GRENOBLE, FRANCE

ACADEMIC PRESS San Diego London New York Sydney

Boston Tokyo

Toronto

This book is printed on acid-free paper.

@

COPYRIGHT 0 1997 BY ACADEMICPRESS ALL RIGHTS RESERVED. NO PART OF THIS PUBLICATION

MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM OR BY ANY MEANS, ELECTRONIC OR MECHANICAL, INCLUDING PHOTOCOPY, RECORDING, OR ANY INFORMATION STORAGE AND RETRIEVAL SYSTEM, WITHOUT PERMISSION IN WRITING FROM THE PUBLISHER.

The appearance of the code at the bottom of the first page of a chapter in this book indicates the Publisher’s consent that copies of the chapter may be made for personal or internal use, or for the personal or internal use of specific clients. This consent is given on the condition, however, that the copier pay the stated per copy fee through the Copyright Clearance Center, Inc. (222 Rosewood Drive, Danvers, Massachusetts 019231, for copying beyond that permitted by Sections 107 or 108 of the US. Copyright Law. This consent does not extend to other kinds of copying, such as copying for general distribution, for advertising or promotional purposes, for creating new collective works, or for resale. Copy fees for pre-1997 chapters are as shown on the chapter title pages; if no fee code appears on the chapter title page, the copy fee is the same as for current chapters. 0080-8784/97 $25.00

ACADEMIC PRESS 525 B Street, Suite 1900, San Diego, CA 92101-4495, USA 1300 Bovlston Street. Chestnut Hill. Massachusetts 02167. USA http://w&w.apnet.com

ACADEMIC PRESS LIMITED 24-28 Oval Road, London NWI 7DX. UK

htlp://www.hbuk.co.uk/ap/

International Standard Book Number: 0-12-752146-1 Printed in the United States of America 97 98 99 00 01 BB 9 8 7 6 5 4 3 2 1

Dedicated to our parents, Anthoula, Christos, Jeanne, and Jean

This Page Intentionally Left Blank

Contents

LISTOF CONTRIBUTORS . . . . . . . . . . . . . . . . . . FOREWORD . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

xi

. . . . .

Xlll

I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Principle of Ellipsometry . . . . . . . . . . . . . . . . . . . . . . . . 111. General Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . IV . Optical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . V . The Complex Dielectric Function . . . . . . . . . . . . . . . . . . . . VJ . Light Penetration . . . . . . . . . . . . . . . . . . . . . . . . . . . VII . Effective Medium Theory . . . . . . . . . . . . . . . . . . . . . . . . VIII . Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I . Gallium Arsenide . . . . . . . . . . . . . . . . . . . . . . . . . . 2.Germanium . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Indium Phosphide . . . . . . . . . . . . . . . . . . . . . . . . . . 4.Silicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 . Silicon-on-Insulator (SOI) Structures . . . . . . . . . . . . . . . . . 6. Separation by IMplantation of Oxygen (SIMOX) . . . . . . . . . . . . 7. Separation by IMpiantation of Nitrogen (SIMNI) . . . . . . . . . . . IX. Sophisticated Multilayer Optical Models . . . . . . . . . . . . . . . . . 1. Profiles with Unknown Depth Variation . . . . . . . . . . . . . . . . 2. Profiles with Known Depth Variation . . . . . . . . . . . . . . . . . X . Closing Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 4 7 9 13 13 15 15 16 18 19 24 26 27 27 29 31 31 34

...

Chapter 1 Ellipsometric Analysis M . Fried. T. Lohner and J . Gyulai

vii

1

viii

CONTENTS

Chapter 2 Transmission and Reflection Spectroscopy on Ion Implanted Semiconductors Antonios Seas and Constantinos Christofides I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1 . General Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. Recent Optical Experimental Studies on Implanted Silicon . . . . . . . . . 1. Phosphorous-Implanted Silicon . . . . . . . . . . . . . . . . . . . . 2. Fourier Transform Infrared Optical Measurements . . . . . . . . . . . IV. Theoretic Background . . . . . . . . . . . . . . . . . . . . . . . . . V . Discussion and Analysis . . . . . . . . . . . . . . . . . . . . . . . . 1. Influence of Annealing Temperature on the Plasma Wavelength . . . . . . 2. Effective Mass vs. Annealing Temperature . . . . . . . . . . . . . . . VI . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

39 40 47 47 49 60 62 62 61 68 69

Chapter 3 Photoluminescence and Raman Scattering of Ion Implanted Semiconductors.Influence of Annealing Andreas Othonos and Constantinos Christofides I. Introduction . . . . . . . . . . . . . . . . . . . . I1. Photoluminescence and Raman Scattering Theory . . . 1. Photoluminescence Theory . . . . . . . . . . . . . 2. Raman Scattering Theory . . . . . . . . . . . . . I11. Photoluminescence and Raman Scattering Techniques. . 1. Common Photoluminescence Techniques . . . . . . 2. Raman Scattering Techniques . . . . . . . . . . . . 3. Time-Resolved Measurements . . . . . . . . . . . 1V. Characterization of Ion-Implanted Semiconductors . . . 1. Photoluminescence Experimental Studies . . . . . . 2. Raman Studies on Ion-Implanted Semiconductors . . V . Summary and Future Perspectives . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . .

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

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

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

73 74 74 77 79 79 80 82 84 84 97 111 112

Chapter 4 Photomodulated Thermoreflectance Investigation of Implanted Wafers Annealing Kinetics of Defects

.

Constantinos Christofides I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Photomodulated Thermoreflectance Theory . . . . . . . . . . . . . . . . 1. Basic Photothermal Equations

2. Three-Dimensional Diffusion

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

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

III . Experimental Methodology 1. Room Temperature Measurements . . . . . . . . . . . . . . . . . . 2. Measurements versus Temperature . . . . . . . . . . . . . . . . . . IV. Experimental Results and Discussion . . . . . . . . . . . . . . . . . . . 1. Characterization of Implanted Wafers . . . . . . . . . . . . . . . . . 2. Influence of Annealing . . . . . . . . . . . . . . . . . . . . . . . .

115 116 116 118 119 119 121 122 122 126

ix

CONTENTS 3. Temperature Influence on the Photothermal Signal . V . Recent Developments . . . . . . . . . . . . . . . . 1. Single-Beam Thermowave Technique . . . . . . . 2. Extension to Two-Layer Model . . . . . . . . . . . VI . Summary and Future Perspectives . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . .

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

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

136 144 144 145 146 147

Chapter 5 Photothermal Deflection Spectroscopy Characterization of Ion-Implanted and Annealed Silicon Films

U. Zammit 1. Introduction . . . . . . . . . . . . . . . . . . . I I . Theory and Experiment . . . . . . . . . . . . . . I11. Results and Discussion . . . . . . . . . . . . . . . 1. The Effect of Implantation Dose . . . . . . . . . 2. Effects of Annealing of Damaged Crystalline Material . 3. Effects of Annealing of Amorphous Material . . . . 1V. Conclusions . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . .

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

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

. . . .

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

. . . .

. . . .

. . . .

151 154 158 158 166 167 174 174

Chapter 6 Photothermal Deep-Level Transient Spectroscopy of Impurities and Defects in Semiconductors Andreas Mandeiis. Arief Budiman and Miguei Vargas I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Physical Foundations of Photothermal Radiometric Deep-Level TransientSpectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . 111. Theory of Photothermal Radiometric Deep-Level Transient Spectroscopy . . IV . Instrumental Foundations of Photothermal Radiometric Deep-Level Transient Spectroscopy: The Lock-In Rate-Window Method . . V . Experiment and Discussion . . . . . . . . . . . . . . . . . . . . . . . 1. Constant-Temperature Photothermal Radiometric Deep-Level Transient Spectroscopy of Silicon . . . . . . . . . . . . . . . . . . 2. Constant Duty-Cycle Photothermal Radiometric Deep-Level Transient Spectroscopy of Semi-Insulating-Gallium Arsenide . . . . . . . . . . . VI . Potential for Ion-Implantation Diagnostics and Conclusions . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .

179

.

181 184

.

188 190

.

191

.

198 207 209

. .

Chapter 7 Ion Implantation into Quantum-Well Structures R. Kalish and S. Charbonneau I. Introduction . . . . . . . . . . . . . . . . . . II . General Background . . . . . . . . . . . . . . . 1. Ion-Implantation-Related Damage . . . . . . 2. Annealing Ion-Implantation-Related Damage . . 3. Evaluation of Structural Modifications . . . . . 4. Evaluation by Optical Techniques . . . . . . . I11 . Ion-Beam-Induced Modifications of QW Structures

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

. .

. . . . . .

. . . . . .

213 216 217 220 221 224 230

CONTENTS

X

1. Point-Defect and Impurity-Induced Heterostructure Interdiffusion . . . . . 2. Threshold Dose for Intermixing of Quantum Wells: Size of Interface Area Affected by Individual Ion Tracks . . . . . . . . . 3. Intermixing of Interfaces Far Beyond the Ion Range: The Ion Channeling Effect . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Realization of Zero- and One-Dimensional Structures by the Use of Focused Ion-Beams (FIBS) . . . . . . . . . . . 5. Defect Diffusion in Ion-Implanted AlGaAs and InP Systems . . . . . . . 6. Implantation Temperature and Dose-Rate Dependence . . . . . . . . . . IV . Future Trends and Applications . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Chapter 8 Ion Implantation and Thermal Annealing of 111-V Compound Semiconducting Systems: Some Problems of III-V Narrow Gap Semiconductors Alexandre M . Myasnikov and Nikolay N. Gerasimenko I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Materials and Impurities . . . . . . . . . . . . . . . . . . . . . . . . 1. Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Impurities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. Ion Implantation . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Range Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Ion Implantation Damage . . . . . . . . . . . . . . . . . . . . . . 3. Ion Implantation Damage in Indium Arsenide . . . . . . . . . . . . . 4. Ion Implantation Damage in Indium Antimonide and Gallium Antimonide . . . . . . . . . . . . . . . . . . . . . . . . . 5. Swelling of Indium Antimonide and Gallium Antimonide . . . . . . . . . IV . Annealing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Damage Annealing . . . . . . . . . . . . . . . . . . . . . . . . . 2 . Capping Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Capless Annealing . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Redistribution of Impurities During Annealing . . . . . . . . . . . . . V . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . INDEX

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

CONTENTS OF VOLUMES IN THISSERIES. . . . . . . . . . . . . . . . . .

231 232 237 238 240 247 248 251

257 258 258 260 261 261 264 266 271 272 279 279 279 281 281 290 291 295 303

List of Contributors

Numbers in parenthesis indicate the pages on which the authors’ contributions begin.

ARIEFBUDIMAN (179), Photothermal and Optoelectronic Diagnostics Laboratories, Department of Mechanical and Industrial Engineering, University of Toronto, Ontario M5S 3G8, Canada S. CHARBONNEAU (2 13), Institute for Microstructural Sciences, National Research Council, Ottawa K I A OR6, Canada CONSTANTINOS CHRISTOFIDES (39, 73, 115), Department of Natural Sciences, University of Cyprus, 1678 Nicosia, Cyprus M. FFUED (l), M T A KFKI Research Institute for Materials Science, H-1525 Budapest, Hungary NIKOLAY N. GERASIMENKO (257), Joint Institute for Semiconductor Research, 103498 Zelenograd, Moscow, Russia J. GYULAI (l), M T A WKI Research Institute for Materials Science, H-1525 Budapest, Hungary R. KALISH(213), Solid State Institute and Physics Department, TechnionIsrael Institute of Technology, Haifa 32000, Israel T. LOHNER (l), M T A KFKI Research Institute for Materials Science, H-1525 Budapest, Hungary ANDREASMANDELIS(179), Photothermal and Optoelectronic Diagnostics Laboratories, Department of Mechanical and Industrial Engineering, University of Toronto, Ontario, M5S 3G8, Canada ALEXANDRE M. MYASNIKOV (257), Institute of Semiconductors Physics, Academy of Sciences of Russia, 630090 Novosibirsk, Russia xi

xii

CONTRIBUTORS

ANDREASOTHONOS (73), Department of Natural Sciences, University of Cyprus, 1678 Nicosia, Cyprus ANTONIOSSEAS (39), Department of Natural Sciences, University of Cyprus, 1678 Nicosia, Cyprus MIGUELVARGAS(179), Photothermal and Optoelectronic Diagnostics Laboratories, Department of Mechanical and Industrial Engineering, University of Toronto, Ontario M5S 3G8, Canada U. ZAMMIT (151), Dipartimento di Ingegneria Meccanica, Universith di Roma “Tor Vergata”, and INFM UNITA’ “Roma 2 Tor Vergata” 00133 Rome, Italy

Foreword

Implantation of impurity atoms for doping semiconductor wafers offers many advantages such as rapidity, mass separation for purity requirements, accuracy and a wide range of doses, flexibility of profile depth, and a control over the amount of ions in a specific region. The bombardment of solids by ion implantation also has been widely used for the study of amorphization and physical analysis of structural disorder introduced by irradiation. One main disadvantage of the ion implantation process is damage introduced into the semiconductor resulting from the energetic character of the process. The consequence of ion bombardment is the amorphization of the semiconductor surface and, at high ion concentrations, the presence of electrical defects such as interstitial impurities, dislocations, grain boundaries, and inhomogeneities. This implantation-induced disorder leads to strong degradation of the electrical features of the materials. It is this degradation that must be dealt with during the device fabrication process. Ion implantation therefore must be followed by one or more annealing processes for the semiconductor to recover its crystallinity and for the doping impurity to become active. In general, thermal annealing in a conventional furnace is currently used to activate the doping such that the thermal diffusion takes place, leading to a serious redistribution of the impurity atoms. This redistribution then causes enlargement of the junctions, which may be prohibitive for the optimal operation of short-channel complementary metal-oxide semiconductor (CMOS) devices. Moreover, these implantation-induced defects strongly modify the recombination properties of the semiconductor that may affect the operation of bipolar devices and p-n junctions. Rapid thermal annealing (RTA) methods have been

...

Xlll

xiv

FOREWORD

introduced to minimize redistribution of the impurities. To this end, an annealing procedure using light or electron beams has been proposed to reduce the annealing duration while maintaining a sufficiently high temperature to activate the doping species. The most commonly used technique is the so-called lamp RTA process in which the thermal heating of the sample is ensured by the heat dissipation of halogen lamps surrounding the wafers. These lamp RTA machines provide very fast heating pulses (approximately 1 to 10sec) that allow the temperature to range from lo00 to 1100°C without significant thermal diffusion. The eight chapters of this book attempt to analyze the experimental results obtained by the four main families of techniques of characterization on implanted wafers: electrical, physicochemical, optical, and thermal wave. It is important to note that this is the first book in the field of ion implantation to review thermal wave studies on implanted materials. Each chapter focuses on the following important effects of ion implantation and the annealing processes: Damage induced (short-range disorder and long-range disorder) by ion implantation; Spatial distribution profiles of the damage; Kinetics of annihilation mechanisms (the defect layer and the amorphous layer); Electrical activation of the implanted impurities. The first three chapters of volume 46 focus on the conventional optical properties of ion implanted materials. In particular, the techniques used to characterize the semiconductor optically after implantation and after annealing are discussed. Chapter 1 deals with the simplest technique, that is, ellipsometry, which allows determination of the thickness and the optical refractive index of the constituent sublayers of the materials. In particular, this can be used to evaluate the amorphous quality of the material versus implantation dose, annealing time, or both, and temperature. Moreover, Chapter 1 provides excellent information about the optical permittivity of the semiconductor layers as a function of the disorder level. Chapter 2 concentrates on the Fourier transform infrared (FTIR) spectroscopy, which is very sensitive to the implantation dose and annealing conditions. In addition, FTIR spectroscopy has the advantage of giving quantitative and qualitative information concerning the four aspects presented previously. The nondestructive and noncontact features of this technique make it very attractive. Measurements by FTIR at low wavelengths can be used to obtain information concerning the influence of implantation dose and annealing temperature on the kinetics of reconstructions. From the interference fringes,

FOREWORD

xv

we can obtain information concerning the amorphous-crystal transition as a function of the annealing temperature. The presence of reflectivity minimum in the case of far-IR measurements is a very practical tool for evaluation of the percentage of the activation of implanted impurities. Additional information concerning the variation of the optical absorption coefficientwith the implantation dose, energy, and annealing temperature can be obtained. Chapter 3 is devoted to Raman spectroscopy and photoluminescence investigation of ion-implanted semiconductors. It is emphasized that Raman spectra obtained from ion-implanted wafers consist of three components that correspond to: (a) the scattering for the crystalline substrate, (b) the amorphous phase, and (c) the mixed amorphous-crystalline phase at the initial amorphous silicon-crystalline silicon interface. This technique can offer important information such as free-carrier concentration, number of impurities in substitutional sites of the implanted lattice, and crystalline-amorphous monitoring. Photoluminescence (PL) measurements also are used to study the effect of implantation and annealing on implanted semiconductors.These measurements are very sensitive to defects in semiconducting crystals. Above all, PL profile measurements reveal the macroscopic defect distribution. Cathodoluminescence is a technique used to characterize the implanted semiconducting layer to determine depth profiles. Chapters 4-6 are concerned with photothermal wave analyses. The noncontact character of the associated techniques makes them particularly attractive for the nondestructive evaluation (NDE) of materials. In Chapter 4, some of the noncontact methods that employ optical excitation are described such as photomodulated thermoreflectance (PMTR), which can detect local variations in the reflectivity of the material caused by thermal effect and plasma free-carrier recombination. These measurements allow the estimation of certain characteristics such as local electronic and thermal transport parameters, which in the case of implanted semiconductors, provide an indication of the local degree of damage as a function of the junction depth. In Chapter 5, photothermal deflection spectroscopy (PDS), which monitors the refractive index gradient in the implanted layer or above its surface, is addressed. The PDS technique has been shown to be a useful tool to investigate the subgap absorption in ion-implanted semiconductor layers. It is a highly sensitive technique and, thus, seems particularly appropriate for this kind of investigation. The analysis of subgap absorption can provide information concerning the concentration and energy of the gap state in implanted semiconductors. Chapter 6 summarizes the potential for photothermal radiometry (PTR), in conjunction with deep level transient spectroscopy (DLTS), to be used as a noncontact technique. The physical principles of this technique are

xvi

FOREWORD

presented first, along with the preliminary results obtained with silicon and gallium arsenide. The instrument setup and experimental characteristics of PTR-DLTS regarding the free-carrier plasma and the lattice thermal transport origin of the signal are presented. Rate-window detection using either a lock-in-amplifier or a boxcar integrator scheme has been evaluated with silicon wafers. The prospects are excellent for this novel technique to be developed to a remote fully noncontact diagnostic tool for real-time in situ characterization. Chapters 7 and 8 are dedicated to some peculiar aspects of ion implantation in quantum wells and compound semiconductors. Since the 1970s, there has been great interest in multilayer structures as a result of the development of molecular beam epitaxy (MBE), metalorganic chemical vapor deposition (MOCVD), and chemical beam epitaxy (CBE). Superlattices and quantum wells therefore have gained considerable importance in electronic and optoelectronic applications. Most of these structures are made of compound semiconductors such as gallium arsenide and indium phosphide. Chapter 7 deals with ion-implanted quantum-well structures. In particular, the physical issues resulting from interdiffusion between the layers, intermixing phenomena, and channelling and disordering effects in heterostructures are addressed. Chapter 8 focuses on the ion implantation and thermal annealing of compound semiconductors. As is well known, after silicon, gallium arsenide is the most important semiconductor material for the fabrication of electronic devices. The unique properties of gallium arsenide of the other III-V semiconductors make them very attractive for applications in optoelectronics and microwave devices. The advantages of III-V materials can be either wider or narrower bandgaps and especially, their direct band-to-band electronic transitions, which provide high quantum efficiency.This chapter therefore is a review of the most recent achievements in ion implantation and the annealing processes for III-V semiconductors. It describes, in particular, the main problems encountered in applications of narrow gap compound semiconductors such as gallium antimonide, indium arsenide, and indium antimonide. CONSTANTINOS CHRISTOFIDES G ~ A GHIBAUDO F ~

SEMICONDUCTORS AND SEMIMETALS. VOL. 46

CHAPTER 1

Ellipsometric Analysis M . Fried T.Lohner J . Gyulai MTA KFKI RFEARCH INSTITVIE FOR MATERIALS SCIENCE BUDAPEST. HUNGARY

1. INTRODUCTION . . . . . . . . . . . . . . . . 11. PRINCIPLE OF ELLIPSOMETRY . . . . . . . . . . . 111. GENERAL REMARKS. . . . . . . . . . . . . . . IV . OPTICAL MODELS . . . . . . . . . . . . . . . V . THECOMPLEX DIELECTRIC FUNCTION . . . . . . VI . LIGHTPENETRATION . . . . . . . . . . . . . . VII . EFFECTIVE MEDIUM THEORY . . . . . . . . . . VIII . EXAMPLES . . . . . . . . . . . . . . . . . . . 1. Gallium Arsenide . . . . . . . . . . . . . . 2. Germanium . . . . . . . . . . . . . . . . . 3. Indium Phosphide . . . . . . . . . . . . . . 4. Silicon . . . . . . . . . . . . . . . . . . . 5. Silicon-on-Insulator (SOI)Structures . . . . . . 6. Separation by IMplantation of Oxygen (SIMOX) I . Separation by IMplantation of Nitrogen (SIMNI) IX. SOPHISTICATED MULTILAYER OPTICAL MODELS . . 1. Profiles with Unknown Depth Variation . . . . . 2. ProJIes with Known Depth Variation . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

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

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

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

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

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

x. CLOSlNG &MARKS . . . . . . . . . . . . . . . . . . . . . . . . . . . References

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

I

.

1 3 4 7 9 13 13 15 15 16 18 19 24 26 21 27 29 31 31 34

Introduction

These days a renaissance of the technique of ellipsometry can be witnessed. This newly activated interest is driven by the demand for rapid. nondestructive analysis of surfaces and thin films. especially films and surfaces occurring in different device technologies. Ellipsometry enables optical constants of materials to be determined with high accuracy and therefore can help to solve a wide variety of problems in different disciplines. 1 Copyright Q 1997 by Academic Press All rights of reproduction in any form reserved. 0080-8784/97125

M. FRIED,T. LOHNER AND J. GYULAI

2

The fact that ellipsometric measurements can be performed in any ambient is a definite advantage over other surface-science techniques for industrial applications. In addition to films in device technologies, ellipsometry is widely used in assessing the quality of oxide coatings and in characterizing such diverse systems as polymers, liquid crystals, and Langmuir-Blodgett films. In ellipsometry, one measures the change in polarization state of a linearly polarized beam of light after nonnormal reflection from the sample to be studied. The term ellipsometry originates from the fact that the light becomes elliptically polarized after the oblique reflection (Fig. 1) due to the different reflection coefficients for p - and s-polarized light ( p stands for parallel and s for perpendicular, “senkrecht” in German). The polarization state can be defined by two parameters, for example, the relative phase and relative amplitude of the orthogonal electric-field components of the polarized light wave. On reflection, both electric-field components are modified in a linear way, and therefore a single ellipsometric measurement provides two independent parameters. By measuring the polarization ellipse, one can determine the complex dielectric function of the sample. Although the principles of ellipsometry were established a long time ago (Drude, 1889a, 1889b, 1890), the technique found its first practical use with the development of the so-called rotating-element ellipsometers and the application computers for solving the complicated complex equations. Because these polarimeter-type ellipsometers measure continuously (each date is measured in a short, fixed

LIGHT

SOURCE elliptically polariz POLARIZER

ANALYZER SAMPLE

FIG. 1. Principle of a rotating analyzer ellipsometer. Riedling (1988). Ellipsometry for Industrial Application. Springer-Verlag, Wien, New York.

3

1 ELLIPSOMETRIC ANALYSIS

period), wavelength scanning, that is, spectroscopic ellipsometry (SE), can be performed.

11.

PRINCIPLE OF

ELLIPSOMETRY

In the following, we outline some results of electromagnetic theory as background to help in understanding of different techniques used to determine optical properties of materials. An incident plane wave propagating in the z-direction of a local orthogonal coordinate system can be represented as (Azzam and Bashara, 1977)

C

where Ex and E, are the complex jield coeficients describing the amplitude and phase dependencies of the projections of Ei(r, t ) along the x and y axes. The complex dielectric constant E = - i E 2 is connected to the complex refractive index N = n - ik through E = N 2 , with E , = n2 - kZ and c2 = 2nk. For absorbing media, k and c2 are positive. If the electromagnetic wave is reflected by a smooth surface, the outgoing wave can be represented in the absence of anisotropic effects in another local coordinate system as

+

(%-,Ex $r,E,) exp

C

where in both local coordinate systems the x axes are in the plane of incidence, the y axis is perpendicular to the plane of incidence, and the z and z’ axes define the plane of incidence. In this approximation, the effect of the surface is described by two coefficients rp and r,. These complex rejection coeficients describe the influence of the sample on the field components parallel (p) and perpendicular (s) to the plane of incidence. Thus four quantities -two amplitudes and two phases -are necessary to characterize the incoming wave completely, and four more are necessary to describe the sample. Consequently, the properties of the sample are obtainable if the properties of both incident and reflected waves are known. ReJectometry and ellipsometry are two techniques for obtaining this information. Neither extracts all the available data; however, each is an incomplete form of a still more general technique, called polarimetry (Hauge, 1980). Using reflectometry, ellipsometry, or polarimetry, one obtains different

4

M. FRIED,T.LOHNERAND J. GYULAI

types of information. Therefore, it is useful to rewrite the four parameters describing the plane waves of Eqs. (1) and (2) into four new parameters by means of the concept of the polarization ellipse. The polarization ellipse is the locus traced out in time by the endpoint of the electric field vector E(r, t ) at any fixed plane z = zi. The polarization ellipse has two attributes that can is a be termed amplitude and shape. The amplitude E = (lEXI2 (Ey12)1/2 scalar whose square is proportional to the energy density of the wave or intensity I. The intensity is the only quantity of interest in a reflectance or transmittance measurement and one of the quantities of interest in a polarimetric measurement. The shape is an intensity-independent, complex quantity, and a possible representation is the polarization state, x = E x / E y . Either intensities or polarization states are measured both for the incident and reflected beams, and sample properties can be determined by taking appropriate ratios. In reflectometry, we distinguish between p- and spolarized light. By Eqs. (1) and (2), the intensity ratios are

+

where the indexes i and r stand for the incident and reflected beam, respectively. The rejectances R , and Rs are the absolute squares of the respective complex reflectance coefficients. At normal incidence, both polarizations are equivalent and R, = R,. In ellipsometry, the ratio of polarization states yields a somewhat different perspective of the sample. By Eqs. (1) and (21,

The complex reflectance ratio p is conveniently expressed as an amplitude ratio tan I(l and a phase difference A. Depolarization or cross-polarization effects determined by polarimetry are important in more complex sample structures and in the presence of scattering (Hauge, 1980), topics that will not be discussed here.

111. General Remarks

After this introduction, one can now make some general remarks about the relative advantages and disadvantages of reflectometry, ellipsometry, and polarimetry (Aspnes, 1985). For isotropic samples, ellipsometry is strictly a nonnormal incidence technique, whereas reflectometry and

1 ELLIPSOMETRIC ANALYSIS

5

polarimetry can be performed at either normal or oblique incidence. Reflectometry deals with intensities and therefore is a power measurement, whereas ellipsometry deals with intensity-independent complex quantities and therefore is more nearly analogous to an impedance measurement. Because of the more complicated quantities involved, an ellipsometer is more a complicated piece of equipment than is a reflectometer, whereas a polarimeter is more complicated than is an ellipsometer. A disadvantage of ellipsometers and polarimeters is that transmission-mode optical elements (e.g., polarizers) are generally required. Therefore, the availability of goodquality transmitting elements poses limitations on the wavelength ranges that can be applied. Ellipsometry undoubtedly is very powerful. First, in any single-measurement operation, two independent parameters are determined simultaneously. For example, both real and imaginary parts of the complex dielectric function E of a homogeneous material also can be obtained together at every wavelength without Kramers-Kronig analysis. Two independent parameters also put tighter constraints on models describing more complicated structures, for example, laminar microstructures. In nonnormal incidence reflectance measurements (with polarized light) two independent parameters R, and R, are also available; however, these parameters can be obtained separately only after additional experimental efforts. Second, intensity fluctuations influence ellipsometric measurements only slightly. Similar insensitivity is experienced on temperature drifts of electronic components and on macroscopic roughness. This latter causes light loss by scattering the incident radiation out of the optical path, which can be a serious problem in reflectometry but not in ellipsometry. Third, accurate reflectometric measurements generally require doublebeam methods. In essence, ellipsometry is a double-beam method in which one polarization component serves as amplitude and phase reference for the other. Finally, p explicitly contains phase information, thus in general, ellipsometry is more sensitive to surface conditions. It is a misleading argument that insensitivity to surface conditions is an advantage of reflectometry, because insensitivity makes it difficult to obtain accurate values of the intrinsic dielectric responses of bulk materials. Ellipsometry can also suffer from artefacts at the surface; however, artefacts occur at the level of the measurment and can be corrected on the spot. In reflectometric measurements, however, artefacts manifest themselves only at the datareduction level. Because reflectometry is discussed in detail elsewhere in this volume (Chapter 2), we now restrict our discussion to ellipsometry. Earlier, the most important instrument was the classic null ellipsometer (Fig. 2). Here, the entrance optics consist of a polarizer and compensator, or quarter-wave plate, and the exit optics consist of a second polarizer, or analyzer. The

6

M. FRIED,T.LOHNERAND J. GYULAI I ANAI Y 7FR

DET

(PCSA) null ellipsometer. FIG.2. Principle of a polarizer-compensator-sample-analyzer Riedling (1988). Ellipsometry for Industrial Application. Springer-Verlag, Wien, New York.

polarizer-compensator combination operates as a general elliptical polarizer. To perform a measurement, the ellipticity is adjusted so that it is exactly canceled by reflection. The reflected beam can then be extinguished by properly rotating the analyzer. The null ellipsometer is the optical analogue of the ac impedance bridge. A null ellipsometer is mechanically simple and still of high accuracy. Above a certain light intensity, the accuracy is determined solely by the mechanical components and their alignment. Detector nonlinearity is not a problem, because it serves only for minimum detection. However, disadvantages have made this device less popular in recent years. For example, during setting the null condition, polarizer and analyzer azimuth settings influence each other. Furthermore, the intensity near zero varies quadratically on both azimuths, and therefore the zero is not well defined. The detector operates at minimum light levels and dark-current and shot noise may be dominant in the signal. The usual way to overcome these problems has been to increase the source intensity. Consequently, most null ellipsometric measurements have been conducted at the mercury green line A = 546.1 nm or, more recently, at the helium-neon (He-Ne) laser line I = 632.8 nm. Therefore, the use of a compensator, a component with a relatively narrow useful wavelength range, has not turned out to be the serious drawback. The other type of ellipsometer is the widely used automatic photometric ellipsometer. Photometric instruments function on entirely different principles than do null ellipsometers. One of the components is a modulator that is used to introduce a time-dependence on the transmitted intensity that is later decoded for the properties of interest. Because most of the light is transmitted, photometric systems can be equipped with relatively weak continuum sources suitable for infrared-visible-near-ultraviolet spectroscopy. In the first design (Cahan and Spanier, 1969), the modulator consisted simply of the final analyzer prism, which was rotated mechanically

1 ELLIPSOMETRIC ANALYSIS

7

to produce a sinusoidal variation of the transmitted intensity (Fig. 1). In addition, the data were obtained automatically, without operator feedback. Other configurations (Muller, 1976; Azzam and Bashara, 1977; Rzhanov and Svitashev, 1979) have also been developed involving photoelastic modulators (Jasperson and Schnatterly, 1969; Bermudez and Ritz, 1978; Jellison and Modine, 1982; Drevillon et al., 1982) and Pockels cells (Billings, 1952) as well as mechanically rotating elements (Hauge and Dill, 1973; Aspnes and Studna, 1975a, 1975b) and also featuring digital encoding and computer analysis of the transmitted intensity. The more elaborate designs (Hauge, 1980; Aspnes and Studna, 1975a) also can distinguish between polarized and unpolarized light. Photometric ellipsometers now totally dominate the field as far as research and spectroscopic applications are concerned, with null ellipsometers now mainly used to perform routine industrial measurements on simple systems, such as determining oxide and nitride thicknesses on semiconductor wafers. Only a few of the many photometric designs have been considered sufficiently practical. The simplest of these are the rotating-analyzer ellipsometer (RAE) (Hauge and Dill, 1973; Aspnes and Studna, 1975a, 1975b) and its complement, the rotating-polarizer ellipsometer (RPE) (Theeten et al., 1980). In these configurations, the entrance and exit optics consist of single polarizing elements. As the names imply, one of these is rotated mechanically while the other is held fixed. The advantages are simplicity and the absence of a compensator; the only wavelength-dependent element is the sample itself. A second advantage is that the transmitted intensity has a very simple Fourier spectrum consisting of a single ac component on a dc background (Fig. 1). Disadvantages include the requirement of either a rigorously unpolarized source for an RPE or a rigorously polarizationinsensitive detector for an RAE. In addition, as with all photometric systems, the detector must be rigorously linear. Finally, these configurations cannot distinguish between circularly polarized or unpolarized light or between the right and left handedness of circularly polarized light, and a significant loss of accuracy occurs for E~ if c2 0.

IV. Optical Models p or 1+5 and A (the ellipsometric angles) only describe the reflection properties of the surface in question. Alone they give no information about the structure or even the composition of the sample. To obtain this information some optical model of the surface must be assumed. Usually enough information about the sample history is available to construct an

8

M. FRIED,T.LOHNERAND J. GYULAI

idealized optical model of it. Most optical models use flat semi-infinite substrates with one or more laminar adherent layers of uniform thickness on the surface. All interfaces are assumed to be sharp, and all layers are assumed to be composed of optically isotropic materials. Vertical inhomogeneities or interface layers can easily be accounted for if II/ and A are calculated from known sample parameters, that is, we can approximate by a stack of sufficiently thin homogeneous layers with suitably graded refractive indices. After a layer model is selected, the ellipsometric equation (Eq. (4b)) describing the ellipsometricproperties of the model may be derived from the Fresnel equations (Azzam and Bashara, 1977). The Fresnel equations are derived directly from the Maxwell equations. They describe reflection and transmission at the interface between two optically isotropic materials.The ellipsometric equation has dependencies in the complex refractive indices and film thicknesses in the model. Generally, the ellipsometric measurements are not evaluated directly. The interpretation is based on a simulation (calculation from the model) and regression program, which minimizes the difference between calculated and measured data. A complicated multilayer structure is modeled as a system built up of plane parallel thin films, each of which can consist of a mixture of two (or more) components. Thus the quality of interpretation depends on the realism of the proposed optical model, on the quality of the reference files for the refractive indices of the different components, on the theory that is used to describe the optical activity of mixed layers (see Part 1.7, Effective-Medium Theory) and last, but not least, on the stability and convergence of the regression algorithm. There are several possibilities to increase the amount of independent information in ellipsometry (e.g., multiple angles of incidence, different ambients, and different layer thicknesses of the same material). The most important is the multiwavelength or SE. However, SE is not an analytical technique that can stand on its own. When using SE, one needs as much information as possible before starting the interpretation of the spectra. SE can be applied in the following cases: Determination of the refractive index as function of the wavelength for mirror polished substrates. The calculated n and k values can be used as reference for the evaluation of measurement of more complex structures Accurate determination of n versus the wavelength and the thickness of a transparent film on a known substrate Computation of N = n - ik versus the wavelength for a layer of known thickness

9

1 ELLIPSOMETRIC ANALYSIS

Comparison of simulated and measured parameters of ideal multilayers based on a proposed structural model (incident angle, layer thicknesses, and materials) Computation of the multilayer thicknesses through linear regression by minimizing the differences between the calculated and recorded spectra. Problem: reference files for refractive indices and the limited number of layers (up to 10) Introduction and determination of composition of mixed layers, to some extent, by the application of effective medium theories Study of surface and interface roughness, both of which can be replaced by a mixed layer Determination of profiles in layers with a continuously varying refractive index, for example, ion implantation damage and deposited layers The performance of the used optical model and the fit procedure can be judged by the normalized mean-squared deviation (unbiased estimator) and the cross-correlation matrix [Azzam and Bashara, 19771. Use of independent cross-checking methods -such as cross-section transmission electron microscopy (XTEM) or Rutherford backscattering spectrometry (RBS) -is also recommended when the optical model is used for the first time for a new type of sample. Examples of sophisticated optical models are presented in the latter part of this chapter.

V. The Complex Dielectric Function It is useful to begin with the ideal case in which the substrate is homogeneous and isotropic and the surface is mathematically sharp. Then E and the normal incidence reflectance R are given by the two-phase (substrate-ambient) model [Azzam and Bashara, 19771: E/E, =

sin24 + sin24tan24[(1 - p ) / ( l

+ p)]’

and na = &:I2 are the where is the angle of incidence, and where n = complex and ordinary indices of refraction of substrate and ambient, respectively. (We assume that the ambient is transparent.) However, in the analysis of ellipsometric data, it is often useful to assume perfection anyway and to convert the measured quantity p into a derived quantity, the pseudo-dielectric function ( E ) , by Eq. (5a). The pseudo-dielectric function

10

M.FRIED,T.LOHNERAND J. GYULAI

necessarily is an average of the dielectric responses of substrate and overlayer, and the accuracy by which it actually represents E depends on the accuracy by which the sample configuration approximates that of the ideal two-phase model. The pseudo-dielectric function is useful simply because it expresses p in a form related to the fundamental quantity of interest. The complex reflectance ratio can always be recovered from ( E ) by inverting Eq. (5a). There are methods to consider the effect of overlayers getting the bulk optical properties of the sample (see, e.g., Aspnes, 1985). A common way to discuss the optical properties of metals and semiconductors is to describe the dielectric function in terms of the electronic band structure of a crystal and to distinguish between bound and free electrons (or free carriers). One can assume that both additively contribute to the E = - k2. Unlike metals, the optical properties of semiconductors are governed by the bound electrons, because the number of free carriers is several orders of magnitude less. Even in the case of heavily doped (3 x lOZ0/cm3)n-type crystalline silicon, Aspnes, Studna, and Kinsbron at 1.5 eV photon energy. (1984) measured a maximum 5% difference in Therefore, above 1.5 eV (in the visible-near-ultraviolet), one can ignore the influence of free carriers for intrinsic silicon. We may therefore consider interband transitions. The values of and cZ can be related by means of the Kramers-Kronig relations, and therefore, it is sufficient to determine eZ(w), for theoretic approach. However, in this case, one must know c2(w) for all the values of w in order to calculate E ~ ( w ) . The dielectric function E ~ ( W for ) a crystalline semiconductor is given by

is the momentum matrix element, c and v denote the conducwhere Pcu(k) tion and valence bands, respectively, E,(k)- E,(k) is the energy separation between the conduction and valence bands, and the integration is performed over the first Brillouin zone. If defects are introduced in the lattice, then new states (but localized ones) have occurred in the original bandgap. There is a lack of a theoretic approach to understanding the optical effects of point defects or defect clusters. (Experimental works use a mixture of amorphous and crystalline materials to interpret measurements on damaged semiconductors.) Most of the work (Tauc, Grigorovici, and Vancu, 1966; Davis and Mott, 1970) deals with amorphous semiconductors, where long-range order is absent but short-range order still exists, meaning that the number of neighboring atoms is the same as in the crystalline structure. Even in this

1 ELLIPSOMETRIC ANALYSIS

11

case the valence and conduction bands retain their meaning; however, but the wave vector loses its meaning. Assuming that the basic volume V contains the same number of atoms in the amorphous as in the crystalline state, and that P,,(k) is independent of k, we obtain

where g, and gv are the densities of states in the conduction and valence bands, respectively. It is apparent that c2(o)is determined by a convolution of densities of states in the conduction and valence bands g, and go for which energy is conserved. Tauc, Grigorovici, and Vancu used parabolic bands, whereas Davis and Mott (1970) used a power function with a suitably chosen exponent for E . The chosen value of this exponent is supported by appropriate plotting of c2(E) against E . Using this type of plot, we can determine the optical energy gap. However, we must note that there is no single type of amorphous silicon (a-Si). Several measurements exist that indicate the properties of implanted a-Si (i-a-Si) differ significantly from those of well-relaxed (annealed) a-Si (r-a-Si) (Fredrickson et al., 1982; Waddell et al., 1984; Fried et af., 1986; Roorda et al., 1991). An addition, we must know the optical constants of different kinds of a-Si to evaluate the optical measurements properly. Earlier measurements showed that the E of i-a-Si and r-a-Si differ from the E of evaporated a-Si (e-a-Si) found in handbooks (e.g., Palik, 1985) as E for a-Si (Fig. 3). If we use this E (e-a-Si) to evaluate SE measurements of ion-implanted silicon, then the fit is very poor. A similar finding is described by McMarr (McMarr, Vedam, and Narayan, 1986; Vedam, McMarr, and Narayan, 1985), who also measured self-implanted fully amorphous silicon and tried to evaluate the spectra modeling of the sample as a mixture of voids and a-Si prepared by low-pressure chemical vapor deposition (LPCVD). The model calculations resulted in a surprising negative (-9%) void fraction. This fact also indicates that the E of LPCVD a-Si must not be used for i-a-Si (Fig. 3). Roorda et al. (1991) investigated different kinds of a-Si. These authors characterized i-a-Si, r-a-Si, ion implanted (re-implanted) well-relaxed amorphous Si (i-r-a-Si) and ion-implanted (with the same ion and dose) crystalline Si (i-c-Si) by widely different techniques, such as calorimetry, Raman spectroscopy, atomic density, and impurity diffusion measurements (Roorda, 1991, and references therein). The authors concluded that the change (structural relaxation) during thermal treatment in i-a-Si is mediated by annihilation of network (“lattice”) defects. This picture (in terms of

12

M. FRIED,T. LOHNERAND J. GYULAI Wavelength [nm] 750

~~

1.5

750

2.0 600

300

450

600

~

2.5

450

3.0

3.5

4.0

300

4.5

Energy [eV]

FIG. 3. Complex dielectric functions of crystalline silicon (c-Si) and different types of amorphous silicon (a-Si). LPCVD, low-pressure chemical vapor deposition; c-Si, crystalline silicon.

annihilation of point defects) is well established by the previously mentioned comparative experiments of i-a-Si, i-r-a-Si, and i-c-Si. The results of our SE measurements support this explanation; that is, the optical change (Fig. 3) can be attributed to the annihilation of point defects. The i-a-Si is the most defected material, and therefore it has the largest number of states in the bandgap. This fact explains why E, is the lowest and the absorption is the highest below 3 eV in the case of i-a-Si. Recently, Adachi (1991) published a relatively simple model that claims to describe at the same time E ~ ( wand ) E ~ ( win) amorphous semiconductors. He assumed parabolic densities of states in Eq. (fib), obtaining a power . a high-energy cut-off, he could obtain E ~ ( w ) function for E ~ ( w )Assuming

13

1 ELLIPSOMETRIC ANALYSIS

using the Kramers-Kronig relations. Introducing phenomenologically a damping (or broadening) effect, E(O) became structureless, which is characteristic of amorphous semiconductors. Fried and van Silfhout (1994) applied Adachi's expression to fit the complex dielectric function of the self-implanted and implanted-annealed (or relaxed) a-Si. The fit resulted in changing optical bandgap (AEH= 0.21 eV) and damping (or broadening, Ar = 0.24 eV) energy. Both changes can be interpreted by the presence of about 1.5% more structural disorder point defects in implanted a-Si compared with relaxed a-Si.

VI. Light Penetration The optical penetration depth (OPD) is defined as A/4zk. The information depth can be defined roughly as 3 times the OPD. The fact that OPD is wavelength-dependent can be exploited as a depth-scan in SE. In the following, the OPD (in nm) is tabulated in the case of different crystalline semiconductors and one kind of a-Si. We must note that in the case of damaged or amorphous materials, the OPD can be less by even one or two orders of magnitude for the longer wavelengths -for example, compare Si and (implanted) a-Si (Table I).

VII. Effective Medium Theory

It is often necessary to determine the optical properties of heterogeneous or composite systems that are more accurately described as mixtures of TABLE I OPTICAL PENETRATION DEPTH I N DIFFERENT CRYSTALLINE SEMICONDUCTORS AND ONEKIND OF AMORPHOUS SILICON Wavelength (nm) 210 300 400 600 800

Si 5.6 5.1

82.2 176.8 10,Ooo

i-a-Si

Ge

GaAs

InP

5.6 1.2 12.1 50.3 212.2

5.9

6.5 11.9 14.8 206.7 744.6

7.1 13.6 18.3 141.7 303.2

6.4 14.4 34.2 196.5

Si, silicon; i-a-Si, implanted amorphous Si; Ge, germanium; GaAs, gallium arsenide; InP, indium phosphide.

M. FRIED,T.LOHNERAND J. GWLAI

14

separate regions of two or more materials, each of which retains its own dielectric identity. This is the objective of effective-medium theory. Examples of composite materials include metal films, which can be described as a heterogeneous mixture of materials and voids owing to the inability of forming grain boundaries in closely packed systems without some loss of material. Other examples include polycrystalline films, amorphous materials, and glasses; in the two latter cases, the inhomogeneity is essentially on the atomic scale and concerns the number of polarizable elements (bonds) per unit volume. A microscopically rough surface can also be considered as a heterogeneous medium, being a mixture of bulk and ambient on a microscopic scale. In this discussion, we assume that the characteristic dimensions of the microstructure are large enough (> 1 or 2 nm) so that the individual regions essentially retain their bulk dielectric responses, but small (< 0.1 or 0.21) compared with the wavelength of light. Then, the macroscopic E and H fields of the Maxwell equations will not vary appreciably over any single region, and quasistatic theory can be used. This avoids complications due to scattering and retardation effects that are dominant in macroscopically inhomogeneous systems. The optical properties of heterogeneous systems are calculated in the same way that macroscopic quantities or observables are calculated in thermodynamics or quantum mechanics. First, the microscopic problem is solved exactly; then the microscopic quantities are averaged to obtain their macroscopic counterparts. The dielectric function is obtained from the macroscopic average electric field E and polarization P according to

D = EE= E + 4nP,

(74

where Axi is the displacement of the charge qi under the action of the local field at q i . It is the appearance of the volume normalizing factor in Eq. (7b) that is responsible for the sensitivity of E to density. Many expressions, some quite elaborate, have been proposed to describe E in terms of the microstructural parameters and the dielectric functions of the constituents (e.g., Hunderi, 1980 and references therein). The simplest general representation of multiphase effective-medium theories is

where E,,

&b

and f, and

fb

are the dielectric functions and volume fractions

1 ELLIPSOMETRIC ANALYSIS

15

of the phase a and b, respectively, and s (= l/z - 1, 0 < z < 1) is the screening parameter; z = 0 means no screening (linear average), whereas z = 1 means maximum screening (reciprocal average). The quantity ch is a host dielectric function that is assigned different values according to the model. To obtain (for two-phase mixtures) the Maxwell-Garnett expressions (Maxwell-Garnett, 1904, 1906), a or b is considered the host medium, therefore &h = E, or E ~ and , one of the terms on the right-hand side of Eq. (8) vanishes. The Bruggeman effective-medium approximation (EMA) (Bruggeman, 1935) is obtained by making the self-consistent choice E,, = E , in which case the left-hand side vanishes. The Lorentz-Lorenz expression (Lorenz, 1880; Lorentz, 1916) is obtained by choosing E,, = 1, that is, empty space. The choice z = 1/3 applies to spherical inclusions appropriate to a heterogeneous system that is macroscopically isotropic in three dimensions. The various effective-medium theories actually differ only in the choice of the host material; however, this choice does imply different microstructures. Thus the Maxwell-Garnett theories, with one medium completely surrounded by another, are appropriate to the cermet or coated sphere microstructures, whereas the Bruggeman theory is appropriate to random or aggregate configurations. The aggregate microstructure can also be considered as an average over grain shapes, and probably more accurately describes most thin films (Aspnes, 1981).

VIII. Examples 1. GALLIUM ARSENIDE The implantation-induced damage and the characteristics of furnace and laser annealing have been studied by single-wavelength (A = 632.8 nm) ellipsometry (Kim and Park 1980). A two-phase optical model was applied for data evaluation. The refractive index and extinction coefficient increase with increasing implant (magnesium (Mg), argon (Ar)) fluence up to l O I 5 ions/cm2. The depth-dependence of the change in the complex index of refraction following removal of a series of thin implanted layers has been obtained as function of ion fluence, flux, and annealing conditions. For high doses, the profiles were found to be in agreement with damage theory. The damage profiles for low doses exhibited a combined effect and could not be adequately described by the damage theory. The regrowth of the damaged crystal by a Q-switched ruby laser is partially achieved only in the outer layer because the optical penetration of the ruby laser radiation is not deep enough for annealing the full implant depth.

M. FRIED,T. LOHNER AND J. GYULAI

16 Oa5

mOaO

a-GaAs [O/OI

l

o

o

p

5 0 t ~ OO

0.0

0.5

1.0

Energy [eVl

1.5

Depth [nm]

( b)

(a) FIG. 4. Infrared spectroscopic ellipsometry measurements for a (100 keV, 10'' ions/cm*) argon ion (Ar +) implantation (a), and the corresponding calculated depth profile (b). a-GaAs amorphous gallium arsenide. Erman, M., and Theeten, J. B. (1983). Analysis of Ion-Implanted GaAs by Spectroscopic Ellipsometry. Surface Sci. 135, 353-373.

Erman and Theeten (1983) demonstrated the use of spectroscopic ellipsometry (0.5 to 5.5eV) for the assessment of implanted layers into semiinsulating gallium arsenide (GaAs) (Figs. 4 and 5). A quantitative evaluation of the amount of lattice damage present in the as-implanted layers, as well as at various stages of thermal annealing, is obtained from detailed analysis of the peak structure in the ellipsometry data around 3 eV (Fig. 5). Using the entire spectrum-0.5 to 5SeV-a multilayer analysis of the as-implanted layers is performed, which gives a depth profile of the lattice damages (a stack of 10 layers is used as a model). These results have been checked by combining spectroscopic ellipsometry with chemical etch. Damage depth profiles and carrier concentration depth profiles have been obtained for ion implantations with an average penetration on the order of 0.1 pm. The annealing of the lattice damage induced by 70 keV beryllium (Be) ion implantation in GaAs has been studied by Chambon et al. (1984) using SE and Raman scattering after each step of isochronal thermal treatment.

2. GERMANIUM Aspnes and Studna (1980a) examined ion-bombarded and annealed (1 11) germanium (Ge) surfaces by SE over the energy range of from 1.5 to 6.0 eV using an RAE. The experiments were conducted in ultrahigh vacuum

a

b B+implanted 10keV 10" ions/cmZ

C

Se+implanted 5OkeV 10" ionslcmz

FIG. 5. Dielectric functions of crystalline gallium arsenide (c-GaAs) (full line) and amorphous gallium arsenide (a-GaAs) (dotted line) (a), derivative of cos A versus energy in the 3 eV region as a function of thermal annealing treatments for implanted GaAs (full line) and c-GaAs reference (dashed line) for boron ion (B') implant (b), and the same for selenium ion (Se') implant (c). Erman, M., and Theeten, J. B. (1983). Analysis of Ion-Implanted GaAs by Spectroscopic Ellipsometry. Surface Sci. 135, 353-373.

18

M. FRIED,T.LOHNERAND J. GYULAI

(UHV) and the samples were characterized by LEED and Auger electron spectroscopy. The energy of the neon or argon beam was adjustable from 100 to 3000 eV. Highly reproducible ellipsometric spectra were obtained after the bombardment and annealing process. Accurate dielectric function spectra are retrieved for crystalline Ge and for the amorphized overlayer formed by ion bombardment. The amorphized overlayer is shown to be 6 & 2% denser than is the UHV evaporated material, independent of bombarding species. Ten minutes of annealing at 620°C was required to obtain a good reconstruction. The dielectric function of the partially annealed overlayer is found to be in good agreement with that of amorphous germanium (a-Ge) film prepared by UHV evaporation. Benourhazi and Ponpon (1992) performed investigations to establish the conditions necessary to produce shallow donor-type layers in Ge. Germanium wafers of (1 11) p-type were implanted with 20-keV phosphorus (P) ions and 30 keV As ions at doses ranging from 10’’ to l O I 5 ions/cm2. The induced damage and its annealing on heat treatment in uacuo have been studied by means of single-wavelength (A = 632.8 nm) ellipsometry and Rutherford backscattering spectrometry (RBS). Benourhazi and Ponpon applied the ambient-substrate model and evaluated the pseudo-dielectric constant from the angles Y and A. The evolution of the extinction coefficient k was used to characterize the crystalline defects as a function of dose and annealing temperature. Depending on the initial damage level, most of the crystalline defects anneal out at temperatures ranging from 200 to 400°C. However, these authors found that residual defects remain up to about 450”C, independent of the implanted dose. 3. INDIUMPHOSPHIDE Damage removal of Se+-implanted indium phosphide (InP) after cw argon laser annealing has been studied by Mizuta and Merz (1983) using single-wavelength (A = 546.1 nm) ellipsometry. A gradual decrease in the extinction coefficient k as a function of laser power is observed for roomtemperature ion-implanted samples, compared with a sharp decrease in k for ion-implanted Si. Sulfur-implanted InP samples were annealed with a cw neodymium: yttrium-aluminium-garnet (Nd: YAG) laser and investigated using ellipsometry (1 = 632.8nm) and Hall measurements by Fremunt et al. (1984). Before laser annealing, the implanted samples were encapsulated by either a PSG (polysilicon glass) film or by a double layer of PSG and silicon dioxide (SiO,). Recovery of the refractive index and extinction coefficient indicated that a double layer is more suitable for capping than is a single layer of PSG film.

1

ELLIPSOMETRIC ANALYSIS

19

4. SILICON

The first paper to deal with the annealing of ion bombardment defects in a semiconductor using ellipsometry was probably that published by Ibrahim and Bashara (1972). These authors studied the surface damage of Si caused by low-energy (150 to 400 eV) Ar ions. Annealing was performed at 500 and 800°C. The presence of an oxide film on Si was accounted for in evaluation of the data from multiple-angle-of-incidence (MAI) single-wavelength (A = 632.8 nm) ellipsometry. Ibrahim and Bashara modeled the damage optically in two ways. In the first model the effect of damage was assumed to modify the substrate uniformly throughout, in the second, giving more information, the damage was supposed to be uniformly confined to a damage layer on top of a single-crystalline substrate. Similar investigations were made using in-situ spectroscopic ellipsometry by Holtslag (1986). Low-energy (0.5 to 2.5 keV) ion bombardment of the noble gases helium, neon argon, and krypton on a clean Si surface were carried out in UHV. During bombardment and later annealing (up to 8 W C ) in situ SE was used to characterize the different processes. Comparing the results with other surface-analysis techniques (Auger electron spectroscopy (AES), RBS, TEM, surface profiling) information was obtained about low-dose and high-dose behavior. Kucirkova (1976) investigated the effect of isochronal annealing in the case of tellurium (Te)-implanted Si. She emphasized that the information obtained on lattice disorder is to a great extent comparable with that of the backscattering technique. An attempt to approximate the implanted and annealed region by one homogeneous absorbing layer has given reasonable values for doses exceeding that of complete amorphization. Decreasing layer thickness values and extinction coefficients were evaluated as a function of increasing annealing temperature. The effect of laser annealing at 530nm was studied by Aspnes et al. (1980b) as a function of the power density for Si implanted with 30 keV Ar ions. Dielectric function spectra were obtained from 1.5 to 6.0 eV using SE. The data were calculated in the two-phase (substrate-ambient) model, which does not take into account the presence of possible surface films. Polycrystalline films resulted at 0.6 J/cm2, whereas TEM showed perfect epitaxy at higher annealing powers. Changes of the amplitude and energy position in the critical point structure were observed in dielectric function data for the epitaxial samples. These changes were due to high concentrations of impurities in the surface region, as was shown by RBS. Annealing of As-implanted Si with a cw carbon dioxide (CO,) laser was investigated by Takai et al. (1980) in comparison with thermal annealing. Ion channeling, single-wavelength ( A = 546.1 nm) ellipsometry, and Hall effect measurements were performed. Before the ellipsometric measurements,

20

M. FRIED,T. LOHNERAND J. GYULAI

the native oxide was etched and the pseudo-dielectric constant was evaluated from the measured data. It was found that the lattice disorder produced during implantation can be completely annealed out by laser annealing with a power density of 500 W/cmz. In the case of high-dose implantation of As ions, a damaged layer located at the interface region between implanted and underlying bulk layers exists after annealing, as was determined by RBS. Optical measurements are not sensitive enough to such defects. Single-wavelength ( A = 546.1 nm) ellipsometry was applied to estimate damage in n-type (100) Si induced by low-dose 75-keV boron (B) implantation (Watanabe et al., 1979). A parameter characterizing the crystal damage is proposed that contains the extinction coefficient. Crystal damage caused by ion implantation at doses as low as 3 x 10" ions/cm2 and the annealing effect on crystal damage of heat treatment in dry nitrogen ambient were detected. Single-wavelength ( A = 546.1 nm) ellipsometry and sheet resistance measurements were used by Nakamura, Gotoh, and Kamoshida (1979) to investigate damage in 1013 to 10l6 ions/cm2 phosphous (P)-ion-implanted Si and recovery of crystallinity by subsequent annealing. The substrate temperature during implantation was set at liquid nitrogen temperature, room temperature, or 250°C. Isochronal annealing was applied in nitrogen ambient at temperatures ranging from 350 to lOOO"C, at 50°C increments for 30 min. An effective complex refractive index was calculated, assuming a 2-nm-thick native SiO,. The refractive index increases as a function of implant dose, except when the dose exceeds l O I 5 ions/cm2. The extinction coefficient exhibits a steady increase with increasing dose and decreasing implant temperature. For the sample implanted over the critical dose, optical constants and sheet resistances recover after annealing at around 5OO0C, indicating the epitaxial regrowth of implantation-induced amorphous layers. The annealing processes of high-dose P+-ion-implanted Si were studied by Watanabe et al. (1980) using single-wavelength ( A = 546.1 nm) ellipsometry. After implantation, the wafers were treated in oxygen plasma to remove surface contaminations that might deposit during implantation. Isothermal annealing was carried out in dry nitrogen ambient. The change in ellipsometric angle (Y,A) during isothermal annealing is measured and compared with that of values calculated using a simple epitaxial regrowth model based on the distribution of the optical constants (Fig. 6). The discrepancy between the experimental and the calculated changes in angle shows the imperfect results of the annealing process at 500 to 550°C. TEM observation confirmed that many crystal microdefects existed in the samples. Solid-phase epitaxial regrowth in ion-implanted Si was studied by Fried et a!. (1986) using RBS and single-wavelength ( A = 632.8 nm) MA1 ellip-

21

1 ELLIPSOMETRIC ANALYSIS

150b,

10

,

I

,

I

,

I

I

,

,

I

,

20

15 [deg I

FIG. 6. Change in ellipsometric angle (Y, A) during isothermal annealing. The dotted curve is based on calculation using a simple epitaxial regrowth model assuming that N = 4.650 - 0.9881. N,.si, amorphous silicon; ai, angle of incidence = 70"; P, phosphorus. Watanabe, K., Motooka, T., Hashimoto, N., and Tokuyama, T. (1980). Ellipsometric Study of Annealing Processes of Phosphorus-Ion-ImplantedLayers of Si. Appl. Phys. Lett. 36,451-453.

sometry. To establish an appropriate optical model and to check the thickness data obtained from ellipsometry, high-depth-resolution backscattering spectrometry was used. A four-phase optical model (air, native oxide, a-Si layer, single crystal bulk) was chosen to describe the as-implanted (N = 4.63 - i0.76) and solid-phase epitaxial regrowth samples (Fig. 7). The so-called thermally stabilized state appears during the recrystallization process after a certain annealing time at a given temperature. The complex refractive index of the thermally stabilized phase was determined; its value was N = 4.55 - i0.35 for the wavelength of the He-Ne laser (632.8 nm). The thickness data of amorphous layers obtained by ellipsometry were found to be in good agreement with the values deduced from backscattering spectrometry.

M. FRIED,T. LOHNER AND J. GYULAI

22

REFRACTIVE INDICES

-

%40

\

/

150' 1

b

FIG.7. Ellipsometric measurements on as-implanted (X,Y and 0)reference and annealed (other symbols) samples. Calculated curves are simulated using the four-phase optical model with corresponding refractive indices. Fried, M., Lohner, T., Vizkelethy, G., Jaroli, E., Mezey, G., and Gyulai, J. (1986). Investigation of Solid Phase Epitaxial Regrowth on Ion-Implanted Silicon by Backscattering Spectrometry and Ellipsometry. Nucl. Instrum. Meth. Phys. Res. B15, 422-424.

Similar measurements also were made by multiwavelength and spectroscopic ellipsometry, (Fried et af., 1987, 1992b). The results, that is, the complex dielectric functions of different kinds of a-Si, are plotted in Figure 3. Using a high-speed ellipsometer, Moritani and Hamaguchi (1985) determined growth rates and growth temperatures during solid-phase epitaxial

23

1 ELLIPSOMETRIC ANALYSIS

29

-

0

28

aJ

?2

+

a

27 26

25 24

168

172

176

1 1

b

FIG. 8. Time-resolved ellipsometric data during argon laser annealing and locus that gives the best fit (using the four-phase model) to the data points (Na.si= 4.80 - i0.68, N,,i = 4.20 - i0.16 and doxidc = 2 nm; the grid is constructed with derideand the temperaturedependence of Nc.si)(a), and depth profiles of growth rate (derived from (a)) and arsenic (As) concentration (b). Moritani, A,, and Hamaguchi, C. (1985). High-speed Ellipsometry of Arsenic-Implanted Si During cw Laser Annealing. Appl. Phys. Lett. 46, 746-748.

24

M. FRIED,T. LOHNERAND J. GYULAI

regrowth (under a cw Ar ion laser annealing) of As-implanted Si applying a four-phase structure model (air, native oxide, a-Si layer, single crystal bulk). With an 8-psec data acquisition time, they could determine several pm/sec growth rates (Fig. S(a)). They observed a good correlation between the depth profiles of growth rate and As concentration (Fig. 8(b)) even at such a high temperature as 1100 K. The recrystallization kinetics of boron fluoride ion-implanted Si have been studied by Holgado et al. (1995) using SE. It has been found that the dielectric constant of the implanted layers depends on the dose and the ion beam current. In order to determine the thickness of the amorphous layer, a four-phase optical model (air, native oxide, implantation-dmaged Si layer, single crystal bulk) was applied. An activation energy of 2.8 eV was obtained for the epitaxial regrowth process. In samples implanted with the largest doses and at low current, a significant decrease of the regrowth rate was detected when the recrystallization front traversed the peak of the impurity distribution. A very interesting application of SE to analyze annealing of a-Si has been presented recently by SOPRA S.A. (French ellipsometer factory). They use single shot excimer laser annealing (SSELA) to achieve high-quality polysilicon layers on low-cost glass substrate. To perform a fast, nondestructive monitoring, real-time SE is used to analyze thin films -thickness, crystallinity, and roughness (Boher et al., 1996; Stehle, 1995). SSELA means a 45 J/shot (three xenon chlorine (XeCl) excimer lasers in a group) every 6 sec. The pulse duraton is 160 nsec. The beam dimension can be more than 200 cm’, which results in very low stress to the optics in terms of peak power and average power. The energy density (more than 200 mJ/cm’) with optimization is enough to produce high-quality polysilicon layers for thin-film transistors for 8-in. active matrix liquid crystals displays by a single shot. Spectroscopic ellipsometry is an excellent tool to determine the layer parameters before and after annealing (Fig. 9). By studying the parameters of thickness, crystallinity, and roughness, optimized conditions -preparation of the a-Si, antireflecting oxide capping layer, energy density, limited heating before and after SSELA -can be established. 5. SILICON-ON-INSULATOR (SOI) STRUCTURES

Silicon-on-insulator (SOI) structures formed by high-dose ion implantation are materials that are very suitable for characterization using SE. The most mature material is SIMOX (Separation by IMplantation of Oxygen), which has been studied in detail with SE by different groups (Narayan et al.,

1 ELLIPSOMETRIC ANALYSIS D+U

WHHC

0.3

0.4

0.5

0.6

25

as deposited afterVEL

0.7

0.8

WI bml FIG. 9. Spectroscopic ellipsometry spectra measured on an a-Si (44 nm)/glass structure before and after optimized single shot excimer laser annealing (SSELA) treatment. In the insert the energy optimization (VEL, very large excimer laser) on two kinds of samples (with and without a 55-nm capping oxide layer) at two temperatures can be seen. RT, room temperature. Boher, P., Stehle, J. L., Stehle, M., and Godard, B. (1996). Single Shot Excimer Laser Annealing of Amorphous Silicon for Active Matrix Liquid Crystal Displays. Appl. Surface Sci. 96-98, 376-383.

1987; Ferrieu et al., 1987; Vanhellemont, Maes, and De Veirman, 1989). Other SO1 structures have been obtained by the implantation of nitrogen (SIMNI) or a combination of nitrogen and oxygen (SIMON). Reports on SE investigations of these types of substrate are scarcer (Fried et al., 1989a, 1989b; Vanhellemont et al., 1990 Lohner et al., 1992).

26

M. FRIED,T.LOHNERAND J. GYULAI

6. SEPARATION BY IMPLANTATION OF OXYGEN (SIMOX)

Continuous buried layers of SiO, can be formed by high-dose ( > lo'* ions/cm2) oxygen ion implantation at energies typically ranging from 50 to 200 keV. Lower doses may not yield a continuous buried oxide layer. The as-implanted material shows a buried damaged layer of which the optical properties can be modeled by a mixture of a-Si and c-Si, or by a mixture of SiO, and damaged Si (Vanhellemont et al., 1991a, 1991b). After hightemperature annealing (> 1ooo"C for several hours), a mixed buried layer consisting of SiO, precipitates in a Si matrix was observed. Accurate composition profiles could be obtained (Fig. 10). Calculation of the integrated SiO, profile in the annealed wafer resulted in a surprisingly good agreement with the nominal implanted dose, using appropriate EMA. For the high-dose samples the SE thickness determinations are very accurate and reliable. The results obtained by other methods (reflectometry,

Depth [nm] FIG.10. Silicon dioxide (SiO,) concentration profiles in a low-dose implanted separation by implantation of oxygen (SIMOX) structure using three different effective-medium approximation (EMA) models. The calculated oxygen ion (Ot) implantation doses are also indicated. Vanhellemont, J., Maes, H. E., and De Veirman, A. (1991a). SpectroscopicEllipsometry Studies of SIMOX Structures and Correlation with Cross-Section TEM. Vacuum 42, 359-365.

1 ELLIPSOMETRIC ANALYSIS

27

XTEM) agree very well (Vanhellemont et al., 1989, 1990, 1991a). An advantage of SE is that one can measure the superficial Si film thickness without having to remove the thermal oxide that is grown in order to thin the Si layer. Contamination during the measurement is also avoided.

7. SEPARATION BY IMPLANTATION OF NITROGEN (SIMNI) SIMNI structures are more difficult to characterize than SIMOX structures, because of the stronger dependence of microstructure on the implantation dose. Fried et al. (1989a, 1989b) studied SIMNI structures implanted with 200 or 400 keV nitrogen ions. The SE data were analyzed in three steps using appropriate optical models and linear regression analysis. A threelayer model (a surface oxide layer, a thick Si layer and a thick nitride layer) was applied with good results first. The second and third models were enhanced with one and two upper and lower interface layers. The fitted parameters were the layer thicknesses and compositions (oxide, nitride, and Si). The results of the fitting using simpler models served as initial values for the more complex models (Fig. 11). The results were compared with data obtained from RBS and XTEM. The sensitivity of the optical model and fitting technique was good enough to distinguish between the Si-rich transition layers near the upper and lower interfaces of the nitride layers, which were unresolvable in RBS measurements. The effect of high-temperature annealing of SIMNI structures produced with 330 keV nitrogen-implantation was investigated by Lohner et al. (1992) using SE, XTEM, and RBS. The implantation was performed at a dose of 1.2 x lo1* ions/cm2 at 500°C into (100) Si. The samples were annealed after implantation at different temperatures (1000 to 1300°C) and durations (0.5 to 5 h): The dependence of the features of the interfaces on the annealing time and temperature was followed with the help of a multilayer optical model and least squares fit. The results showed that the higher the temperature and the longer the annealing, the more compact the nitride layer.

IX. Sophisticated Multilayer Optical Models

Determining depth profiles by SE is an iteration method (Vanhellemont and Roussel, 1992). A simple three-layer model is used to start interpretation of the spectra (Fig. 12(a)). It is assumed that the structure consists of a layer embedded in the substrate that is covered by a thin native oxide. As a first

model 1

model 3

model 2

1 0

d

= 0.001

-

3 0 c

00 Ephoton [ e V 4.13

i.7e

h, CQ

1

4 13

1 . 7 8 , ,4.13

-

calculated

xxx

erDcrirnental

Q yl

"

-10

-1.0

300

COO

500

600

70' WL[nrnl

WLtnrnl

F

FIG. 11. Plot of COSA and tan$ versus the wavelength of a separation by implantation of nitrogen (SIMNI) (200 keV, 75 x IOi7N+/m2, 600°C implanted into silicon, annealed at 1300"C, 2 h) sample fitted using three-layer, five-layer, and seven-layer optical models, respectively. The unbiased estimator is u. Fried, M., Lohner, T., de Nijs, J. M. M., van Silfhout, A., Hanekamp, L. J., Laczik, Z., Khanh, N. Q.. and Gyulai, J. (1989a). Nondestructive Characterization of Nitrogen-Implanted Silicon-on-Insulator Structures by Spectroscopic Ellipsometry. J. Appl. Phys. 66, 5052-5057; Fried, M., Lohner, T., de Nijs, J. M. M., van Silfhout, A., Hanekamp, L. J., Khanh, N. Q., Laczik, Z., and Gyulai, J. (1989b). Non-Destructive Characterization of Nitrogen-Implanted Silicon-on-Insulator Structures by Spectroscopic Ellipsometry. Materials Sci. Engrg. B2, 131- 137.

29

1 ELLIPSOMETRIC ANALYSIS

1

3 layers with variable thickness

2

12 layers, 10 with fixed thickness

FIG. 12. Schematic overviews of the procedure to determine a depth profile of a layer with varying refractive index, buried in a substrate. Vanhellemont, J. and Roussel, Ph. (1992). Characterizationby Spectroscopic Ellipsometry of Buried Layer Structures in Silicon Formed by Ion Beam Synthesis. Materials Sci. Engrg. BlZ, 165-172.

approximation it is assumed that the buried layer is a mixed layer with a homogeneous refractive index that is either fixed or fitted during the regression. The regression analysis will yield an approximate thickness of the buried layer and of the top layers. The next step is to replace this three-layer model by a more complex model (Fig. 12(b)). Two approaches can be followed depending on the information that is available on the functional relation between the concentration and the depth below the surface. 1.

PROFILES WITH UNKNOWN

DEPTH VARIATION

The buried layer is subdivided into a number of thin, mixed layers of equal (and generally fixed) thickness. The total thickness should be the same as that for the three-layer model (Fig. 12(b)). In most cases the number of layers cannot be larger than 10, because of the growing cross-correlations of the unknown parameters. During the regression analysis the thicknesses of the top layers are allowed to vary, whereas for the other mixed layers, only the concentration of the second component is allowed to float. In some cases an additional improvement (after the fit) can be obtained by readjusting the layer thicknesses within the buried layer, whereby thinner layers are used in depths at which the composition is changing rapidly and thicker layers are used in depths with small gradients. The drawback of this approach is that the number of variable parameters is limited, and therefore the depth accuracy is also limited.

M. FRIED,T.LOHNER AND J. GYULAI

30

1

15.5 keV

a

I

SE

I

‘

1

75 keV

fli ,..”’!

3

E

Y

-:4E15

v)

0 0 > 1

I

10’8

b .-C

C

e

0 .+

2

@

111

E V

C

0 u I

0

200

400

600

800

Depth [nml FIG. 13. Damage profiles derived from spectroscopic ellipsometry (SE) spectra of hydrogen-implanted wafers (a), and the hydrogen profiles obtained with X-ray diffraction (XRD) (strain) and secondary ion mass spectroscopy (SIMS) (dotted line), respectively (b). Vanhellemont, J., Roussel, Ph., and Maes, H. E. (1991b). Spectroscopic Ellipsometry for Depth Profiling of Ion Implanted Materials. Nucl. Instrum. Meth. Phys. Res. B55, 183-187.

Erman and Theeten (1983) demonstrated the use of this type of layer model for the assessment of implanted layers into semi-insulating GaAs (Fig. 4). The disorder distribution seemed to be extremely dependent on the ion implantation experimental conditions (temperature of the substrate, ion beam current). The evolution of the disorder profiles also can be studied as the effect of thermal annealing. The application of the method is illustrated for low-dose SIMOX in the Section 1.8.6. Another good example is hydrogen implantation into Si (Vanhellemont et al., 1991b). The damage profiles were obtained with a 1Zlayer model assuming a native oxide, an undamaged top Si layer, and 10

1 ELLIPSOMETRIC ANALYSIS

31

mixed layers of crystalline Si and voids of fixed thickness but varying composition. The profiles obtained with SE correspond very well with those obtained independently with X-ray diffraction (XRD) (strain profile) and SIMS (chemical profile), see Fig. 13. By using an overall regression using the same multilayer model to interpret the measurements made at different angles, an additional known (and strong) parameter can be entered in the fit.

2.

PROFILES WITH

KNOWNDEPTH VARIATION

In many cases, for example, for ion-implantation-induced damage or composition profile, a functional relation between the damage concentration and the depth below the surface is known. In such cases it is advantageous to use this relation as it will strongly reduce the number of unknown parameters that must be calculated through the regression program. In an early work, Adams and Bashara (1975) used numeric fitting of MA1 measurements on P+ implanted Si, assuming that both the n and the k profile exhibit Gaussian distribution. The complex refractive index profiles generated from these measurements were in good agreement with those obtained by an anodization-stripping method. Furnace-annealed specimens showed effects (evident in the k profiles) of residual damage even after a treatment of 650°Cfor 2 h. For ion implantation damage one can also use, for example, coupled half-Gaussians to describe the damage profile (Fried et al., 1991, 1992a, 1992c; Lohner et al., l993,1994a), where only four unknown parameters (the center, the height, and two standard deviations) must be determined to describe the damage (or compositional) profile. (The method can be used for MAI-ellipsometry as well as for SE measurements.) The number of the layers can then be chosen to obtain the required depth resolution (taking into account the required calculation time). From the first approximation (the three-layer model) an estimation of starting values (even for the top layers) can be obtained. The applications and the model are shown in Figs. 14(a) and (b), respectively.

X. Closing Remarks Having studied this chapter, one should appreciate that the important prerequisite for investigating defect annealing by ellipsometry is the detailed optical characterization of the damaged material, for example, the

M. FRIED,T.LOHNERAND J. GYULAI

, l l l l r l l l l l

1

1

I

1

1

I

'

1

I

8

I

symbols - measurement lines - calculation 0.6 -

angle of incidence: 70'

-

FIG. 14. (a) Results of spectroscopic ellipsometry (SE) fitting for phosphorus (P)-implanted Si samples. The insert shows the deduced damage profiles from SE (histograms) and Rutherford backscattering spectrometry (symbols). (b) The applied optical model. a-Si, amorphous silicon; c-Si, crystalline silicon. Lohner, T., Toth, Z., Fried, M., Khanh, N. Q., Yang, G . Q., Lu, L. C., Zou, S. C., Hanekamp, L. J., van Silfhout, A., and Gyulai, J. (1994a). Comparative Investigation of Damage Induced by Diatomic and Monoatomic Ion Implantation in Silicon. Nucl. Instrum. Meth. B85, 524-527; Fried, M., Lohner, T., Aarnink, W. A. M., Hanekamp, L. J., and van Silfhout, A. (1992a). Nondestructive Determination of Damage Depth Profiles in Ion-Implanted Semiconductors by Spectroscopic Ellipsometry Using Different Optical Models. J. Appl. Phys. 71, 2835-2843.

1 ELLIPSOMETRIC ANALYSIS

33

FIG. 14. (Continued)

knowledge of the complex dielectric function of both the undamaged and the damaged material. Consequently, another essential precondition for analyzing defect annealing using ellipsometry is that the difference between complex dielectric functions of undamaged and damaged crystal should be sufficiently great. Throughout this chapter, emphasis has been placed on the need for appropriate optical models to interpret experimental data on defect annealing. Experimental works use a mixture of amorphous and crystalline materials to interpret measurements on damaged semiconductors. A few investigations report that by doing so we systematically overestimate the amount of point defects (Lohner et al., 1994a, 1994b). Systematic and detailed study of point-defect annealing is one of the promising new areas of investigation. However, this can only be done by using complementary analytical methods, such as RBS channeling. We may confidently look forward to engaging the newly emerging infrared ellipsometry in this field (Roseler, 1990; Londos et al., 1994a, 1994b).

ACKNOWLEDGMENTS This work was partially supported by grants from the Hungarian Science Foundation under contracts OTKA-TO17344 and OTKA-F4378. The technical assistance of Mrs. R. Majthenyi is gratefully acknowledged.

34

M. FRIED,T. LOHNER AND J. GYULAI

REFERENCES Adachi, S. (1991). Optical Dispersion Relations in Amorphous Semiconductors. Phys. Rev. B 43, 12316-12321. Adams, J. R., and Bashara, N. M. (1975). Determination of the Complex Refractive Index profiles in P + Ion Implanted Silicon by Ellipsometry. Surface Sci. 49, 44-458. Aspnes, D. E., and Studna, A. A. (1975a). High Precision Scanning Ellipsometer. Appl. Optics 14,220-228. Aspnes, D. E. (1975b). Photometric Ellipsometer for Measuring Partially Polarized Light. J. Optical SOC. America 65, 1274-1278. Aspnes, D. E., and Studna, A. A. (1980a). An Investigation of Ion-Bombarded and Annealed (11 1) Surfaces of Ge by Spectroscopic Ellipsometry. Surfme Sci. 96, 294-306. Aspnes, D. E., Celler, G. K., Poate, J. M., Rozgonyi, G. A,, and Sheng, T. T. (1980b). Characterization of Laser-Annealed Si by Spectroscopic Ellipsometry, Rutherford Backscattering, and by Transmission Electron and Optical Microscopy. Proceedings of the Symposium on Laser and Electron Beam Processing of Electronic Materials, (C. L. Anderson, G . K. Celler, and G . A. Rozgonyi, eds.) Proc. Vol. 80-1, The Electrochemical Society, Princeton, NJ, 414-420. Aspnes, D. E. (1981). Microstructural information from optical properties in semiconductor technology. SPIE Proc. 276, 188-195. Aspnes, D. E., Studna, A. A., and Kinsbron, E. (1984). Dielectric Properties of Heavily Doped Crystalline and Amorphous Silicon from 1.5 to 6.0eV. Phys. Rev. B 29, 768-779. Aspnes, D. E. (1985). Accurate Determination of Optical Properties by Ellipsometry. In Handbook of Optical Constants ofsolids. (E. D. Palik, ed.) Academic Press, Boston, MA, Chapt. 5. Azzam, R. M. A., and Bashara, N. M. (1977). Ellipsometry and Polarized Light. North-Holland, Amsterdam. Benourhazi K., and Ponpon, J. P. (1992). Implantation of Phosphorus and Arsenic Ions in Germanium. Nucl. Instruments Meth. Phys. Res. B71,406-41 1. Bermudez, V. M., and Ritz, V. H. (1978). Wavelength-Scanning Polarization-Modulation Ellipsometry: Some Practical Considerations. Appl. Optics 17, 542-552. Billings, 9. H. (1952). The Electro-optic Effect in Uniaxial Crystals of the Dihydrogen Phosphate (XH,PO,) Type IV. Angular Field of the Electro-optic Shutter. J. Optical SOC. America 42, 12-20. Boher, P., Stehle, J. L., Stehle, M., and Godard, B. (1996). Single Shot Excimer Laser Annealing of Amorphous Silicon for Active Matrix Liquid Crystal Displays. Appl. Surfnce Sci. 96-98, 376-383. Bruggeman, D. A. G. (1935). Berechnung verschiedener physikalischer Konstanten von heterogenen Substanzen I. Dielektrizitatskonstanten und LeitFahigkeiten der Mischkorper aus isotropen Substanzen. Ann. Physik (Leipzig) 24, 636-664. Cahan, 9. D., and Spanier, R. F. (1969). A High Speed Precision Automatic Ellipsometer. Surface Sci. 16, 166-176. Chambon, P., Erman, M., Theeten, J. B., Prevot, B., and Schwab, C. (1984). Spectroscopic Ellipsometry and Raman Scattering Study of the Annealing Behavior of Be-Implanted GaAs. Appl. Phys. Lett. 45, 390-392. Davis, E. A., and Mott, N. F. (1970). Conduction in Non-Crystalline systems. V. Conductivity, Optical Absorption and Photoconductivity in Amorphous Semiconductors. Philos. Mag. 22, 903-922. Drevillon, B., Perrin, J., Marbot, R., Violet, A., and Dalby, J. L. (1982). Polarization Modulated Ellipsometer Using a Microprocessor System for Digital Fourier Analysis. Rev. Sci. Instrum. 53,969-977.

1 ELLIPSOMETRIC ANALYSIS

35

Drude, P. (1889a). Oberflachenschichten. Ann. Phys. Chem. 36, 532. Drude, P. (1989b). Oberflachenschichten. Ann. Phys. Chem. 36, 865. Drude, P. (1890). Bestimmung der optischen Constanten der Metalle. Ann. Phys. Chem. 39,481. Erman, M., and Theeten, J. B. (1982). Multilayer Analysis of Ion Implanted GaAs Using Spectroscopic Ellipsometry. Surface Interface Anal. 4, 98- 108. Erman, M., and Theeten, J. B. (1983). Analysis of Ion-Implanted GaAs by Spectroscopic Ellipsometry. Surface Sci. 135, 353-373. Ferrieu, F., Vu, D. P., DAnterroches, C., Oberlin, J. C., Maillet, S., and Grob, J. J. (1987). Characterization of the Silicon-on-Insulator Material Formed by High-dose Oxygen Implantation Using Spectroscopic Ellipsometry. J . Appl. Phys. 62, 3458-3461. Fredrickson, J. E., Waddell, C. N., Spitzer, W. G., and Hubler, G. K. (1982). Effects of Thermal Annealing on the Refractive Index of Amorphous Silicon Produced by Ion Implantation. Appl. Phys. Lett. 40, 172-174. Fremunt, R., Hirayama, Y., Arai, F., and Sugano, T. (1984). Characterization of laser annealed InP with ellipsometry and Hall-effect measurement. Appl. Phys. Lett. 44, 530-532. Fried, M., Lohner, T., Vizkelethy, G., Jaroli, E., Mezey, G., and Gyulai, J. (1986). Investigation of Solid Phase Epitaxial Regrowth on Ion-Implanted Silicon by Backscattering Spectrometry and Ellipsometry. Nucl. Instrum. Meth. Phys. Res. B15, 422-424. Fried, M., Lohner, T., JBroli, E., Vizkelethy, G., Kotai, E., Gyulai, J., Birb, A,, Adam, J., Somogyi, M., and Kerkow, H. (1987). Optical Properties of Thermally Stabilized Ion Implantation Amorphized Silicon. Nucl. Instrum. Meth. Phys. Res. B19t20, 577-581. Fried, M., Lohner, T., de Nijs, J. M. M.,van Silfhout, A., Hanekamp, L. J., Laczik, Z., Khanh, N. Q., and Gyulai, J. (1989a). Nondestructive Characterization of Nitrogen-Implanted Silicon-on-Insulator Structures by Spectroscopic Ellipsometry. J. Appl. Phys. 66, 50525057. Fried, M., Lohner, T., de Nijs, J. M. M., van Silfhout, A., Hanekamp, L. J., Khanh, N. Q., Laczik, Z., and Gyulai, J. (1989b). Non-Destructive Characterization of Nitrogen-lmplanted Silicon-on-Insulator Structures by Spectroscopic Ellipsometry. Materials Sci. Engrg. B2, 131-137. Fried, M., Lohner, T., Jaroli, E., Hajdu, C., and Gyulai, J. (1991). Nondestructive Determination of Damage Depth Profiles in Ion-Implanted Semiconductors by Multiple-Angleof-Incidence Single-Wavelength Ellipsometry. Nucl. Instrum. Meth. Phys. Res. B55, 257260. Fried, M., Lohner, T., Aarnink, W. A. M., Hanekamp, L. J., and van Silfhout, A. (1992a). Nondestructive Determination of Damage Depth Profiles in Ion-Implanted Semiconductors by Spectroscopic Ellipsometry Using Different Optical Models. J. Appl. Phys. 71, 2835-2843. Fried, M., Lohner, T., Aarnink, W. A. M., Hanekamp, L. J., and van Silfhout, A. (1992b). Determination of Complex Dielectric Function of Ion-Implanted and Implanted Annealed Silicon by Spectroscopic Ellipsometry. J. Appl. Phys. 71, 5260-5262. Fried, M., Lohner, T., Jaroli, E., Khanh, N. Q., Hajdu, C., and Gyulai, J. (1992~).Nondestructive Determination of Damage Depth Profiles in Ion-Implanted Semiconductors by Multiple-Angle-of-Incidence Single-Wavelength Ellipsometry Using Different Optical Models. J. Appl. Phys. 72, 2197-2201. Fried, M., and van Silfhout, A. (1994). Optical Dispersion Relations in Two Types of Amorphous Silicon Using Adachi’s Expression. Phys. Rev. B 49, 5699-5702. Hauge, P. S. and Dill, F. H. (1973). Design and Operation of ETA, an Automated Ellipsometer. IBM J. Res. Development 17, 472-489. Hauge, P. S. (1980). Recent Developments in Instrumentation in Ellipsometry. Surface Sci. 96, 108-140.

36

M. FRIED,T.LOHNERAND J. GWLAI

Holgado, S., Martinez, J., Garrido, J., Piqueras, J. (1995). Regrowth-Process Study of Amorphous BF Ion-Implanted Silicon Layers Through Spectroscopic Ellipsometry. Appl. Phys. A 60,325-332. Holtslag, A. H. M. (1986). Noble-Gas Ion Bombardment on Clean Silicon Surfaces Studied Using Ellipsometry and Desorption. Ph.D. thesis, Twente University, the Netherlands. Hunderi, 0. (1980). Optics of Rough Surfaces, Discontinuous Films and Heterogeneous Materials. Surface Sci. 96, 1-31. Ibrahim, M. M., and Bashara, N. M. (1972). Ellipsometric Study of 400eV Ion Damage in Silicon. Surface Sci. 30,632-640. Jasperson, S. N., and Schnatterly, S. E. (1969). An Improved Method for High Reflectivity Ellipsometry Based on a New Polarization Modulation Technique. Rev. Sci. Instrum. 40, 761-767. Jellison, G. E., Jr., and Modine, F. A. (1982). Optical Constants for Silicon at 300 K and 10 K Determined from 1.64 to 4.73 eV by Ellipsometry. J. Appl. Phys. 53, 3745-3753. Kim, Q., and Park, Y. S. (1980). Ellipsometric Invesitgation of Ion-Implanted GaAs. Surface Sci. 96, 307-318. Kucirkovh, A. (1976). Ellipsometric Study of Tellurium Implanted Silicon. Radiation Efects 28, 129- 131. Lohner, T., Skorupa, W., Fried, M., Vedam, K., Nguyen, N., Grotzschel, R., Bartsch, H., and Gyulai J. (1992). Comparative Study of the Effect of Annealing of Nitrogen-Implanted Silicon-on-Insulator Structures by Spectroscopic Ellipsometry, Cross-Sectional Transmission Electron Microscopy and Rutherford Backscattering Spectroscopy. Materials Sci. Engrg. B12, 177-184. Lohner, T., Fried, M.,Gyulai, J., Vedam, K., Nguyen, N. V., Hanekamp, L. J., and van Silfhout, A. (1993). Ion-Implantation-Caused Special Damage Profiles Determined by Spectroscopic Ellipsometry in Crystalline and in Relaxed (Annealed) Amorphous Silicon. Thin Solid Films 233, 117-121. Lohner, T.,Tbth, Z., Fried, M., Khinh, N. Q., Yang, G. Q., Lu, L. C., Zou, S. C., Hanekamp, L. J., van Silfhout, A., and Gyulai, J. (1994a). Comparative Investigation of Damage Induced by Diatomic and Monoatomic Ion Implantation in Silicon. Nucl. Instrum. Meth. Bs5,524-527. Lohner, T., Kbtai, E., Khinh, N. Q., Tbth, Z., Fried, M., Vedam, K., Nguyen, N. V., Hanekamp, L. J., and van Silfhout, A. (1994b). Ion-Implantation Induced Anomalous Surface Amorphization in Silicon. Nucl. Instrum. Meth. -5, 335-339. Londos, C. A., Georgiou, G. I,, Fytros, L. G., and Papastergiou, K. (1994). Interpretation of Infrared Data in Neutron-Irradiated Silicon. Phys. B 50, 11531-1 1534. Lorentz, H. A. (1916). Theory of Electrons. 2nd ed., Teubner, Leipzig. Lorenz, L. V. (1880). Uber die Refractionsconstante. Ann. Phys. Chem. (Leipzig) 11, 70-103. Maxwell-Garnett, J. C. (1904). Colours in Metal Glasses and in Metallic Films. Philos. Pans. 203, 385-420. Maxwell-Garnett, J. C. (1906). Colours in Metal Glasses, in Metallic Films and in Solutions. Philos. Pans. R. SOC.205,237-288. McMarr, P. J., Vedam, K., and Narayan, J. (1986). Spectroscopic Ellipsometry: A New Tool for Nondestructive Depth Profiling and Characterization of Interfaces. J. Appl. Phys. 59, 694-701. Mizuta, M., and Merz, J. L. (1983). CW Laser Annealing Behavior of Se+-implanted InP Investigated by Ellipsometry. Appl. Phys. Lett. 43, 375-377. Moritani, A., and Hamaguchi, C. (1985). High-speed Ellipsometry of Arsenic-Implanted Si During cw Laser Annealing. Appl. Phys. Lett. 46, 746-748. Muller, R. H. (1976). Present Status of Automatic Ellipsometers. Surface Sci. 56, 19-36.

1 ELLIPSOMETRIC ANALYSIS

37

Nakamura, K., Gotoh, T., and Kamoshida, M. (1979). Characterization of "P+-Implanted Si Layers by Ellipsometry. J. Appl. Phys. 50, 3985-3989. Narayan, J., Kim, S. Y., Vedam, K., and Manukonda, R. (1987). Nondestructive Characterization of Ion-Implanted Silicon-on-Insulator Layers. Appl. Phys. Lett. 51, 343-345. Palik, E. D., editor (1985). Handbook of Optical Constants of Solids. Academic Press, Orlando, FL. Riedling, K. (1988). Ellipsometry for Industrial Application. Springer-Verlag, Wien, New York. Roorda, S., Sinke, W. C., Poate, J. M., Jacobson, D. C., Dierker, S., Dennis, B. S., Eaglesham, D. J., Spaepen, F., and Fuoss, P. (1991). Structural Relaxation and Defect Annihilation in Pure Amorphous Silicon. Phys. Reu. E 44, 3702-3725. Roseler, A. (1990). Infrared Spectroscopic Ellipsometry. Akademie-Verlag, Berlin. Rzhanov, A. V., and Svitashev, K. K. (1979). Ellipsometric Techniques to Study Surfaces and Thin Films. Adu. Electronics Electron Phys. 49, 1-84. Stehle, J. L. (1995). Single Shot Excimer Laser Annealing. SEMICON KANSAI 95 (FDP Technology Symposium) 22-23 June 1995, Osaka, Japan. Takai, M., Tsien, P. H., Tsou, S. C., Roschenthaler, D., Ramin, M., Ryssel, H., and Ruge, I. (1980). CW C0,-Laser Annealing of Arsenic Implanted Silicon. Appl. Phys. 22, 129-136. Tauc, J., Grigorovici, R., and Vancu, A. (1966). Optical Properties and Electronic Structure of Amorphous Germanium. Phys. Status Solidi. 15, 627-637. Theeten, J. B., Chang, R. P. H., Aspnes, D. E., and Adams, T. E. (1980). In Situ Measurement and Analysis of Plasma-Grown GaAs Oxides with Spectroscopic Ellipsometry. J . Electrochem. SOC.127, 378-385. Vanhellemont, J., Maes, H. E., and De Veirman, A. (1989). Spectroscopic Ellipsometry and Transmission Electron Microscopy Study of Annealed High-Dose Oxygen Implanted Silicon. J. Appl. Phys. 65, 4454-4456. Vanhellemont, J. Colinge, J. P., De Veirman, Van Landuyt, J., Skorupa, W., Voelskow, M., and Bartsch, H. (1990). Non-destructive Characterisation of Buried Insulator Structures Using Spectroscopic Ellipsometry and Reflectometry and Correlation with Other Analytical Techniques. Proceedings of the 4th International Symposium on Silicon-on-Insulator Technology and Devices (D. N. Schmidt, ed.). The Electrochemical Society Proceedings, Vol. 90-6, 187-195. Vanhellemont, J., Maes, H. E., and De Veirman, A. (1991a). Spectroscopic Ellipsometry Studies of SIMOX Structures and Correlation with Cross-Section TEM. Vacuum 42, 359-365. Vanhellemont, J., Roussel, Ph., and Maes, H. E. (1991b). Spectroscopic Ellipsometry for Depth Profiling of Ion Implanted Materials. Nucl. Instrum. Meth. Phys. Res. B55, 183-187. Vanhellemont, J. and Roussel, Ph. (1992). Characterization by Spectroscopic Ellipsometry of Buried Layer Structures in Silicon Formed by Ion Beam Synthesis. Materials Sci. Engrg. B12,165-172. Vedam, K., McMarr, P. J., and Narayan, J. (1985). Nondestructive Depth Profiling by Spectroscopic Ellipsometry. Appl. Phys. Lett. 47, 339-341. Waddell, C. N., Spitzer, W. G., Fredrickson, J. E., Hubler, G. K., and Kennedy, T. A. (1984). Amorphous Silicon Produced by Ion Implantation: Effects of Ion Mass and Thermal Annealing. J. Appl. Phys. 55, 4361-4366. Watanabe, K., Miyao, M., Takemoto, I., and Hashimoto, N. (1979). Ellipsometric Study of Silicon Implanted with Boron Ions in Low Doses. Appl. Phys. Lett. 34, 518-519. Watanabe, K., Motooka, T., Hashimoto, N., and Tokuyama, T. (1980). Ellipsometric Study of Annealing Processes of Phosphorus-Ion-Implanted Layers of Si. Appl. Phys. Lett. 36, 451-453.

This Page Intentionally Left Blank

SEMICONDUCTORS AND SEMIMETALS,VOL. 46

CHAPTER 2

Transmission and Reflection Spectroscopy on Ion Implanted Semiconductors Antonios Seas Constantinos ChristoJdes DEPARTMENT OF NATURAL ScENCES UNIVERSITY OF CYPRUS

NICOSIA, CYPRUS

I. INTRODUCTION

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

11. GENERAL OVERVIEW . . . . . . . . . . . . . . . . . . . . . 111. RECENTOPTICAL EXPERIMENTAL STUDIESON IMPLANTED SILICON.

. . . . . . . . . . .

1. Phosphorous-Implanted Silicon . . . . . . . . . . . . . . . . . . . . 2. Fourier Transform Infrared Optical Measurements . . . . . . . . . . . . IV. THEORETIC BACKGROUND. . . . . . . . . . . . . . . . . . . . . . . . V. DISCUSSION AND ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . 1. Influence of Annealing Temperature on the Plasma Wavelength . . . . . . . 2. Effective Mass versus Annealing Temperature . . . . . . . . . . . . . . VI. S U M M A R Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

39 40 47 47 49 60 62 62 67 68 69

I. Introduction During the last decades various types of measurements have been used for the characterization of implanted wafers (Christofides, 1992). Generally, the aim of experimental characterization of such materials is the study of effects concerning the induced damage (short and long-range disorder) due to the ion implantation and the annihilation processes that take place during annealing (Gibbons, 1968, 1972). Optical spectroscopic methods have been adopted by several groups for the characterization of implanted semiconducting materials because of the advantage of their giving quantitative and qualitative information concerning the defects state and annihilation process in implanted (unannealed and annealed) semiconducting wafers. In addition, these methods represent nondestructive and contactless techniques. In 1970, Crowder et al. used 39 Copyright 0 1997 by Academic Press All rights of reproduction in any form reserved. 0080-8784/97 $25

40

A. SEASAND C. CHRISTOFIDES

interference phenomena observed in the optical absorption spectra to estimate the depth of the damaged layer in silicon (Si) due to heavy implanted ions. Similar experiments also were performed by Kurtin, Shifrin, and McGill (1969) and Hart and Marsh (1969). During the mid 1970s, Kachare et al. (1976a, 1976b) performed normal incidence reflection and transmission measurements on gallium arsenide (GaAs) and gallium phosphide (Gap)-implanted wafers at high doses = 1 x 1017cm-2) and high energies (E = 3 MeV). By using the same technique, Wang et al. (1985) studied the effect of annealing on the optical properties of implanted germanium (Ge). Infrared (IR) studies on heavily implanted phosphorus (P) ( E = 0.2 to 2.7 MeV and @ = 10l6 to l O I 7 P+/cm2) also were performed to study the thickness and the refractive indices of such implanted layers (Hubler et al., 1979a; Hubler, Malmberg, and Smith, 1979b). Infrared spectroscopy on beryllium (Be)-implanted GaAs was carried out by Kwun et al. (1979). Refractive index profiles and range distribution of Si implanted with high-energy nitrogen has been studied by Hubler et al. (1979a, 1979b). Other researchers such as Fredrickson et al. (1982), Waddell et al. (1982), and Spitzer et al. (1977) have used similar techniques for the characterization of implanted semiconducting wafer. Brown et al. (1981) performed electroreflectance measurements on ion-implanted GaAs, and Brierley, Lehn, and Grabinski (1988) performed IR transmission measurements on Si+-implanted GaAs for mapping the implanted dose distribution. An excellent review paper on the optical effects of high-energy implantations in semiconductors is that by Tatarkiewicz (1989). This chapter covers several aspects concerning optical characterization of ion-implanted wafers and the influence of annealing on their optical properties. When possible, quantitative characterization of the implanted wafers and estimations concerning the activation of implanted impurities versus annealing temperature is presented. After a brief review of results obtained between the 1970s and the 1980s, we concentrate on recently obtained results. In Part I1 we review some key results that played a significant role in the development of optical studies on implanted semiconductors. Part I11 summarizes recent experimental results, and Part IV presents a useful theoretic background. The theoretic model is used in Part V to estimate the degree of activation of implanted impurities. A general summary and future perspectives are presented in Part VI. We apologize to those scientists in the field whose work we might have somehow omitted. 11. General Overview

Reflection is a powerful nondestructive technique for the determination of many physical parameters in implanted semiconductor wafers. In a series of

2 REFLECTIONSPECTROSCOPY ON ION IMPLANTEDSEMICONDUCTORS 41

papers, Kachare et al. (1974, 1976a; 1976b), Kwun et al. (1979), Hubler et al. (1979), Fredrickson et al. (1982), and Waddell et al. (1982) studied the reflection from various highly implanted semiconducting wafers. Using a simple model they tried to fit the interference fringes observed in the measured IR reflection spectrum to derive information such as refractive indices of the implanted and nonimplanted layers and layer thickness. In the model, they assumed that the implanted semiconductor consisted of three layers (layer 1: surface and near-surface region; layer 2 heavily implanted region; Layer 3: substrate) and by considering the dielectric properties of the various layers using the classic dispersion theory for a damped harmonic resonance, they tried to reproduce the measured spectral behavior. The real part of the complex dielectric constant is given by the expression

&’(V)

N

n2 = A

+ R -Bv + V , ” -SV ’ ~

where the first two terms on the right-hand side of Eq. (1) represent the refractive index on the low-frequency side of the fundamental absorption edge, and the remaining term estimates the contribution to the refractive index from the IR resonance at v, (Kachare et al., 1976a). In Eq. (l), v is the frequency, v, is the resonance frequency, S is the resonance strength, A, B, and R are constants, and n is the refractive index. The reflectivity was computed using the expression given by Heavens (1964), which considers two absorbing layers on a nonabsorbing layer. Kwun et al. (1979) performed optical studies on high-fluence Be-implanted GaAs by using IR reflection and transmission measurements. The GaAs (100) wafers were implanted in the range of 1 x 1015 to 1 x 10l6Bef/cm2 and annealed between 500 and 800°C. Infrared reflection measurements were obtained in the range of 320 to 7600cm-’ by using a single beam spectrometer. Figure 1 shows reflection spectra of Be-implanted GaAs ((D = 1 x 10” Be+/cmz) as a function of photon frequency for various annealing temperatures. We note that all spectra show interference effects except the nonimplanted samples. Figure 2 shows typical IR transmission data obtained for the GaAs wafer implanted at 1 x 10l6Be+/cm2. The interpretation of the previous results comes from a simple reflection model also presented previously, which is able to take into account interference effects. In the case of the nonannealed samples it has been found that the implantation produces a layer of significantly larger refractive index than do the nonimplanted materials, while the refractive index increases with increasing implantation dose. After annealing at 500”C, a free-hole plasma layer contribution to the electric susceptibility was observed by analyzing reflection interference effects. It is important to note that similar

A. SEASAND C. CHRISTOFIDES

42

4u i$ d-

m -a -

z 0

W

a Be -GOAS 280 KeV Ix 10'5/cmf

FIG. 1. Reflection of beryllium (Be)-implanted gallium arsenide (GaAs) as a function of photon frequency for various annealing temperatures for a fluence of Q, = 1 x 10l5Be+/cm2 The scale is indicated by a data point at the top of the scale. Kwuan, S., Spitzer, W. G., Anderson, C. L., Dunlap, H. L., and Vaidyanathan, K. V. (1979). Optical Studies of BeImplanted GaAs. J . Appl. Phys. 50, 6873-6880.

qualitative results have been obtained by Hubler et al. (1979) for Si wafers. However, in the case of annealed samples it was found that after annealing the refractive index is almost identical to that of the crystalline substrate. The absorption coefficient of the implanted layers are calculated from the transmission expression T=

(1 - R)' exp[-(add 1 - R2 exp[ -2(add

+ ~,t,)] + a,t,)]

where R is reflection of the GaAs at a given wavelength, ad and a, are the mean absorption coefficient of the implanted layer and substrate, respectively, and d and t, represent the thickness of the implanted layer and substrate, respectively. From transmission measurements and by knowing the various thicknesses of the samples and the optical characteristics of the substrate, we can obtain the optical absorption spectra of the implanted films. Figure 3 presents absorption spectra obtained by using optical transmission data. We note that add increases with increasing annealing temperature.

2 REFLECTIONSPECTROSCOPY ON ION IMPLANTED SEMICONDUCTORS 43

WAVE NUMBER (cm-3 FIG. 2. Transmission of beryllium (Be)-implanted gallium arsenide (GaAs) as a function of photon frequency and annealing temperature for a fluence equal to 1 x 10l6 ions/cm*. Kwuan, S., Spitzer, W. G., Anderson, C. L., Dunfap, H. L., and Vaidyanathan, K. V. (1979). Optical Studies of Be-Implanted GaAs. J . Appl. Phys. 50, 6873-6880.

The optical properties of As-implanted and annealed Si were studied by Wagner and Schaefer (1979) using IR spectroscopy. These authors presented several experimental results with both qualitative and quantitative analysis. They showed how nondestructive IR optical measurements can be used to monitor several properties, such as sheet resistance, electrical activation, and concentration, using (100) p-type Si samples implanted with As between 1 x 1015 and 2 x 10l6 As/cmZ. Figures 4(a) and 4(b) show the variation of the optical reflection and transmission versus wavelength for various implantation doses. The strong dependence of the reflectance and transmittance on the implanted dose is clearly shown. The influence of annealing on the optical reflectance and transmittance is presented in Figs. 5(a) and 5(b). We note that while annealing strongly influences the reflectance, it does not significantly affect the transmittance. Wagner and Schaefer (1979) used the theoretic model of Schumann and Phillips (1967) (see Part IV) for the interpretation of these results.

44

A. SEASAND C. CHRISTOFIDES

FIG.3. Absorption spectra versus photon frequency for beryllium (Be)-implanted gallium arsenide (GaAs) for a fluence of 1 x 10l6ions/cm* at various annealing temperatures. Kwuan, S., Spitzer, W. G., Anderson, C. L., Dunlap, H. L., and Vaidyanathan, K. V. (1979). Optical Studies of Be-Implanted GaAs. J . Appl. Phys. 50, 6873-6880.

Several IR measurements were performed by Engstrom (1980) on Si wafer implanted with boron (B) at 35 keV (doses: 1 x 1014 and 1 x 10l6B/cm2). After implantation the samples were laser annealed. Reflection and transmission measurements in the range of 2.5 to 20 pm were obtained by using Fourier transform infrared (FTIR) spectroscopy. Figures 6(a) and 6(b) present reflectance and transmittance spectra, respectively. Note the strong influence of implantation dose both on reflection and transmission. From these results, Engstrom (1980) put forth the following conclusions: (1) the relaxation time is independent of the implantation dose and (2) the optical properties are linearly affected by implantation dose. Infrared reflectance studies of amorphous silicon (a-Si) produced by Si ion implantation have indicated that after annealing at 5WC, the asimplanted a-Si forms an anneal stabilized state whose IR reflectance behavior is distinct from both the unannealed as-implanted a-Si obtained after epitaxial regrowth (Fredrickson et al., 1982). Differential reflectometry was used by Hummel et al. (1988) to identify whether an implanted layer is crystalline, damaged crystalline, or amorphous. These authors performed optical differential reflectometric measurements on Si-implanted Si wafers (dose: 1 x 1015ions/cm2; energy 60 to 180 keV). Figure 7 presents a series of differential reflectograms for various implantation energies. We note that for all presented implantation energies,

2 REFLECTIONSPECTROSCOPY ON

ION IMPLANTED SEMICONDUCTORS

45

FIG. 4. Influence of the implanted dose on the infrared (a) reflectance and (b) transmittance spectra (curve A: 1 x lot5;curve B 6 x lOI5; curve C: 8 x curve D 1 x lot6 and curve E: 2 x 10l6As/cm2.As, arsenic. Wagner, H. H., and Schaefer, R. R. (1979). Contactless Probing of Semiconductor Dopant Profile Parameters by IR Spectroscopy. J. Appl. Phys. SO, 26972704.

we can distinguish three characteristic Si interband transition peaks near 3.4, 4.2, and 5.6 eV. According to the authors, some implanted-induced damage occurred for all implantation energies. Hummel et al. (1988) introduced the idea that the type of implantation damage depends on the implantation energy. These authors also showed that by exploiting the

A. FIG.5. Influence of 20-min (curve A) and 180-min (curve B) annealing time on their (a) reflectance and (b) transmittance spectra (implantation dose: 8 x lot5 As+/cmZ).As, arsenic. Wagner, H. H., and Schaefer, R. R. (1979). Contactless Probing of Semiconductor Dopant Profile Parameters by I R Spectoscopy. J. Appl. Phys. 50, 2697-2704.

A. SEASAND C. CHRISTOFIDES

46 4 .O

0.8 W

u 0.6

+ f u) a a l-

0.4

o.2 0

0.2 II

1

-

Re. 6. Transmittance and reflectance spectra of boron-implanted, laser-annealed silicon. Implant doses in ions/cm2 are shown at the right of the graphs. At these wavelengths, the spectra of an unimplanted sample is virtually indistinguishable from the sample implanted with 1 x loL4ions/cm2. Engstrom, H. (1980). Infrared Reflective and Transmissivity of BoronImplanted, Laser-Annealed Silicon. J . Appl. Phys. 51, 5245-5249.

intensity of interband transitions, we can determine the thickness of the implanted-induced damaged layer over a submerged amorphous layer. In addition, they showed a direct relation between interference effects and the thickness of the implanted amorphized layer.

2 REFLECTION SPECTROSCOPY ON IONIMPLANTED SEMICONDUCTORS 47

60 KoV

I

I

800 1

1.5

1

I

SO0 I

I

1

2

l

I

a

400

I

3

1

4

1

1

5 6 E(oV)

FIG. 7. Differential reflectograms of silicon ion-implanted wafers. The implantation energies are shown for each curve. The fluence is 1 x 10" ion/cmz.The ordinate is constant for all spectra. The individual curves have been shifted for clarity. The zero point for each curve is shown by the dashed lines. Hummel, R. E., Xi, W., Holloway, P. H., and Jones, K. A. (1988). Optical Investigations of Ion Implant Damage in Silicon. J . Appl. Phys. 63,2591.

111. Recent Optical Experimental Studies on Implanted Silicon

1. PHOSPHOROUS-IMPLANTED SILICON

Two-inch-diameter Si wafers (100) lightly doped with boron (20 to 25 51 cm) were implanted with phosphorus at various doses (@ = 1 x 10' to 1 x 10l6P+/cm2and E = 150 keV) and energies ( E = 20 to 180 keV and @ = 5 x l O I 4 P+/cmz) through a thin oxide layer at room temperature.

48

A. SEASAND C. CHRISTOFIDES TABLE I

JUNCTION’S DEPTHS OF SAMPLES IMPLANTED WITH 150 keV AT VARIOUS DOSES (@ (P+/cmz)), UNANNEALED AND ANNEALED AT VARIOUS ANNEALING TEMPERATURES ( Ta(“C)) @(p+/cm2)

T,(“C) NA 800 850 900 950 loo0 1100

1 x loi3 (W1)

1 x loi4

0.45 0.45 0.48 0.46 0.50 0.48 0.60 0.53 0.60 0.63 0.68 1.32 0.80

0.48 0.50 0.64 0.49 0.65 0.50

(W2)

0.80

0.57 0.85 0.72 0.90 1.55 1.15

5 x loi4 (W3)

1 x loi5 (W4)

5 x loi5 W5)

1 x loi6 04‘6)

0.50 0.52 0.65 0.53 0.65 0.54 0.70 0.60

0.51 0.53 0.63 0.52 0.67 0.55 0.70 0.60 0.75 0.78 0.80 1.74 1.52

0.52 0.63 0.75 0.75 0.80 0.90 0.90 0.97 1.10 0.97

0.53 0.63 0.69 0.80

0.80

0.75 0.90

1.70 151

1.30

2.03 1.81

0.80

1.02 1.01 1.45 1.50 1.90 1.50 2.40 2.60

The values of the first lines for each annealing temperature were obtained from ID-SUPREM 111 simulation (I&); the second lines (bold face) were obtained from spreading resistance measurements (dR).

TABLE I1 JUNCTION’S DEPTHS (IN pm) OF SAMPLES IMPLANTED AT VARIOUS IMPLANTATIONENERGIES WITH DOSE 5 x loi4 (P’ (P’/cm2), UNANNEALED AND ANNEALED AT VARIOUS TEMPERATURES T, (“C) 140 (W11)

NA 800 850 900 950 1000 1100

0.12 0.25 0.24 0.23 0.32 0.55 1.56

0.16 0.16 0.22 0.22 0.33 0.55 1.56

0.26 0.26 0.27 0.29 0.37 0.38 1.59

0.37 0.38 0.38 0.40 0.47 0.65 1.63

0.48 0.48 0.48 0.50 0.56 0.73 1.68

0.50 0.50 0.51 0.53 0.56 0.73 1.70

0.57 0.57 0.58 0.59 0.65 0.81 1.75

These values were obtained from spreading resistance measurements.

After implantation, the wafers were cut along the crystallographic axes into several 1 x 1 cmz samples, which were then thermally annealed isochronally at various temperatures, T, between 300 and 1100°C for 1 h in an inert nitrogen atmosphere. After annealing, the oxide overlayer was etched away and the samples were used for FTIR spectroscopic characterization. Tables

2 REFLECTIONSPECTROSCOPY ON ION IMPLANTEDSEMICONDUCTORS 49

I and I1 give several characteristics of the implanted and annealed samples such as implantation dose, energy, and implanted layer thickness, as obtained by simulation and spreading resistance measurements Stanford University PRocess Engineering Model (SUPREM) (Christofides et al., 1994; Seas et al., 1995). 2. FOURIER TRANSFORM INFRARED OPTICAL MEASUREMENTS a.

Measurement between 0.75 and 4.0 pin

Figures 8(a) to 8(e) present several FTIR reflective spectra obtained at room temperature in the range of 0.75 to 4 pm ( NN 14,000 to 2500 cm- l). As expected, for lightly implanted samples (1 x 1013P+/cm2) the difference between the reflection spectra of low and high annealing temperatures is not significant (Christofides et al., 1994). Figure 8(a) presents spectra obtained from Si implanted at higher doses (1 x 1014P+/cm2) and annealed at various temperatures. The low-annealed sample (R300) already shows a certain trend of differentiation vis-a-vis the rest of the spectra (see the inset in Fig. 8(a)). For samples implanted at a slightly higher dose, 5 x 1014P + / cm2 (Fig. 8(b)), unannealed and low annealed, the increase of the fringe amplitude can be seen. The fringes of these samples and those presented in Figs. 8(b) to 8(e) indicate that the materials implanted at high doses present a bilayer form, whereas the implanted and highly annealed materials seem more homogeneous. For implantation doses equal to and higher than the critical dose Qc (ac= 5 x l O I 4 P+/cm2; Prussin, Margolese, and Tauber, 1985), we note that all the samples, except from those that have been annealed at very high temperatures (900 and 1lOOOC), present interference fringe patterns due to the amorphous-crystalline (a-c) interface (Fig. 8(c) to 8(e)). It is also important to note that the fringe separation changes slightly with increasing implantation dose, which is due to the fact that there is an increase in the difference between the reflective index of the two layers. In addition to the fringe patterns that arise from interferences between light reflected by the front-implanted surface and the light reflected by the amorphized disorder-to-crystalline interface, Figs. 8(d) and 8(e) show some other very interesting features. There appears to be a significant decrease in the reflection spectrum, and sometimes reflection minima, for the highly annealed samples at long wavelengths (between 3 and 4 pm). This phenomenon does not appear in the samples implanted at a lower dose. It is also important to note that in this IR region, the variation of the reflectivity of the implanted and very highly annealed samples undergoes an anomalous dispersion. The same phenomenon was also noted by Wang et al. (1985) in implanted wafers and was explained by the theory of plasma effect. In Figs. 8(d) and 8(e) we note that a characteristic minimum appears only for the

A. SEASAND C. CHRISTOFIDES FTlR REFLECTION MESUREMENTSOF SAMPLE 2

Implanted 1 ~ 1 0 ' ~

0.5

1.0

1.5

2.0

2,s

3.0

3.5

WAVELENGTH (mlcmns)

m R REFLECTION MESUREMENTS OFSAMPLE 3 Implantad 5x10"

I

0.5

\\

1.0

1.5

2.0

2.5

3.0

3.5

4.0

WAVELENGTH (mlcmns)

FIG.8. Fourier transform infrared (FTIR) reflection measurements of phosphorus-implanted silicon (Si) wafers at various doses, CJ(P'/cm2): (a) 1 x (b) 5 x loL4;(c) 1 x (d) 5 x loL5;and (e) 1 x (f) FTIR transmission measurements of phosphorus-implanted Si wafers at doses cD(P'/cm2): 1 x The implantation energy is 150keV. The samples were annealed isochronally (1 h) at various temperatures. The dashed line indicates a nonimplanted sample.

2 REFLECTIONSPECTROSCOPY ON IONIMPLANTEDSEMICONDUCTORS 51 FTlR REFLECTION MESUREMENTS OF SAMPLE 4

Implanted 1 ~ 1 0 ' ~

20

~

I

WAVELENGTH (microns) FTlR REFLECTION MESUREMENTS OF SAMPLE 5

Implanted 5 ~ 1 0 ' ~

45

E. W

i

W

a

30

25

20

WAVELENGTH (microns)

FIG. 8. (Continued)

A. SEASAND C. CHRISTOFIDES FTlR REFLECTION MESUREMENTSOF SAMPLE 6 Implanted IX~O'' 5(1

45

I

1.o

1.5

2.0

2.5

3.0

3.5

4.0

WAVELENGTH (mlcrons) FTlR TRANSMISSION MESUREMENTS OF SAMPLE 6 Implanted 1x10'' 35

25

15

10

5

0 0.5

1.0

1.5

2.0

2.5

WAVELENGTH (mlcrons)

FIG. 8. (Continued)

3.0

3.5

4.0

2 REFLECTIONSPECTROSCOPY ON ION

IMPLANTED SEMICONDUCTORS

53

samples annealed at 900°C (R900). One would also expect to see such a minimum on the highly annealed samples (1l00OC) because it is well known that these possess more free carriers. Although this is true, Table I shows that these carriers are diffused in a very large thickness in the material, and thus the number of free carriers per unit volume is lower than in the R900 sample. Finally, for the two samples annealed at 900"C, note that the reflectivity minimum appears at lower wavelength, Ap (x3.1 pm), for the sample implanted at higher dose (@ = 1 x 10l6P+/cm2) than for the sample implanted at lower dose (@ = 5 x 10'' P+/cm2), Ap x 3.5 pm. In any case, confusion should be avoided between reflectivity minima and interference fringes because for samples annealed at high temperatures (> 600 to 700°C) the a-c interface disappears. Therefore, no possibility exists of observing any interference fringes from the samples annealed at 900 and 1100°C from our samples in Fig. 8(e). Finally, Fig. 8(f) presents the transmission spectra of samples implanted at 1 x 10'6P+/cm2. We note that in the spectral range of 1 to 4 p m for samples annealed at 900 and llOO°C, both transmission ( T ;Fig. 8(f)) and reflection ( R ;Fig. 8(e)) decrease drastically. In fact, as T R A = 1 at 3.1 pm, we can have A = 1 - T - R x 0.69, which is due to a very high absorption. This low absorbance can be explained only by the free-carrier absorption phenomenon.

+ +

b. Fourier Transform Infrared Measurement between 3 and 25 pm

On all the implanted layered Si wafers, we measured the reflection and transmission as a function of wavelength in the spectral range of 2.5 to 25pm. To do this we used a Fourier transform infrared (FTIR) spectrophotometer, with a wavelength resolution of 8 cm- '. Spreading resistance measurements are used to obtain the profile of the implanted impurities for all the P-implanted, nonannealed and annealed Si wafers (Othonos et al., 1994). Van der Pauw (Hall effect) measurements were performed for the high-dose implantation wafer series (W6) to better understand the underlying annealing kinetics and to complement data concerning the conductive effective mobilities. These measurements are reported in Tables I to IV. Figures 9(a) to 9(f) present several FTIR transmisson spectra, obtained at room temperature, in the range of 2.5 to 25pm. As expected, for lightly implanted samples (W1 and W2) the difference between the transmission spectra of low and high annealing temperatures is not significant. O n those spectra we note that the influence of annealing temperature for each dose of implantation plays a role only at long wavelengths. The solid line with no marks refers to a nonimplanted sample (NI). In fact, for low-dose implantation (W1 and W2)-that is,

54

A. SEASAND C.CHRISTOFIDES

FIG. 9. Fourier transform infrared (FTIR) transmission spectra of phosphorus-implanted (b) 1 x lOI4; (c) 5 x lOI4; (d) silicon wafers at various doses, cP(Pt/cm2): (a) 1 x 1x (e) 5 x and (f) 1 x

below the critical amorphization dose- the various spectra do not show any significant variation in the wavelength range that the measurement was performed. However, for higher implantation doses (W3-W6) -that, is around and over @,-the samples annealed at high temperatures present significant differences at wavelengths even shorter than 5 pm. In the case of

2 REFLECTIONSPECTROSCOPY ON

ION IMPLANTED SEMICONDUCTORS

55

TABLE 111 CONCENTRATION OF FREECARRIERS FOR SAMPLES IMPLANTED WITH 150 keV AT VARIOUSDOSES (W1 to W6), NONANNEALED AND ANNEALED AT VARIOUS TEMPERATURES (T, ("C) N (P'/cm3)

I x 1019 (W4i ____

~~

800 850 900 950

1000 1100

2.08 2.00 1.66 1.66 1.47 1.25

1 x lozo (W6i

1.56 1.54 1.25 1.18 1.11 1.15

0.77 0.77 0.7 1 0.62 0.55 0.33

1.59 1.49 1.43 1.33 1.25 0.66

6.61 6.25 5.55 4.54 3.85 2.76

1.45 1.25 0.99

0.66 0.67 0.38

N was obtained for each annealing temperature from spreading resistance measurements by assuming that all implanted impurities are electrically active.

the wafer series W5 and W6, note that the annealing temperature T, monotonically influences the variation of transmission versus 1. For the heavily implanted samples the transmission spectra change monotonically with annealing temperatures. For example, in Figs. 9(c) to 9(f9 we can observe that the transmission for highly annealed samples (over 600°C) decreases with increasing T, because of free-carrier absorption. For the sample W6 series (see Fig. 9(f)) annealing at 1100°C and for A > 7 pm, the transmission spectrum is practically zero mainly due to high reflection (0.8) and free-carrier absorption. Figures 10(a) to 10(f) present FTIR reflection spectra obtained at room temperature. Figures lO(a) and 10(b) show the behavior of a series of low-implanted and annealed samples (W1 and W2). The influence of annealing temperature on the reflectivity can be realized from Figs. 10(a) and 10(b). We note that the low-dose samples do not present any minima, whereas the samples implanted around and over the critical implantation dose (W3 to W6) present minima. These wavelength minima depend on the implantation dose and, of course, on the annealing temperature. The plasma wavelength minimum Ap is a function of the concentration of free activated carriers N , which depends on the annealing temperature. This dependence can be expressed as (Kireev, 1975)

Ap(Ta)cc

J 16n2e2cZN(T,) m*(T,)

(3)

56

A. SEASAND C . CHRISTOFIDES

FIG.10. Fourier transform infrared (FTIR) reflection spectra of phosphorus-implanted silicon wafers at various doses, @(Pf/cm2): (a) 1 x (c) 5 x 1014; (d) (b) 1 x 1 x 10”; (e) 5 x 10”; and (f) 1 x

where m* is the conductive effective mass, and the constants e and c are the electron charge and light velocity in vacuum, respectively. Figure 1O(f) presents FTIR reflection measurements of heavily implanted, nonannealed and annealed Si layers (W6: 1 x 1OI6P’/cm’). Again, note the shift of the reflection minimum due to the variation of the free-carriers

2 REFLECTIONSPECTROSCOPY ON

57

ION IMPLANTED SEMICONDUCTORS

TABLE IV MOBILITYAND ESTIMATED FREE-CARRIER CONCENTRATION OF VARIOUS SILICON SERIESW6 FOR DIFFERENT ANNEALING TEMPERATURES

T,("C) Ir (cm2/Vs)

N (P+/cm3)

700

BOO

72 69 1.1 x lozo 1.02 x lozo

850

I1 1 x lozo

900

950

87 87 1.4 x loL9 1.4 x l O I 9

WAFERS OF

1050

80 1.7 x 10''

concentration as a function of the annealing temperature. This shift easily can be explained using Eq. (3) and Table IV. Table IV presents the concentration of free carriers for all samples calculated using the experimentally determined layer thickness and implantation dose and the electrical experimental results. For the nonannealed and low annealed samples, no minima are present at least in the spectral range under investigation. Howarth and Gilbert (1962) showed that in P-doped Si, reflection minima can be observed in the range of 2 to 20 pm for doses over 7 x lo1' Pf/cm3. Figures 11 and 12 present transmission and reflection FTIR spectra for samples implanted with various energies at a dose equal to 5 x 1014P+/cm2 (see Table 11, wafers W7-Wl2) (Seas et al., 1995). As expected, the implantation energy does not play a significant role either in the transmission or on the reflection spectra. In addition, samples implanted around QC,at the low implantation energy of 20 keV (W7), and annealed at various temperatures show the same behavior as those implanted at energies almost one order of magnitude higher. Figures 13(a) and 13(b) show the reflection and transmission coefficients at 25 pm as a function of the annealing temperature. It is evident from Fig. 13 that for low-dose implantation (wafer series W1 and W2), the optical transmission and reflection are independent of the annealing process because there is no significant change in the concentration of free carriers with annealing. However, the wafer series (W3) implanted at 5 x 10'4P+/cm2 shows a slight increase in reflectivity for annealing temperatures over 700°C. For wafers implanted at a high dose (W4 to W6), where the implantation is over mC,the reflection coefficient changes drastically. The reflection coefficient varies from 0.35 for the nonannealed (or low-annealed) layers, to almost 0.90 in the case of samples annealed at over 800°C. In Fig. 13(a) it is important to note that samples implanted with a dose of 5 x 1015P+/cm2 (W5) reach a high reflectivity after annealing at 80O0C, whereas samples implanted at the highest dose (W6: 1 x 10'6P+/cm2) reach the maximum reflectivity at lower temperatures. Similar behavior was observed by Engstrom (1980) for B-implanted and laser-annealed Si layers.

58

A. SEASAND C. CHRISTOFIDES

FIG. 11. Fourier transform infrared (FTIR) transmission spectra of phosphorus-implanted silicon wafers at various implantation energies, E (keV): (a) 20, (b) 30, (c) 60, (d) 100, (e) 150, and (f) 180.

All previous comments concerning the reflection coefficient are also valid for the transmission coefficient. In Fig. 13(b) we show the transmission coefficients versus the annealing temperature plotted on a logarithmic scale for presentation reasons. The transmission measurements clearly show the distinction between the two samples (W5 and W6) implanted at high doses.

2

REFLECTIONSPECTROSCOPY ON

59

ION IMPLANTED SEMICONDUCTORS

0.55 L

0.5

i

t w8

i

g 0.45

% s

0.4

* 0.35 03 0

15 20 Wavelength (microns) 5

10

0.55 1 051

25

-

0.25 t . - - .' . - .- ' . - - 0 5 10 15 20 Wavelength (microns)

1

25

1

w9

1

Wavelength (microns) 0.51

g

z 8

5

0.45

wll

1

055 r 051

1

w12

0.4

0.35 0.3

0

5

1 0 1 5 2 0 2 Wavelength (microns)

5

FIG.12. Fourier transform infrared (FTIR) reflection spectra of phosphorus-implanted silicon wafers at various implantation energies, E (keV): (a) 20, (b) 30, (c) 60,(d) 100, (e) 150, and (f) 180.

For example, after annealing at 1100°C we note that the transmission at 25 pm for the W6 series is only 0.08, whereas under the same conditions the corresponding wafer of W5 series presents a transmission almost one order of magnitude higher. Wafers implanted with a smaller dose (W1 series), present a transmission that is more than two orders of magnitude higher than the heavily implanted samples.

60

A. SEASAND C. CHRISTOFIDES

FIG. 13. (a) Reflection coefficient and (b) transmission coefficient,each as a function of the annealing temperature for various implantation doses.

IV. Theoretic Background The reflectivity minimum of FTIR spectra associated with plasma resonance has been an important tool for the determination of free-carrier concentration in semiconductors for many years. A great deal of experimental and theoretic work has been carried out to improve and increase the accuracy of the calculation. One can cite, for example, the classic theoretic

2 REFLECTIONSPECTROSCOPY ON ION IMPLANTEDSEMICONDUCTORS

61

approach of Smith (1959) and the interesting development to this approach made by Schumann and Phillips (1967). The principal equations of Smith (in MKS) for the expression of the optical coefficient n and k were written as

2 n k = L op,E,m*

(1 + ) z

w2z2

where E~ is the permittivity due to the lattice, cr, is the dc conductivity, pa is the dc mobility, m* is the conductivity effective mass, and z is the relaxation time. Schumann and Phillips (1967) have shown, under the assumption of nondegenerate statistics, that the previous two equations can be written as

where J(D) and 40)are nondimensional and depend on the product d z 2 , N is the carrier concentration, 1 is the wavelength, pa is the dc resistivity, which is also a function of N (like po), c is the velocity of light in vacuum, D is a function of N and A, and r(4) is the gamma function. According to Seeger (1988), the approximation of small damping, where oz >> 1, is valid for most semiconductors even in the far-IR spectrum. As pointed out by Schumann and Philips (1967) this is the used most approximation and leads to J ( D ) = 40)= I. Taking into account this approximation from the system of Eqs. (5.1) and (5.2), we can evaluate the index of refraction and the extinction coefficient k, each as a function of wavelength 1. The reflectivity was calculated from the well-known expression R=

(n - 1)2 (n + 1)’

+ k2 + k2

The minima in the reflectivity obtained from the derivative of the previous equation are plotted in Fig. 14 (solid line):

A. SEASAND C. CHRISTOFIDES

62

100

W3 - W6

10

1' . . lo'*

. . ' . . . I

1019

'

.

.

1020

. ' " ' I

'

'

..'-

Id'

FIG. 14. Mobility as a function of the logarithm of the carrier concentration. min. refl., minimum reflectivity.

For the numeric calculation, the effective mass is assumed to be 0.26. The value of resistivity as a function of carrier concentration is obtained from various sources (Chapman et al., 1963; Irvin, 1962). It is important to note that this model is valid for concentrations ranging between 1 x l O I 9 and 1 x 102'cm-3. When this model is used with our implanted unannealed and annealed results the following three points must be taken into account: 1. The dc conductivity a, and the mobility, p, are dependent of the annealing temperatures (Christofides, Guibaudo, Jaouen, 1987, 1989a, 1989b). 2. The conductivity effective mass m* depends on the degree of inhomogeneity of the material. 3. E is not constant for implanted Si annealed at various temperatures.

V. Discussion and Analysis 1. INFLUENCE OF ANNEALING TEMPERATURE ON THE PLASMA WAVELENGTH Figures 8(b) to 8(e) clearly show that the increase of annealing temperature provokes a decrease of the fringe amplitude, which indicates that the

2 REFLECTION SPECTROSCOPY ON ION IMPLANTED SEMICONDUCTORS 63

index of refraction was decreased during the isochronal annealing. In fact, the difference between the refractive index of implanted layers and substrate disappears as the annealing temperature increases. Samples implanted with 5 x l O I 4 P+/cm2 (Fig. 8(b)) need only an annealing of up to 550°C for 1 h to make the fringes disappear, whereas annealing over 800°C is necessary in the case of samples implanted at high doses. Using the data presented in Table I and the relation

the concentration of free carriers, N , is calculated. This concentration N assumes that all the implanted carriers are electrically active. The constant d is the average value [d zz 1/2(d, + d,)] where d , and d, are the thicknesses of implanted layers obtained from spreading resistance and SUPREM I11 simulation of the junction depth, respectively (Othonos et al., 1994). In Table 111, we can find the concentration of free carriers for several implanted and highly annealed samples. Using Eq. (3) and Table I, it is easy to understand why 1, (of sample R900, 1 x 10l6 P+/cmZ)is smaller than 1, (of sample R900, 5 x 1015Pf/cm2). The plasma wavelength 1, of the samples annealed at 1100°C are presented in longer wavelengths because their carriers per volume are smaller due to their high diffusion in the wafer. Table V reports some experimental and theoretic data concerning 8 samples annealed at temperatures over 800°C. In Table V, we also report the values of the “spreading” and “SUPREM” concentrations, the freeTABLE V VARIOUS VALUES FOR EIGHTSAMPLES IMPLANTEDAND ANNEALED AT DIFFERENT CONDITIONS @ (P +/cm2)

5

x 1014

1 5 1 1 1

x x x x x x

1 5x

T,(“C)

1015 1015 1OI6

1016

10l6 loL6 1 0 1 ~

1100 1100 1100 1100 lo00 900 800 900

N , (cm-3)

N , (cm-’)

3.31 x 6.58 x 2.76 x 4.16 x 6.67 x 0.99 x 1.43 x 5.56 x

2.94 x 5.74 x 2.46 x 3.86 x 5.26 x 0.97 x 1.59 x 5.56 x

10” 10” 1019

10” 10” lo2’ 10” 10”

x,annealing temperature; N , ,

10” 1ol8

10” 10”

N ( ~ m - ~ ) A,(m) 3.11 x 6.13 x 2.60 x 4.00 x

lo’* 10”

1019 1019

1019

5.88 x 1019

10’’

0.98 x lo2’ 1.51 x lo’’ 5.56 x 1019

10”

10”

12.8 10.5 6.3 5.0 5.2 3.3 3.1 3.6

spreading free-carrier concentration ( %@,hin);N , , A,, plasma wavelength taking into account the average free-carrier concentration.

@, implantation dose;

SUPREM free-carrier concentration ( %@/d>; N , average free-carrier concentration (%@id);

64

A. SEASAND C. CHRISTOFIDES

carrier impurities concentration N (obtained from Eq. (8)), and the experimental plasma wavelength A, (Christofides et al., 1994). Figure 14 presents the plasma wavelength of the minimum reflectivity as a function of carrier concentration for n-type Si (Seas et al., 1995). The solid line is obtained from the theoretic model presented in Part IV. In Fig. 14, we can see the good fitting of our experimental data. The good agreement of the theory with the data shows that w2z2>> 1 is a good approximation for implanted and highly annealed Si wafers. However, from these data one can conclude that only after annealing at 800°C is there almost complete activation of implanted impurities. The theoretic curve (Schumann and Phillips, 1967) of Fig. 14, which was the main test of these experimental results, was calculated with well-determined p,,,'E and m* at each concentration. Here, we consider p,, cL, and m* as constant with annealing temperature over than 800°C. The assumption of constant mobility is justified since the mobility varies slightly in the range of 800 to 1100°C (Christofides et al., 1987; 1989a, 1989b), as indicated in Table IV. The variation of the mobility of the various wafers also was observed from the Hall effect. A small deviation of our experimental points from the theoretic curve in Fig. 14 is expected for three main reasons: 1. Annealing at 800°C is probably not sufficient for a complete recrystallization. 2. There is a slight variation of mobility due to the fact that the plasma wavelength in implanted materials and materials annealed at low temperatures, which are not completely recrystallized, shifts to a longer wavelength. 3. The Schumann and Phillips (1967) model assumed that the numeric value of m* was 0.26, whereas in $2 of Part I1 it is shown that the effective mass for the sample series W3 to W6 varies. It is also important to note that in Fig. 10(f) even annealing at 700°C is not sufficient to activate the implanted impurities. In fact, according to Eq. (8), where a complete activation is assumed, this sample possesses 1.88 x 10'' cm-3 carriers. In the case in which 1% of these impurities are activated, there would appear a reflectivity minimum for A, = 20 pm, according to the Schumann and Phillips model. This has not happened in our case. However, this conclusion is not completely realistic. As shown in Eq. (3), A, does not only depend on the concentration N but also on the dc reflectivity and mobility and on the effective mass. In fact, according to the theoretic curve presented in Fig. (14), in the case in which 1% of the carriers are activated (1.88 x 10'' cmP3),there corresponds a plasma wavelength of 20pm. How-

2 REFLECTIONSPECTROSCOPY ON ION IMPLANTEDSEMICONDUCTORS65

ever, it should not be ignored that this theoretic curve corresponds to completely crystalline Si. Christofides et al. (1987, 1989a) and Othonos et al. (1994) have shown, by performing electrical measurements, that dc conductivity and mobility vary even for annealing temperatures up to 700°C. In our case, annealing at 700°C is probably not sufficient for a complete recrystallization. However, the value of the effective mobility taken for this model (n-type c-Si) may be different from that of an inhomogeneous material. Thus the plasma wavelength in implanted materials and at low temperature materials annealed, which are not completely recrystallized, shifts to a longer wavelength. Electrical activation lower than 1%, no doubt, is an underestimation. Another point, which we believe is an additional important test of the theoretic model adopted in this study, is to check whether the obtained carrier density corresponds to each experimental point I , of Fig. 14 by using the following well-known relation: 1

Po =epo N

(9)

The dc mobility can be calculated by using the carrier density given in Table V and the spreading resistance data for the resistivity. Figure 15 shows the mobility as a function of the carrier density. The space included between the two solid lines corresponds to experimental data found by several researchers (Masetti, Severi, Solmi, 1983). There is close agreement between the experimental points obtained in this work and previously published data. Minor disagreement exists in the literature between experimental data obtained for samples annealed at temperatures lower than 1100°C. Figure 16 presents the reflection minimum as a function of the annealing temperature for the implanted series W3 to W6 (Seas et al., 1995; Christofides, Seas, and Othonos, 1995). We note that 1, remains almost constant versus the annealing temperature for samples implanted with a lower dose (series W3), whereas 1, increases versus implantation dose, in qualitative agreement with Eq. (3). Samples implanted at higher doses present reflectivity minima at shorter wavelengths. For example, the highly implanted samples (series W6) present a reflectivity minimum around 3 pm after an annealing at 800°C. It is also important to note that the minimum reflectivity for the four wafers implanted at around (Dc (W3) and over (Dc (W4 to W6) increases as a function of the annealing temperature. From Eq. (3) and Tables I11 and IV, it is easy to understand why I , increases with T,, with a decrease in the volume of free-carrier concentrations due to their diffusion under the surface of the wafer.

A. SEASAND C . CHRISTOFIDES

66

01 I8

'

19

.

Log(N)

. . 'ao'

. ".' 21

'

. * . '

22

[N: cm"]

FIG. 15. The wavelength minimum reflectivity I , versus the logarithm of the carrier concentration N (solid line). Schumann, Jr., P. A., and Phillips, R. P. (1967). Comparison of Classical Approximations to Free Carrier Absorption in Semiconductors.Solid State Electron. 10, 943.

5

.g e,

*oh

800

-

900

1000

Iloc

Aanealing Temperature ( "C) FIG. 16. The wavelength minimum reflectivity. I , as a function of annealing temperature for four different implantation doses.

2 REFLECTIONSPECTROSCOPY ON ION IMPLANTED SEMICONDUCTORS67

2. EFFECTIVE MASSVERSUS ANNEALING TEMPERATURE This section describes the variation of the effective mass m* as a function of the free-carrier concentration. An attempt to study the influence of the annealing temperature on the effective mass of free carriers in implanted and annealed Si wafers is presented. As is well known, the effective mass theory for doped semiconductors cannot be applied directly to disordered or amorphous semiconductors (the cases of unannealed and partially annealed semiconductors) because it is formulated in terms of the momentum-space wave functions of the crystal. High doses produce the incorporation of species other than the introduced impurities, such as vacancies, bivacancies, vacancy complexes, and structural disorder. In the case of amorphous materials, there have been numerous discussions as to whether the effective mass concept has any realistic meaning. Street (1991) and Kivelsen and Gelatt (1979) strongly support the idea that m* remains a significant physical parameter even in the case of amorphous materials. Our optical measurements indicate the influence of all the previous effects, and such measurements can result in useful information. In the following, we use the results presented in Fig. 16 concerning reflection minima to get an idea of how the conductive effective mass changes for the various samples of series W3 to W6. From Eq. (3) we define the ratio

in order to study the effect on the effective conductive mass as a function of the annealing temperature. Therefore, the ratio of the conductive effective masses can be written as follows:

It is obvious that due to the distribution of carrier density in the implanted layer it is impossible to determine an exact value for concentration of carriers and there is always an error associated with such measurements. The two different methods used to estimate the carrier concentration in the implanted layer indicate that there is an error of approximately 30%, which is shown in Fig. 17 as error bars. Figure 17 presents the variation of the normalized effective conductive mass of free carriers as a function of the

68

A. SEASAND C. CHRISTOFIDES 2.5

w4

10

1

0.0

800

900

0.0

lo00

1100

800

Annealing temperature 2.5 2.0

F

1

0.5

0.0

lo00

1100

Annealing Temperature ( "C) 1

2.5

c W6

w5

t

900

T T

& u 800

900

law,

0.0

1100

Annealing Temperature ( "C)

800

sa,

loo0

1100

Annealing Temperature ( "C)

FIG. 17. Normalized ratio of conductive effective masses as a function of the annealing temperature (TJ

annealing temperature. We note that the effective mass of the heavily implanted samples changes with annealing temperature. It is accepted that annealing of ion-implanted wafers leads to annihilation of defects, which, in turn, leads to a decrease in carrier scattering and therefore to a change of the effective mass. The variation of m* as a function of the annealing temperature shows the sensitivity that this physical parameter has to annealing.

VI. Summary

In conclusion, optical reflection and transmission measurements can be used to obtain information concerning the influence of implantation dose and annealing temperature on the kinetics of reconstructions. From the

2 REFLECTIONSPECTROSCOPY ON ION IMPLANTEDSEMICONDUCTORS 69

interference fringes, we can obtain information concerning the amorphouscrystalline transition as a function of the annealing temperature. From the presence of the absorption of free carriers in such samples, we can obtain information concerning the electrical activation of these carriers. The main results of our study can be summarized as follows. 1. FTIR spectroscopy is a promising technique toward nondestructive evaluation of implanted materials. 2. This technique is sensitive to the implantation dose and to the annealing temperature. 3. The free-carrier absorption phenomenon is sensitive in the case of heavily implanted and highly annealed samples. 4. The minimum reflectivity is a practical reference for the evaluation of the percentage of the activation of free carriers. 5. For implantation doses lower than the critical dose Oc,the energy of implantation does not influence the optical properties. 6. The classic model by Schumann and Phillips is in good agreement with our experimental results. The fit of the plasma wavelength as a function of the concentration of free carriers shows that an annealing of 800°C or more for 1h is sufficient for a complete activation of carriers. 7. As expected from the theory, all the highly annealed samples present a wavelength at which we have a minimum in the reflectivity. In the case of samples implanted at temperatures lower than Oc, A, remains constant with the annealing temperature, whereas for samples implanted with high doses, the plasma wavelength increases with the annealing temperature. 8. The effective mass has been studied as a function of the annealing temperature. Our analysis suggests that m* is influenced by the annealing temperature due to the annealing kinetics of defects. More research on this subject is needed to verify and further develop understanding in this area of semiconductors.

REFERENCES Brierley, S. K., Lehn, D. S., and Grabinski, A. K. (1988). Implant-Dose Mapping Using Infrared Transmission.J . A p p l . Phys. 63, 5085-5087. Brown, R. L., Schoonveld, L., Abels, L. L., Sundararn, S., and Raccah, P. M. (1981). Electroreflectance of Ion-Implanted GaAs. J . A p p l . Phys. 52, 2950-2957. Chapman, P. W., Tufte, 0. N., Zook, J. D., and Long, D. (1963). Electrical Properties of Heavily Doped Silicon. J . A p p l . Phys. 34, 3291. Christofides, C. (1992). Annealing Kinetics of Defects of Ion-Implanted and Furnace-Annealed Silicon Layers: Thermodynamic Approach. Sernicond. Sci. Technol. 7, 1283- 1294.

70

A. SEASAND C. CHRISTOFIDES

Christofides, C., Guibaudo, G., and Jaouen, H. (1987). Etude de silicium implante a I’arsenic par effet de transport. Influence du recuit thermique. Revue Phys. Appl. 22,407-412. Christofides, C., Jaouen, H., and Guibaudo, G. (1989a). Electronic Transport Investigation of Arsenic-Implanted Silicon. 1. Annealing Influence on the Transport Coefficients. J . Appl. Phys. 65,4832-4839. Christofides, C., Guibaudo, G., and Jaouen, H. (1989b). Electronic Transport Investigation of Arsenic-Implanted Silicon. 11. Annealing Kinetics of Defects. J . Appl. Phys. 65,4840-4844. Christofides, C., Othonos, A,, Bisson, M. Boussey-Said, J. B. (1994). Optical Spectroscopy on Implanted and Annealed Silicon Wafers: Plasma Resonance Wavelength. J. Appl. Phys. 75, 3377-3384. Christofides, C., Seas, A,, and Othonos, A. (1995). Reconstruction Mechanisms in Ion Implanted and Annealed Silicon Wafers. Defects Difusion Forum 117, 45-64. Crowder, B. L., Title, R. S., Brodskey, M. H., and Pettit, G. D. (1970). ESR and Optical Absorption Studies of Ion-Implanted Silicon. Appl. Phys. Lett. 16, 205-208. Engstrom, H. (1980). Infrared Reflective and Transmissivity of Boron-Implanted, LaserAnnealed Silicon. J . Appl. Phys. 51, 5245-5249. Fredrickson, J. E., Waddell, C. N., Spitzer, W. G., and Hubler, G. K. (1982). Effect of Thermal Annealing of the Refractive Index of Amorphous Silicon Produced by Ion Implantation. Appl. P h p . Lett. 40, 172- 174. Gibbons, J. F. (1968). Ion Implantation in Semiconductors. Part I: Range Distribution Theory and Experiments. Proc. IEEE 56, 295-320. Gibbons, J. F. (1972). Ion Implantation in Semiconductors. Part 11: Damage Production and Annealing. Proc. IEEE 60, 1062-1096. Hart, R. R., and Marsh, 0. J. (1969). Changes of Optical Reflectivity (1.8 to 2.2eV) Induced by 40-keV Antimony Ion Bombardment of Silicon. Appl. Phys. Lett. 14, 225-226. Heavens, 0.S. (1964). Optical Properties of Thin Films. Butterworths, London, 77-79. Howarth, L. E., and Gilbert, J. F. (1962). Solid State Commun. 10, 236-237. Hubler, G. K., Waddell, C. N., Spitzer, W. G., Fredrickson, J. E., Prussin, S., and Wilson, R. G. (1979a). High-Fluence Implantations of Silicon: Layer Thickness and Refractive Indices. J . Appl. Phys. 50, 3294-3303. Hubler, G. K., Malmberg, P. R., and Smith, T. P. (1979b). Refractive Index Profiles and Range Distributions of Silicon Implanted with High-Energy Nitrogen. J . Appl. Phys. 50, 71477155. Hubler, G. K., Malmberg, P. R., Waddell, C. N., Spitzer, W. G., Fredrickson, J. E. (1982). Electrical and Structural Characterization of Implantation Doped Silicon by Infrared Reflection. Radiation Efects 60, 35-47. Hummel, R. E., Xi, W., Holloway, P. H., and Jones, K. A. (1988). Optical Investigations of Ion Implant Damage in Silicon. J. Appl. Phys. 63, 2591-2594. Irvin, J. C. (1962). Resistivity of Bulk Silicon and of Diffused Layers in Silicon. Bell Systems Tech. J . 41, 387-410. Kachare, A. H., Spitzer, W. G., Euler, F. K., and Kahan, A. (1974). Infrared Reflection of Ion-Implanted GaAs. J. Appl. Phys. 45, 2938-2946. Kachare, A. H., Spitzer, W. G., Fredrickson, J. E., and Euler, F. K. (1976a). Measurements of layer thicknesses and refractive indices in high-energy ion-implantation GaAs and Gap. J. Appl. Phys. 47, 5374-5381. Kachare, A. H., Spitzer, W. G., and Fredrickson, J. E. (1976b). Refractive Index of IonImplanted GaAs. J. Appl. Phys. 47,4209-4212. Kireev, P. (1975). The Physics of Semiconductors, Mir, Moscow. Kivelsen, S., and Gelatt, Jr., C. D. (1979). Phys. Rev. B19, 5160-5177.

2 REFLECTIONSPECTROSCOPY ON ION IMPLANTED SEMICONDUCTORS71 Kurtin, S., Shifrin, G . A., and McGill, T. C. (1969). Ion Implantation Damage of Silicon as Observed by Optical Reflection Spectroscopy in the 1 to 6 e V Region. i p p l . Phys. Lett. 14,223-225. Kwun, S., Spitzer, W. G., Anderson, C. L., Dunlap, H. L., and Vaidyanathan, K. V. (1979). Optical Studies of Be-Implanted GaAs. J . Appl. Phys. 50, 6873-6880. Masetti, G., Severi, M., and Solmi, S. (1983). IEEE Trans. Electron Devices, ED-30, 764-769. Othonos, A., Christofides, C., Boussey-Said, J., and Bisson, M. (1994). Raman Spectroscopy and Spreading Resistance Analysis of Phosphorus Implanted and Annealed Silicon. J . Appl. Phys. 75, 8032-8038. Prussin, S., Margolese, D., and Tauber, R. N. (1985). Formation of Amorphous Layers by Ion Implantation. J . Appl. Phys. 57, 180-185. Seas, A., Eleftheriou, M.-E., Christofides, C., and Theocharis, C. R. (1995). Infrared Spectroscopy and Electrical Characterization of Phosphorus Implanted and Annealed Silicon Layers. Nucl. Instrum. Meth. Phys. Res. B103,46-55. Seeger, K. (1988). Semiconductor Physics, Springer Series in Solid State Sciences 40 (M. Cardona, P. Fudle, K. von Klitzing and H.-J. Queisser, eds.). 4th ed. Springer-Verlag, Berlin, New York, Pans, Tokyo, Chapt. 11. Smith, R. A. (1959). Semiconductors, Cambridge University Press, Cambridge. Spitzer, W. G., Waddell, C. N., Narayanan, G. H., Fredrickson, J. E., and Prussin, S. (1977). Free-Carrier Plasma Effects in Ion-implanted Amorphous Layers of Silicon. Appl. Phys. Lett. 30,623-626. Street, R. A. (1991). Hydrogenated Amorphous Silicon, Cambridge University Press, Cambridge, Chapt. 5. Tatarkiewicz, J. (1989). Optical Effects of High Energy Implantations in Semiconductors. Phys. Status Solidi (b) 153, 11-47. Schumann, Jr., P. A,, and Phillips, R. P. (1967). Comparison of Classical Approximations to Free Carrier Absorption in Semiconductors. Solid State Electron 10, 943-948. Ure, R. W. (1972). Semiconductors and Semimetals. Thermoelastic effect in 111-V compounds. R. K. Willarson and A. C. Beer, eds.), Vol. 8, Transport and optical phenomena. Academic Press, New York, 86. Waddell, C. N., Spitzer, W. G., Hubler, G . K., and Fredrickson, J. E. (1982). Infrared Studies of Isothermal Annealing of Ion-Implanted Silicon: Refractive Indices, Regrowth Rates, and Carrier Profiles. J . Appl. Phys. 53, 5851-5862. Wagner, H. H., and Schaefer, R. R. (1979). Contactless Probing of Semiconductor Dopant Profile Parameters by IR Spectroscopy. J . Appl. Phys. 50, 2697-2704. Wang, K.-W., Spitzer, W. G., Hubler, G . K., and Donovan, E. P. (1985). Effect of Annealing on the Optical Properties of Ion-implanted Ge. J . Appl. Phys. 57, 2739-2751.

This Page Intentionally Left Blank

SEMICONDUCTORS AND SEMIMETALS,VOL. 46

CHAPTER 3

Photoluminescence and Raman Scattering of Ion Implanted Semiconductors. Influence of Annealing Andreas Othonos Constantinos Christojdes DEPARTMENT OF NATURAL SCIENCES OF CYPRUS UNIVERSITY CYPKUS NICOSIA,

1. INTRODUCTION. . . . . . . . . . . . . . . . . . . 11. PHOTOLUMINESCENCE AND RAMANSCATTERING THEORY.

. . . . . . . . .

. . . . . . . . . . 1. Photoluminescence Theory . . . . . . . . . . . . . . . . . . . . . . 2. Raman Scattering Theory . . . . . . . . . . . . . . . . . . . . . . . 111. PHOTOLUMINESCENCEAND RAMANSCATTERING TECHNIQUES . . . . . . . . . 1. Common Photoluminescence Techniques . . . . . . . . . . . . . . . . . 2. Raman Scattering Techniques . . . . . . . . . . . . . . . . . . . . . 3. Time-Resolved Measurements . . . . . . . . . . . . . . . . . . . . . IV. CHARACTERIZATION OF ION-IMPLANTED SEMICONDUCTORS . . . . . . . . . . 1. Photoluminescence Experimental Studies . . . . . . . . . . . . . . . . 2. Ramun Studies in Ion-Implanted Semiconductors . . . . . . . . . . . . . V. SUMMARY AND FUTURE PERSPECTIVES. . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

13

14 14 17 79 19

80 82

84 84 91 111 112

I. Introduction Ion implantation is a key technology in the preparation of doped semiconductors with controlled impurity profiles for device applications. The key in designing and manufacturing such highly sophisticated semiconductors lies in the ability of characterization and probing of these materials. There are many characterization techniques with advantages and disadvantages. Among these methods, optical techniques stand out because they require little and often no sample preparation. In general, the sample is unaltered and the measurement itself does not cause damage. The optical 13 Copyright Q 1997 by Academic Press All rights of reproduction in any form reserved. 0080-8784/97 $25

14

A. OTHONOS AND C.

CHRISTOFIDES

beam, which generally is a laser beam, can easily be manipulated. Even more important, it can be focused down to the size of the wavelength of light, probing submicron areas of the semiconductor. In addition, it is possible to probe the material as the light propagates into the sample. The penetration depth of the light depends on the wavelength and the sample properties, so that the probing depth can range from nanometers to micrometers. Two of the main optical characterization techniques are photoluminescence (PL) and Raman scattering (RS). In PL, light separates the charge carriers within the band or impurity structure of a semiconductor, whose later recombination produces characteristic emission. In RS, the energy of an incoming photon is altered by nonlinear interaction with phonons, carriers, or impurities in the material, to produce a frequency-shifted outgoing photon. The introduction of lasers, with their abilities to deliver high peak powers and to focus the beam to submicron areas, has opened new ways of conditioning semiconducting materials designated for highly sophisticated technology. For example, pulses from a powerful laser are used to anneal semiconducting materials after ion implantation. Characterization of ionimplanted semiconductors can also be done, nondestructively, using laser light. Photoluminescence is one such technique in which excitation of the semiconductor with laser light results in the emission of light. Light scattering is also a very powerful technique for studying semiconductors and the effect of ion implantation in these materials. In particular, Raman scattering is a widely used technique for the characterization of semiconductors (Kachare et ai., 1976; Beserman and Bernstein, 1977; Balkanski, Morhange, and Kanellis, 1981; Forman, Bell, and Myers, 1981; Kirillov, Powell, and Nodul, 1985; Fortner and Lannin, 1988). In this chapter we give an overview of the application of PL and RS on ion-implanted semiconductors. In Part I1 we begin with basic theoretic background and then describe the common experimental techniques for measuring both PL and RS. In Part I11 we give a summary of some of the work carried out over the last few years using PL and RS in ion-implanted semiconductors. Finally, a summary is presented in Part IV.

11. Photoluminescence and Raman Scattering Theory 1. PHOTOLUMINESCENCE THEORY

Optical excitation of a semiconductor and subsequent emission of radiation by this system is referred to as photoluminescence (PL) (Pankove,

3 PHOTOLUMINESCENCE AND RAMANOF ION IMPLANTEDSEMICONDUCTORS75

1975). Emission of radiation may be considered, in simple terms, as the inverse of the absorption process. An excited electron occupying an energy state higher than it would under equilibrium conditions, makes a transition to an empty lower-energy state and all or most of the energy difference between the two states can be emitted as electromagnetic radiation. We begin by discussing the various absorption processes in semiconductors. This will help to show why and how PL occurs. When a semiconductor is excited by an electromagnetic radiation where the energy quantum is larger than its bandgap, an electon-hole pair is generated. This fundamental absorption, where an electron is excited from the valence to the conduction band of a semiconductor, is strongly affected whether the bandgap is direct or indirect. In a direct gap semiconductor such as gallium arsenide (GaAs), the electron at the valence band maximum is excited via a vertical transition (in k space) to the conduction band minimum, because the incoming photon has negligible momentum compared with that of an electron (during an electron-hole generation). Energy conservation requires that

where, fio, is the photon energy, and E , and Ei are the final and initial states, respectively. For indirect materials such as silicon (Si) and germanium (Ge), the electron requires additional momentum to reach the conduction band minimum. This is achieved with the interaction of a phonon. The energy conservation is then given by the following expression:

where hQ is the energy of the phonon, and the plus and minus signs correspond to the phonon emission and absorption, respectively. The need of a phonon to conserve momentum in an indirect semiconductor makes the absorption less probable than in direct absorption. After absorption of electromagnetic radiation by a semiconductor, the generated electron-hole pairs recombine through either radiative or nonradiative processes. In a direct-gap semiconductor the radiative recombination transition, just as in the absorption, is vertical and the conservation of energy is given by Eq. (1). For an indirect-gap material a phonon must be involved in the transition, with the energy of the emitted photon given by Eq. (2). The probability of radiative transition in direct-gap semiconductors is much higher than it is for indirect-gap material. For example, the probability of radiative transition in Si is only lo-’ that of direct-gap material such as indium antimonide (InSb).

76

A. OTHONOS AND C . CHRISTOFIDES

We should point out that although radiative emission appears to be the inverse of absorption, there is one important difference. In absorption all the electronic states whose energy difference satisfies the conservation law participate in the process, which leads to broad spectral features. In the luminescence, however, the recombining electrons and holes both have well-defined energies. The result is that the PL features are narrow, thus making this technique a better tool for characterization than is absorption spectroscopy. Other radiative processes include radiative recombination of excitons. In a pure semiconductor the Coulomb attraction between the generated electron and hole can bind them into quasi-hydrogen molecules called an exciton. High doping reduces the probability of exciton formation because the free charges tend to screen out the Coulomb interaction. An exciton supports a set of bound energy levels such as those in a hydrogen atom with a binding energy of a few MeV. In a direct gap semiconductor, the energy of the photon emitted when the exciton collapses radiatively is (Elliott, 1957; Pankove, 1975)

where E,, is the binding energy of the exciton. For an indirect semiconductor the emitted photon energy is (Elliott, 1957; Pankove, 1975)

The behavior of excitons can be important in PL and may indicate the quality of the sample. In samples that are not very pure, Coulomb forces from donor, acceptor, or neutral impurities attract the free excitons to form bound excitons. These bound excitons still recombine and emit photons with energies that satisfy Eqs. (3) and (4). However, the photons emitted are of lower energy than those from a free exciton, because there is an additional localization energy involved. Radiative transitions can also take place between impurity states and the valence and conduction band. These transitions are less likely than band-to-band transitions; however, they still are strong enough to be considered. Transitions such as donor to acceptor are also possible. This variety of transitions means that PL spectra from samples with impurities are typically complex and involve combination of band-to-band excitonic and impurity features.

3 PHOTOLUMINESCENCE AND RAMAN OF ION IMPLANTED SEMICONDUCTORS77

2. RAMANSCATTERING THEORY When light encounters the surface of a semiconductor, most of it is reflected, transmitted, absorbed, or Rayleigh scattered because of first-order elastic interaction with electrons, phonons, and impurities. Under these interactions there is no change in the incident photon frequency. However, a very small part of the light interacts inelastically with phonons, producing scattered photons whose frequencies are shifted. This frequency-shifted light is called Raman scattered radiation. The scattered photons either gain energy by absorbing a phonon (anti-Stokes shift) or lose energy by emitting a phonon (Stokes shift). Conservation of energy and momentum requires that (Cardona, 1975)

where wi and o,are the incoming and scattered photon frequencies respectively, q, and qi are the incoming and scattered photon wave vectors, respectively, and R and k are the phonon frequency and wave vector, respectively. Although RS is a weak process, lasers provide sufficient power to measure the spectra. The full theory of RS is complex, which is the main reason detailed line-shape analysis is not often used for this kind of spectroscopy. However, we may use a simple classic picture to understand the basic process. The shifted RS photons can be described as side bands at the phonon frequency, which arise from the nonlinear interaction between the radiation and the semiconductor lattice (Cardona, 1975). A zone center phonon produces a lattice deformation that preserves the translational symmetry. We can therefore calculate the dielectric constant as a function of phonon coordinates u. Expanding the dielectric constant as a function of u, we have (Cardona, 1975)

The derivatives in Eq. (6) define the first- and second-order Raman tensor and can be simply obtained from band-structure dielectric constant calculation. For an incident electric field E , exp( - iwt) and a phonon coordinate u = u, exp[ _+ iRt], Eq. (6) gives the induced dipole moment to first order in u, (scattering b y one phonon):

P(w, & 0)

5

d& du

- u,E,

exp[i(w,

+ R)t]

(7)

78

A. OTHONOS AND C. CHRISTOFIDES

The plus and minus signs in front of R correspond to the anti-Stokes and Stokes radiation, respectively. Thus the intensity of the scattered dipole radiation, which is the sum of intensities of the Stokes and the anti-Stokes lines for the single phonon scattering, is proportional to

This equation contains the thermal average of the phonon displacement (u:). The square dependence of the already small polarization term d&/du suggests that RS is indeed a very weak process. Of the two Raman single-phonon bands, the Stokes modes are the stronger. The mode strengths depend on the number of phonons available, which is simply a count of the distribution of normal mode harmonic oscillators in the lattice at a temperature. The results can be derived from the matrix element for photon creation or annihilation, which depends on the number of phonons. From the Planck distribution function, the relative strength of lines can be calculated as

zanti-Stokes ~

IStokes

[ &]

= exp -

(9)

Clearly, as the temperature approaches zero the anti-Stokes line vanishes and the Stokes line dominates. The RS process comes from the interaction of the electric field vector and the polarization vector, and thus the direction of the electric field relative to the crystal geometry defines the strength of the Raman signal. For crystals with the diamond structure and in the widely used backscattering experimental geometry, only LO phonons are seen when a (100) surface is examined; only TO phonons appear from (1 10) surfaces; and both TO and LO modes appear from ( 1 11) surfaces. These rules apply for ideal crystals and surfaces but break down in disordered or polycrystalline materials. Thus RS can yield the crystal symmetry or give information about the degree of disorder. We should point out that there is a situation in which the weak RS can be enhanced. For example, if the energy of the exciting photon matches a fundamental gap in the semiconductor, the energy transferred to the lattice increases dramatically and so does the Raman signal.

3 PHOTOLUMINESCENCE AND RAMANOF ION IMPLANTED SEMICONDUCTORS 79

111. Photoluminescence and Raman Scattering Techniques

In this section we describe briefly some of the most common techniques used in PL and RS measurements for the characterization of semiconducting materials. We hope this will give the reader a better idea of the advantages and limitations of the techniques.

1. COMMON PHOTOLUMINESCENCE TECHNIQUES The excitation source required for PL measurements may be any source that generates photons with sufficient energy to excite electron-hole pairs. This source could be as simple as an arc lamp, or a cw laser source for steady-state measurements, or more complicated pulsed lasers for transient measurements. The pulsed lasers may be Q-switched lasers generating nanosecond pulses or state-of-the-art mode-locked lasers generating picosecond and femtosecond pulses. The complexity of the experiment increases with the complexity of the source. In general, transient measurements are more sophisticated than are cw measurements due to the additive parameter of time. For steady-state measurements where cw sources are used, the experiments are straightforward. The source of excitation, which in many cases is a cw argon ion laser, is used to excite electron-hole pairs (although titanium: sapphire, dye-laser, and diode lasers may be used as well). The P L signal is then collected and spectrally analyzed with a spectrometer equipped with a sensitive detector appropriate for the spectral region of interest. To improve sensitivity, the excitation source is usually modulated and the signal is processed with a phase-sensitive amplifier (lock-in amplifier). The PL sensitivity can also be increased, or the measurement time decreased, if the dispersive grating monochromator is replaced by a Michelson interferometer (in other words replacing the grating spectrometer with a Fourier transform infrared (FTIR) spectrophotometer). Figure 1 shows a schematic of a typical experimental setup used in obtaining PL measurements. Photoluminescence measurements are generally performed below room temperature. Cooling of the semiconductor samples results in sharper, more readily identifiable peaks. Lower temperatures reduce the thermal broadening of the excited carrier energies, which at temperature T is roughly k,T This gives a significant broadening that at room temperature is 25 MeV, but reduces to 6 MeV at 77 K and less than 1 meV at liquidhelium temperatures.

80

A. OTHONOS AND C. CHRISTOFIDES

FIG. 1. Schematic of a typical photoluminescence setup using a laser as an excitation source and a spectrometer to analyze the luminescence. The spectrometer may be a dispersive grating spectrometer or a Fourier transform infrared (FTIR) spectrophotometer.

2. RAMANSCATTERING TECHNIQUES One of the main factors influencing the design of RS experiments is that the weak Raman signal is spectrally very near the exciting laser light. Thus the weak Raman phonon peaks must be measured against a background of intense Rayleigh scattering. Raman measurements require strong laser source, well-designed optics to filter out the Rayleigh signal, and sensitive detector to record a very weak signal. Most experiments are carried out with argon ion or krypton ion lasers because these lasers deliver watts of power in many lines from ultraviolet (UV) to the red end of the spectrum. Available dye lasers and tunable solid-state lasers may be used to achieve continuous tuning when required. A typical Raman experimental setup is shown in Fig. 2. The laser is focused on the semiconductor samples with a lens and the signal usually is collected in a backscattering configuration. This Raman signal is directed into a double (and sometimes a triple) monochromator to discriminate the unwanted Rayleigh light and spectrally resolve the signal. Since the Raman shift is small compared with the excitation frequency, the scattered light

3 PHOTOLUMINESCENCE AND RAMAN OF ION IMPLANTED SEMICONDUCTORS 81

FIG.2. Schematic of a typical experimental setup used in Raman scattering experiments. An objective lens is used to focus the laser beam onto the sample and the backscattered Raman signal is directed into a triple-pass spectrometer and then analyzed with a charge-coupled device (CCD) array.

remains within the same frequency range. For excitation in the visible range, standard photomultiplier tubes work very well as detectors. However, over the past few years Si charge-coupled device (CCD) arrays have become the detector of choice for RS in the visible spectrum. Using a CCD array as a detector at the output of the spectrometer, a complete spectral profile of the Raman phonon peak can be obtained without scanning the spectrometer. The spectral width that the CCD is able to capture is typically 5 to 30nm but mostly depends on the resolution required by the experiment. The main reason that the RS is normally performed with laser sources that emit visible or UV radiation (as opposed to infrared (IR) radiation) is that the scattered intensity is proportional to o4 (see Eq. (8)). However, there is an advantage to operating in the IR region. Fluorescence, which may sometime overwhelm the Raman signal, is less prevalent in the near-IR than in the UV-visible region. Raman systems incorporating Fourier methods offer enough sensitivity to compensate for the loss in scattering intensity. A typical system uses as a source a neodymium: yttrium-alumi-

82

A. OTHONOS AND C . CHRISTOFIDES 4.00

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.95 1.00 1.05 1.10

Photon Energy (eV)

FIG.3. Raman and photoluminescence spectra from a silicon sample (unimplanted unannealed and unimplanted annealed at 1100°C) and obtained with a Fourier Transform-Raman Fourier transform infrared spectrophotometer. Othonos, A,, and Christofides. C. (1995a). Photoluminescence Measurements on Phosphorus Implanted Silicon: Annealing Kinetics of Defects. J . Appl. Phys. 78, 796-800.

num-garnet (Nd:YAG) laser operating at 1.064pm, and instead of the double or triple monochromator, a Fourier spectrometer collects the radiation. A special filter rejects the strong Rayleigh scattering light before it enters the interferometer. The detector most often is a cooled indiumgallium-arsenic (InGaAs) unit. Figure 3 shows the RS spectra of a Si sample obtained using such technique. One of the strengths of RS, when polarization analysis is included, is its sensitivity to sample orientation. This means, however, that in order to obtain and analyze the spectra, the relation between the incident light and the sample surface must be carefully established.

3. TIME-RESOLVED MEASUREMENTS Time-resolved PL and RS measurements may be divided into transient and excite-probe techniques. Transient techniques refer to situations in which the required time resolution may be achieved from the temporal response of the detectors or by simply placing a switch (usually a pockel

3

PHOTOLUMINESCENCE AND RAMANOF ION IMPLANTEDSEMICONDUCTORS 83

FIG.4. Time-resolved measurements. (a) A schematic of a pump-probe experimental setup used for time-resolved photoluminescence studies. (b) A schematic of pump-probe experiments used for time-resolved optical phonon studies in Raman scattering.

84

A. OTHONOS AND C . CHRISTOFIDES

cell) in front of the detector. For example, resolution of nanoseconds may be achieved with a fast response detector such as a photomultiplier or Si. If time resolution of a picosecond or subpicosecond is required, an exciteprobe technique must be used (Othonos et d.,1991). With the development of laser sources and detectors, time-resolved measurements are becoming more widely accepted and are seeing use as an optical characterization technique in ion-implanted wafers. Typical schematics of experimental setups for time-resolved measurements are presented in Figs. 4(a) and 4(b).

IV. Characterization of Ion-Implanted Semiconductors Photoluminescence has become a very powerful tool in detecting impurities and defects in semiconductors. A given impurity produces a set of characteristic spectral features that serve to identify it. In some cases PL may be used to measure impurity concentrations and determine semiconductor bandgaps. The half-widths of PL peaks can also be used as indicators for sample quality and crystallinity. Similarly, RS parameters -such as frequency shift, intensity, line shape, line width, and polarization behavior -may be used to characterize the lattice, impurities, and free carriers in a semiconductor. The intensity of the Raman phonon peak gives information about the crystallinity of the sample, which is very useful in ion implantation and subsequent annealing of semiconducting wafers. The ability to change probing depth by simply changing the excitation wavelength makes RS a very powerful technique in studying semiconductor materials. In addition, the combination of laser light with Raman microprobe techniques gives a very good spatial resolution, which is required in many advanced applications of semiconductor devices. Here we give some specific examples of PL and RS studies of ion-implanted semiconductors performed over the last two decades.

EXPERIMENTAL STUDIES 1. PHOTOLUMINESCENCE a. InJuence of the Impluntution Impurities and General Comments

Nakashima, Shiraki, and Miyao (1979) have used PL to study laserannealed phosphorus-implanted and nonimplanted Si. Phosphorus ions with an energy of 50 keV were implanted into boron-doped (100) Si (20 to 30Qcm) at doses ranging from 1 x 1OI2to 8 x 10l5cm-2. Laser annealing was carried out using a single Q-switched ruby laser pulse (25nsec pulse

3 PHOTOLUMINESCENCE AND RAMAN OF ION IMPLANTED SEMICONDUCTORS 85 40

30

Ruby Laser Power

h

-? CJ

>-

t

cn Z

E20

z

w

U

z

W V Y,

W

Z 5 10

3

0 0

1 DISTANCE ( c m )

2

FIG. 5. Room temperature photoluminescence profiles of laser-annealedsamples irradiated with phosphorus ions (P') with a dose of 8 x lOI5cm-* at 50keV. a.u., arbitrary units. Nakashima, H., Shiraki, Y., and Miyao, M. (1979). Photoluminescence Study of Laser Annealing in Phosphorus-Implantedand Unimplanted Silicon J. Appl. Phys. 50, 5966-5969.

duration and energy densities from 0.5 to 1.5 J/cmZ). A cw argon ion laser was used to optically excite the sample, and room temperature PL profiles were obtained by moving the samples over the annealed and unannealed regions. The PL profiles (see Fig. 5) of the laser-annealed sample irradiated with phosphorus ions at a dose of 8 x 10'5cm-2 had strong PL emission, whereas PL could not be detected on the implanted and unannealed region. Since there is a damaged layer at the surface of the unannealed region, the excited carriers recombine nonradiatively in this layer. However, at the surface of the annealed region, there are no such damaged layers. Therefore, the excited carriers can diffuse into the bulk and recombine radiatively both at the surface and in the bulk. The PL of the region annealed at 1.4 J/cm2 is as high as that of bulk Si, indicating complete recovery of crystal quality.

86

A. OTHONOS AND C. CHRISTOFIDES

However, no PL was detected from thermally annealed implanted samples. Therefore, even if the electrical activity of the implanted ions is recovered by thermal annealing, crystal quality is partially recovered. As seen in the same Fig. 5, recovery of PL intensity was not complete in the region annealed with 0.8 J/cmZ, even though electrical measurements showed complete recovery of electrical activity of the implanted ions. It is obvious from these measurements that PL gives a higher threshold energy for complete recovery annealing than do electrical measurements. It is interesting to point out that in their work, Nakashima, Shiraki, and Miyao (1979) observed a dramatic reduction of the PL intensity in unimplanted Si that was laser annealed. Surface spreading resistance measurements carried out on this sample showed that laser annealing increases the surface resistance of the unimplanted sample. Mooney et al. (1978) reported that deep-level defects were detected in laser-annealed, unimplanted samples using deep-level transient spectroscopy (DLTS). These deep-level defects may shorten minority carrier lifetime and reduce PL intensity carrier mobility and carrier concentration. Pankove and Wu (1979) studied hydrogenated ion-implanted crystalline silicon (c-Si) using PL. In their work, c-Si was implanted at room temperature using 2ESif ions with implantation energies ranging from 40 to 280 keV. The damaged layer was hydrogenated by exposing these samples to atomic hydrogen at 300°C. Photoluminescence of the samples was measured at liquid nitrogen temperature using an argon ion laser at 488 nm as the excitation source. A typical emission spectrum is shown in Fig. 6 together with emission spectra from c-Si and hydrogenated amorphous Si (a-Si:H). These data show an emission peak at 0.99 eV. Identical spectra were obtained at all implantation doses up to 1 x 10'4cm-2 and did not appear to depend on the chemical nature of the ion. In fact, similar spectra were obtained with aluminum, hydrogen, phosphorus, arsenic, fluorine, or neon implantation into Si. Pankove and Wu (1979) also studied the dependence of luminescence intensity on the implantation dose. It appears that the luminescence efficiency increases gradually for doses up to 1 x 1013cm-2and remains nearly constant up to 1 x 1014cm-2 at which point it decreases abruptly. Note that below 1 x 1014cm-', which is below the critical amorphization dose, the material is still a single crystal. At higher doses a broad emission peak at 1.15eV was obtained that corresponds to the luminescence of a-Si:H. The gradual increase in efficiency with implantation dose suggests an increase in the number of luminescence centers. The saturation of efficiency implies that these centers are so efficient that a concentration of lo'* cm-3 (assuming a depth of implantation of 1 x lo-' cm) is sufficient to recombine all the excess carriers. Adding more centers beyond the critical concentration of 1 x 10l8cmP3 cannot increase the efficiency.

3

PHOTOLUMINESCENCEAND

RAMANOF ION IMPLANTEDSEMICONDUCTORS 87

T H I S WORK

E,

-

( 0 -Si:

H1

FIG. 6. Emission spectra at 78 K from hydrogenated ion-implanted silicon (Si), from crystalline Si (c-Si), and from hydrogenated amorphous Si (a-Si:H). The energy gap is E,. The numbers labeling the peaks are the external emission efficiencies. Pankove, J. I., and Wu, C. P. (1979). Photoluminescence from Hydrogenated Ion-Implanted Crystalline Silicon. Appl. Phys. Lett. 35, 937-939.

On the contrary, above 1 x l O I 9 cmP3,concentration quenching sets in. At this high density, the defects are on the average 46A apart, and thus may form centers for nonradiative recombination that compete efficiently with the radiative process. It is believed that the luminescence seen at 0.99 eV was due to a structural damage induced by ion implantation rather than to a chemical doping effect, since the spectrum did not depend on the chemical species of the ion. Swenson, Luke, and Hengehold (1983) have used both PL and cathodoluminescence (CL) to study thallium (T1)-implanted Si. These experiments were performed using ion-implanted (1 11) and (lOO> Si wafers at energies of 280 and 560keV and with fluences of 1 x 10l2, 1 x and 1 x l O I 4 ions/cm2. The samples were excited with the 488-nm Ar' laser line and the luminescence was spectrally analyzed with a spectrometer. This was the first luminescence study of radiation damage by heavy ion implantation in Si. Previous studies included boron, phosphorus, argon, arsenic, carbon, oxygen, silicon, and neon, and the defect line introduced by these light ion implants was identical to that produced by the T1 implants. This suggests that implant defect lines are independent of implanted species. The unannealed defect luminescence (Fig. 7(a)) is removed after an annealing at

88

A. OTHONOS AND C. CHRISTOFIDES

energy (eV)

FIG.7. (a) Photoluminescence spectra from unannealed implanted silicon:thallium (Si:TI) at 24 K; (b) Photoluminescence from ion-implanted Si:TI at 30 K after annealing at 550, 550 + 650,550 + 750, and 550 + 950°C. Swenson, 0.F., Luke, T. E., and Hengehold, R. L. (1983).Luminescence Study of Thallium Implanted Silicon. J . Appl. Phys. 54, 6329-6335.

500°C and replaced by a broad structure containing many lines, as seen from the PL spectra in Fig. 7(b). This broad structure introduced by annealing at 550"C, in turn, was replaced by a sharper peak after subsequent annealing at 650°C. The defect luminescence reduced considerably when the second annealing was carried out at 750°C and all defect luminescence disappeared after a second annealing at 950°C. In addition to TI becoming optically active after annealing, it also has been shown to be electrically active. Measurements made on these samples using Rutherford backscattering spectrometry (RBS) and secondary ion mass spectroscopy (SIMS) indicate that a high concentration of T1 is retained in the Si lattice after annealing at 940°C. Chang et al. (1989) carried out PL at room temperature in Si-implanted n-type, and semi-insulating wafers of single-crystal indium phosphide (InP) and investigated this as a technique for characterizing the quality of ion-implanted and annealed materials. These authors found that the intensity of the luminescence is a good indicator of the quality of the annealed samples. Chang et al. (1989) reported results about the efficiency of impurity

3 PHOTOLUMINESCENCE AND RAMANOF ION IMPLANTEDSEMICONDUCTORS 89

cn

c .-

c

-8

C

0

2

4

6

8 10 12 14 Annealing tirne(sec)

16

18

20

FIG. 8. Variation of the integrated photoluminescence response with annealing time for rapid thermal annealing of indium phosphide at 800°C (curves (a) and (b)) and 700°C (curves (c) and (d)). Chang, R., Lile, D. L., Singh, S., and Hwang, T. (1989). Photoluminescence Studies of Si Implanted InP. J . Appl. Phys. 66,3753-3757.

activation and associated thermal degradation of InP, resulting from various annealing conditions. The InP wafers were implanted Si atoms 7 degrees off-axis at room temperature, using a dose of either 5 x 10'3cm-2 at 150keV or 5 x 1014cm-2 at 200keV. After implantation some of the samples were capped with 0.2-mm-thick layer of silicon dioxide (SiO,) to retain InP sample integrity during annealing. In Fig. 8 curves (a) and (c) show the PL response of InP implanted with 5 x 1014 Si+/cmz at 200 keV and capped with SO, as a function of annealing time at annealing temperatures of 800 and 700"C, respectively. Curves (b) and (d) represent PL from the nonimplanted samples that had undergone identical processing to the implanted samples. There are some interesting observations to be made from this data. The PL signal for the nonimplanted material began to decrease almost immediately, possibly due to the degradation of the InP at the surface region. Although this degradation would also be expected for the implanted samples, it is offset initially by an increasing PL signal due to the activation of the implant, which both increases the carrier concentration and reduces the implant-induced defect density at the surface region. At long annealing times, this thermally driven increase in material quality and carrier density slows down and would be expected to be overwhelmed by the continuing thermal degradation resulting in an overall decline in PL. It

90

A. OTHONOS AND C . CHRISTOFIDES

is worth noting that for both annealing temperatures the implanted samples never reach a PL response as large as that of the nonimplanted starting material, suggesting that even capped short-time annealing cannot recover the quality of the original semiconductor. Further investigation by this group using progressive etching of the implanted layers made possible the use of PL response to spatially profile the activation and the degradation of the wafers. The results suggest that annealing is not able to restore the original quality of the unimplanted semiconductor and that even short transient annealing cycles can lead to detectable thermally induced material degradation that extends many micrometers into the InP. Thompson, Barbara, and Ridgeway (1992) have performed PL study of Indium Phosphite:Iron after implantation with carbon, silicon, germanium, tin, indium, and phosphorus ions with a projected range of lpm. Their study concentrated on the 0.75-eV emission band. These authors concluded that this emission was due to a donor-acceptor-pair center comprised of a group IV element and a defect. The defect related to this center is formed during annealing of implantation damage and is enhanced by deviation in stoichiometry toward excess In.

b. Influence of Implantation Dose and Energy Recently, Othonos and Christofides (1995a) performed PL at room temperature on a set of phosphorus-implanted Si wafers (Table I). These wafers were implanted at different doses ranging from 1 x 1013ions/cm2 to 1 x 10l6 ions/cm2 at an implantation energy of 150 keV. In addition, a set of wafers having a constant doping concentration were implanted at different implantation energies ranging from 20 keV to 180 keV at constant dose of 5 x IOl4 P+/cm2. The phosphorus implantation was performed through a thin oxide layer at room temperature. Some of the samples were then annealed isochronically at various temperatures ranging from 300 to 1000°C for 1hr in an inert nitrogen atmosphere. In these PL experiments,

TABLE I PHOSPHORUS-IMPLANTED SILICON SAMPLES AT VARIOUSDOSES AND ENERGIES @(P+/cm2) 120 keV E (keV) 1 x 10i4~m-2

1 x loi3 (W1)

1 x loi4 (W2)

5 x loi4 (W3)

1 x loi5 (W4)

5 x toi5 (W5)

1 x 10'6 (W6)

20 (w7)

30 (W8)

60 (W9)

100 (W10)

140 (W11)

180 (W12)

3 PHOTOLUMINESCENCE AND RAMANOF ION IMPLANTED SEMICONDUCTORS 91 8.0

7.0

,

1

I

w1

60O0C

4W0C NA

80O0C 0.0

-1

4

0.90

0.95

1 .oo

1.05

1.10

1.15

Photon Energy (eV)

8.0

w6

7.0

2

6.0 :

4.0 v

<

C

-+! 2.0 ; 1.0 1

0.0 :

Photon Energy (eV) FIG. 9. Photoluminescence measurements from phosphorus-implantation unannealed (NA) and annealed samples for various temperatures of annealing. arb, arbitrary. Othonos, A., and Christofides, C. (1995a). Photoluminescence Measurements on Phosphorus Implanted Silicon: Annealing Kinetics of Defects. J . Appl. Phys. 78, 796-800.

488- and 1064-nm excitation sources were used in conjunction with an FTIR spectrometer to spectrally analyze the signal. Figure 9 shows PL spectra obtained with 1.06-pm excitation for samples implanted at two different doses (at 150 keV) and annealed isochronically at different temperatures. The spectra appear to be broad (100 MeV), which is typical of electron-hole recombination at room temperature. The peak value, which occurs around 1.08pm, is simply due to the indirect-gap

92

A. OTHONOS AND C. CHRISTOFIDES

transition of electrons from the minimum to the top of the valence band involving a phonon for momentum conservation. The general shape of the curve remains the same for all annealing temperatures except at the highest temperature, 1100°C. At this temperature the luminescence signal reduces to zero and the only feature that remains is a sharp peak corresponding to the Raman signal from Si (Othonos et al., 1994). The main difference in room temperature PL measurements in all the samples presented in Fig. 9 is the total integrated intensity of the signal. It appears that the various defects in the samples affect the luminescence signal (Nakashima, Shiraki, and Miyao, 1979;Nakashima and Shiraki, 1978). In this case of near-bandgap excitation (1.06pm) it is important to note that the PL signal is generated over the entire thickness of the bulk, because the penetration depth at this wavelength is on the order of the sample thickness. Thus the PL signal is a combination of a signal from both bulk and damaged implanted layer. The unexpected decrease of the PL signal at high annealing temperatures led to the investigation of nonimplanted samples. Figure 3 illustrates the PL signal in the case of two nonimplanted wafers (unannealed and annealed at 1l0OoC).We note that an annealing at 1100°C results in no PL signal. It seems that the induced PL signal is mainly due to the bulk. The fact that luminescence can be observed on the highly annealed ion-implanted samples indicates that thermal annealing was not totally effective in restoring crystal quality. This was also shown by Nakashima, Shiraki, and Miyao (1979) and Swenson, Luke, and Hengehold (1983). Figure 10(a) shows the total integrated luminescence as a function of annealed temperature for samples implanted at various doses at 150keV. In Figure 10 the integrated luminescence from the nonimplanted Si (nonannealed) sample is taken as unity. As seen in Fig. 10(a), the low annealed samples have a decreasing PL signal with increasing implanted dosage. It is obvious that the PL efficiency of the Si samples deteriorates severely due to the phosphorus ion bombardment. It is believed that the damaged layer due to the implantation at the surface of the unannealed samples caused the excited carriers to recombine nonradiatively, reducing the luminescence normally seen from the Si substrate. For the sample implanted at a dose of 1 x 1015P+/cm2,the luminescence behavior is more complicated. At the two annealed temperature extremes the behavior is similar to the other samples and, in particular, to the highest doped samples. However, the luminescence has a large dip around 600°C. This phenomenon is known as negative annealing and also has been observed recently while photothermal reflectance measurements were performed on the same samples (Seas and Christofides, 1995). The negative annealing around 600°C is due to the formation of complex defects around this annealing temperature, which are responsible for the trapping of minority carriers. In the case of implantation

3

PHOTOLUMINESCENCE

AND

RAMANOF ION IMPLANTEDSEMICONDUCTORS93

0.8 ,

200

400

600 800 Annealed Temperature ("C)

1000

1200

lo00

1

0.8 .

-: ; -

- j-

$0.6 c

-

E

;

B

:

" I -~

'S0.4 1 0.2

a ' : 0.0 4 200

400

600

800

Annealed Ternmrature (OC)

FIG. 10. Integrated photoluminescence intensity of phosphorus-implanted silicon wafers versus annealing temperature obtained from various implanted doses and energies: (a) W1, W2, W4, and W6; (b) W7, W10, and W12. The excitation laser wavelength is 1.06pm. am, arbitrary units. Othonos, A., and Christofides, C. (1995a). Photoluminescence Measurements on Phosphorus Implanted Silicon: Annealing Kinetics of Defects. .I AppL . Phys. 78, 796-800.

above the amorphization dose, 45-8i-diameter clusters are formed and may be centers for recombination. Looking at the data of Fig. lO(a), it appears that the curves may be divided into three well-defined annealing temperature ranges: 300 to 400°C; 400 to 700°C; and 700 to 1100°C. In each stage the PL mechanism depends on the implantation-induced disordering, which depends on the implanta-

94

A. OTHONOS AND C. CHRISTOFIDES

tion dose because that determines the degree of the induced disorder and amorphization. In the first range (300 to 400"C), the PL signal increases dramatically with annealing temperature, especially for the highly implanted samples. This change in PL intensity shows the sensitivity of the PL technique to the annihilation of defects in this temperature range. In the range of 400 to 700"C, we note that the four samples implanted at various doses present different behaviors. The PL of the sample series W1 increases with annealing temperature and reaches a maximum at 600°C after which it decreases again. Recall that samples lightly implanted at 1 x 1013 are much below the critical amorphization dose (ac= 5 x loi4 P+/cmZ)(Prussin, Margolese, and Tauber, 1985). Thus from 300 to 600°C we can achieve at least 90% of activation of free carriers while the layer becomes completely recrystallized. The wafer series W2 needs annealing at 700°C to achieve a maximum PL signal, because these samples need higher annealing temperature for restoration of defects due to the higher implantation dose. We note that the PL of the wafer series W4 varies in a strange way with annealing temperature. The series W6 has no minimum because highly implanted samples present some self-annealing phenomena (before any temperature annealing). This can be explained by the fact that some annihilation mechanisms occur, in the case of high implantation doses, that lead to self-annealing of lattice damage from the heat generated during ion implantation (Gibbons, 1968). This also was observed in the past by Seas and Christofides (1995) and Ishikawa, Yoshida, and Inoue (1981). Annealing from 700°C (for W1) or 800°C (for W2-W6) to 1100°C results in a decrease of the luminescence, reaching a negligible value at the highest annealing temperature. Figure 10(b) presents the total integrated PL signal as a function of annealed temperature for samples implanted at various energies: 20, 100, and 180keV at a constant dose of 5 x 10i4P+/cmZ.We note that for low annealing temperatures ( < 550°C) the integrated signal decreases with increasing implantation energy. For these three samples the signal reaches a maximum at 750°C and then decreases again with increasing annealing temperature. One important observation that can be made for all samples presented in Fig. 10, is their behavior at the two different extremes of the annealed temperature. The decrease of the PL of the nonannealed samples is believed to be due to the formation of defects near the surface. However, the decrease in luminescence at the highest annealing temperature is probably due to the same thing that causes reduction of luminescence of laser-annealed nonimplanted samples (Nakashima, Shiraki, and Miyao, 1979). Deep-level transient spectroscopy (Mooney et al., 1978) reveals deep-level defects in laser-annealed nonimplanted samples. These deep-level defects may shorten minority carrier lifetimes and reduce PL intensity.

3 PHOTOLUMINESCENCE AND RAMANOF ION IMPLANTED SEMICONDUCTORS95

Annealed Temperature ("C)

Annealed Temperature ("C)

FIG. 11. Integrated photoluminescence intensity of phosphorus-implanted silicon wafers versus annealing temperature obtained from various implanted doses and energies: (a) W1, W2, W4, and W6; (b) W7, W10, and W12. The excitation laser wavelength is 0.488pm.a.u., arbitrary units. Othonos, A., and Christofides, C. (1995a). Photoluminescence Measurements on Phosphorus Implanted Silicon: Annealing Kinetics of Defects. J . Appl. Phys. 78, 796-800.

The PL signal obtained with 0.488-pm laser excitation appears to have the same spectral profile as does the 1.06-pm excitation PL signal. This is not surprising since only low excitation densities are used, thereby eliminating bandgap re-normalization effects and any thermal effects. Figure ll(a) presents the integrated PL signal versus annealing temperature in the case in which the excitation beam is 0.488pm. Clearly, the PL behavior is

A. OTHONOS AND C . CHRISTOFIDES

96

300% W

A

400% 5OO0C '

"I

'

,

" " " I

I013

. .

.

, , , . . I

1014

,

, , , , , , I

1ole

10'6

@( P+/CmZ)

,,,I, ,? 0.1

1 " " 1 " " 1 ' " ' 1 " l ~ ' ' ~ I ' ' ~ ' I ' ~ ~ ~ I ~

0

20

40

60

80

100

120

140

160

180

200

E(keV)

FIG. 12. Dependence of the integrated photoluminescence on implanted samples for various annealing temperatures. (a) Dose. (b) Energy. P', phosphorus ions. Othonos, A., and Christofides, C. (1995a). Photoluminescence Measurements on Phosphorus Implanted Silicon: Annealing Kinetics of Defects. J . Appl. Phys. 78, 796-800.

different from the case of near-bandgap excitation. We note that the PL is zero for low annealing temperature (<800°C). The same phenomenon is also seen in Fig. ll(b) where the behavior of the PL versus annealing temperature is presented for the constant implantation dose samples (W7, W10 and W12; see Table I). We also note that the total normalized integrated signal is smaller than the one obtained in near-bandgap excitation. The differences in the PL signal may be attributed to the probing depth at 0.488 pm, which is less than 1pm. This means that less light reaches the bulk to induce the PL signal.

3

PHOTOLUMINESCENCE AND RAMANOF ION IMPLANTED SEMICONDUCTORS97

Figure 12(a) presents the dependence of the integrated PL intensity versus implantation dose at 150 keV. These measurements were performed with 1.06-pm laser excitation. The same behavior also was obtained by Summers and Miklosz (1973) in the case of implanted GaAs. The variation versus dose illustrates the high sensitivity of radiative recombination to damage introduced by ion implantation. In fact, the PL intensity decreases with dose. At an annealing temperature of 300"C, the PL varies by almost a factor of 3 from the lowest (1 x 1013P+/cm2) to the highest dose (1 x 10'6P+/cm2). Figure 12(b) presents the integrated PL signal as a function of implantation energy at a constant dose. We note that PL decreases with increasing implantation energy and this can be correlated with the increase in damage overlap as the ion range increases. This agrees with the results presented by Summers and Miklosz (1973) in the case of implanted GaAs.

2. RAMAN STUDIES IN ION-IMPLANTED SEMICONDUCTORS a. General Observation on Raman Signal

Raman spectroscopy has become one of the most widely accepted techniques for the characterization of semiconductor materials. In particular, its application as a nondestructive technique for examining damage that accompanies ion implantation is well established. Raman scattering can measure the damage due to any kind of ion implantation and can determine the effectiveness of annealing. Lattice disorder created during ion implantation can be examined using RS. As crystalline materials become amorphous, the intensity of the RS signal decreases. Several years ago, Morhange, Beserman, and Balkanski (1974) performed Raman studies on implanted Si with ions at different dosages and gave quantitative criteria. Bourgoin, Morhange, and Beserman (1974) compared the Raman data with electron paramagnetic resonance, RBS, and reflectance. These authors concluded that the Raman techniue gave similar results. Sharp phonon modes, which are characteristic of crystalline materials, change to broad peaks if the material becomes amorphous. Thus Raman spectra easily distinguish between crystalline and amorphous semiconducting materials. Hesse and Compaan (1979) performed RS with near-resonance excitation in cuprous oxide (Cu,O) implanted with helium, sodium, or cadmium ions. Their work on this semiconductor has shown that the use of excitation energy near resonance with the excitation states at the band edge can yield greatly enhanced sensitivity to lattice damage induced by ion implantation and can provide a useful monitor of the repair of this damage during annealing sequences.

98

A. OTHONOS AND C . CHRISTOFLDES

Laser light excitation near resonance was used for three main reasons. 1. In the region of strong band-to-band absorption, the light penetration depth is very short (25 nm), and thus the laser probe penetrates only the implanted region. 2. The strong coupling between the radiation field and the electronic states produces four to five orders of magnitude enhancement in the Raman cross section, allowing a good Raman signal at low laser powers even though the scattering volume in the experiment was as small as 5 x i0-10cm-3. 3. The width of the excitation states near the band edge appeared to be very sensitive to damage so that the intensity of the Raman peak shows significant change for doses as low as 1 x 10" ions/cm2. Figure 13 shows some of the Raman spectra as a function of light frequency shift for various Cd+-implanted doses. It is clear that with increasing implantation dose, the line structure of the Raman spectra broadens. These broad features in the heavily damaged crystal arise from RS, which refects the single-phonon density of states and arises from the breakdown of wave vector conservation in the damaged crystal. The sharp peaks that remain near 155 and 220cm-' may arise from some penetration of the 488-nm laser light into the undamaged substrate. Similar behavior in the Raman spectrum was also observed for excitation at 476.5 nm. Annealing experiments of ion-implanted Cu,O with helium, sodium, and cadmium all showed a similar annealing stage near 270"C, which is well below the melting temperature of 1050°C. During these experiments, at least 99% of the damage implantation that was produced was removed. Engstrom and Bates (1979) examined boron-implanted laser-annealed Si using RS, in order to study in detail the structure and properties of laser-annealed Si. Numerous reports demonstrate that annealing of implanted semiconductors by laser irradiation offers a number of advantages over thermal annealing, some of which are: (a) the laser intensity used for annealing may be selected to avoid heating the substrate, which would modify its electrical characteristics; and (b) a greater reduction of lattice damage may be achieved (Narayan, Young, and White, 1978). The concentration of electrically active impurities can be made to exceed the solubility limit of the host (Khaibullin et al., 1977). In their experiments (Engstrom and Bates, 1979), Si samples lightly doped with phosphorus (1 x 1013cm-2) were implanted with "B to doses of approximately 1 x 10'6ions/cm2 at 35 keV. Thermal annealing was done in a helium atmosphere by annealing the samples at 900°C for 30 min, followed by heating at 1100°C for another 30 min. Laser annealing was accomplished with a single pulse from a

3 PHOTOLUMINESCENCE AND RAMANOF ION IMPLANTED SEMICONDUCTORS99

FREQUENCY (10~crn-1)

200

,

1

400

1

1

600

1

1

800

FREQUENCY SHIFT (m-') FIG. 13. Raman spectra obtained with 100mW power at 488nm, showing the effects of increasing cadmium implantation dose: (a) unimplanted, (b) 1.5 x lo", (c) 1.5 x and (d) 5.3 x 1014 ions/cm2. Hesse, J. F., and Compaan, A. (1979). Resonance Raman Studies of Annealing in He-, Na-, Cd-Implanted Cuprous Oxide. J . Appl. Phys. 50, 206-213.

Q-switched ruby laser (693.4nm). The pulse energy density was 1.7 J/cm2, and the duration was approximately 40nsec. Figure 14 shows some of the Raman spectra of the boron-implanted Si samples at 77 K. The spectrum labeled (a) in Fig. 14, which corresponds to the unannealed sample, shows a strong background due to the first-order scattering induced by lattice damage. This scattering is most intense for energy shifts near 480cm-', reflecting the peak in the phonon density of states in Si. The Si phonon peak remains visible because the probing penetration depth is greater than the depth of the damaged region caused by ion implantation. Figures 14(b) and 14(c) show Raman data for the thermally and laser-annealed samples, respectively. In both cases there is a reduction in background scattering

100

A. OTHONOS AND C . CHRISTOFIDES

FIG. 14. Raman spectra of "B implanted silicon at 35 keV and 77 K: (a) unannealed, (b) thermally annealed, and (c) ruby-laser annealed. The silicon optic mode is at 523.5cm-', and the boron local mode is at 620cm-'. The laser output is 620mW at 488nm. Engstrom, H., and Bates, J. B. (1979). Raman Scattering from Boron-Implanted Laser-Annealed Silicon. J . Appl. P h p . 50,2921-2925.

3

PHOTOLUMINESCENCE AND RAMAN OF

ION IMPLANTED SEMICONDUCTORS 101

due to lattice damage and to the local vibration mode of substitutional boron atoms visible at 630cm-’. However, there is a significant difference between Figs. 14(b) and 14(c). The phonon mode of the thermally annealed sample is quite broad compared with pure Si or the laser-annealed samples. In addition, the optic phonon mode of the laser-annealed samples shows a pronounced “shoulder” on the low-energy side. Both of these effects were analyzed assuming a Fano-type interaction between the discrete optic mode and the continuous valence band states. It has been shown through their analysis that measurements of the Si phonon optic mode line shape provide a convenient means of determining the peak boron concentration in boron-implanted laser-annealed samples. However, this method does not work for thermally annealed samples, possibly due to the existence of residual lattice damage. Raman spectra from Si- and tin (Sn)-implanted GaAs was studied by Nakamura and Katoda (1982). Variation of Raman spectra from Si- and Sn-implanted GaAs with various doses was studied. The samples used in their experiments were chromium (Cr)-doped GaAs implanted with Si and Sn ions to doses ranging from 2 x 10I2 to 1 x 1 0 I 6 cm-’ at room temperature. GaAs implanted with Sn showed three types of Raman spectra corresponding to single crystalline,amorphous, and mixed states, depending on ion dose. However, GaAs implanted with Si showed only one type of Raman spectra, which corresponded to those obtained from low-dose Sn-implanted GaAs. A model of the disordered region including a partly crystalline layer, an amorphous layer, or both, was derived according to the Raman spectra. Kim et a!. (1993) studied the effects of Ga ion implantation into GaSb with fluence ranging from 1 x 10l2 to 5 x 1014cm-2 and subsequent thermal annealing effects by means of RS. Figure 15(a) shows Stokes Raman spectra from the GaSb samples. Spectra (1) corresponds to the unimplanted sample. In the backscattering geometry, only the LO phonon mode (observed at 236 cm- l ) is allowed, whereas the TO phonon mode is forbidden by selection rules (Loudon, 1964). The other spectra in Fig. 15 correspond to Stokes Raman data observed for 100 keV Ga ion-implanted GaSb with a different fluence. The arrows indicate phonon frequencies at various points of the Brillouin zone for crystalline GaSb. With increasing fluence, the LO phonon mode broadens asymmetrically and is shifted to lower energy. In addition, it becomes weaker while the TO mode becomes stronger. These effects are attributed to the disordering of the crystalline structure. At fluences above 5 x 10’ cm-2 both LO and TO phonon modes disappear, which means that GaSb crystal was transformed into an amorphous state by Ga ion implantation over the probing penetration depth. It is interesting to note that this threshold is smaller than the one observed for InP (1 x l O I 4 cm-’).

A. OTHONOS AND C. CHRISTOFIDES

102

1 -

I

I

I

I

GaSb :Ga min

t

!z

v)

z w I-

z

4 - L -

z a

2a

5x1 013~m'2

lxlo'2an'2

m -. 2 1

100

150

200

250

300

~

i0

FIG.15. (a) Raman spectra of (100) gallium antimony (GaSb), (1) unimplanted, and (2)-(6) Ga+-implanted at various fluences. (b) Raman spectra of GaSb implanted with 100 keV G a ions at various fluences followed by furnace annealing at 400°C for 15 min. arb., arbitrary. Kim, S. G., Asahi, H., Seta, M., Takizawa, J., Emura, S., Soni, R. K., Gonda, S., and Tanoue, H. (1993). Raman Scattering Study of the Recovery Process in Ga Ion Implanted GaSb. J . Appl. Phys. 14, 579-585.

3

PHOTOLUMINESCENCE AND RAMAN OF

ION IMPLANTEDSEMICONDUCTORS 103

Raman scattering as a function of annealing was also investigated. For fluences of 1 x 10” to 5 x l O I 4 cm-2, the intensity of LO phonon mode of GaSb increased with increasing annealing temperature, annealing time, or both. However, for a fluence of 5 x 1014cm-2, the LO and TO phonon modes were not observed by the conventional furnace annealing while they were observed by rapid thermal annealing (RTA). New phonon modes around 114 and 150cm-’ were observed for annealing above 400°C by furnace annealing and above 500°C by RTA (see Fig. 15(b)). These two modes were related to Sb-Sb bond vibrations ( E , and A , , modes). The recovery of the LO phonon mode intensity was low. The TO phonon mode remained even after annealing. This was due to the out-diffusion of Sb and the weak bond strength of Ga-Sb. For samples implanted with high doses (over 5 x l O I 4 cm-2), no healing of the damaged layers was noticeable after

-

I

.-

n ?

-

T

-

1

1

-

-

-

1

Ion In@ooicd Rrpion

OLSTANCE (5pml Idiv) FIG. 16. (a) Schematic of the implanted silicon crystal. (b) Integrated intensity of the crystalline Raman components as a function of the position. arb., arbitrary; c-Si, crystalline silicon. Mizoguchi, K., Nakashima, S., Fujii, A. Mitsuishi, A,, Morimoto, H., Onoda, H., and Kato, T. (1987). Characterization of Silicon Implanted with Focused Ion Beam by Raman Microprobe. J p n . J. Appl. Phys. 26,903-907.

104

A. OTHONOS AND C. CHRISTOFIDES

furnace annealing or RTA. This is believed to be due to polycrystallization of the sample. The use of RS to examine Si processed with focused ion beams is illustrated in Fig. 16 (Mizoguchi et al., 1987). A focused ion beam with energy 200 keV and diameter 0.1 to 0.2 pm formed implanted regions in Si,

5Pm FIG.17. Raman intensity image of the laser annealed polysilicon island. The extension of the flat intensity region along the annealing direction indicates grain growth. a.u., arbitrary units; SiO,, silicon dioxide. Nakamura, T., and Katoda, T. (1982). Raman Spectra from Si and Sn Implanted GaAs. J . Appl. Phys. 53,5870-5872.

3 PHOTOLUMINESCENCE AND RAMAN OF ION IMPLANTEDSEMICONDUCTORS 105

as shown in Fig. 16(a). The researchers scanned a Raman microscope across the sample and measured the integrated intensity of the 520 cm- Si phonon peak versus position as shown in Fig. 16(b). The change at each boundary between implanted and unimplanted regions appears clear with a strong from the nonimplanted and contrast between the intensities I,,,,, and limp implanted region, respectively. Nakashima and Hangyo (1989) reviewed the use of Raman methods in examining silicon-on-insulator (SOI) geometries. Nakashima et al. (1986) measured the Raman intensity from laser re-crystallized islands of Si on SO,. Figure 17 shows two-dimensional data for the Raman intensity scattered from a sample with one direction along the sweep direction of the laser annealing beam.

6. Infruence of Implantation Dose and Energy on Raman Signal Raman spectroscopy and spreading resistance analysis of phosphorusimplanted and annealed Si wafers was used by Othonos et al. (1994) and Christofides, et al. (1995). In these works, a large matrix of implanted Si wafers was characterized by using Raman spectroscopy and spreading resistance measurements (see Table I and Table I in Chapter 2). The phosphorus-implanted Si wafers were implanted with various energies and doses, ranging below and above the critical dose of amorphization (ac= 1 x 1015P+/cm2). The broad feature that is typical of amorphous materials disappears after annealing at 800°C (for 1 hr) where annihilation of many kinds of defects (interstitials, vacancies, point defects, complexes, dislocation lines and loops, and full amorphized layers) and re-crystallization take place (Boltaks, 1977; Christofides, Jaouen, and Ghibaudo, 1989; Christofides, 1992). It should be noted that in amorphous materials there is a breakdown of the RS selection rules because of the absence of well-defined crystal momentum. Raman spectra can no longer relate to the allowed optical phonons at the center of the Brillouin zone but, rather, to a convoluted function of the density of phonon states of the amorphous layer. This is another reason RS is an excellent technique for nondestructive monitoring of crystalline-amorphous transitions. Two sets of samples were thermally annealed isochronically at temperatures ranging from 300 to 1100°C. Room temperature RS was performed using a krypton ion laser at 530.9nm. Typical RS spectra from samples having been implanted at constant energy and constant dose are shown in Figures 18(a) to 18(d) and 19(a) to 19(c), respectively. Clearly evident is the characteristic Si phonon peak that occurs at approximately 64.5 MeV (520 cm - I). Special attention has been given to the amorphous-crystalline

A. OTHONOS AND C. CHRISTOFIDES

106

Implanted I ~ l O ' ~ c m at - * E 4 5 0 keV

r

2 $

2

I

I

sM)o

4KlO

3000

2000

Implanted IxlO'' m2 at E=150 keV

I

(b)

Energy in meV

FIG.18. Raman spectra from phosphorus (€')-implanted, unannealed and annealed, silicon (b) samples. The implantation energy is 150 keV and the implanted doses are (a) 1 x 1 x 1014, (c) 1 x 1015,and (d) 1 x 1016Pt/cm2.Othonos A., et al. (1994). Raman Spectroscopy and Spreading Resistance Analysis of Phosphorus Implanted and Annealed Silicon. J . Appl. Phys. 75, 8032-8038.

3 PHOTOLUMINESCENCE AND RAMAN OF ION IMPLANTED SEMICONDUCTORS 107

0 200

1W

Implanted 1 ~ 1 cm2 0 ~ at~E=150 keV

I

6wo 5000

4000

3000

2000

1WO

0

100

FIG. 18. (Continued)

A. OTHONOS AND C. CHRISTOFIDES Implanted 5 ~ 1 0at' ~E=20keV

g c

-

rn 4W0 .P

40

45

55

50

65

80

70

80

75

Energy in meV

(a)

Implanted 5 ~ 1 0at ' ~€=I00 keV

40

(b)

45

50

55

80

85

70

75

W

Energy in meV

FIG. 19. Raman spectra from phosphorus (P)-implanted, unannealed and annealed, silicon samples at room temperature. The implanted dose is 5 x 1014P+/cm2 and the implantation energy are (a) 20, (b) 100, and (c) 180 keV. Othonos A,, et al. (1994). Raman Spectroscopy and Spreading Resistance Analysis of Phosphorus Implanted and Annealed Silicon. J . Appi. Phys. 75, 8032-8038.

3

PHOTOLUMINESCENCE AND

RAMANOF ION IMPLANTED SEMICONDUCTORS

109

Implanted 5x10'' at E=180 keV

40 (C)

45

50

55

80

65

70

75

80

Energy in meV FIG. 19. (Continued)

transition occurring at various annealing temperatures. It is interesting to point out that for the samples annealed at high temperatures (800 to 1lOOOC), the phonon peak intensity (PPI) does not vary significantly with annealing temperature. Contrary to heavily implanted samples, the PPI signal for lightly implanted samples is almost constant with annealing temperature. However, the presence of an amorphized layer in the unannealed and low-temperature annealed samples results in a considerable decrease of the PPI. Since the implantation energy is constant, the damage should occur approximately over the same volume for all samples. Therefore, the PPI should decrease with an increasing dose. Figure 20 clearly indicates such behavior since for high doses we can only see a broadband Raman signal attributed to amorphization. Figure 21 shows the plot of the PPI as a function of the annealing temperature for the various implantation energies. In Fig. 21, we can see that the sharp Raman PPI decreases as the energy of ion implantation increases from 20 to 100keV. However, for high implantation energies (140 to 180 keV), the variation of the RS is not monotonic, as expected. This is because the implanted layers contain two types of damages: (1) amorphiz-

110

A. OTHONOS AND C . CHRISTOFIDES 6000 I

150 KeV 5000

4000

~

1

3000 I

2000 -

loo0 1

0-

f

10"

10'3

~0'4

10"

10"

LOG[ DOSE (m2) ]

FIG. 20. Summary of the Raman peak intensity as a function of implantation dose for various annealing temperatures. The implantation energy is 150 keV. Othonos A., et al. (1994). Raman Spectroscopy and Spreading Resistance Analysis of Phosphorus Implanted and Annealed Silicon. J. Appl. Phys. 75, 8032-8038.

ation of the layer, including dislocation lines and loops of the lattice and (2) interstitial implanted impurities. A closer look at Fig. 21 suggests that at low implantation energies (20 to 100 keV) the Raman signal is due to the global implanted volume since the probing depth is smaller than the depth of the implanted layer (Othonos et al., 1994). For higher implanted energies (140 to 180 KeV) where the thickness of the implantation layer becomes larger than the Raman probing depth, some of the implanted impurities are no longer within the implantation depth, and their contribution will become less important. In fact, once the implanted ions are deeper than the probing depth, their contribution to the Raman signal will only be from the amorphization of the Si crystal. Continuing their previous work, Othonos and Christofides (1995b) recently used a multiple wavelength RS technique to characterize the same set of phosphorus-implanted Si wafers. In this work, several excitation wavelengths ranging from 458 to 752.5 nm were used to probe phosphorusimplanted Si wafers at room temperature. Some of the key observations in this work include the following:

3 PHOTOLUMINESCENCEAND RAMANOF IONIMPLANTED SEMICONDUCTORS 111

0

A

v 0

2OKeV MKeV 6oKeV 1OOKeV lYKeV

18oKev

FIG.21. Summary of the Raman peak intensity as a function of implantation energy for phosphorus samples implanted at 5 x I O l 4 Pf/cmz and annealed at 1000°C. Othonos A., et al. (1994). Raman Spectroscopy and Spreading Resistance Analysis of Phosphorus Implanted and Annealed Silicon. J. Appt. Phys. 75, 8032-8038.

1. Damage caused by ion-implantation with small doses was minor, and even low-temperature annealing restored the samples back to their full crystalline structure. 2. For those samples with doses less than the critical amorphization dose, RS data revealed damage depths that extended over a few microns. 3. Most of the damage in these samples was near the sample surface. 4. Damage induced deeper into the sample recovered at lower temperature annealing than did damage caused at shallow depths.

V. Summary and Future Perspectives Photoluminescence and RS techniques are very useful for studying the degree of inhomogeneity of implanted materials as well as the kinetics of annihilation of damage with annealing temperature. The main results of PL and RS studies on Si-implanted materials over the last decades can be

112

A. OTHONOS AND C. CHRISTOFIDES

summarized as follows: 1. After annealing at 700"C, we can observe with both PL and RS techniques strong modification of the material. However, total reconstruction can be achieved only at temperatures above 800°C. 2. All the samples, independent of the implantation dose, reach the re-crystallization stage for annealing temperatures higher than 800°C; the re-crystallization process occurs smoothly for all the samples implanted with doses lower than the critical one. The annealing process in these samples is governed by different kinetics, which have more to do with the migration of interstitial impurities and vacancies than with the amorphous-crystalline transition (Said et al., 1990). 3. Ion implantation damage to the samples is apparent even at low implantation energy, (a1 low as 20 keV). 4. The PL signal is highly influenced by ion implantation conditions (dose and energy) and by the annealing temperature. 5. The total PL signal is a contribution from the bulk and the amorphization layer. 6. The PPI seems to be very sensitive even for slightly disordered samples such as the ones implanted at low doses (1 x 1013P+/cm2). 7. There is a strong dependence of the PPI on the optical penetration depth of the laser beam. 8. The Raman signal is more sensitive to local defects than to spreading resistance measurements.

REFERENCES Balkanski, M., Morhange, J. F., and Kanellis, G. (1981). Light Scattering from Laser Annealed Ion Implanted Semiconductors. J. Raman Spectroscopy 10, 240-245. Boltaks, B. (1977). Difusion and Point Defects in Semiconductors. Mir, Moscow. Beserman, R., and Bernstein, T. (1977). Raman Scattering Measurement of the Free-Carrier Concentration and of the Impurity Location in Boron-Implanted Silicon. J. Appl. Phys. 48, 1548-1550. Bourgoin, J. C., Morhange, J. F., and Beserman, R. (1974). On Amorphous Layer Formation in Silicon by Ion Implantation. Radiation Efects 22, 205-208. Cardona, M. (1975). Light Scattering in Solids. Topics in Applied Physics (M. Cardona, ed.) Vol. 8. Springer-Verlag, New York, 1-20. Chang, R., Lile, D. L., Singh, S., and Hwang, T. (1989). Photoluminescence Studies of Si Implanted InP. J. Appl. Phys. 66,3753-3757. Christofides, C. (1992). Annealing Kinetics of Defects of Ion-Implanted and Furnace-Annealed Silicon Layers: Thermodynamic Approach. Semicond. Sci. Techno/.7, 1283- 1294. Christofides, C., Jaouen, H., and Ghibaudo, G. (1989). Electronic Transport Investigation of Arsenic-Implanted Silicon. I. Annealing Influence on the Transport Coefficients. J . Appl. Phys. 64, 4832-4838.

3 PHOTOLUMINESCENCE AND RAMAN OF ION IMPLANTEDSEMICONDUCTORS 113 Christofides, C., Seas, A., and Othonos, A. (1995). Reconstruction Mechanisms in Ion Implanted and Annealing Silicon Wafers. Defects Diyusion Forum 117, 45-64. Elliott, R. J. (1957). Intensity of Optical Absorption by Excitons. Phys. Rev. 108, 1384-1389. Engstrom, H., and Bates, J. B. (1979). Raman Scattering from Boron-Implanted LaserAnnealed Silicon. J . Appl. Phys. 50, 2921-2925. Forman, R. A., Bell, M. I., Myers, D. R. (1981). Comments on “Raman Scattering from Boron-Implanted Laser Annealed Silicon.” 1. Appl. Phys. 52, 4337-4339. Fortner, J., and Lannin, J. S. (1988). Structural Relaxation and Order in Ion-Implanted Si and Ge. Phys. Reu. B 37, 10154-10158. Gibbons, J. F. (1968). Ion Implantation in Semiconductors. Part I: Range Distribution Theory and Experiments. Proc. l E E E 56,295-320. Hesse, J. F., and Compaan, A. (1979). Resonance Raman Studies of Annealing in He-, Na-, Cd-Implanted Cuprous Oxide. J . Appl. Phys. 50, 206-213. Ishikawa, K., Yoshida, M., and Inoue, M. (1981). Secondary Defects of Ast Implanted Silicon Measured by a Thermal Wave Technique. Japan. J . Appl. Phys. 26, L1089-L1091. Kachare, A. H., Cherlow, J. M., Yang, T. T., Spitzer, W. G., and Euler, F. K. (1976). Infrared Reflection and Raman Scattering of Ion-Implanted Nitrogen in Gallium Phosphide. J . Appl. Phys. 47, 161-173. Khaibullin, J. B., Shtyrkov, E. I., Zaripov, M. M., Galjautdinov, M. F., and Zakirov, G. C. (1977). Utilization Coefficient of Implanted Impurities in Silicon Layers Subjected to Subsequent Laser Annealing. Sou. Phys. Semicond. 11, 190-192. Kim, S. G., Asahi, H., Seta, M., Takizawa, J., Emura, S., Soni, R. K., Gonda, S., and Tanoue, H. (1993). Raman Scattering Study of the Recovery Process in Ga Ion Implanted GaSb. J . Appl. Phys. 74, 579-585. Kirillov, D., Powell, R. A,, and Hodul, D. T. (1985). Raman Scattering Study of Rapid Thermal Annealing of As+-Implanted Si. J . Appl. Phys. 58, 2174-2179. Loudon, R. (1964). The Raman Effect in Crystals. Adu. Phys. 13,423-482. Mizoguchi, K., Nakashima, S., Fujii, A. Mitsuishi, A., Morimoto, H., Onoda, H., and Kato, T. (1987). Characterization of Silicon Implanted with Focused Ion Beam by Raman Microprobe. Japan. J . Appl. Phys. 26, 903-907. Mooney, P. M., Young, R. T., Karins, J., Lee, Y. H., and Corbett, J. W. (1978). Defects in Laser Damaged Silicon Observed by DLTS. Phys. Status Solidi. A&, K31 -K34. Morhange, J. F., Beserman, R., and Balkanski, M. (1974). Raman Study of the Vibrational Properties of Implanted Silicon. Phys. Status Solidi A23, 383-391. Nakamura, T., and Katoda, T. (1982). Raman Spectra from Si and Sn Implanted GaAs. J . Appl. Phys. 53, 5870-5872. Nakashima, S., and Hangyo, M. (1989). Characterization of Semiconductor Materials by Raman Microprobe. I E E E J . Quantum Electronics 25, 965-975. Nakashima, S., Mizoguchi, K., Inoue, Y., Miyauchi, M., Mitsuishi, A., Nishimura, T., and Akasaka, Y. (1986). Raman Image Measurements of Laser Recrystallized Polycrystalline Si Films by a Scanning Raman Microprobe. Japan. J . Appl. Phys. 25, L222-L224. Nakashima, H., and Shiraki, Y . (1978). Photoluminescence Observation of Swirl Defects and Gettering Effects in Silicon at Room Temperature. Appl: Phys. Lett., 33, 257-258. Nakashima, H., Shiraki, Y., and Miyao, M. (1979). Photoluminescence Study of Laser Annealing in Phosphorus-Implanted and Unimplanted Silicon. J . Appl. Phys. 50, 5966-5969. Narayan, J., Young, R. T., and White, C. W. (1978). A Comparative Study of Laser and Thermal Annealing of Boron-Implanted Silicon. J . Appl. Phys. 49, 3912-3917. Othonos, A,, and Christofides, C. (1995a). Photoluminescence Measurements on Phosphorus Implanted Silicon: Annealing Kinetics of Defects. J . Appl. Phys. 78, 796-800. Othonos, A,, and Christofides, C. (1995b). Multi-Wavelength Raman Probing of Phosphorus Implanted Silicon Wafers. Nuclear Instrument and Methods in Physics Research B 117,367-374..

114

A. OTHONOS AND C. CHRISTOFIDES

Othonos, A., Christofides, C., Boussey-Said, J., and Bisson, M. (1994). Raman Spectroscopy and Spreading Resistance Analysis of Phosphorus Implanted and Annealed Silicon. J . Appl. Phys. 75,8032-8038. Othonos, A., Van Driel, H. M., Young, J. F., and Kelly, P. J. (1991). Correlation of Hot-Phonon and Hot-Carrier Kinetics in Ge on a Picosecond Time Scale. Phys. Rev. B 43,6682-6690. Pankove, J. I. (1975). Optical Processes in Semiconductors. Dover Publications, New York. Pankove, J. I., and Wu, C. P. (1979). Photoluminescence from Hydrogenated Ion-Implanted Crystalline Silicon. Appl. Phys. Lett., 35, 937-939. Prussin, S., Margolese, D., and Tauber, R. N. (1985). Formation of Amorphous Layers by Ion Implantation. J . Appl. Phys. 57, 180-185. Said, J., Jaouen, H., Ghibaudo, G., and Stoemenos, I. (1990). Electrical and Physical Investigation of Defect Annihilation in Arsenic Implanted Silicon. Phys. Status Solidi 117, 99-104. Seas, A., and Christofides, C. (1995). Photothermal Reflectance Investigation of Implanted Silicon: The Influence of Annealing. Appl. Phys. Lett., 66, 3346-3348. Summers, C. J., and Miklosz, J. C. (1973). Photoluminescence of Ion-Implantation-Damage n-Type GaAs. J. Appl. Phys. 44, 4653-4656. Swenson, 0. F., Luke, T. E., and Hengehold, R. L. (1983). Luminescence Study of Thallium Implanted Silicon. J. Appl. Phys. 54, 6329-6335. Thompson, T. D., Barbara, J., and, Ridgway, M. C. (1992). The Origin of the 0.75eV Photoluminescence Emission Band in Ion-Implanted InP. J . Appl. Phys. 71, 6073-6078.

SEMICONDUCTORS AND SEMIMETALS. VOL. 46

CHAPTER 4

Photomodulated Thermoreflectance Investigation of Implanted Wafers. Annealing Kinetics of Defects Constunt inos Christofides DEPARTMEW OF NATURAL SCIENCES OF CYPRUS UNIVERSITY NICOSIA, CYPRUS

I. INTRODUCTION .

.

.

.

. .

.

. .

.

.

.

.

. .

.

. . .

Ill. EXPERIMENTAL METHODOLOGY . . . . . . . . . . 1. Room Temperature Measurements . . . . . . . 2. Measuremenls versus Temperature . . . . . . . IV. EXPERIMENTAL RESULTSAND DISCUSSION . . . . . 1. Characterizationof Implanled Wafers . . . . . . 2. Influence of Annealing . . . . . . . . . . . . 3. Temperature Influence on the Photothermal Signal V. RECENTDEVELOPMENTS . . . . . , . . . . . . . 1. Single-Beam Thermowave Technique . . . . . . 2. Extension to the Two-Layer Model . . . . . . . VI. SUMMARY AND FUTURE PERSPECTIVES . . . . . . . References . . . . . . . . . . . . . . . . . .

. . . . . . . . .

. .

.

.

. . . . . . .

11. PHOTOMODULATED THERMOREFLECTANCE THEORY . . . . . . . . . . . . . . 1. Basic Photothermal Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Three-Dimensional Difusion

. .

. . . . . . . . . . . . . ... . . . . . . . . . . , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.

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

.

. .

.

. . . .

.

.

. .

115 116 116 118 119 119 121 122 122 126 136 144 144 145 146 147

I. Introduction During the last decade, various photothermal techniques have been developed for the characterization of implanted semiconductors. The noncontact character of these techniques makes them particularly attractive for the nondestructive evaluation of ion-implanted materials (Rosencwaig, 1982). One of the most popular thermal-wave techniques is photomodulated thermoreflectance (PMTR), which is able to detect the local variations in the reflectivity of semiconducting materials (Rosencwaig et al., 1985; Guidotti and van Driel, 1985; and Smith et al., 1985a, 1985b). 115 Copyright 0 1997 by Academic Press All rights of reproduction in any form reserved. 0080-8784/97 $25

116

C . CHRISTOFIDES

Today, research into PMTR covers several subjects of materials characterization, focusing primarily on: (a) ion-implanted monitoring and annealing kinetics of defects, (b) etch monitoring, (c) imaging of subsurface defects in metal lines, (d) polycrystalline and amorphous silicon (a-Si) thin films, and (e) metallization monitoring. The technique is particularly useful because of its high spatial resolution, its nondestructive and noncontact nature, and its reproducibility and rapidity. This technique also is noncontaminating and requires no special sample preparation. The features listed give the technique the advantage for direct measurements during the line fabrication process of wafers in high-technology industry. In fact, very few techniques of semiconductor characterization possess these features and this is the main reason for the introduction and use of PMTR in the semiconductor industry. Smith, Rosencwaig, and Willenborg (1985a) applied the new thermalwave photothermal reflectance for dose monitoring of ion-implanted semiconducting wafers. Although nominally a thermal-wave method, PMTR has been conventionally and increasingly used as an electron-hole plasma-wave probe for semiconducting substrate characterization. During the last few years, several workers reported numerous studies using this technique (Smith et al., 1986; Kirby, Larson, and Liang, 1987; Schuur et al., 1987; Wendman and Smith, 1987; Uchitomi et al., 1988; Smith et al., 1985a, 1985b; Vitkin, Christofides, and Mandelis, 1989, 1990; Christofides, Vitkin, and Mandelis, 1990; Forget, Fournier, and Gusev, 1993; Nestoros et al., 1995; and Seas and Christofides, 1995). In this chapter, we focus on PMTR investigations of implanted materials performed in the 1990s. Note that progress achieved before 1990 has already been reported in other works (Rosencwaig, 1987, 1994). Any repetition of results that already have been reviewed by other authors will be avoided. Part I1 of this chapter is devoted to the presentation of the relevant PMTR theory. After a brief outline of the basic theory, a summary of some important equations used for the analysis of the photothermal results is presented. Part 111 introduces the PMTR experimental setup as well as some details concerning implanted samples, and in Part IV, which is the main part of this chapter, the most recent experimental results are reviewed. Finally, Part V gives a summary, concluding remarks, and future perspectives of the PMTR technique. I truly apologize to those authors whose work may have been inadvertently omitted.

11. Photomodulated Thermoreflectance Theory

1. BASICPHOTOTHERMAL EQUATIONS The mechanism of PMTR signal in semiconductors can be understood in terms of the induced modulation of the refractive index. In general, the

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS

117

PMTR signal can be divided into two main contributions, one proportional to the surface temperature (thermal-wave effect: ST) and the other proportional to the photoinduced changes in the free-carrier density (plasma-wave effect: SN).Thus the total PMTR-induced signal S , can be expressed as the sum of the previous two components (Rosencwaig, 1987): .

where R is the reflectivity at temperature T; C T and C N are the temperature and plasma coefficients, respectively; and A R , AT, and AN are the local variations in reflectivity, temperature, and plasma density, respectively. In order to evaluate A T and A N and obtain the induced photoreflectance signal, we must first solve the thermal and plasma diffusion equations for an isotropic semi-infinite medium (Sablikov, 1987):

a(W -at

AN

- DEV2(AN)- -

z

ano A T +-+ h e i o t aT z

where D, and DT are the electronic (ambipolar) and thermal diffusivities, respectively, z is the recombination lifetime, no is the equilibrium free carrier density, CD is the incident photon flux, w is the angular frequency (w = 27cA where f is the modulation frequency), x is the thermal conductivity, E, is the bandgap energy, hv is the photon energy of the pump beam, t is the time, and a is the optical absorption coefficient of the sample. It has been proved that thermal activation (the third term on the right-hand side of Eq. (2)) is negligible in the case of relatively low temperatures, and when free carriers (at equilibrium) no satisfy the following condition (Vasilev and Sandomirskii, 1984; Sablikov, 1987):

where K , is the Boltzmann constant, S* is the normalized surface recombination velocity (S* = spE/DE,s is the surface recombination velocity and p E is the plasma diffusion length). At room temperature and for a frequency of 300 Hz ( w t = O.l), the previous equation corresponds to an electron density no close to 1 x 1019cm-3,whereas f = lOOkHz (or= 100) corresponds to no = 1 x 1022cm-3. Thus it is obvious that Eq. (4) is easily satisfied except in the case of heavily doped semiconducting samples. In the

118

C. CHRISTOFIDES

E

2 +

0.001

1 0

50

100

150

200

250

300

Temperature (K) FIG. 1. Thermal diffusion length pT = (DT/nJ)”z versus temperature (solid line) of a nonimplanted crystalline silicon sample at f = 100 kHz. The constant value is the effective radius L, = (as + (dashed line), which determines whether a one- or a three-dimensional thermal diffusion should be used. Nestoros, M., Forget, B. C., Christofides, C., and Seas, A. (1995). Photothermal Reflection versus Temperature: Quantitative Analysis. Phys. Rev. B 51, 14115-14123.

case where the condition in Eq. (4) is satisfied, the system is decoupled and can be solved easily. Recently, a rigorous analytic solution has been obtained for the coupled Eqs. (2) and (3) (Mandelis, Nestoros, and Christofides, 1996). Therefore, a model considering only one-dimensional diffusion of heat could be sufficient to fit the experimental data only at high temperatures. In Fig. 1, it can be seen that p T which equals ( D , / z ~ ) ” ~becomes equal to the effective radius, L, around 280 K (Nestoros et al., 1995).The same argument also can be made about the plasma diffusion length p E , which generally is larger than pT.

2. THREE-DIMENSIONAL DIFFUSION A three-dimensional solution to the diffusion equations is needed to evaluate the photothermal signal. Considering the axial symmetry of the problem, a solution is easily obtained by using the Hankel transformation (Fournier and Boccara, 1989). The plasma density AN and the temperature AT integrated over the surface of the probe beam are

4 PHOTOMODULATED THERMOREKECTANCE OF IMPLANTED WAFERS a

1d l

1 19 (5)

[(hv-EJ -

where 1 is the spatial frequency introduced by the Hankel transformation. In the integrals, we define iEand iTas the plasma and the thermal-wave vectors:

where a, and ap are the radii of probe and pump beams, respectively. Finally, @ is given by the relation @=

2(1 - R)P hvna,2

where P is the incident power. In addition, it is important to note that in the case of thin films the term 1 - R is not completely correct. As indicated in recent publications (Christofides et al., 1991; and Mandelis et al., 1993), a more precise representation of this term should consider the incident power at the air-film interface, the reflection coefficient of the interface and also the multiple reflections in the various layers. In the analysis herein, multireflection phenomena and optical interference phenomena are not considered because they result in a change in (9. 111. Experimental Methodology

1. ROOMTEMPERATURE MEASUREMENTS In this section, the most common experimental setup used for PMTR measurements by several groups around the world is described. Figure 2(a)

120

C. CHRISTOFIDES

FIG.2. Experimental setup for laser-induced photomodulated thermoreflectance (PMTR) measurements (a) Room temperature measurements; the photothermal reflectance setup: L, Lens; AO, acousto-optic modulator; M, mirror; DM, dichroic mirror; AL, achromatic lens; S-B, soleil babinet; P, cube beam polarizer; NB, narrow band filter. Seas and Christofides, 1995). (b) Photomodulated thermoreflectance (PMTR) versus temperature measurements. He:Ne, he1ium:neon; PBS, polarization beam splitter; Si, silicon. Vitkin, I. A. Christofides, C., and Mandelis, A. (1990). Photothermal Reflectance Investigation of Processed Silicon. 11. Signal Generation and Lattice Temperature in Ion-Implanted and Amorphous Thin Layers. J. Appl. Phys. 67, 2822-2830.

4

PHOTOMODULATED

THERMOREFLECTANCE OF IMPLANTED WAFERS

121

presents a typical PMTR setup. This apparatus is similar to the ThermaProbe 150 system, first introduced by Opsal, Rosencwaig, and Willenborg (1983) and then employed by several other researchers (Rosencwaig et al., 1985; Guidotti and van Driel, 1985; Christofides, Vitkin, and Mandelis, 1990); Forget, Fournier, and Gusev, 1993). Recently, Seas and Christofides (1995) used it to study the annealing kinetics of defects in phosphorus-implanted materials. The periodic sample heating is obtained using the output radiation of a laser, which was modulated by an acoustooptic modulator. The pump power is usually between 20 to 100mW and normally is focused at the sample surface. The spot size of both pump and probe beams were adjusted to be several micrometers. The change in the reflectivity due to the periodic heating of the pump beam is detected using a few mW from he1ium:neon (He:Ne) laser (632.8nm) and measured with a silicon photodiode. The photodetector output is monitored with a computer-controlled lock-in amplifier at various modulation frequencies.

VERSUS TEMPERATURE 2. MEASUREMENTS

The cryogenic part of the experimental system, which is presented at the bottom of Fig. 2(b), has been used to study the PMTR response as a function of temperature, down to the 20K range (Vitkin, Christofides, and Mandelis, 1990). In order to perform a temperature study, the semiconductor wafer was placed in the experimental chamber of a helium-cooled expander module. Optical access was available through a vacuum-sealed quartz window. The operating pressures within the chamber were obtained with a mechanical pump. The experimental temperature range, as monitored by a gold-iron: constantan thermocouple with 0.1 K accuracy, was 20 to 300K. As the chamber temperature was lowered at a rate of 2.5 K/min, the thermocouple voltage was periodically probed by the computer. Once the voltage corresponded to a temperature difference of 1 K, several data points from the fast lock-in amplifier were recorded and averaged. Thus the in-phase and quadrature components of the photodetector output, and the corresponding chamber temperature, were stored in the computer at regular 1 K intervals. Frequency-scanned studies were performed by changing the sinusoidal output of the acousto-optic modulated wave form generator. The output frequency was measured with a frequency meter, to an accuracy of 0.01%. Section 3 of Part IV presents several PMTR results obtained as a function of ambient temperature.

122

C. CHRISTOFIDFE

IV. Experimental Results and Discussion 1. CHARACTERIZATION OF IMPLANTED WAFERS a. Injluence of Implanted Impurities

It is important to note that PMTR measurements are highly sensitive to the type of impurities implanted (Rosencwaig, 1994). Figure 3(a) shows the variation of the PMTR with ion implantation dose for three different impurities: phosphorus, arsenic, and boron. In Fig. 3(a), we can see the nonmonotonic behavior of arsenic (As) and phosphorus-implanted wafers. According to Rosencwaig, this behavior is due to the high-dose implantation of impurities for a subsurface amorphous layer. Similar measurements concerning the type of the implanted impurities also have been performed by Anjum et al. (1991). These authors show that As ions produce a PMTR signal that is larger when compared with the one induced by phosphorus ion implantation. In Fig. 3(b), we can see the induced PMTR signal versus the implantation dose for the two impurities.

b. Influence of Implantation Dose and Energy This section presents experimental data obtained from PMTR measurements on phosphorus-implanted Si. Silicon wafers, lightly doped with boron ( p = 20 to 25 R cm), were implanted with phosphorus at various doses at 15OkeV: @ = 1 x 1013(Wl); 1 x 1014 (W2); 5 x 1014 (W3); 1 x loi5 (W4); 5 x 1015 (W5); and 1 x 10I6 (W6) (Seas and Christofides, 1995). The phosphorus ion implantation process was performed through a thin oxide layer at room temperature. Some samples were then annealed isochronally for 1 h at various temperatures from 300 to 1100°C in nitrogen atmosphere. After annealing, the oxide overlayer was etched away and the samples were used for the experiments. Figure 4(a) presents the variation of the PMTR signal amplitude as a function of the implantation dose over the range 1 x 1013to 1 x 10I6P+/cm2 for the nonannealed samples at various modulation frequencies (Seas and Christofides, 1995). We note, in general, an increase of the photothermal signal as a function of the implanted dose except for the very high doses of 1 x loi6 P+/cm2.In fact, this is the main problem in industrial applications. This phenomenon was aloso observed in the past by Christofides, Vitkin, and Mandelis (1990) and Ishikawa, Yoshida, and Inque (1987). This can be explained by the fact that some annihilation mechanisms take place in the case of high implantation doses, which leads to a self-annealing of lattice damage via the heat generated during ion implantation (McFarlane and

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS

I

I

1-11

lc+12

I

1-14

1-13

123

1

1-15

1-16

Ion Implant dace 12000

10000 7

-: -

rn 0.485MeVP

8000

4.=n

-

rn

6000

CD C

Q

$

4000

2oooi :

10’2

rn

1013

1014

Dose (atoms/crn2)

FIG. 3. Variation of the photomodulated thermoreflectance (PMTR) signal with ion implant dose. Note nonmonotonic behavior at high dose for arsenic (As) and phosphorus (P) implants due to the presence of an amorphous silicon layer. a u , arbitrary units. (a) Rosencwaig, A. (1994). Thermal Wave Monitoring and Imaging of Electronic Materials and Devices. In Non-Destructive Eualuation. Progress in Photochemical and Photoacoustic Science and Technology Series, Vol. 11. (A. Mandelis, ed.) PTR Prentice Hall, Englewood Cliffs, NJ, Chapt. 4.; (b) Anjum, M., Sandhu, G . S., Cherekdjian, S., and Weisenberger, W. (1991). Thermal Wave Characterization of Silicon Implanted with MeV Phosphorus Ions. Nucl. Instrum. Meth.

B55.266-268.

C . CHRISTOFIDES

124

f = 1 MHz

Implantation ~ o s e(P+ /an2)

01

Id

.

. . ' . . ' . I

.

' ' ' . . . - I

'

....-"I

lo4 16 Modulation Frequency (Hz)

106

FIG.4. Photomodulated Thermoreflectance(PMTR) amplitude: (a) versus implantation dose. for unannealed phosphorus (P)-implanted silicon samples at five different modulation frequencies; and (b) versus modulation frequency for wafers W1 to W6. (After Seas, A., and Christofides, C. (1995). Photothermal Reflectance Investigation of Implanted Silicon: The Influence of Thermal Annealing. Appl. Phys. Lett. 66,3346-3348.

Hess, 1981). Figure 4(b) clearly indicates the self-annealing process of wafer W6 when the signal is plotted as a function of the modulation frequency. We note that for each frequency the PMTR signal of wafer W6 is lower than that of wafers W5 and W4, which were implanted at lower doses. The same phenomenon of self-annealing was recently reported by Othonos and Christofides (1996), who performed Raman spectroscopy on the same samples. The decrease of the PMTR signal amplitude at high doses was also observed and discussed by Wurm et al. (1988). These authors claim that the decrease of PMTR amplitude at high doses is due to the increasing thickness

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS

125

of the amorphous layer with implantation dose. This interpretation is improbable since the thicknesses of the amorphous layers of these samples, which were evaluated in the past with the 1D-SUPREM (Stanford University PRocess Engineering Model) I11 simulation package (Othonos et al., 1994; Christofides et al., 1994), vary slightly from 0.45 to 0.53pm with implanted dose: Measurements on W5 and W6 indicate thicknesses of 0.52 and 0.53 pm, respectively. With respect to the frequency dependence, we can also point out the decrease of the PMTR amplitude signal with increasing modulation frequency (see Fig. 4(b)). The dependence of the PMTR signal on modulation frequency is obvious and is in agreement with several simulations and experimental results presented in the past (Vitkin, Christofides, and Mandelis, 1990; Smith, Rosencwaig, and Willenborg, 1985a; and Nestoros et al., 1995). As expected and reported several times, an increase of the phototherma1 signal amplitude with increasing damage (short- and long-range disorder) was observed for all modulation frequencies. 4000 J

1x10~~

A 5~10’~

-

v

3000 -

1x10’~

6 5~10’~ 0 1x10’2

? m

A

v

al

3 .?=

-

E a - 2000 -

A

a

m

A

Ul

i7j K I-

B

A

v

m

A

A

1000 -

i o

~ 0.0

r

7 1.o

~

~

l 2.0

~

~

l

~

3.0

Energy (MeV)

FIG. 5. Photomodulated Thermoreflectance (PMTR) amplitude as a function of phosphorus implantation energy at various doses. am., arbitrary units. (After Anjum, M., Sandhu, G. S., Cherekdjian, S., and Weisenberger, W. (1991). Thermal Wave Characterizationof Silicon Implanted with MeV Phosphorus Ions. Nucl. Instrum. Meth. B55,266-268.

1

126

C. CHRISTOFIDES

Finally, it is also important to note that Anjum et al. (1991) have used PMTR for the characterization of implanted Si wafers at high energy (0.5 to 3MeV). The sensitivity of the PMTR signal to implantation dose and energy and to implant angle also has been studied by these authors. It has been demonstrated that PMTR signal is sensitive to energies up to 3 MeV for phosphorus ions implanted in Si. Figure 5 presents the PMTR signal as a function of implantation energy for various implantation doses. We note that the sensitivity of the PMTR signal on the implantation energy is small in the case of low doses (1 x 10” ions/cm2) (Anjum et al., 1991).

2. INFLUENCEOF ANNEALING a. Annealing Kinetics of Defects The annealing behavior of ion-implanted semiconductor wafers, in most cases, is strongly dependent on the implantation dose. Usually, with increasing implantation dose, higher annealing temperatures are necessary to achieve a high recrystallization and a given level of electrical activation. During the last decade it has been observed (Ryssel and Ruge, 1986) that at very high doses, where amorphization occurs, annealing from 300 to 1100°C does not lead to monotonic annihilation of defects. On the contrary, we have an increase of disorder with increasing annealing temperatures where complex phenomena sometimes take place. Figures 6(a) and 6(b) present the PMTR amplitudes versus the annealing temperature for the Si implantation doses of phosphorus-implanted Si samples at two different modulation frequencies: (a) 400kHz and (b) 900kHz. The thermal diffusion lengths in the case of Si for these two modulation frequencies are 2.8 and 1.9 pm, respectively. The choice of modulation frequency was based on thermal diffusion length because in the case of 900kHz we probe closer or inside (Othonos et al., 1994) the implanted layer, whereas in the case of 400 kHz we probe deeper into the bulk of the wafer but close enough to the interface. Modulation frequencies smaller than 400kHz were not used because in such cases, the major contribution to the PMTR signal is due to the bulk (substrate) of the wafer and does not have any significant interest for the study of the annealing kinetics of defects. In Fig. 6, for each implantation dose, 13 different isochronal annealings were performed to obtain a global view of the annealing process and annihilation mechanisms. One of the first observations we can make is the fact that the relative variation of the PMTR amplitude versus the annealing temperature is the same for both frequencies. Thus all the comments that follow are valid for both cases. In addition, it

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTEDWAFERS 0.7

127

I

500 700 900 Annealing Temperature (“c)

1100

FIG. 6. Photomodulated Thermoreflectance (PMTR) amplitude versus annealing temperature for the silicon implantation doses of phosphorus-implantedsilicon wafers at (a) 400 kHz and (b) 9 0 k H z . Seas, A., et al. (1995). Photothermal Reflectance Investigation of Implanted Silicon: The Influence of Thermal Annealing. Appl. Phys. Lett., 66, 3346-3348.

is obvious from Figs. 6(a) and 6(b) that the annealing curves are composed of five well-defined stages in the following annealing temperature ranges: 300 to 400°C; 400 to 550°C; 550 to 700°C; 700 to 800°C; and finally 800 to 1100°C. In each stage a different annihilation process takes place. This depends on the implantation dose because it is the dose that determines the degree of the induced disorder and amorphization. It is important to note that even for the low-dose implanted wafers W1 and W2 the PMTR technique is sensitive enough for monitoring the annihilation process of point defects as a function of the annealing temperature. In fact, in the case of Si wafers implanted with low doses there are only

128

C. CHRISTOFIDES

point defects in the lattice and there is no long-range disorder. Wafers W3 and W4 were implanted at doses around the critical amorphization dose 0 = 5 x 1014P+/cmz (Prussin, Margolese, and Tauber, 1985). This implantation dose does not provoke a completely amorphized layer, and generally induces point defects. However, contrary to wafers implanted at high doses (W5 and W6), the PMTR amplitude decreases in the range of annealing temperatures of 300 to 550"C, implying a significant decrease in local disorder. According to Boltaks (1977), by annealing up to 500°C many kinds of point defects such as phosphorus vacancies and divacancies can be annihilated. In addition, by annealing from 550 to 700°C the PMTR amplitude again increases due to the formation of phosphorus complexes. This phenomenon is known as negative annealing and was described previously by Gibbons (1972). The highly implanted wafers W5 and W6 present a completely different behavior from the other four wafers (W1 to W4). The two wafers implanted at high doses (over the critical amorphiz= 5 x 10'4cm-2; Prussin, Margolese, and Tauber, 1985) ation dose of possess completely amorphized layers, a fact that leads to long-range disorder and thus to a strong PMTR amplitude. A strong modification of the PMTR signal occurs between 700 and 800°C. It seems that in this range a phase transition (amorphous-to-crystalline) takes place. It is also important to note that the PMTR amplitude presents only a small change in the case of samples annealed at high temperatures, that is, 800 to 1100°C. Another important observation for wafers W5 and W6 is that the PMTR signal amplitude is higher for the lower implanted sample (W5). This was expected and was discussed in detail in Fig. 4 in terms of the self-annealing process. Nevertheless, at high annealing temperatures (800 to 1100°C) the W6 presents a higher relative signal than do the lower implanted samples, which is usually due to the annihilation of dislocations (Christofides, 1992). On wafer W6, thermal annealing from 300 to 700°C leads to an increase of the photothermal signal, increasing the damage. In the case of phosphorusimplanted and amorphized Si wafers in this range, only amorphous clusters become more complex with annealing. This leads to a higher degree of disorder and inhomogeneities in the material.

b. Thermodynamic Approach As is well known, high implantation doses lead not only to the creation of many localized defects but also to the formation of an amorphous layer (Prussin, Margolese, and Tauber, 1985). Annealing is expected to repair the damage caused by ion implantation primarily (Gibbons, 1972), and it also is expected to decrease the randomness of the impurity distribution. Table

4 PHOTOMODULATED THERMORFTLECTANCE OF IMPLANTED WAFFNS

129

TABLE I DEFECTS INDUCED REQUIRED FOR

ION IMPLANTATION: ANNEALING TEMPERATURE THEIR ANNIHILATION AND ACTIVATION ENERGIES BY

T.(“C) Type of Damage Vacancy, interstitial Vacancy, arsenic Divacancy Layer completely “amorphous” Layer incompletely “amorphous Line dislocations Loop dislocations

t = 30 to 60min

4 (ev)

- 180

0.2-0.5 0.2-0.8

200 300 550-650 700 800-1000 800- lo00

1

2-3 3 -4 5-8 5-8

I (Boltaks, 1977; Christofides, 1992; Tamura, 1973) presents the various types of defects, the annealing temperatures required to annihilate them (if the annealing time is between 30 and 60 min), and their activation energies. According to Gibbons (1972), the effect of annealing depends strongly on the implantation dose (D and on the relation between this dose and the critical amorphization dose (Dc (Prussin, Margolese, and Tauber, 1985 and Christel, Gibbons, and Sigmon, 1981). Figure 7 presents QC as a function of the atomic number Z at room temperature. In addition to the annealing temperature T,, the time t also is an important parameter in annealing processes to obtain activation (recrystallization) of the impurities (implanted layer). The arrow in Fig. 7 shows that higher implantation doses are required to achieve amorphization of the layer for temperatures higher than 300 K. An existing model has been used to determine the activation energy of the local annealing recovery mechanism. The basis of this model is the law of mass action (LMA). Wagner and Schottky (1930) were the first to apply a thermodynamic method and the LMA to the investigation of the equilibrium properties of insulating and semiconducting solids. The importance of using these statistical methods with semiconductors has been pointed out by several authors (Fuller, 1959). Ryssel and Ruge (1986) have developed the principal equations and have shown that the concentration of ions electrically implanted in wafers can be expressed by a first-orderreaction kinetic equation. In 1956, Reiss, Fuller, and Morin showed that, even in degenerate semiconductors in which classic statistics are not valid, the distribution law may still apply to the heterogeneous equilibrium. The kinetics of the annihilation of the damaged layer, as will be shown, is consistent with a local annealing process in which the observed variation of the electrical, optical, or thermal properties could be described by the

C. CHRISTOP~DES

130

*’

I , Energy of Implantation: E= few to 300 keV Silicon Target

1

h

& 3 4

T>300K

U

1A

IB

00

P

Sb-

AS I

20

I

I

40 60 Atomic Number:Z

I

80

FIG.7. Critical implantation dose mC,as a function of the atomic number Z . Christofides, C. (1992) Annealing Kinetics of Defects of Ion-Implanted and Furnace-Annealed Silicon Layers: Thermodynamics Approach. Semicond. Sci. Technol. 7, 1283-1294.

relaxation-type relationship (Fuller, 1959; Ryssel and Ruge, 1986), which is characterized by a unique activation energy and prefactor:

where t can be an electrical or optical coefficient or finally the PMTR amplitude signal. One of the previous properties of the annealed wafer at a given annealing temperature is t(T,), tt is the coefficient corresponding to the implanted wafer annealed at high temperature (approximately 1 lO0C) for a very long time ( t = 1 h), and z is the relaxation time of the local annealing process (Fuller, 1959; Ryssel and Ruge, 1986):

where E, is the activation energy, which depends mainly on the type of

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS

131

defect (Christofides, Jaouen, and Ghibaudo, 1987, 1989b; Christofides, 1992) characteristic of the relaxation process, K , is the Boltzmann constant, and to is a characteristic time. Essentially, the recrystallization of the layer by the migration of the implanted ions and displaced Si atoms, which forms interstitial (electrically inactive) to substitutional (electrically active) sites, has been modeled (Gibbons, 1968). In order to understand the fundamental physics shown in Eqs. (9) and (lo), we express the annealing behavior of the implanted ions by the first-order thermodynamic reaction-kinetic equation: dCi _ dt

Ci

--

z

where Ci is the concentration of the implanted ions at electrically inactive sites (trapped at an interstitial site). It is important to note that in the case in which more complex defect rearrangements take place (bound in a complex or localized within a cluster), Eq. (1 1) cannot satisfy our semiquantitative model because z becomes more complex. The concentration of the electrically active ions, C,, as a function of the annealing temperature T, and the time t is determined as follows (Fuller, 1959):

where C,(O, 0) is the number of ions that are already electrically active without annealing (this depends on the implantation temperature and dose CD); Ci(O, 0) is the number of ions that are not active without annealing; Ci(t, T,) is the instantaneous density of inactive ions, which decreases exponentially with T, and t, as suggested by Eqs. (10) and (1 1); and C,.,,, is the maximum concentration of electrically active ions, which should occur in the limit of long times and high annealing temperatures. After a simple integration, we can write (Fuller, 1959)

We can now replace the concentration C by any physical quantity that is directly or indirectly related to C. However, the previous hypothesis can be made only under the assumption that the various optical and electronic coefficients are mainly dominated by the concentration of electrically active impurities. Annealing of defects into other types (i.e., complex defects) and, especially in the case of negative annealing, C no longer reflects the

132

C. CHRISTOFIDES

“experimental reality.” By taking into account the previous hypothesis in the case of electrical transport coefficients, for example, we can easily show that the resistivity of the implanted materials depends strongly on the concentration of the electrically active impurities. This effect has been explored during the last few years by Christofides et al. (l987,1989a, 1989b). The concentration C has also been replaced by the PMTR signal and an activation energy has been found that was close to those found by using electrical property analysis (Christofides, Vitkin, and Mandelis, 1990; Vitkin, Christofides, and Mandelis, 1989, 1990). In addition, as described in Chapter 5 (Volume 4 9 , Boussey-Said replaced C by electronic factors, such as Hall mobility and resistivity. For the case of PMTR investigation, the PMTR signal monitors the restoration of the damage layer via its electronic and thermophysical properties, and thus is an indicator of the electronic activation of impurities. Let us now turn our attention to the quantitative analysis of the PMTR data. In fact, in order to evaluate the PMTR signal (one-dimensionalmodel; Doka, Miklos, and Lorincz, 1988) and the influence of the annealing temperature on it, we present the general PMTR expression (Eq. (15)), which can be related to the annealing temperature (Rosencwaig, 1987; Vitkin, Christofides, and Mandelis, 1990; Wagner and Mandelis, 1991):

where I) depends on the real and imaginary parts of the refractive index and on the dielectric constant. The sample density is p,, and the modulation frequency is f. Finally, C, is the specific heat at constant volume and x the thermal conductivity. These terms have been taken to be dependent on the annealing temperature. In the case of low frequency, the thermal wave contribution is the dominant one. As was proved recently by Wagner and Mandelis (1991), the assumption made in Eq. (15) is absolutely right: According to these authors, the Drude component of the PMTR signal in the implanted Si wafers is expected to be much less important than in the case of crystalline silicon (c-Si). Conversely, the thermal-wave component becomes much more important in implanted Si than in c-Si. In fact, ( C , X ) ” ~ increases when the material becomes more amorphous (Vitkin, Christofides, and Mandelis, 1990). As a result, this factor is greater for the case of amorphous silicon (a-Si) than for a-Si. In order to use the thermodynamic model described previously, several PMTR experiments have been performed. An experimental batch consisting of (loo), p-type (6 R cm) Si wafers implanted through a thick oxide layer, with arsenic (As) (a = 5 x 1014As+/cmZ;implantation energy at 150 keV)

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS

133

20 200 400 600 800 ANNEALING TEMPERATURE T,,("C,

FIG. 8. Variation of the Photomodulated Thermoreflectance(PMTR) signal as a function of the annealing temperature for arsenic-implantedsilicon (Si) wafers. arb., arbitrary. Christofides, C., et al. (1990) Photothermal Reflectance Investigation of Processed Silicon. I. Roomtemperature Study of the Induced Damage and of the Annealing Kinetics of Defects in IonImplanted Wafers. J. Appl. Phys. 67, 2815-2821.

at room temperature was used to test the foregoing PMTR concepts and applications to ion-implanted materials. On removal of the silicon dioride (SiO,) overlayer, some of the samples were thermally annealed isochroiially (1 h) at different temperatures (400 to 800°C). Figure 8 shows the PMTR signal as a function of the annealing temperature, T, over the temperature range of 400 to 800°C (Christofides, Vitkin, and Mandelis, 1990); also shown are the signal levels for the unimplanted and unannealed samples. It is well-known that annealing at the relatively low temperature of 400°C causes a noticeable decrease in local disorder (Boltaks, 1977). Annealing at 400°C can annihilate many kinds of point defects, such as As vacancies and divacancies. Also noticeable in Fig. 8 is the increase of the PMTR signal with T, between 400 and SWC, which may be associated with the formation of As complexes. Higher annealing temperatures ( > SSOOC) are required to dissociate these As multivacancies. Also evident from Fig. 8 for higher annealing temperatures ( >60O"C), is the signal that approaches that of the unimplanted Si, indicating a high degree of restoration of crystallinity. In fact, according to Table I, only line and loop dislocations could survive after an annealing at 800°C. It seems from the similarity of the PMTR signal of the samples at 800°C and the

134

C. CHRISTOFIDES

unimplanted sample that the dislocation-type defects do not significantly influence the PMTR signal. That is, whether dislocations are present predominantly in the bulk of the wafer-where the surface-sensitive PMTR technique is not affected by them-or at the surface, their effect on the local optical, thermal, and electronic parameters is not sufficient to influence the observed signal. This may be so because the dislocation density is too low or individual dislocations are too small for the beam-averaged signal to “see” them. A PMTR study as a function of modulation frequency could give a more definitive answer as to whether the effects of the crystalline bulk are significant. Vitkin, Christofides, and Mandelis (1990) have performed several studies on implanted layers (a = 1 x lo1’ to 1 x 1OI6 ions/cm’) as a function of the modulation frequency and environmental temperature (10 to 300K). The kinetics of the annihilation of the damaged layer, as monitored by the PMTR method, is consistent with a local annealing process in which the observed photothermal reflectance signal could be described by a relaxation-type relationship (Fuller, 1959; Reiss, Fuller, and Morin, 1986; and Christofides, Vitkin, and Mandelis, 1990) in which 5 of Eq. (9) can be replaced by the PMTR signal:

where AR(T,) is the PMTR signal from the wafer annealed at a given temperature, ARi is the signal for the As+-implanted sample, AR, is the signal corresponding to the unimplanted wafer (i.e., equivalent to a very high annealing temperature). Rearranging Eq. (16), the reduced phototherma1 signal Q a R could be defined as

QAR

ARi - AR, = exp AR(T,) - AR,

(:)

From Eqs. (10) and (17), then,

Thus plotting the double logarithm of Q against (To)-’ should yield a straight line of slope - E,/k,. To construct this plot, we must ignore the experimental point in the negative annealing regime, because this “global

4 F’HOTOMODULATED THERMORE~ECTANCE OF IMPLANTED WAFERS 135

Isochronal Annealing: t= 1 hour 2.0 -

1.0-

b -2, W

Ea

0

= 0.25eV

0.2 -1

0.1 I

I

I

I

I

I

I

2.0

1.8

1.6

1.4

1.2

1.0

0.8

1000/Ta (KI) FIG. 9. Determination of the activation energy of the local annealing process with photomodulated thermoreflectance (PMTR) results. The region corresponding to the negative annealing regime is indicated by a broken line. Christofides, C. (1992). Annealing Kinetics of Defects of Ion-Implanted and Furnace-Annealed Silicon Layers: Thermodynamics Approach. Semicond. Sci. Technol. I, 1283- 1294.

amorphization” process is not governed by Eq. (17). Performing these appropriate calculations, the plot shown in Fig. 9 was obtained (squares), where the range between 450 to 550°C is indicated by a broken line (Christofides, 1992). The measurement of the slope yields E , close to 0.15 eV. The small value of E , is most likely attributable to a local migration of point defects, and it appears that this technique is indeed sensitive to local rearrangement of defects such as the migration of interstitial atoms. In Fig. 9, we can see the results of the application of the local relaxation model to some experimental PMTR data obtained from other experiments (Ishikawa, Yoshida, and Inque, 1987) with samples implanted at 40 keV with 4 x 10l5As+/cm2 and annealed isochronally for 20min in nitrogen atmosphere. This activation energy is three times higher than the one obtained by the same technique (Christofides, Vitkin, and Mandelis, 1990); this variation may be due to the differences in dose and annealing treatment of these two groups of samples.

136

C. CHRISTOFIDES

INFLUENCE ON 3. TEMPERATURE

THE PHOTOTHERMAL

SIGNAL

a. Dependence on Temperature Nearly all of the parameters mentioned in Eqs. (5) and (6)-s, z, D,, D,, CE, C,, x, and ,+exhibit a temperature dependence that affects the photomodulated reflectance signal. For the thermal parameters, empirical expressions of their temperature dependencies have been used (Nestoros et al., 1995). The analysis of this section concentrates on the three main electronic parameters D,, s, and z, and on the plasma and thermal coefficients C N and C,, respectively. Other parameters used in the numeric fit of the experimental data such as the energy gap and optical absorption coefficient were taken from the literature (Sze, 1965; Pankove, 1971; Smith, 1978). The electron and hole mobilities (pe and ph) of semiconducting materials (Pierret, 1987), are related to the ambipolar electronic diffusivity by the Einstein relation (Sze, 1965): a,

where e is the electron charge. In the case of implanted wafers it has been shown that the mobilities remain constant, or exhibit only weak temperature dependence, over the temperature range of the experiments (Christofides et al., 1987, 1989b). Still, the diffusivity will keep a linear temperature dependence due to the presence of T in Eq. (19). For Si-doped samples, a temperature dependence must be introduced for lifetime and surface recombination velocity (Bebb and Williams, 1972):

where fl is a factor inversely proportional to the trap energy level. In the case of phosphorus impurities, fi was calculated to be around 0.1 meVThis relationship is approximately consistent with the T P 2 dependence observed for Shockley-Read-Hall recombination in heavily doped Si (Dziewior and Schmid, 1977). Furthermore, in the case of intrinsic Si wafers the recombination time is mainly due to deep level traps, since p is inversely proportional to the trap energy. In this case, /3 is close to 0.01 meV-l. Thus T becomes almost insensitive to temperature change and therefore is taken to be constant.

'.

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS 137

The surface recombination velocity s is a phenomenologic parameter that expresses the probability of recombination at the surface of the sample that is similar to t, except for the fact that z is inversely proportional to the recombination probability. Low recombination is expressed by high values for t and small values for s. Considering this assumption, the temperature dependence of s is taken to be the inverse of that of z (Nestoros et al., 1995):

Theoretic expressions for the thermal coefficient CT and the plasma coefficient C , can be obtained; these are expressed as a function of the index of refraction n in order to obtain their temperature dependence. In the case of the thermal coefficient, the relation between R and n gives (Rosencwaig, 1987)

The Drude effect gives the theoretic expression for the plasma coefficient C , (Drude, 1900; Rosencwaig, 1987):

where 1, is the optical wavelength of the probe laser, m* is the electron effective mass, and c is the velocity of light. The temperature dependence of the index of refraction of Si can be expressed as follows (Tomita, 1986):

+ "

n = n, exp K,T

]]

+ T+T,

a 1.9595 - Ego mT2

(24)

where the values of the various constants are n, = 3.6, y = 1.67, a = 0.1004, m = 4.73 x Ego = 1.52, and T, = 636. Although the refractive index itself varies only slightly with temperature, over the range of our experiments, its derivative with respect to temperature is not negligible. Figure 10 presents the normalized temperature variation of C T / C T J O O and CN/CN300 (at 300K). The variations of the two coefficients are opposite. This effect, favoring the plasma contribution at low temperatures, is important and must be taken into account in the theoretic fitting of the PMTR signal versus ambient temperature. Finally, the band-filling contribution depends on temperature only through the 1/R term. This dependence is very small:

138

C. CHRISTOFIDES

1.2

0

50

100 150 200 Temperature (K)

250

300

FIG. 10. Normalized thermal and plasma coefficients versus temperature. Nestoros, M., et a[. (1995). Photothermal Reflection versus Temperature: Quantative Analysis. Phys. Rev. B 51, 14115-14123.

R(300) = 0.3493, and R(77) = 0.3433 (Sze, 1965; Tomita et al., 1986) and the band filling was considered to be constant with temperature. b. Experimental Results and Theoretic Fittings In order to examine the plasma and thermal contribution of the PMTR signal versus ambient temperature, experiments were performed with highly implanted Si at two different modulation frequencies: 3 and 100 kHz. Wafers of n-type c-Si (100) were implanted with phosphorus ions with three different doses at energy E = 150 keV: 1 x lo”, 1 x 1014, and 1 x 10l6P+/cmZ. One p-type crystalline Si wafer (100) ( p = 6Rcm) also was used in this study. The apparatus used for the PMTR studies versus ambient temperature has been extensively described in $2 of Part 111. For this study, a supplementary cryogenic system was added to perform measurements from 40 to 300 K (Vitkin, Christofides, and Mandelis, 1990). This setup is shown in Fig. 2(b). The various characteristics of the experimental constants are given in Table I. The experimental results are displayed in Figs. 1 l(a) and 1 l(b) (Nestoros et al., 1995). Figure 1 l(a) presents experimental data and theoretic fittings (with Eqs. (l), (9,and (6)) obtained for p-type c-Si. The values used for the various electronic and thermal parameters are given in Table 11. Note the decrease of the PMTR signal with increasing ambient temperature. The dependence of the PMTR signal on temperature is obvious. This was confirmed with theoretic simulations by Nestoros et al. (1995). For tempera-

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS 139 : n-TypeC-Si

c 3 2.5 7 9

3kHz

J

Temperature (K)

5

0

0

50

100 150 200 Temperature (K)

250

300

FIG. 11. Experimental results and theoretic fittings (solid lines) of the photomodulated thermoreflectance (PMTR) amplitude as a function of temperature at two different modulation frequencies (3 and 100 kHz) for (a) p-type nonimplanted crystalline silicon (c-Si); and (b) silicon implanted at a high dose (@ = 1 x 10l6 P+/cm-'). P', phosphorus ions; a.u., arbitrary units. Nestoros, M., et al. (1995). Photothermal Reflection versus Temperature: Quantative Analysis. Phys. Reo. B 51, 14115-14123.

tures over 150K,where the thermal contribution is dominant, we see a reduction of this contribution with increasing modulation frequency. Another important point is the discrepancy between theory and experimental data at temperatures lower than 170 K. Figure ll(b) presents experimental and theoretic data obtained from a highly damaged implanted wafer: The modulation frequency influences the photothermal signal in the same way as in the case of p-type c-Si (Fig. ll(a)). With respect to the signal level, as expected, the signal is much higher due to the presence of disorder. The disproportionality in the frequency dependence and the relatively high

140

C. CHRISTOFIDES TABLE I1 EXPERIMENTAL CONSTANT PARAMETERS ~~

~

~

~

~

Symbols

Experimental Constants

Values

f

Modulation frequency Pump beam intensity Pump beam energy Pump beam radius Probe beam radius Probe beam wavelength Sample reflectivity

3 and 100 kHz 25mW 2.548 eV 15 pm 15 pm 632.8 nm 0.35

@ hv aP

as ).o

R ~

~~

~

(Reprinted from Nestoros, M., Forget, 8. C., Christofides, C., and Seas, A. (1995). Photothermal Reflection versus Temperature: Quantitative Analysis. Phys. Rev. B 51, 141 1514123; with permission.)

importance of the plasma effect at low temperatures led Nestoros et al. (1995) to perform all the experimental studies at the modulation frequency of 100 kHz. Figure 12 presents the experimental results and theoretic fittings for the implanted samples (Nestoros et al., 1995);the agreement between the theoretic Eqs. (5) and (6) and the experimental data is excellent. The values used for the various electronic and thermal parameters are given in Table 111. As expected and reported on several times in the past (Guidotti and van

-

3

i

5 2.5 4 .-2

2

-i 1.5 m

I

.-8

1

0.5

OO

50

100

150

200

250

300

Temperature (K) FIG. 12. Experimental results and theoretic fittings of the photomodulated thermoreflectance (PMTR) amplitude as a function of the ambient temperature for phosphorus-implanted silicon samples: 1 x lo’’, 1 x and 1 x 10l6 P+/cmZ(f = 100kHz). am., arbitrary units. (Reprinted from Nestoros, M., Forget, B. C., Christofides, C., and Seas, A. (1995). Photothermal Reflection versus Temperature: Quantitative Analysis. Phys. Rev. E 51, 141 15- 14123; with permission.)

4

PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS

141

TABLE I11 THERMAL ELECTRONIC, AND OPTICAL FITTEDPARAMETERS Impurity

Phosphorus

@, (P+/cm2) a, (cm ') fie (cm2/vs)

1x 6x 5 2 0.1 1x 4x

~

(cm2/vs) Do (cm2/sec) do) (set) s(0) (cm/sec) fih

10'6

lo4

10-9 105

Phosphorus 1 x 1014 6 x lo3 30 10 0.4 s x 10-9 2 x 105

Phosphorus

Boron

1 x 10l2 6 x lo3 437(T/300)-0.3 135(T/300)-0~3 1 3 x 10-6 1.8 x 105

6 x lo3 Pierret, 1987 Pierret, 1987 1 7.7 x 1 0 - ~ 3.87 103

(Reprinted from Nestros, M., Forget, B. C., Christofides, C., and Seas, A. (1995). Photothermal Reflection versus Temperature: Quantitative Analysis. Phys. Rev. B 51, 141 15-14123; with permission.)

Driel, 1985; Vitkin, Christofides, and Mandelis, 1990; Smith, Rosencwaig, and Willenborg, 1985a), an increase of the signal with increasing damage (short- and long-range disorder) was observed. In all cases the signal is dominated by the thermal contribution for ambient temperatures over 200 K. Concerning the relative amplitudes of the curves presented in Fig. 12, the behavior of the PMTR signal cannot be distorted by any optical or thermal interference phenomena. In the case of low implantation, layers and substrates possess similar indices of refraction and thermal properties, whereas the highly implanted sample is out of any interference fringes. In fact, according to the theoretic model by Wurm et al. (1988), we will observe interference phenomena for implanted layers with thicknesses less than 0.4pm. Experimental evidence by the same authors, however, indicated that such interferences will be observed in the case of implanted layers of thicknesses less than 0.3 pm. Moreover, Wurm et al. (1988) show that the interference fringes for samples implanted at energies over 125 keV cannot appear due to the annihilation of the sharp interface between the amorphous implanted layer and the crystalline substrate. In the present case, the implantation energy was 150 keV and the thickness of the highly implanted layer was approximately 0.53 pm (Othonos et al., 1994). Therefore, thermal interference phenomena are excluded. It is important to remind the reader that the PMTR experiments performed by Nestoros et al. (1995) were as a function of ambient temperature. Thus it needs to be determined whether the thermal expansion effects could change the thickness of the damaged layer as the temperature decreases from 300 to 40K. According to Touloukian et al. (1973), such a temperature variation will change the thickness by less than 7nm. This change is negligible and is not sufficient to give rise to interfer-

142

C. CHRISTOFIDJB

ence fringes, in agreement with out theoretic predictions and experimental results. It is also important to note that the PMTR signals of the implanted samples 1 x 10l2 and 1 x 10'"P+/cm2 reach minima around 120 and 140K, respectively, whereas the highly implanted sample (1 x 1016P+/cm2) presents a completely different behavior (Fig. 12). This phenomenon, which was confirmed by recent results obtained at room temperature (Wagner, 1993) and which was observed by Nestoros et al. (1995) for the first time on implanted Si, is directly related to the degree of damage in the implantation layer. The thermal and plasma contributions have different temperature dependencies. Therefore, it is expected that photothermal measurements versus ambient temperature will show a passage, or shift, from one contribution to the other when their relative amplitudes change. The presence of the minimum indicates that the two added contributions are of opposite sign (or, in other words, out of phase by 180") (Fournier and Forget, 1991). By introducing the Drude effect to the previous equations an excellent fitting is permitted. Figures 13(a), 13(b), and 13(c) present the experimental results and theoretic fitting of the PMTR signal versus temperature as well as the thermal and plasma contributions. In Figs. 13(a) and 13(b), we clearly see that for the two samples implanted at 1 x 10" and 1 x 10'"P+/cm2, the minimum occurs for the temperature at which the thermal and plasma contributions are approximately equal. Conversely, the sample implanted at 1 x 10l6 P+/cm2 (Fig. 13(c)) does not show this minimum. As a result we may conclude that the Drude contribution cannot account for this feature of the PMTR signal at low temperatures. This can be explained since 1 x 1OI6P+/cm2 is above the critical dose (OC= 5 x 10'" P+/cm2;Prussin, Margolese, and Tauber, 1985) for amorphization of Si by phosphorus, and the Drude effect is expected to be negligible in amorphous semiconductors (Wagner, 1993). It is therefore impossible to fit the low-temperature data in the same way as in the case of higher temperatures (Figs. 12(a) and 12(c)). In any case, we must consider a plasma-wave contribution of the same sign as that of the thermal-wave contribution (Mandelis and Wagner, 1996). From Figs. 13(a) to 13(c), we can point out the following two important remarks: (1) the samples implanted at low dose present contributions (thermal and plasma) of opposite signs, which lead to PMTR minima; and (2) the higher implanted samples are dominated by the thermal contribution, as expected. In any case, if there is an additional contribution (by another mechanism), it must be of the same sign (e.g., band-filling phenomena; Mandelis and Wagner, 1996). Let us now discuss the values used for the various parameters in theoretic fittings. Table I11 presents thermal, electronic, and optical fitted parameters.

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS

0.01

' '

0

'

"

'

50

'

' "

' . ..' .

100

150

' ' '

' ' ..

200

143

-

. ' . (a) 250 300 '

'

Temperature (K)

T:

10

Q W

D

FIG. 13. Experimental results and theoretic fitting of the photomodulated thermoreflectance (PMTR) amplitude versus temperature. The thermal and plasma contributions are also and (c) 1 x 10l6 P+/cm2(f = 100 kHz). P+, phosphorus presented. (a) 1 x 10"; (b) 1 x ions a.u., arbitrary units. (Reprinted from Nestoros, M., Forget, B. C., Christofides, C., and Seas, A. (1995). Photothermal Reflection versus Temperature:Quantitative Analysis. Phys. Reo. B 51, 14115-14123; with permission.)

144

C . CHRISTOFIDES TABLE IV THERMAL, ELECTRONIC, AND OPTICAL PARAMETERS AT 300 K

Impurity

Phosphorus

Phosphorus

Phosphorus

@, (P+/cmZ)

1 x 1016

u (cm-’)

1.4 x 10’ 5 2 0.14 2.1 x 1 0 - ’ O 1.89 x 106 1.9 x 2 x 10-4

1 1014 1.4 x 103 30 10

1 x 10’2 1.4 x 103 437 135 1.38 6.3 x 8.5 x 10’ 1.9 x 2 10-4

K (cm2/vs) (cm2/Vs) D , (m2/sec) T (sec) s (cm/sec)

Ph

c,(m-3)

c, u - 7

0.55 1.1 x 10-9 9.4 x los - 6 x lo-’’ 2 x 10-4

Boron 1.4 x 103

Pierret, 1987 Pierret, 1987 1.38 7.7 10-4 3.87 x 103 1.6 x 1.8 x 10-4

(Reprinted from Nestoros, M.. Forget, B. C., Christofides, C., and Seas, A. (1995). Photothermal Reflection versus Temperature:QuantitativeAnalysis. Phys. Rev. B 51, 14115-14123; with permission.)

The variation of those parameters with increasing damage is consistent with the deterioration of the electronic, thermal, and optical properties. Table IV presents the previous properties at 300K for all the samples used in this study. We note that as expected, in the case of p-type doped (3 x 10’’ ~ m - ~ ) e-Si, z is equal to 770psec (Sze, 1965; Pierret, 1987) which is on the same order of magnitude as is generally measured and presented in the literature (Li, 1980). Also, as expected, the lifetime decreases as damage increases, that is, as the implantation dose increases. The same comments also can be made for the surface recombination velocity. Table IV also presents C, and C,. The data for C, are in good agreement with the literature and C, is on the order of low4 (Huldt, Nilsson, and Svantesson, 1979; Weakliem and Redfield, 1979). However, as discussed earlier for C,, the sign changes in the case of shift from the crystalline to the amorphous material. Note that the thermal diffusivity decreases by one order of magnitude from its crystalline to amorphous phase (from 1.38 to 0.14 cm2/sec),which is in good agreement with the values in the literature (Touloukian, 1973).

V. Recent Developments 1. SINGLE-BEAM THERMOWAVE TECHNIQUE Finally, it is important to mention that the use of PMTR for the characterization of implanted semiconductors still presents an excellent area

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS

145

=%-beam splitter

1

.

locusq lens

.

diaphragm

ottenuotw

J reference detector FIG. 14. Experimental setup applied for single-beam reflection thermowave analyses: after passing the acousto-optical modulators (AOM), different diffraction orders were taken as partial beams to avoid optical interference causing intermodulation. Ar, argon. From Wagner, M., and Geiler, H. D. (1991). Single-BeamThermowave Analysis of Ion Implanted and Laser Annealed Semiconductors.Meas. Sci. Techno[.2, 1088-1093.

for research and development. Currently, several scientists are working on experimental development and on extending the present theoretic knowledge. For example, it will be a definite omission not to point out the development of the single-beam thermowave technique based on twofrequency hetero-dyne modulation of a pump laser beam used for both excitation and detection of the photothermal response in a sample (Fig. 14) (Wagner and Geiler, 1991; Wagner, Winkler, and Geilen, 1991). This technique provides frequency transformation to the low-frequency region. The single-beam technique proves to yield various types of information concerning ion-implantation-induced defects and annealing influences on the wafers under investigation. Instrumentally, it enjoys an improved signal-to-noise ratio (SNR) over PMTR. 2. EXTENSION TO THE TWO-LAYER MODEL Recently, Christofides et al. (1996) presented a complete theoretic analysis of the laser PMTR signal from a two-layer semiconducting wafer. This analysis is of relevance especially in the case of heavily implanted wafers resembling a two-layer system (bulk and implanted film). It is shown that the electronic and thermal properties of a thin surface layer may be determined by using the measured induced PMTR signal. By performing several numeric simulations, Christofides et al. (1996) show the influence of

146

C. CHRISTOFIDES

the various electronic, optical, and thermal parameters of the two-layer samples on the PMTR signal. In these simulations, the special case of the implanted and disordered layers have been taken into account. The influence of the upper layer (implanted) and the influence of the substrate on the signal are discussed and parameter regimes identified, where the characterization of the thin overlayer may be possible using this technique. The extension of the two-layer model to the variation of optical, thermal, and electronic properties as a function of annealing temperature may yield depth profilometric analysis on long- and short-range disorder in implanted and annealed (or unannealed) semiconduction, if the relevent substrate properties are known or can be measured.

VI. Summary and Future Perspectives Photomodulated thermoreflectanceis an important nondestructive evaluation technique for studying the annihilation process of implanted wafers due to thermal annealing. It has been shown that the PMTR technique is sensitive to various types of short- and long-range disorder annihilation processes. In this chapter, we have presented PMTR studies on ion-implanted semiconducting wafers. The crystallinity restoration effectiveness of shorttime, high-temperature annealing in implanted layers has been examined. In addition, we have reported the influence of thermal annealing at different temperatures in order to study the annihilation of defects induced by As and phosphorus ion implantation. Quantitative discussions relating the magnitude of the PMTR signal to the local conditions of the layer (e.g., presence and types of defects, inhomogeneous-to-amorphous transition) also have been performed. It also has been shown that PMTR is an interesting nondestrucive evaluation technique for studying optoelectronic parameters over a wide temperature range. The free-carrier contribution to the PMTR signal becomes more important as temperature decreases. Furthermore, the nature of this contribution changes when amorphization occurs in the implanted layer. The physical mechanism then switches from Drude to the band-filling effect. Experimental evaluations of the temperature dependence of the lifetime and surface recombination velocities have been obtained with thermal-wave techniques. The dependencies of thermal and plasma temperature coefficients offer the advantage of decoupling thermal and plasma contributions even at low modulation frequencies. The variation of the thermal diffusion length with temperature and its effect on the PMTR versus

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS 147

temperature signal have been discussed. The main results of this chapter can be summarized as follows. 1. The strong influence of the implant dose on the PMTR thermal-wave signal is related to the degree of the induced disorder in the damaged layer; comparison with studies performed in the plasma-wave regime suggests that ion implantation has a larger effect on electronic properties than on the thermal properties of the system. 2. The strong increase of the PMTR signal around the critical dose and its subsequent saturation indicates the crystalline-to-inhomogeneousto-amorphous transformation of the implanted layer. 3. The large effect of the annealing temperature on the PMTR signal of the implanted layers also is observed. Annealing periods longer than 15 min appear to be necessary to recrystallize heavily damaged layers, even at T, = 1110°C. The activation energy of the local annealing recovery mechanism in As-damaged Si has been found to be E = 0.15 eV. 4. It seems that the 100 kHz PMTR signal (induced by a laser beam of 30 pm spot size) is not very sensitive to the presence of dislocation-type defects. 5. There is a strong variation of the PMTR signal with ambient temperature. 6. Contributions of the substrate material to the PMTR versus the ambient temperature signal have been detected and have been explained in terms of the frequency and temperature dependence of ,+. This suggests some depth-profiling applications as a function of experimental temperature. 7. The PMTR signal temperature dependence may become a useful tool in elucidating the thermal and optical phenomena in ion-implanted semiconductors; useful comparisons with results from experiments involving carrier transport may be drawn vis-a-vis the effects of implantation damage on carrier and phononlike properties in semiconductors.

REFERENCES Anjum, M., Sandhu, G . S., Cherekdjian, S., and Weisenberger, W. (1991). Thermal Wave Characterization of Silicon Implanted with MeV Phosphorus Ions. Nucl. Instrum. Meth. B55,266-268. Bebb, H. B., and Williams, E. W. (1972). Semiconductors and Semimetals. (R. K. Willardson and A. C. Beer, eds.) Transport and Optical Phenomena, Vol. 8. Academic Press, New York, London, 262.

148

C. CHRISTOFIDES

Boltaks, B. (1977). Difusion and Point Defects in Semiconductors. Mir, Moscow, Chapts. 2 and 3. Christel, L. A., Gibbons, J, F., and Sigmon, T. W. (1981). Displacement Criterion for Amorphization of Silicon during Ion-Implantation. J. Appl. Phys. 52, 7143-7146. Christofides, C. (1992). Annealing Kinetics of Defects of Ion-Implanted and Furnace-Annealed Silicon Layers: Thermodynamics Approach. Semicond. Sci. Technol. 7, 1283-1294. Christofides, C., Jaouen, H., and Ghibaudo, G . (1987). Etude de silicium implante a I'arsenic par l'effet de transport. Influence du recuit thermique. Rev. Phys. Appl. 22,407-412. Christofides, C., Jaouen, H., and Ghibaudo, G. (1989a). Electronic Transport Investigation of Arsenic Implanted Silicon. I. Annealing Influence on the Transport Coefficients. J . Appl. Phys. 65,4832-4839. Christofides, C., Ghibaudo, G., and Jaouen, H. (1989b). Electronic Transport Investigation of Arsenic Implanted Silicon. 11. Annealing Kinetics of Defects, J . Appl. Phys. 65,4840-4844. Christofides, C., Vitkin, I. A,, and Mandelis, A. (1990) Photothermal Reflectance Investigation of Processed Silicon. I. Room-temperature Study of the Induced Damage and of the Annealing Kinetics of Defects in Ion-Implanted Wafers. J. Appl. Phys. 67, 2815-2821. Christofides, C., Mandelis, A., Engel, A,, Bisson, M., and Harling, G. (1991). Quantitative Photopyroelectric Out-of-Phase Spectroscopy of Amorphous Silicon Thin Films Deposited on Crystalline Silicon. Canad. J. Phys. 69, 317-323. Christofides, C., Othonos, A., Bisson, M., and Boussey-Said, J. (1994). Optical Spectroscopy on Implanted and Annealed Silicon Wafers: Plasma Resonance Wavelength. J . Appl. Phys. 75, 3377-3384. Christofides, C., Diakonos, F., Seas, A., Christou, C., Nestoros, M., and Mandelis, A. (1996). Two Layer Model for Photomodulated Thermoreflectance of Semiconductor Wafers. J. Appl. Phys. 80, 1713-1725. Doka, O., Miklos, A., and Lorincz, A. (1988). Signal Generation in Optically Detecting Thermal-wave Instruments. J. Appl. Phys. 63, 2156-2158. Drude, P. (1900). Zur Elektronentheorie der Metalle. Ann. Phys. (Leipzing) 1, 566-613. Dziewior, J., and Schmid, W. (1977). Auger Coefficients for Highly Doped and Highly Excited Silicon. Appl. Phys. Lett., 31, 346-348. Forget, B. C., Fournier, D., and Gusev, V. E. (1993). Nonlinear Recombinations in Photoreflectance Characterization of Silicon Wafers. Appf. Surfac. Sci. 63, 255-259. Fournier, D., and Boccara, A. C. (1989). In Photothermal Investigation ofsolids and Fluids (JA Sell ed.) Academic Press, New York, 36-81. Fournier, D., and Forget, B. C. (1991). Thermal Wave Probing of the Optical, Electronic and Thermal Properties of Semiconductors. J . de Phys. IVC6, 241-252. Fuller, C. S. (1959). Defect Interactions in Semiconductors. In Semiconductors (N. B. Hannay, ed.). Reinhold, New York, Chapt. 5. Gibbons, J. F. (1968). Ion Implantation in Semiconductors. Part I: Range Distribution Theory and Experiments. Proc. IEEE 56,295-320. Gibbons, J. F. (1972). Ion Implantation in Semiconductors. Part 11: Damage Production and Annealing. Proc. IEEE 60, 1062-1096. Guidotti, D., and van Driel, H. M. (1985). Spatially Resolved Defect Mapping in Semiconductors using Laser-modulated Thermoreflectance. Appl. Phys. Lett., 47, 1336- 1338. Huldt, L., Nilsson, N. G., and Svantesson, K. G. (1979). The Temperature Dependence of Band-to-Band Auger Recombination in Silicon. Appl. Phys. Lett., 35, 776-777. Ishikawa, K., Yoshida, M., and Inque, M. (1987). Secondary Defects of AS* Implanted Silicon Measured by Thermal Wave Technique. Japan. J. Appl. Phys. 26, L1089-L1091. Kirby, B. J., Larson, L. A., and Liang, R. (1987). Nucl. Instrum. Meth. B21, 550. Li, H. H. (1980). Refractive Index of Silicon and Germanium and its Wavelength and Temperature Derivatives. Journal of Physical and Chemical Reference Data, 561-658.

4 PHOTOMODULATED THERMOREFLECTANCE OF IMPLANTED WAFERS 149 McFarlane, R. A., and Hess, L. D. (1980). Photoacoustic Measurements of Ion-Implanted and Laser-annealed GaAs. Appl. Phys. Lett., 36, 137-139. MacFarlane, G . G., McLean, T. P., Quarrington, J. E., and Roberts, V. (1958). Fine Structure in the Absorption-edge Spectrum of Si. Phys. Rev. 111, 1245-1254. Mandelis, A., Nestoros, M., and Christofides, C. (1996). Effect of Thermal Excitation on the PMTR Signal Generation. J . Appl. Phys. submitted. Mandelis, A., Vanniasinkam, J., Budhudu, S., Othonos, A., and Koktas, M. (1993). Absolute Nonradiative Energy-Conversion-Efficiency Spectra in Ti3':A1,0, Crystals Measured by Noncontact Quadrature Photopyroelectric Spectroscopy. Phys. Rev. B 48, 6808-6821. Mandelis, A., and Wagner, R. (1996). Quantitative Deconvolution of Photomodulated Thermoreflectance Signals from Si and Ge Semiconductors. Japan. J. Appl. Phys. 35,1786- 1797. Nestoros, M., Forget, B. C., Christofides, C., and Seas, A. (1995). Photothermal Reflection versus Temperature: Quantitative Analysis. Phys. Rev. B 51, 14115-14123. Opsal, J., Rosencwaig, A., and Willenborg, D. (1983). Thermal-Wave Detection and Thin-Film Thickness Measurements with Laser Beam Deflection. Appl. Opt. 22, 3169-3179. Othonos, A., and Christofides, C. (1996). Multi-wavelength Raman Probing of Phosphorus Implanted Silicon Wafers. Nucl. Instrum. Meth. B 117, 367-374. Othonos, A., Christofides, C., Said, J. B., and Bisson, M. (1994). Raman Spectroscopy and Spreading Resistance Analysis of Phosphorus Implanted and Annealed Silicon. J. Appl. Phys. 75,8032-8038. Pankove, J. I. (1971). Optical Processes in Semiconductors. Dover Publications, New York, 27. Pierret, R. F. (1987). Modular Series on Solid State Devices. In Volume VI Advanced Semiconductor Fundamentals (R. F. Pierret and G. W. Neudeck, eds.) Addison-Wesley, Reading, MA, 188-192. Prussin, S., Margolese, D. I., and Tauber, R. N. (1985). Formation of Amorphous Layers by Ion Implantation. J . Appl. Phys. 57, 180-185. Reiss, H., Fuller, C. S., and Morin, F. J. (1956). Chemical Interactions Among Defects in Germanium and Silicon. Bell Systems Tech. J . 35, 535-636. Rosencwaig, A. (1982). Thermal-Wave Imaging. Science 218, 223-228. Rosencwaig, A. (1987). Thermal Wave Characterization and Inspection of Semiconductor Materials and Devices. In Photoacoustic and Thermal Wave Phenomena in Semiconductors (A. Mandelis, ed.) North-Holland, New York, Chapt. 5. Rosencwaig, A. (1994). Thermal Wave Monitoring and Imaging of Electronic Materials and Devices. In Non-Destructive Evaluation. Progress in Photothermal and Photoacoustic Science and Technology Series, Vol. 11 (A. Mandelis, ed.) PTR Prentice Hall, Englewood Cliffs, NJ, Chapt. 4. Rosencwaig, A,, Opsal, J., Smith, W. L., and Willenborg, D. L. (1985). Detection of Thermal Waves Through Optical Reflectance. Appf. Phys. Letz., 46, 1013- 1015. Rosencwaig, A., Opsal, J., Smith, W. L., and Willenborg, D. L. (1986). Detection of Thermal Waves Through Modulated Optical Transmittance and Modulated Optical Scattering. J. Appl. Phys. 59, 1392-1394. Ryssel, H., and Ruge, I. (1986). Ion Implantation. John Wiley & Sons, New York. Sablikov, V. A. (1987). Photothermal Displacement of a Semiconductor Surface. Sou. Phys. Semicond. 21, 1319-1322. Schuur, J., Waters, C., Maneval, J., Tripsis, N., Rosencwaig, A., Taylor, M., Smith, W. L., Golding, L., and Opsal, J. (1987). Relaxation of Ion Implant Damage in Silicon Wafers at Room Temperature Measured by Thermal Waves and Double Implant Sheet Resistance. Nucl. Instrum. Meth. B21, 554-558. Seas, A,, and Christofides, C. (1995). Photothermal Reflectance Investigation of Implanted Silicon: The Influence of Thermal Annealing. Appl. Phys. Lett., 66, 3346-3348.

150

C. CHRISTOFIDES

Smith, R. A. (1978). Semiconductors, 2nd ed. Cambridge University Press, Cambridge, UK, 321. Smith, W. L., Rosencwaig, A,, and Willenborg, D. L. (1985a). Ion Implant Monitoring with Thermal Wave Technology. Appl. Phys. Lett., 47, 584-586. Smith, W. L., Taylor, M. W., and Schuur, J. (198513). Ultra-High-Resolution Dose Uniformity Monitoring with Thermal Waves. SPIE Proc. 530, 201-205. Smith, W. L., Rosencwaig, A., Willenborg, D. L., Opsal, J., and Taylor, M. W. (1986). Ion Implant Monitoring with Thermal Wave Technology. Solid State Technol. 29, 85-92. Sze, S. M. (1981). Physics ofSemiconductor Devices (Second Edition). John Wiley & Sons, New York. Tamura, M. (1973). Secondary Defects in Phosphorus-Implanted Silicon. Appl. Phys. Lett., 23, 651-653. Tomita, T., Kinosada, T., Yamashita, T., Shiota, M., and Sakurai, T. (1986). A New NonContact Method to Measure Temperature of the Surface of Semiconductor Wafers. Japan. J . Appl. Phys. 25, L925-L927. Touloukian, Y . S., Powell, P. W., Ho, C. Y., and Nicolaou, M. C. (1973). Thermophysical Properties of Matter. Vol. 10. IFI/Plenum, New York, Washington, 160. Uchitomi, N., Mikami, H., Toyoda, N., and Nii, R. (1988). Experimental Study on the Correlation Between Thermal-Wave Signals and Dopant Profiles for Silicon-Implanted GaAs. Appl. Phys. Lett., 52, 30-32. Vasilev, A. N., and Sandomirskii, V. B. (1984). Photoacoustic Effects in Finite Semiconductors. Sou. Phys. Semicond. 18, 1095-1099. Vitkin, 1. A., Christofides, C., and Mandelis, A. (1989). Laser-Induced Phothermal Reflectance Investigation of Silicon Damaged by Arsenic Ion Implantation: A Temperature Study. Appl. Phys. Lett., 54, 2392-2394. Vitkin, I. A., Christofides, C., and Mandelis, A. (1990). Photothermal Reflectance Investigation of Processed Silicon. I1 Signal Generation and Lattice Temperature in Ion-Implanted and Amorphous Thin Layers. J . Appl. Phys. 67, 2822-2830. Wagner, R. E. (1993). Quantitative Photomodulated Optical-Reflectance Studies of Silicon and Germanium Semiconductors, Ph.D. thesis, University of Toronto, Toronto, Ontario, Canada. Wagner, C., and Schottky, Z. (1930). Phys. Chem. B. 11, 163. Wagner, M., and Geiler, H. D. (1991). Single-Beam Thermowave Analysis of Ion Implanted and Laser Annealed Semiconductors. Measurement Sci. Technol. 2, 1088- 1093. Wagner, R., and Mandelis, A. (1991). A Generalized Calculation of the Temperature and Drude Photo-Modulated Optical Reflectance Coefficients in Semiconductors. J . Phys. Chem. Solids 52, 1061- 1070. Wagner, M., Winkler, N., and Geiler, H. D. (1991). Single-Beam Thermowave Analysis of Semiconductors. Appl. Surface. Sci. 50, 373-376. Weakliem, H. A., and Redfield, D. (1979). Temperature Dependence of the Optical Properties of Silicon. J . Appl. Phys. 50, 1491-1493. Wendman, M. A., and Smith, W. L. (1987). Thermal Wave Implant Dosimetry for Process Control on Product Wafers. Nucl. Instrum. Meth. B21, 559-562. Wurm, S., Alpern, P., Savignac, D., and Kakoschke, R. (1988). Modulated Optical Reflectance Measurements on Amorphous Silicon Layers and Detection of Residual Defects. Appl. Phys. A47, 147-155.

SEMICONDUCTORS AND SEMIMETALS, VOL. 46

CHAPTER 5

Photothermal Deflection Spectroscopy Characterization of Ion-Implanted and Annealed Silicon Films U.Zammit DIPARTIMENTO DI INCEGNZRJA

MECCANICA

UNIVERSIT~ DI ROMA ‘TonVERCATA” AND

INFM UNITA’ “ROMA2 TORVERGATA”

ROME,ITALY

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ill. RESULTSAND DISCUSSION . . . . . . . . . . . . 1. The Effect of the Implantation Dose . . . . . .

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

151 154 158 158

2. Effects of Annealing of Damaged Crystalline Material . . . . . . . . . . 3. Effects of Annealing of Amorphous Material. . . . . . . . . . . . . . . IV. CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

167 174 174

I. INTRODUCTION

11. THEORY AND EXPERIMENT.

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

166

I. Introduction

Gap states in disordered and amorphous semiconductors directly affect their optical and electronic transport properties. It therefore is most important to study the nature, number, and energy distribution of such states, which generally originate from absence of long-range order and also from dangling bonds, defects, and impurities present in the material. In ion-implanted semiconductors the gap states are associated with radiation-induced damage that, in silicon (Si), has been extensively studied for 30 years. Today, interest mainly is directed toward the understanding and application of the damage accumulated during ion implantation of the material because of its technologic implications for device fabrication and, recently, for defect engineering. In the former case the damage must be removed to achieve adequate electrical activation of the implanted atoms and carrier mobility, whereas the latter procedure refers to the controlled 151 Copyright 0 1997 by Academic Press All nghts of reproduction in any form reserved. 0080-8784197$25

u. ZAMMIT introduction of ion-induced damage to achieve the desired properties of the material (Wong et af., 1988; Lu et al., 1989; Tamura, Ando, and Onyu, 1991; Schreutelkamp et al., 1991). This is obtained by engineering the interactions between the induced defects and implanted atoms through a suitable choice of implantation conditions and annealing cycles and therefore derives from a thorough knowledge of damage nucleation and evolution during implantation and subsequent annealing cycles. The study of defects in amorphous Si (a-Si) obtained by ion implantation very recently has also drawn considerable attention because of the role defects play in the processes taking place during annealing of the material, which are referred to as structural relaxation. Amorphous Si is generally viewed as a continuous random network of tetrahedrally coordinated, covalently bonded atoms (Polk, 1971). On annealing, the positions of all the atoms in the network evolve, giving rise to variations of the average network parameters, such as reduction of the tetrahedral bond-angle distortion and therefore of the average strain in the material (see, e.g., Roorda et al., 1991 and references therein). In ion-implanted a-Si there have been indications that this process is accompanied by annihilation of defects of the same kind as those observed in damaged crystalline material (Coffa et al., 1992). Point defects such as vacancies and vacancy-impuritycomplexes, in fact, have been detected in ion-implanted a-Si (Van de Hoven, Liang, and Niesen, 1992). Defects in ion-implanted semiconductors, as well as in hydrogenated amorphous Si (a&: H), have been studied using a variety of techniques, such as electron paramagnetic resonance (EPR), luminescence, photoconductivity, electrical conductivity, capacitance, and optical absorption (Brower, Vook, and Borders, 1970; Amer and Jackson, 1984; Fritzsche, 1985). Except for the optical absorption, however, all the techniques are sensitive to the Fermi energy position, detect only radiative transitions, or require special kind of doping; therefore, they are not sensitive to all defects. Conversely, subgap absorption is sensitive to the energy and number of all defects, and thus provides a versatile tool for their study. In ion-implanted Si, for example, EPR has certainly been of major importance since it has provided the microscopic information at the atomic level that has led to the identification of several defects in irradiated Si. Among the intrinsic ones, such defects include the divacancy (Watkins and Corbett, 1965), the four-vacancy (Brower, 197 1) and the five-vacancy complexes (Lee and Corbett, 1973) and the di-interstitial (Lee, Gerasimenko, and Corbett, 1976). Electron paramagnetic resonance has also enabled determination of the absolute concentration of the observed defects. Subgap absorption has also proved useful in studying irradiation-induced defects in Si. Absorption bands at 1.8, 3.3, and 3.9 pm have been associated with the various charge states of the divacancy, the most prominent intrinsic defect being stable at room temperature (Fan and Ramdas, 1959; Cheng et

5 ION-IMPLANTED AND ANNEALED SILICON FILMS

153

al., 1966; Cheng and Lori, 1968). Moreover, for the known divacancy concentration determined by EPR and the absorption measured at the 1.8-pm band, it was possible to correlate the magnitude of the absorption with the divacancy concentration (Cheng and Lori, 1968). This has enabled monitoring of the divacancy concentration in ion-implanted Si by subgap absorption as a function of both the implantation conditions and the annealing cycles (Stein et al., 1970). With respect to EPR detection, the determination of the concentration of defects by optical absorption measurements is less subject to indetermination that may arise in connection with changes in the charge state of the defect. Such changes, associated with Fermi energy level variations occurring during ion implantation (Sonder and Templeton, 1963), may turn the defects diamagnetic, thus making them undetectable by EPR. The detection of very low levels of absorption, such as the one typical in the subgap region of thin films of semiconductors, requires very sensitive techniques. Conventional transmission and reflectance methods enabled the detection, in 1-pm-thick films of a-Si: H, of absorption levels of 50 to 100cm-' (see, e.g., Freeman and Paul, 1979; Cody et al., 198la), whereas subgap absorption associated with gap states may be considerably smaller. This limitation has finally been overcome with the introduction of photothermal techniques based on the measurement of the thermal energy deposited in the material of interest as electromagnetic radiation is absorbed. This direct method for measuring optical absorption has enabled the achievement of sensitivity levels that, depending on the detection configurto lop7 ation for detecting the photothermal signal, ranged from PI = (p is the absorption coefficient and 1 the sample thickness). The gas microphone photoacoustic (PA) detection configuration was first used for measurements in semiconductors. Rosencwaig reported measurements in powdered gallium phosphide (Gap) samples in 1977 (Rosencwaig, 1977) in the spectral region above the bandgap. PA subgap absorption measurements were reported later in bulk GaAs to detect the extrinsic absorption of chromium (Cr) (Eaves, Vargas, and William, 1981) and in ion-implanted gallium arsenide (GaAs) films (Morita and Sato, 1983). It was with the development of the photothermal deflection spectroscopy (PDS) configuration (Boccara, Fournier, and Badoz, 1980), however, that subgap absorption measurements in semiconductor thin films received a great impulse due to its considerably larger sensitivity, with respect to PA, and to the significantly smaller influence of the stray signal arising from light scattered by sample inhomogeneities and hitting the PDS cell walls (Yasa, Jackson, and Amer, 1982). Great effort has been devoted to the study of a-Si: H films by (Jackson and Amer, 1982), who showed that the concentration of dangling bonds, determined by an optical sum rule from the absorption data, compared favorably with that obtained from EPR measurements over

-

154

U. ZAMMIT

four orders of magnitude. Moreover, the absorption spectra enabled the derivation of the features in the electronic density of states of the material (Amer and Jackson, 1984). Finally, the PDS measurements enabled the detection of surface and interface states in a : Si-H films, both from results obtained as a function of film thickness (Jackson et al., 1983) and from the analysis of the interference-inducedoscillations in the spectra (Grill0 and De Angelis, 1989; Amato et al., 1990). The first subgap absorption results obtained by PDS measurements of ion-implanted Si and GaAs films were first reported in 1989 (Luciani et al., 1989). More detailed results over a wider spectral region were later reported (Zammit et al., 1991) and allowed description of the possible changes induced in the distribution of gap states of the material with increasing ion dose (Zammit et al., 1992). Moreover, absorption studies performed in a-Si:H by a conventional transmission technique (Cody et al., 1981b) showed that even the absorption values in the vicinity of the band-edge region-so as to enable the determination of the band-edge slope and optical gap of the material-are very useful because they allow analysis of the average degree of disorder present in the material. This circumstance makes absorption measurements particularly useful for studying structural relaxation in ion-implanted a-Si since they enable the detection of the reduction of both the defect concentration and the average strain in the material, which are known to accompany the structural relaxation process. Strain reduction by optical absorption measurements also has been detected in ion-irradiated hydrogenated amorphous carbon (a-C :H) (Compagnini et al., 1995). In this chapter, a review is given of the results obtained in implanted Si samples over a wide spectral region that extends from the visible to deep and into the subgap region of the samples as a function of implantation conditions and annealing cycles (Zammit et al., 1994a, 1994b). In order to have access to the absorption data in the spectral region above the band-edge region of the samples, the investigations were carried out on thin-film samples lying on transparent substrates. Implantation thus was carried out on silicon on sapphire (SOS) films and on a-Si films sputtered on quartz.

11. Theory and Experiment

In the PDS configuration, the heat generated by the radiation that is absorbed by the sample diffuses into the surroundings and induces a temperature gradient in the medium in front of the sample. A probe laser

5

ION-IMPLANTED AND ANNEALED SILICON FILMS

155

FIG. 1. Schematic representation of experimental setup (see text).

beam, grazes the sample surface and then meets a region with a refractive index gradient -which is associated with the temperature gradient -and therefore undergoes a deflection that is detected by a position sensor, as sketched in Fig. 1. If the incident radiation is intensity-modulated, then the probe beam experiences synchronous periodic deflection, the corresponding signal of which can be fed to a lock-in amplifier that can measure the amplitude and the phase of the signal. The theory for the PDS signal is now well established (Jackson et al., 1981). When the sample thickness is much where smaller than the thermal diffusion length in the sample, 1, = D is the thermal diffusivity and f the modulation frequency, and if both the deflecting medium and the sample’s substrate absorption are negligible, then the PDS signal amplitude S is proportional to the sample absorptance A = 1 - T - R, where T is the transmittance and R the reflectance. If the signal is normalized with respect to a value obtained in the spectral region where the sample is optically opaque (optical saturation) and assuming no or weak wavelength dependence of the reflectance, then we obtain

mJ

from which the sample absorption coefficient readily can be obtained (Amer and Jackson, 1984). The angular brackets indicate spatial averaging along the probe beam. When thin films are considered, however, removal of the interference-induced oscillations in the absorption spectra is sometimes

u. ZAMMIT

156

obtained by averaging also with respect to such oscillations. Such a procedure is not always advisable, particularly when we want to detect some structures in the absorption spectra (such as the absorption bands at 1.8 pm) associated with divacancies in ion-implanted Si films since they would become altered when performing the averaging. In such a case the effects of the oscillations must be removed in some other way. A very interesting procedure was proposed (Ritter and Weiser, 1986), which was based on the analysis of both the transmittance and absorptance spectra. It was shown that the oscillations in T and A are coherent and that the ratios between the interference-induced maxima and minima in T and A are approximately equal

and #Imin are the absorption coefficients at the wavelength at where which the interference maximum and minimum, respectively, are observed, and R, and R, are the film-front medium and film-substrate reflection coefficients, respectively. If the ratio T/A is considered, then the oscillations should cancel out. In fact, we obtain T A

--x

1-R, exp(b1) - R, exp( - f l l ) - (1 - R,)

(3)

which is free from interference-induced oscillations and can be solved easily to obtain b. It also has been shown that the cancellation of the oscillations in the ratio TIA is complete only provided the film volume absorption is much larger than the localized absorption associated with surface or interface states (Grillo and De Angelis, 1989; Amato et al., 1990). This is because the presence of such states would increase the ratio Amax/Aminbut not Tmax/Tmin. This circumstance, in fact, was exploited to measure the ratio of bulk-to-surface-stateabsorption in a-Si :H and to determine whether the localized states originated from the front surface of the sample or from the film-substrate interface. Here we measure the ratio of the simultaneously determined A and T spectra to eliminate the interference effects and set up the experiment to perform this kind of measurement, as sketched in Fig. I The broadband light source consists of a 1OOO-W Xenon (Xe) lamp monochromatized by a 0.25-m monochromator. The light is chopped by a mechanical chopper, which operates at 20Hz and has an aluminized blade. During the first half

5 ION-IMPLANTED AND ANNEALED SILICON FILMS

157

of each cycle, the blade allows the light to reach the PDS cell that contains the sample. During the second half of the cycle, the blade reflects the light onto a PA cell containing a black absorber. This enables us to perform the normalization, with respect to the spectral intensity distribution of the broadband source, at the same time as the absorption measurements. The PDS cell is contained in a compact PDS assembly (Charbonnier and Fournier, 1986) and carbon tetrachloride is (CCl,) used as the deflecting medium. Finally, a second PA cell with a black absorber is placed behind the PDS cell to measure, once again simultaneously with the absorption measurements, the sample transmittance. The refracting optics and the cell windows are made of infrasil quartz, which is transparent even in the 2.6 to 2.8pm range. Thus it was possible to carry out measurements from the visible-ultraviolet (UV) up to well inside the infrared (IR) region. Since the maximum efficiency of the Xe lamp is in the 800 to 1200 nm range and since the Xe lamp has a poorer output at wavelengths outside this range, we have used two gratings, blazed at 400 nm and at 2 pm, to optimize the transfer efficiency of the monochromator in the UV and IR regions. This combined with the progressive increase of the monochromator slit aperture with increasing wavelength-as shown in Fig. 2-to compensate for the power reduction, enabled us to achieve a light source with an adequate intensity distribution to perform measurements in the 300 to 3400 nm range, and with a value of AL/A, which is fairly uniform. The two gratings were contained in a double grating holder in the monochromator and could be interchanged during the measurement at a suitable wavelength. Given the wide spectral range, a normalization run -with a black absorber in the PDS cell -was

1 (nm) FIG. 2. Spectral distribution of the intensity of the broadband source used in the experiment, together with the bandwidth (AA) in the different spectral regions.

U.ZAMMIT

158

performed to compensate for the wavelength-dependent performance of the reflective and refractive optics. The impIanted samples consisted of 800-nm-thick crystalline SOS films and 400-nm-thick a-Si films sputtered on quartz. On each sample, Si ions were implanted at 150 keV and 300 keV (dose ratio 0.4: 1) to obtain a fairly uniform damage profile that extended over 450 nm in the SOS and over the entire thickness in the sputtered films. Self-ion-irradiation was chosen to eliminate the effects arising from chemically dissimilar ions on the damage formation and evolution. 111. Results and Discussion

1. THEEFFECTOF T H E IMPLANTATIONDOSE Figure 3 shows the spectra relative to the samples implanted with increasing ion dose (dose values refer to the sum of the doses implanted at 150 and 300keV on each sample) at 300K, while Fig. 4 refers to the corresponding results for 80 K implantation. The lower absorption observed for the higher implantation temperature reflects the effect of the partial annealing of the damage that occurs during 300 K implantation. It should be remarked that the layer implanted at 80K with a dose of 8.4 x l O I 4 cm-' consisted of a uniform amorphous layer, as determined by electron diffraction (not shown), whereas the other implantations led to the formation of damaged crystalline material or a mixture of the two materials.

0.5

1

E eV

2

3

FIG.3. Dose dependence of the absorption skectb of silicon on sapphire (SOS) films implanted at room temperature. (From Zammit, U., (1994). Phys. Rev. B49, 14322, with permission.)

5 ION-IMPLANTED AND ANNEALED SILICON FILMS

159

n

<E 103:: 7

0.5

1 E (ev)

2

3

FIG.4. Dose dependence of the absorption spectra of 80 K implanted silicon on sapphire (SOS) films. (From Zammit, U., (1994). Phys. Reo. B49, 14322, with permission.)

The spectra show a progressive increase in absorption with dose all over the investigated spectral region, due to the progressive accumulation of damage, as long as the implanted layer consisted of damaged crystalline material. When, with further dose increase, amorphous material also began to form in the layer (2.8 x 1015cmP2 at 300 K in Fig. 3, and 8.4 x l O I 4 at 80 K in Fig. 4),an attenuation or a complete disappearance of the absorption band, which is peaked at 0.69 eV (1.8 pm) occurs. The 0.69-eV absorption band has already been investigated thoroughly and its origin well established (Fan and Ramdas, 1959; Watkins and Corbett, 1965; Cheng et al., 1966; Cheng and Lori, 1968). Correlation of the optical absorption investigations, carried out as a function of implantation conditions and annealing cycles, against EPR results enabled the association of the absorption band with the divacancy defect. Moreover, optical absorption measurements performed on implanted samples stressed along specific directions so as to introduce dichroism in the absorption bands has confirmed that the defect responsible for the absorption has the same atomic symmetry as the one determined by EPR for the divacancy (missing neighboring atoms are along [lll] directions) (Watkins and Corbett 1965; Cheng et al., 1966). Finally, once again through correlation of absorption studies carried out on implanted samples of different doping (such as to induce different charge states of the divacancy) with EPR results, it was possible to establish that the 0.69-eV band is associated with divacancies in their single negatively, neutral, and single positively charged states, and that divacancies also give rise to absorption bands at 0.32eV (3.9pm) when

U. ZAMMIT

single positively charged and at 0.37 eV (3.3 pm) when double negatively charged (Fan and Ramdas, 1959; Cheng et al., 1966). Our results show progressive growth and subsequent quenching of the absorption band at 0.69 eV with implantation dose. It should be pointed out that the other two previously mentioned bands would not be observable in our samples since they would require the Fermi energy EF to be positioned below ( E , + 0.21) eV and above ( E , - 0.4) eV, respectively (Cheng et al., 1966) ( E , and E , are, respectively, the bottom of the conduction band and the top of the valence band). In our intrinsic SOS material E F is situated in midgap vicinity ( E , + 0.55 eV). According to a previously adopted procedure (Stein et al., 1970), we have calculated the absorption due to the band alone (B,,) by subtracting from the absorption spectra the contribution due to the background, determined by fitting the data away from the band. Figure 5(a) shows the results for the 300K implantation where the data relative to 8.4 x 1014cm-2 (not shown in Fig. 3) also are reported. Correspondingly, Fig. 5(b) shows the results relative to 80 K implantation. No divacancy absorption could be detected for the samples with the largest degree of damage, that is, 80K implantation with 2.8 x 1014 and 8.4 x lOI4 cm-’, where the implanted layer consisted mainly of amorphous material. From By, we have calculated the concentration of divacancies according to the relation, suggested by Stein et al. (1970): N , , ( ~ m -= ~ 7.7 ) x 10l6 PVYmax, where /?vv,,x is the maximum value of by, in the band, and the average number of divacancies produced per ion R,, = N,,/lO, where O is the ion dose and 1 is the implanted layer thickness. Figure 6 reports the dose dependence of such quantities. N , , initially increases, reaches a maximum value of 10” cm-3 at a dose of 2 x 1014cm-2,and then gradually becomes quenched for doses exceeding such a value. Conversely, R,, initially shows a slight decrease, which then becomes substantial for doses exceeding 2 x 10’4cm-2. A similar result concerning N , , had previously been obtained in bulk implanted Si where the maximum attained value was -7 x 1019cm-3 (Stein et al., 1970). Considering an average value of 1.5 f 0.5 keV for the divacancy production energy (Stein et al., 1970), this corresponds to an average energy deposited per unit volume of lozokeV/cm3, an order of magnitude smaller than the average value required to induce complete amorphization in the implanted layer. The mechanism that has been suggested as being responsible for the decrease of the detected divacancies is self-quenching due to interaction among divacancies or interaction with other defects that anneal in parallel with the divacancies. In fact, when a sample that had been implanted with a dose value such as to induce quenching of P,, was annealed at temperatures that eliminate divacancies (200 to 230°C) and was then re-implanted

-

-

5

ION-IMPLANTED AND ANNEALED SILICON FILMS

161

1200 n

E

Y

7

1000

800 600 400

200 0 0.4

0.5

0.6

0.7

0.8

0.9

0.8

0.9

E (ev)

0.4

0.5

0.6

0.7 E (ev)

FIG.5. Dose dependence of the divacancy absorption bands for silicon on sapphire (SOS) films: (a) room temperature implantation and (b) 80 K implantation. (From Zammit, U., (1994). Phys. Rev. B49, 14322, with permission.)

with lower doses, the divacancy absorption and production rate obtained were similar to those obtained when implanting virgin crystalline material (Stein et al., 1970). Regarding the origin of the quenching process, it has been suggested that once a critical concentration of defects is exceeded, interaction of the divacancies with other divacancies, defects, or both would induce, with respect to isolated divacancies, changes in their wave functions and energy levels or give rise to new complexes altogether, thus changing their optical and their EPR response (Brower and Beezhold, 1972). This is likely to occur for doses such that overlapping of the damage regions from the different ions becomes substantial, due to the considerable reduction

U. ZAMMIT

162

10l2

10 l3

10 l4 10 l5 Dose (1/cm2)

10l6

FIG.6. Dose dependence of the average number of divacancies produced per ion (Rvv), divacancy concentration (Nvv), and band-edge inverse logarithmic slope (E,) for SOS films implanted at room temperature. (From Zammit, U., (1994). Phys. Reo. B49, 14322, with permission.)

induced in the average distance between the divacancies, and would cause the production rate of divacancies to become nonlinear with respect to the implantation dose. A considerable reduction in the average number of divacancies produced per ion would then be expected, as we report in Fig. 6 for doses exceeding the one corresponding to the maximum divacancy concentration. Quenching of divancies as detected by EPR has been previously reported but for concentration exceeding only 1019cm-3 (Brower and Beezhold, 1972). A word of caution, however, is needed when EPR detection is involved. In fact, the reduction in the observed concentration may be caused by a change in the charge state of the divacancy, which would turn diamagnetic and thus become nondetectable by EPR. Accumulation of irradiation damage in Si is known to shift the Fermi energy progressively toward midgap (Sonder and Templeton, 1963) and could well cause a change of the divacancy charge state from singly negative or singly positive to neutral. Conversely, a reduction of the detected N , , value could not be attributed to the P doping of the Si used in Brower and Beezhold (1972), because P doping in Si is known to enhance the divancy production by trapping the interstitial Si (Svensson and Lindstrom, 1992). Thus, in this respect, divacancy detection through its 0.69 eV absorption band is of greater flexibility with respect to EPR,since it remains unaffected when the Fermi energy position ranges between ( E , - 0.4) eV and E , (Cheng et al., 1966).

5 ION-IMPLANTEDAND ANNEALEDSILICON FILMS

163

E (W FIG. 7. Exponential fits in the band-edge regions of spectra of Fig. 3. (From Zammit, U., (1994). Phys. Rev. B49, 14322, with permission.)

In Fig. 6 we also report the dose dependence of the Urbach parameter (inverse-logarithmicslope) E,, which characterizes the exponential dependence of the absorption in the band-edge region (fl = fl, exp[E/E,]), where absorption depends on the distribution of band-tail states. The E, values have been obtained by fitting the absorption data with such an exponential dependence in the band-edge regions of the spectra, as shown in Fig. 7. The dose dependence of E, closely correlates to that of N,,, which suggests that the electronic states associated with the divacancies strongly affect the population of band-tail states. This is consistent with the suggestion that the divacancy-associatedstates, from which the optical transitions originate, lie in the valence-band tail (Cheng et al., 1966). In Fig. 8 we show that the Po values we have obtained from the fit correlate very closely with the corresponding E , values. This should not be unexpected since it has been shown that, when assuming nonconservation of momentum in the optical transitions and contribution to the optical absorption due to a single band-tail region, flo is proportional to E , (Pankove, 1965). The first assumption should be justified for disordered semiconductors;the second can be considered valid, bearing in mind that in disordered Si (a-Si: H) it has been found that the conduction band-tail region is considerably steeper than that of the valence band (Tiedje et al., 198 1). Nonetheless, despite the strong correlation between such parameters, no unambiguous linear dependence between them can be obtained, as shown in the inset of Fig. 8, where we have also included the data relative to the 80 K, 8.4 x 10'4cm-2 implantation (asterisk), which led to the formation of a uniform amorphous layer.

U. ZAMMIT

164

f&(l/cm)

Eo(W

0.30 - 0.28 0.26

.’

6 Wcm)

40 -

,

0

0.22 0.20 0.18 -

20 *r---4

10’’

10l4 10’” Dose (1/cm*)

10’”

FIG.8. Dose dependence of /lo(see text) and E , for silicon on sapphire films implanted at room temperature. The mutual dependence of the parameters is shown in the inset where the asterisk refers to the 80 K 8.4 x 10’4cm-2 implantation. (From Zammit, U., (1994). Phys. Rev. B49, 14322, with permission.)

In Fig. 9 we report (BE)”’ versus E in order to probe the effect of the implantation conditions on the value of the optical gap EG, and on the dependence of the absorption coefficient on EG near the band-edge region. For unimplanted crystalline Si, near the band-edge region, the absorption coefficient is given by (Lynch, 1985)

where no is the refractive index, E,, is a phonon energy, and T is the temperature. Such an expression represents the well-known phonon-assisted (absorption or emission) nature of the indirect optical transition in Si between two electronic states for photon energies in the vicinity of the indirect bandgap. For low enough values of the absorption, where direct transitions are not involved, the versus the E plot should yield a dual linear-dependent region, which is evident in the inset of Fig. 9. The linear fits through the two linear-dependent regions yield intercepts on the abscissa whose arithmetic mean provide the value of EG, while the half dzerence value between such intercepts should yield Eph(Lynch, 1985). From the fits we obtain E, = 1.04 k 0.1 eV and E,, = 0.04 If: 0.1 eV. These values are not too different from those obtained by MacFarlane and Roberts (1955) at 300 K, that is, EG = 1.08 eV and E,, 0.05 eV (an equivalent phonon

-

5 ION-IMPLANTEDAND ANNEALED SILICON FILMS

165

? ! r

FIG.9. Dose dependence of (flE)’/’ versus E plots for implanted silicon on sapphire (SOS) films. Crystalline material spectrum is shown in the insert. (From Zammit, U., (1994). Phys. Rev. B49, 14322, with permission.)

temperature of 600 K). According to MacFarlane and Roberts (1955), E,, coresponds to the energy of the longitudinal acoustic wave with momentum equal to the momentum of the electrons at the minimum of the conduction band, which in Si is about 7/9 of the distance from the center to the edge of the Brillouin zone in the [00l] direction. According to the data obtained by neutron scattering and reported by Brockhouse (1959) such energy corresponds to l O I 3 Hz or 0.041 eV, showing the physical soundness of these and the results of MacFarlane and Roberts (1955). For amorphous Si, however, for absorption values on the order of 104cm-’, the energy dependence of /Iis given by the Tauc formula: BE = K ( E - E,)’ (Mott and Davies, 1979a), where K is a constant. A (BE)”’ versus E plot should then yield a linear dependence whose intercept on the abscissa provides E,. This is what we obtain for the fully amorphized layer (8.4 x lOI4 cm-2 in Fig. 9) for which E , = 1.15eV. This value is larger than the one for the crystalline material but smaller than those obtained for other kinds of a-Si, such as those obtained by room temperature sputtering (1.26eV) (Lewis, 1972) and in a-Si:H (1.5 to 1.6eV) (Mott and Davies, 1979a). In Fig. 9, a good linear fit with the same angular coefficient as in the previous case is also obtained for the sample implanted at 300 K with a dose of 2.8 x 1015cm-2 as it consisted predominantly of amorphous material with some residual damaged crystalline material. For all the other (lower) implantation doses, which gave rise to layers with predominantly heavily damaged crystalline material, no specific dependence of on E and E , can be predicted, and thus no comparison with theoretic expressions can

166

U. ZAMMIT

be performed. It can only be observed that, as expected, for increasing implantation dose the behavior of the data shifts progressively from that predicted for crystalline material toward that predicted for amorphous material. 2. EFFECTSOF ANNEALING OF DAMAGED CRYSTALLINE MATERIAL Figure 10 shows the spectra relative to the sample implanted at 300K with a dose of 2.8 x 10i4cm-2, which had originally shown the largest subgap absorption associated with divacancies and other defects, following isochronal annealing subsequently carried out for 15 min at increasing temperatures. The annealing temperature was increased in steps of 40 K between 393 and 433 K; 30 K between 433 and 463 K; 20 K between 463 and 523 K; and 50 to 70 K between 523 and 803 K. For the sake of clarity in the graphical representation, we only report spectra relative to some of the intermediate annealing stages. Nevertheless, it is clear from the reported spectra that the absorption decreases all over the investigated spectral region with increasing annealing temperature. However, while an annealing at 803 K is sufficient to recover the absorption typical of the unimplanted material in the region above the band edge, this does not occur in the subgap region where the absorption is also affected by extended defects, such as dislocations, that require much larger temperatures (1300 to 1400 K) to anneal out completely.

O

As implanted

* 433 K 1o4 0 463 K

h

E0

\

c o2 =1

-

403 K 523K 663 K

' 803K

0 Unlrnplanted

1 oo

0.3

0.5

1.o

E W)

2.0

FIG. 10. Annealing temperature-dependence of the absorption spectra of silicon on sapphire (SOS) film implanted at room temperature with a dose of 2.8 x 1014cm-Z. (From Zammit, U., (1994). Phys. Reo. B49, 14322, with permission.)

5 ION-IMPLANTEDAND ANNEALED SILICON FILMS

0.3

1-

0.8-

9

NdNO Eo

0.60.4 -

0.2-

167

. .--

- - a-- - -

.

0.

-

0.25

-

0.2

-

0.15 0.1

In the subgap region it can be observed that the band edge progressively sharpens with increasing annealing temperature and that the divacancy band is no longer detectable for annealing temperatures over 483 K. This latter aspect is reported in greater detail in Fig. 11, which shows, once again, that there is a strong correlation between the annealing behavior of the relative divacancy concentration and the Urbach parameter values, confirming that the divacancies strongly affect the population of band-tail states. Once the divacancies anneal out, the E , value remains constant around 0.14eV up to the maximum annealing temperature reached in this work. This value is still considerably larger than the one we had obtained for unimplanted material (0.07 eV) in the plot shown in Fig. 7, and it is probably associated with residual point defects, point-defect complexes, and strain induced by extended defects that survive up to much higher temperatures.

3. EFFECTSOF ANNEALING OF AMORPHOUS MATERIAL The annealing on amorphous material leading to structural relaxation was carried out in SOS implanted at 80 K with 5 x 10” cm-’ at 8 0 K, on a film sputtered at 350 K and annealed at 800 K for 2 h to induce complete relaxation and then implanted in the same conditions as the SOS sample, and on a film sputtered at 350K. The aim was to investigate and compare the structural relaxation processes in a-Si samples obtained by different methods and to emphasize the role of the point defects and point-defect

168

U. ZAMMIT

573 K 15 min-

803 K 15 min-

loo

I

0.3

803K2hr.

0.5

1

2

e

3

EW) FIG. 12. Annealing temperature and duration dependence of the absorption spectra of sputtered ion-implanted amorphous silicon film.

complexes that, as stated earlier on, should be found in a-Si obtained by ion implantation. Each sample was isochronally annealed for 15 min at temperatures of 473,573, and 803 K. The samples were finally annealed at 803 K for 2 h to induce maximum relaxation in the material, avoiding any crystallization. Figure 12 shows as an example the changes with annealing that take place in the spectra relative to the sputtered implanted sample (similar changes occurring in the other samples), and Fig. 13 shows some of the results obtained for the implanted SOS and the sputtered samples. The spectra for these samples extend almost down to 0.35 eV (3.5 pm). It should be pointed out that, despite the fact that the spectra extend to regions where the lamp throughput becomes poorer and where some of the samples show low absorption, the signal-to-noise ratio (SNR) for any of the reported data never went below 5. Absorption bands below 0.5eV can be observed that show smaller absorption after annealing. Concerning the origin of such bands, they lie in a spectral range where multiphonon absorption has been observed for Si (Hordvick and Skolnik, 1977). Moreover, we have observed these bands (not shown) also in crystalline (unimplanted) bulk Si and in crystalline SOS, but with lower absorption levels. Their absorption is thus sensitive to the presence of defects, although not entirely associated with them. Similar multiphonon absorption features have also been observed in the range of 2.8 to 6pm for diamond where the transitions involved are, moreover, defect-assisted (Smith and Taylor, 1962). Then a possible tentative explanation for the bands we observe in Si is that they also may be due to multiphonon defect-assisted transitions. More investigations are needed to further clarify this issue.

5 ION-IMPLANTED AND ANNEALED SILICON FILMS

169

1o5

lo4 n

lo3

\

1o2

E 0

Y F

E

-

S as implanted a-Si as sputtered + Sputtered a-Si + 803 K 2hr.

1o1

1oo

0.3

0.5

2

1 E(W

3

FIG. 13. Absorption spectra of as implanted silicon on sapphire (SOS), as sputtered and sputtered and annealed amorphous silicon (a-Si) films.

Our spectra show that, with the progressive increase in heat treatment, the absorption values decrease over all the investigated spectral region, whereas the absorption edge progressively shifts to larger energies and sharpens, as shown in greater detail in Figs. 14 and 15 where the changes in the band-edge slopes and in the Tauc plots, respectively, are reported for the implanted sputtered sample alone. The changes induced in the Urbach parameter and optical gap by the annealing cycles are reported in Figs. 16 and 17 for all the samples. The annealing of the amorphous material is thus

lo5

lo4 E

lo3

As implanted+

2 lo2

473 K 15 min-

lo1

803 K 15 min-

573 K 15 min-

Y

803K2hr.

1oo 1

135

2

2 5

E(W FIG.14. Exponential fits in the band-edge regions of the spectra of Fig. 12.

170

U. ZAMMIT

573 K 15 min803 K 15 min-

FIG. 15. Tauc plots (see test) of spectra of Fig. 12.

shown to lead to an increase in the optical gap and a decrease in the Urbach parameter (steeper band-edge slope). The width of the exponential tail (the inverse of the Urbach parameter) in crystalline semiconductors is generally accepted to provide a direct measure of the temperature-induced disorder (Cody et al., 1981b), because the random electric fields that it generates broaden the energy spectrum for the electronic transitions (Mott and Davies, 1979b). In amorphous semicon-

FIG. 16. Optical bandgap E , and E , as a function of annealing conditions for the implanted silicon on sapphire (SOS) and sputtered amorphous silicon (a-Si) films.

5 ION-IMPLANTED AND ANNEALED SILICON FILMS

171

E,(W 0.18 .

15 min isochronal

FIG. 17. Optical bandgap E , and E , as a function of annealing conditions for the unimplanted sputtered amorphous silicon (a&) films.

ductors there is an additional nonthermal (structural) component to disorder that should also affect the width of the exponential tail in the sense that this component should increase the width with increasing internal strain (Cody et al., 1981b). Thus the plausible general relation between the Urbach parameter and disorder in the material is given by Cody et al. (1981b) as

where K is a constant, ( U z ) is the mean atomic displacement from their equilibrium position, T is the temperature, and X is a parameter describing structural disorder. An analogous extension of the dependence of the optical gap from structural disorder as well as thermal disorder, which is known to reduce the gap value (Allen and Cardona, 1981) leads to (Cody et al., 1981b)

where E,(O,O) is the zero temperature and zero strain-gap value, D is a second-order deformation potential, and ( U 2 ) , represents the zero point uncertainty in the atomic positions. It is then evident that a reduction in the structural component of the disorder (strain) should lead to a decrease in the Urbach parameter and to

172

U. ZAMMIT

an increase in the optical gap of the material. Finally, the values of the two quantities should be linearly related by:

The changes occurring in the two parameters after annealing cycles thus indicate that the structural relaxation phenomenon is accompanied by reduction of strain in the material. A reduction in strain during structural relaxation in ion-implanted a-Si also has been detected by Raman spectroscopy (see Chapter 4) measurements and also has been related to the relaxation enthalpy detected by calorimetric measurements (Roorda et al., 1991). Moreover, such kind of behavior of E , on annealing has also been reported during structural relaxation of a-Si films obtained by vacuum evaporation at room temperature and then annealed for 2 h at 773 K (Lewis, 1972). On annealing, E , changed from 1.26 to about 1.5eV, a value that is very close to those we obtain for our relaxed a-Si samples (1.4 to 1.5 eV). Similar behaviors of both E , and E , also have been obtained in other cases in which structural disorder was reduced. A sharpening of the band edge and an increase of the optical gap, in fact, has been shown to be associated with a reduction of structure-related disorder in a-Si :H (Cody et al., 1981b), where such disorder was controlled by the H content in the material. Moreover, the relation between the changes in the two quantities have been found to be in good agreement with Eq. (8) insofar as the values of the parameters that appear in the equation and were determined from the experimental data to be physically reasonable. In fact, a linear fit to their E , versus E , data yielded a slope of approximately 6.2. From consideration reported in (Cody et al., 1981b) values of 0.08 A’ for (UZ),and of 17 meV for E,(O, 0) are used in their data analysis, leading to D = 16 eV/A2, a value of the same order of magnitude as similar deformation potentials obtained in crystalline Ge (4 to 5eV/A2) (Allen and Cardona, 1981). This confirms the consistency of their approach. Finally, a value of 2.0eV was obtained for E,(O,O), which is stated to represent the upper limit for the bandgap of the a-SiH, family of materials. Concerning our a-Si samples it is evident that, apart from the region below 0.5 eV where some differences appear, the spectra are very similar in the two implanted samples before annealing (and substantially different from the sputtered unimplanted one), independent of the fact that the samples have been obtained by implanting c-Si or relaxed a-Si. The same applies to the values E , and E,. Thus it is the implantation-induced defects in the two amorphous matrices that mainly affect their optical (physical) properties. Also, after structural relaxation induced by annealing at 803 K for 2 h, the optical properties of all three samples (not shown for implanted

5

2

ION-IMPLANTED AND

1.7 1.6 1.5 1.4 1.3 1.2 1.1

ANNEALEDSILICON FILMS

SOS 5E15 Sputtered 5E15

173

Sputtered

I 0.1 0.12 0.14 0.16 0.18 0.2

I " '

E, (ev) FIG. 18. E , versus E , plots for the annealed amorphous silicon (a-Si) films. SOS, silicon on sapphire.

SOS) become very similar; however, unlike the implanted samples, the values of E , and E , that are obtained for the sputtered unimplanted a-Si are not the maximum and minimum, respectively, that are obtained during the annealing cycle (currently, no explanation for this is available). In Fig. 18, we report the E , versus E , plot that we obtain in the case of ion-implanted and sputtered a-Si samples. The results indicate a linear dependence between the two parameters for all our samples, as predicted by Eq. (8). The value of the slope of the linear fit curve ranged between 3.7 and 4.2, yielding D values that ranged between approximately 10 and 11 eV/A2, similar to the value reported for a-Si:H. Even the value we obtain for EG(O,O), ranging between 1.8 and 1.9 eV, is similar to the a-Si : H value. The strain dependence behaviors of both E , and E G that we find in our three a-Si samples therefore are similar and are close to those reported for a-Si :H. In fact, the similarity in the properties of a-Si obtained by different methods, which we have detected in the similarity of the strain-associated behaviors of E,, goes further when we consider the similar results of calorimetric measurements performed during the relaxation of such films, bearing in mind that, as stated earlier, the calorimetric results also have been related to the strain reduction in the films. Thus the results we have obtained indicate that structural relaxation of a-Si leads to a material with physical properties that turns out to have little dependence on its preparation conditions. Moreover, the absorption data in the vicinity of the band-edge region confirm the strain reduction occurring during the structural relaxation process. As far as the role of defects during structural relaxation of implanted a-Si, comparing the subgap spectra of the three a-Si samples we note that before

174

U. ZAMMIT

annealing and in the subgap region where the absorption is entirely related to defects (above OSeV), the absorption values of the two implanted samples are very similar and considerably larger (by about a factor of three) than are those relative to the unimplanted sputtered film. This indicates that the excess absorption in the implanted sputtered film with respect to the unimplanted sputtered film is associated with implantation-induced defects. After annealing, the subgap absorption values become very similar in all three films, meaning that annihilation of the implantation-induced defects has occurred during structural relaxation of a-Si obtained by ion implantation. This points in the same direction as the results obtained by metal solubility and diffusion (Coffa et al., 1992). IV. Conclusions

In this chapter we have demonstrated the utility of optical absorption measurements for the study of implantation-induced damage in thin films of ion-implanted Si in the spectra range extending from the region above the bandgap down to the subgap region of the material. In this respect the use of the photothermal deflection spectroscopy (PDS) technique has been shown to be particularly adequate to give access to energies far into the subgap region of Si. We have monitored the various stages of formation and quenching of divacancies as a function of implantation conditions and annealing cycles through their 1.8-pm absorption band. The studies revealed that the divacancies strongly affect the population of band-tail states, that they reach a maximum concentration of 10'' cm- before formation of any amorphous material in the implanted layer, and that they anneal out completely over a very narrow temperature range around 230°C. The study of structural relaxation in ion-implanted a-Si gave indications that the process is accompanied both by a reduction in the average strain in the material and with the annihilation of implantation-induced defects. Moreover, it is shown that the structural relaxation process leads to an amorphous material whose optical properties are independent of its preparation method. REFERENCES Allen, B., and Cardona, M. (1981). Theory of the Temperature Dependence of the Direct Gap of Germanium. Phys. Rev. B 23, 1495. Amato, G., Benedetto, G., Boarino, L., and Spagnolo, R. (1990). Photothermal Deflection of Surface States in Amorphous Silicon Films. Appl. Phys. A 50, 507.

5 ION-IMPLANTEDAND ANNEALED SILICON FILMS

175

Amer, N., and Jackson, J.D. (1984). Optical Properties of Defect States in a-Si:H. In Semiconductors and Semimentals. (J. I. Pankove, ed.) Vol. 21B. Academic, New York, 83. Boccara, A. C., Fourneir, D., and Badoz, J. (1980). Thermo-optical Spectroscopy: Detection by the “Mirage” Effect. Appl. Phys. Lett. 36, 130. Brockhouse, B. N. (1959). Lattice Vibrations in Silicon. Phys. Rev. Lett. 2, 256. Brower, L. K. (1971). Structure of Multiple-vacancy (Oxygen) Centres in Irradiated Silicon. Radiation Eflects 8, 213. Brower, K. L., and Beezhold, W. (1972). Electron Paramagnetic Resonance of the Lattice Damage in Oxygen-implanted Silicon. J . Appl. Phys. 43, 3499. Brower, K. L., Vook, F. L., and Borders, J. A. (1970). Depth Distribution of EPR Centres in 400 keV 0’ Ion-Implanted Silicon. Appl. Phys. Lett. 16, 108. Charbonnier, F., and Fournier, D. (1986). Compact Design for Photothermal Deflection (Mirage): Spectroscopy and Imaging. Rev. Sci. Instrum. 57, 1126. Cheng, L. J., Corelli, J. C., Corbett, J. W., and Watkins, G . D. (1966). 1.8, 3.3 and 3.9pm Bands in Irradiated Silicon: Correlations with the Divacancy. Phys. Rev. 152, 761. Cheng, L. J., and Lon, J. (1968). Characteristics of Neutron Damage in Silicon. Phys. Rev. 171, 856.

Cody, G. D., Tiedje, T., Abeles, B., and Goldstein, Y. (1981b). Disorder and the Optical Absorption Edge of Hydrogenated Amorphous Silicon. Phys. Rev. Lett. 47, 1480. Cody, G., Wronsky, C. R., Abeles, B., Stephens, R., and Brooks, B. (1981a). Optical Characterisation of Amorphous Silicon Hydride Films. Solar Cells 2, 227. Coffa, S., Poate, J. M., Jacobson, D. C., Frank, W., and Gustin, W. (1992). Determination of Diffusion Mechanisms in Amorphous Silicon. Phys. Rev. B 45, 8355. Compagnini, G., Zammit, U., Madhusoodanan, K. N., and Foti, G . (1995). Disorder and Absorption Edges in Ion Irradiated Hydrogenated Amorphous Carbons. Phys. Reo. B 51, 11168.

Eaves, L., Vargas, H., and William, P. J. (1981). Intrinsic and Deep-level Photoacoustic Spectroscopy of GaAs (Cr) and other Bulk Semiconductors. Appl. Phys. Lett. 38, 768. Fan, H. Y., and Ramdas, A. K. (1959). Infrared Absorption and Photoconductivity in Irradiated Silicon. J . Appl. Phys. 30, 1127. Freeman, E. C., and Paul, W. (1979). Optical Constants of rf sputtered Hydrogenated Amorphous Si. Phys. Rev. B 20, 716. Fritzsche, T. H. (1985). Density-of-states Distribution in the Mobility Gap of a-Si : H. J . Non-Crystalline Solids 77i78, 273. Grillo, G., and De Angelis, L. (1989). Surface States and In-depth Inhomogeneity in a-Si: H Thin Films: Effects on the Shape of the PDS Sub-gap Spectra. J . Non-Crystalline Solids 114, 750.

Hordvick, A,, and Skolnik, L. (1977). Photoacoustic Measurements of Surface and Bulk Absorption in HF/DF Laser Window Materials. Appl. Optics 16, 2919. Jackson, W. B., and Amer, N. M. (1982). Direct Measurement of Gap State Absorption in Hydrogenated Amorphous Silicon by Photothermal Deflection Spectroscopy. Phys. Reo. B 25, 5559. Jackson, W. B., Amer, N. M., Boccara, A. C., and Fournier, D. (1981). Photothermal Deflection Spectroscopy and Detection. Appl. Optics 20, 1333. Jackson, W. B., Biegelsen, D. K., Nemanich, R. J., and Knights, J. C. (1983). Optical Absorption Spectra of Surface or Interface States in Hydrogenated Amorphous Silicon. Appf. Phys. Lett. 42, 105. Lee, Y. H., and Corbett, J. W. (1973). EPR Studies of Neutron-irradiated Silicon: A Negative Charge State of a Nonplanar Five-vacancy Cluster ( V J . Phys. Rev. B 8, 2810. Lee, Y. H., Gerasimenko, N. N., and Corbett, J. W. (1976). EPR Studies of Neutron-irradiated Silicon: A Positive Charge State of the (100) Split Di-intersitial. Phys. Rev. B 14, 4506.

176

U. ZAMMIT

Lewis, A. (1972). Evidence for the Mott Model of Hopping Conduction in the Anneal Stable State of Amorphous Silicon. Phys. Rev. Lett. 29, 1555. Lu, W. X., Oian, Y. H., Tian, R. H., Wang, Z. L., Schreutelkamp, R. J., Liefting, J. R., and Saris, F. W. (1989). Reduction of Secondary Defect Formation in MeV B+ Ion-implanted Si (100). Appl. Phys. Lett. 55, 1838. Luciani, L., Marinelli, M., Zammit, U., Pizzoferrato, R., Scudiei, F., and Martellucci, S. (1989). Subgap Absorption Spectra of Ion Implanted Si and GaAs Layers. Appl. Phys. Lett. 55, 2745. Lynch, D. W. (1985). Interband Absorption -Mechanisms and Interpretation. In Handbook of Optical Constants of Solids (E. D. Palik, ed.) Academic Press, New York, 197-198. MacFarlane, G. G., and Roberts, V. (1955). Infrared Absorption of Silicon Near the Lattice Edge. Phys. Rev. 98, 1865. Morita, M., and Sato, F. (1983). Photoacoustic Spectra on Laser-annealed GaAs Surfaces. Japan. J. Appl. Phys. 22 (suppl. 3), 199. Mott, N. F., and Davies, E. A. (1979a). In Electronic Processes in Noncrystalline Materials. Clarendon, Oxford, 387. Mott, N. F., and Davies, E. A. (1979b). In Electronic Processes in Noncrystalline Materials. Clarendon, Oxford, 276. Polk, D. E. (1971). Structural Model for Amorphous Silicon and Germanium. J. NonCrystalline Solids 5, 365. Pankove, J. I. (1965). Absorption Edge of Impure Gallium Arsenide. Phys. Rev. A 140, 2059. Ritter, D., and Weiser, K. (1986). Suppression of Interference Fringes in Absorption Measurements on Thin Films. Opt. Commun. 57, 336. Roorda, S., Sinke, W. C., Poate, J. M., Jacobson, D. C., Dierker, S., Dennis, B. S., Eaglesham, D. J., Spaepen, F., and Fuoss, F. (1991). Structural Relaxation and Defect Annihilation in Pure Amorphous Silicon. Phys. Rev. B 44, 3702. Rosencwaig, A. (1977). In Optoacoustic Spectroscopy and Detection (Yoh-Han Pao, ed.) Academic Press, New York, 193-239. Schreutelkamp, R. J., Custer, J. S., Liefting, J. R., Lu, W. X., and Saris, F. W. (1991). Preamorphisation Damage in Ion-Implanted Silicon. Materials Sci. Rep. 6, 275. Smith, S. D., and Taylor, W. (1962). Optical Phonon Effects in the Infra-red Spectrum of Acceptor Centres in Semiconducting Diamond. Proc. Phys. Soc., 79, 1142. Sonder, E., and Templeton, L. C. (1963). Gamma Irradiation of Silicon. 11. Levels in N-Type Float-zone Material. J . Appl. Phys. 34, 3295. Stein, H. J., Vook, F. L., Brice, D. K. Borders, J. A., and Picraux, S. T. (1970). Infrared Studies of the Crystallinity of Ion Implanted Si. Radiation Efects 6, 19. Svensson, B. G., and Lindstrom, J. L. (1992). Generation of Divacancies in Silicon by MeV Electrons: Dose Rate Dependence and Influence of Sn and P. J. Appl. Phys. 72, 5616. Tamura, M., Ando, T., and Onyu, K., (1991). MeV Ion-induced Damage in Si and its Annealing. Nucl. fnstrum. Meth. Phys. Res. B59160,572. Tiedje, T., Cebulka, J. M., Morel, D. L., and Abele, B. A. (1981). Evidence of Exponential Band Tails in Amorphous Silicon Hydride. Phys. Rev. Lett. 46, 1425. Van den Hoven, G. N., Liang, 2 . N., and Niesen, L. (1992). Evidence of Vacancies in Amorphous Silicon. Phys. Rev. Lett. 68, 3714. Watkins, G. D., and Corbett, J. W. (1965). Defects in Irradiated Silicon: Electron Paramagnetic Resonance of the Divacancy. Phys. Rev. 138, A543. Wong, H., Cheng, N. W., Chu, P. K., Liu, J., and Mayer, J. W. (1988). Proximity Gettering with Mega Electron-volt Carbon and Oxygen Implantations. Appl. Phys. Lett. 52, 1023. Yasa, Z. A., Jackson, W. B., and Amer, N. M. (1982). Photothermal Spectroscopy of Scattering Media. Appl. Optics 21, 21.

5 ION-IMPLANTEDAND ANNEALEDSILICONFILMS

177

Zammit, U., Gasparrini, F., Marinelli, M., Pizzoferrato, R.,Agostini, A., and Mercuri, F. (1991). Ion Dose Effect in Sub Gap Absorption Spectra of Defects in Ion Implanted GaAs and Si. J . Appl. Phys. 70, 7060. Zammit, U., Madhusoodanan, K. N., Scudieri, F., Mercuri, F., Wendler, E., and Wesch, W. (1994a). Optical Absorption Study of Structural Relaxation of Ion Implanted a-Si. Phys. Rev. B 49, 2163. Zammit, U., Madhusoodanan, K. N., Marinelli, M., Scudieri, F., Pizzoferrato, R., Mercuri, F., Wendler, E., and Wesch, W. (1994b). Optical Absorption Studies of Ion Implantation Damage in Si on Sapphire. Phys. Rev. B 49, 14322. Zammit, U., Marinelli, M., Pizzoferrato, R.,and Mercuri, F. (1992). Gap States Distribution of Ion Implanted Si and GaAs from Subgap Absorption measurements. Phys. Rev. B 46, 7515.

This Page Intentionally Left Blank

SEMICONDUCTORS AND SEMIMETALS, VOL. 46

CHAPTER 6

Photothermal Deep-Level Transient Spectroscopy of Impurities and Defects in Semiconductors Andreas Mandelis A rief Budiman Miguel Vargas* hOTOTHERhuL A N D

OFTOELECTRONIC DIAGNOSTICS LABORATORIES

DEPARTMENT OF MECHANICAL AND h l X W " D L ENGINEERING ONTARIO, CANADA UNlvotSlTY OF TORONTO,

1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . OF PHOTOTHERMAL RADIOMETRICDEEP-LEVEL 11. PHYSICAL FOUNDATIONS

TRANSIENT SPECTROSCOPY . .

. .

179

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

181

111. THEORY OF PHOTOTHERMAL RADIOMETRICDEEP-LEVEL TRANSIENT SPECTROSCOPY.184

FOUNDATIONS OF PHOTOTHERMAL RADIOMETRIC DEEP-LEVEL IV. INSTRUMENTAL TRANSIENT SPECTROSCOPY: THE LOCK-INRATE-WINDOW METHOD. . . . . . . V. EXPERIMENT AND DISCUSSION . . . . . . . . . . . . . . . . . . . . . . 1. Constant-Temperature Photothermal Radiometric Deep-Level Transient Spectroscopy of Silicon . . . . . . . . . . . . . . . . . . . . . . . . 2. Constant Duty-Cycle Photothermal Radiometric Deep-Level Transient Spectroscopy of Semi-Insulating- Gallium Arsenide . . . . . . . . . . . . VI. POTENTIAL FOR ION-IMPLANTATION DIAGNOSTICS AND CONCLUSIONS . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

188 190

191 198 207 209

I. Introduction Deep-level transient spectroscopy (DLTS) (Lang, 1974) has become a valuable and indispensable tool in semiconductor manufacturing and characterization, especially for the study of deep and shallow impurity and defect levels in these materials (Miller, Ramirez, and Robinson, 1975; Lang, 1979; Auret and Nel, 1988, 1991). The nature of the original DLTS as introduced by Lang (1974) was an all-electrical technique, called E-DLTS in this work. It is best applicable to junction devices and depletion-layer forming geometries: A nonequilibrium occupation of impurity or defect levels is achieved by *Onleave from Instituto Fisica Universidad Guanajuato Universidad de Guanajuato, Mexico.

179 Copyright 0 1997 by Academic Press All rights of reproduction in any form reserved. 0080-87841971625

180

A. MANDELIS, A. BUDIMANAND M. VARGAS

a voltage pulse, followed by restoration of equilibrium through a thermal emission carrier (electron or hole) transient from the levels following the cut-off of the pulse. Return to equilibrium through thermal emission is usually monitored via junction capacitance, conductance, or current decay transients. Since the emergence of E-DLTS in the 1970s, it has become desirable to develop less intrusive DLTS methodologies, primarily to preserve sample integity, eliminate the requirement for electrical probes and contacts to the semiconductor and, as a result, to increase the speed of the technique for on-line characterization and nondestructive evaluation (NDE), while exploiting novel capabilities, such as real-time impurity or defect scanning imaging at the wafer level. An optical version of the technique termed O-DLTS, or DLOS, was introduced (Martin et al., 1980; Chantre, Vincent, and Bois, 1981), which utilizes an optical source for photoexcited carrier creation and the same electrical (capacitance, current, or conductance) transient detection. By its very nature, O-DLTS is a contacting technique in which the probe-device (or probe-wafer) interface remains the same as in E-DLTS, even though the electrical voltage pulse has been replaced by optical pulse excitation., For this reason, in more recent years many laboratories around the world guided their efforts toward developing entirely noncontacting DLTS methodologies. Broadly speaking, two such techniques have been introduced and show good promise toward their on-line implementation for semiconductor manufacturing and device fabrication. The surface photovoltage (SPV) DLTS developed by Lagowski, Edelman, and Morawski (1992a) is an extension of an earlier surface photovoltage spectroscopy introduced by Lagowski, Balestra, and Gatos (1972). The laser microwave (LM) DLTS was first introduced by Fujisaki, Takano, and Ishiba (1986) and was further developed by Shimura's group (Kirino et al., 1990). Another O-DLTS technique based on detection of carrier absorption from changes in a transmitted optical beam (Dobrilla and Blakemore, 1985) proved to be too sensitive to fluctuations in the intenity of the transmitted light caused by surface and volume imperfections (Desnica et a/., 1991). Thus additional optical filters and several additional measurements are usually needed for reliable data to improve the selectivity of the technique in the (unavoidable) presence of background signals. For this reason, there appears to be little recent activity associated with this O-DLTS technique, which will not be discussed further in the present context. Regarding the former two noncontacting DLTS techniques, although they are definite improvements over the conventional DLTS methodologies, they tend to be quite restrictive in their applications scope. For instance, LM-DLTS requires for detection the presence of free-carrier concentrations within limits determined to be (Fujisaki, Takano, and Ishiba, 1986) 8.3 x 10" < N < 2.6 x l O I 3 ~ m - ~ . Furthermore, the spatial resolution of LM-DLTS is limited by the micro-

6

IMPURITIES AND

DEFECTS IN SEMICONDUCTORS

181

wave probe-beam spot size (on the order of 10 to 20 mm; Fujisaki, Takano, and Ishiba, 1986), which broadens (in a convolution sense) the probed surface area. The resulting effective area is usually too large for device structure imaging compared with the laser carrier photoinjection beam spot size (on the order of 50 to 100 pm and possibly much less: 1 to 5 pm). On the other hand, SPV-DLTS is virtually a contacting technique, owing to the requirement for extreme closeness (0.2 to 0.5 mm; Lagowski, Morawski, and Edelman, 1992b) between the wafer surface and the transparent conducting pick-up electrode required for adequate capacitive coupling. Furthermore, the surface photovoltage technique may possibly exhibit practical difficulties in performing rapid scanning imaging of the wafer surface because this technique requires perfect constancy of the wafer surface-to-electrode distance to yield meaningful relative signals. Nevertheless, Lagowski, Edelman, and Morawski (1992a) have successfully demonstrated the potential of SPV-DLTS for performing limited resolution scanning imaging on an n-type gallium arsenide (GaAs) wafer. The major improvement of the foregoing LM-DLTS and SPV-DLTS techniques over the contacting counterparts is the removal of the requirement for the presence of a junction or Schottky barrier at the probed surface, which allows use of noncontacting techniques with unprocessed substrates and at all levels of processing before the fabrication of working devices. From the point of view of materials physics, the development of noncontacting DLTS methodologies heralds the dawn of a more thorough understanding of near-surface physics in impurity and defect semiconductors because this development eliminates the (usually unquantifiable) perturbation of the instrumental probe on the electronic behavior of the semiconductor. Especially, for the main topic of this volume-ion implantation-there is a strong thrust to develop novel nondestructive techniques to characterize the energies of defect structures at, or near, the wafer surface within the implantation depth range. Motivated by these powerful considerations in favor of noncontacting DLTS techniques, in this chapter the development and the first few exploratory applications of a novel all-optical, improved methodology, called infrared photothermal radiometric deep-level transient spectroscopy (PTR-DLTS),will be described, and its potential for ion-implantation defect diagnostics discussed.

-

11. Physical Foundations of Photothermal Radiometric Deep-Level Transient Spectroscopy Infrared photothermal radiometry (PTR) has recently been applied to the analysis of electronic materials (Sheard, 1987) as an alternative to purely

182

A. MANDELIS, A. BUDIMAN AND M. VARGAS

optical infrared (IR) techniques that have been used for some time to probe electron-hole plasma effects (Spitzer and Fan, 1957; Spitzer and Whelan, 1959), carrier homogeneity (Edwards and Maker, 1962), and implantation parameters, including damage (Kachare et al., 1974; Sopori, 1985). Sheard, Somekh, and Hiller (1990) demonstrated that optically generated carriers increase the IR (blackbody) emission from the surface of an electronic solid. The measured laser-beam-modulation-frequency-scanned signal was shown to carry information on the recombination and diffusion of the free-carrier plasma (Sheard, Somekh, and Hiller, 1990; Little, Crean, and Sheard, 1990). According to a statement of Kirchhoffs law reflecting conservation of energy, at thermodynamic equilibrium the rate of emission of blackbody radiation from the surface and throughout the bulk of a material is exactly equal to the rate of absorption of radiation incident on the material per wavelength interval.

Therefore, the IR emission spectrum for a de-excitation process in a semiconductor can be obtained directly from its (usually better known) absorption spectrum. Although the foregoing statement is strictly valid under conditions of thermodynamic equilibrium, in practice it has been found to be of broader validity (Baltes, 1976) and applicability to several nonequilibrium processes in semiconductors (Ulmer and Frankl, 1968; Cho and Davis, 1989). In PTR detection it is rather the reverse application of Kirchhoffs law that is exploited: Optical absorption of a laser beam in the ultraviolet (UV), visible or near-IR (super-bandgap) spectral ranges results in electronic excitation, primary ultrafast energy transfer, and intraband thermalization processes in the excited state (carrier-carrier, carrier-lattice, carrier-defect-impurity collisions), followed by much longer interband recombination kinetics (e.g., band-to-band and band-to-defect). It is precisely this latter mechanism that PTR probes: The blackbody photon flux emission in the bulk and on the surface of a semiconductor due to energy release via recombination leading to elementary carrier de-excitation, integrated over the emission depth of the material, is equivalent t o surface and bulk absorption of the same infrared photon flux integrated along the thickness of the same material, in agreement with Kirchhoffs law. This equivalence has been further assumed to be valid under dynamic recombination conditions (Cho and Davis, 1989). Under general nonthermal equilibrium conditions between a solid and its environment, it has been established that the emissivity is exclusively due to spontaneous emission of the freely radiating body and is consequently completely determined by the statistical thermodynamics of the material states, whereas the temperature character-

6 IMPURITIESAND DEFECTS IN SEMICONDUCTORS

183

king the ambient radiation field is irrelevant to the validity or invalidity of Kirchhoffs law (Baltes, 1976). Nevertheless, for a freely radiating body, Kirchhoffs law strictly holds only when the material states of the solid obey the thermal equilibrium condition, that is, the thermodynamic temperature of the atomic (harmonic) oscillator ensembles that make up the material is well defined. Under dynamic recombination conditions, Kirchhoffs law is valid to a very good approximation provided that the relaxation time of the recombining carriers is long compared with the emission time of a vibrational quantum (on the order of ps), as the system continuously adjusts itself to the loss of energy via spontaneous emission. This condition is definitely valid for the time ranges involved in current state-of-the-art PTR-DLTS (1 psec or over), and ensures that at each instant of the transient event the material is effectively at a quasi-equilibrium state represented by the instantaneous transient temperature, so that Kirchhoffs law applies for all time intervals At between the fastest possible successive measurements during an experiment. Based on the foregoing quasi-equilibrium description of the IR radiation field, the interception of back-propagating blackbody radiation power arriving at the surface of a semiconductor yields information entirely equivalent to absorption of forward-propagating IR radiation power via carrier generation (e.g., band-to-band, defect-to-band) and may be used to study this process. The excess energy due to ultrafast carrier thermalization or direct lattice absorption and the subsequent increase in lattice temperature is also detected as an additional increase in blackbody emission, in agreement with Planck's radiation law. Given that the generation of this thermal energy is usually instantaneous on the time scale of the carrier recombination lifetime, this radiation source is, in principle, always superposed on the photocarrier (plasma) relaxation-emission source, but with a generally different time constant governed by conduction heat transfer in the lattice and into the ambient. The evolution of the superposition radiation emitted by a semiconductor and collected within a fixed solid angle in the surrounding medium is focused on, and monitored by, an appropriate PTR detector. The main advantage of PTR detection of optoelectronic phenomena over conventional optical techniques is that the former does not require sample preparation with a resolution approaching the optical diffraction limit. The advantage of PTR over other photothermal techniques (Mandelis, 1987), lies in its ability to distinguish more easily between the foregoing thermal and plasma transport processes, with clear domination of the signal by the latter processes in the case of high-quality electronic materials, due to the fact that free-carrier emission prevails over thermal emission of IR radiation (Sheard, 1987). Our frequency-domain studies have shown that, as a general rule, the emission of IR radiation from the optically injected free carriers is

184

A. MANDELIS, A. BUDIMANAND M. VARGAS

dominant for intrinsic and moderately doped material, thus allowing a less ambiguous interpretation of the electronic properties than do conventional photothermal techniques, such as laser thermoreflectance.The thermal-wave emission increases progressively with doping density and eventually can dominate the signal for strongly extrinsic semiconductors. This characteristic advantage of PTR with intrinsic and nearly intrinsic materials has been strengthened through time-gated and narrow band filtered transient excitation and detection instrumentation (Mandelis, Bleiss, and Shimura, 1993). On the other hand, laser thermoreflectance, for instance, exhibits strong overlap between electron-hole plasma and thermal-wave contributions of similar strengths to the signal up to the MHz range, and therefore sophisticated modeling and experimental procedures are required for realistic deconvolution of the experimental data (Mandelis and Wagner, 1996). Historically, the first time-resolved application of PTR to a semiconductor silicon (Si) wafer was reported by Cho and Davis (1989). In common with its other photothermal counterparts in the pulsed mode, the timeresolved method exhibits severe overlap of the photoexcited free-carrier density and lattice thermal-wave effects in processed electronic materials. In principle, nevertheless, this mode is preferable to the frequency-domain detection (Sheard, Somekh, and Hiller, 1990), as electronic relaxation processes can be isolated and interpreted in terms of simple system timedelay constants, instead of time-multiplexed frequency responses that require Fourier transformation and deconvolution. In reality, however, the temporal separation and isolation of different, partially overlapping processes may be difficult, tedious, and even ambiguous in the worst case. The ability to separate out individual (opto-) electronic relaxation processes is central to any DLTS scheme probing the excitation and decay dynamics of particular impurity or defect centers and leading to the determination of an activation energy jingerprint for the correct diagnostics of the defect or impurity in question (Lang, 1979). This important feature was introduced into the PTR methodology by coupling time-gated semiconductor photoexcitation and photothermal detection with a lock-in amplifier rate-window (RW) scheme (Mandelis and Chen, 1992; Mandelis et al., 1995).

111. Theory of Photothermal Radiometric DeepLevel Transient Spectroscopy The configuration of a one-dimensional mathematical model used in PTR-DLTS is shown in Fig. 1. The semiconductor sample is assumed to be semi-infiniteand is irradiated with a repetitive square laser pulse of duration

6 IMPURITIES AND DEFECTS IN SEMICONDUCTORS

185

FIG. 1. Schematic configuration of the photothemal radiometric deep-level transient spectroscopy (PTR-DLTS)signal model involving a square laser pulse of duration T~ and repetition period To and a semi-infinite photoexcited semiconductor.

t pand period To. The photon energy hv of the laser beam is assumed to be greater than the semiconductor bandgap E,, and the excess minority density is small compared with the majority carrier density (low-injection limit). The photoexcited carrier density AN(z, t ) is given by solving the carrier continuity equation (McKelvey, 1966):

a2AN(z,t ) -~ AN(z, t ) z

az2

t) + G(z, t ) = aAN(z, at

(1)

Here D (m2/sec) is the ambipolar diffusion coefficient, z(s) is the carrier bulk lifetime, and G(z, t ) (m-3 sec-') is the carrier generation rate, which is given by t

c tp

t

>tp

G(z, t ) = The square laser pulse of intensity lo(W/m2) illuminates the sample surface and is absorbed throughout its bulk with an absorption coefficient tl (m- ') and quantum efficiency r]. The most usual initial and boundary conditions are, respectively, AN(z, 0) = 0

(3)

186

A. MANDELIS,A. BUDIMANAND M. VARGAS

and aAN(0, t )

aZ

= sAN(0, t )

(4)

where s is the carrier surface recombination velocity in (mjsec). The former (initial) condition indicates that there is no significant excess free-carrier concentration before the onset of the laser pulse. The initial-value and boundary-value problem in Eqs. (1)-(4) has been solved in the Laplace domain and inverted (Chen et al., 1993) to yield the free photoexcited carrier dependence on the depth coordinate z and on time. When evaluated at the surface ( z = 0), the excess carrier density becomes qlo AN(0, t ) = -

hv

1

&-A

In Eq. ( 5 ) the following definitions have been made:

is an optoelectronic diffusion time constant, a measure of the time required for a carrier to diffuse to a depth in the semiconductor equal to the optical absorption length l/a. Similarly,

D =-

ts

SZ

(7)

is a time constant due to recombination at surface defect states, which is dependent on the surface recombination velocity and the carrier diffusion coefficient. W(z) = exp(z2)erfc(z)

(8)

is a function encountered in time-domain diffusion-type problems (Mandelis

6

IMPURITIES AND

DEFECTS IN SEMICONDUCTORS

187

and Chen, 1992), where erfc(z) is the complementary error function defined by

For opaque semiconductors in which z, << z and z, << T simplified to

~ Eq. ,

(5) can be

In the special case of very low surface recombination velocity (s -+ 0, i.e., >> T ) , Eq. (10) can be simplified further:

T,

The I R radiometric signal is proportional to an integral over the thickness of the semiconductor of the photogenerated excess carrier density AN(z, t). This integral gives the integrated free-carrier absorbance (Sheard, 1987). The result of the integration can be shown to be (Mandelis, 1995) proportional to AN(0, t):

where C is an instrumental constant independent of any photoexcited carrier characteristic time constants, and [ is a semiconductor constant related to the static optical parameters of the free-carrier plasma absorption (Smith, 1978):

Here q is the elementary charge, ilis the wavelength of the carrier-exciting radiation, c is the speed of propagation of light in the semiconductor medium of refractive index n, c0 is the permittivity of free space, m is the free electron mass, and p is the carrier mobility. It is evident from Eqs. (11) to (13) that the PTR signal is a purely exponential function dependent on the

188

A. MANDELIS, A. BUDIMANAND M. VARGAS

bulk lifetime z only, when the surface recombination velocity is very low (high-quality surface). An additional assumption built into Eq. (1 1) is that the bulk lifetime is small compared with the electronic diffusion transit time across the thickness L of the semiconductor (Chen et al., 1993): 4L2 D

z << -

This condition has been found to be valid (z << 290 psec) for normal thicknesses of Si wafers (- 500 pm) with lifetime z < 20 psec and on taking the diffusion coefficient 34.4 cm2/sec for electrons, or 12.3 cm2/sec for holes (Wolf, 1969). This lifetime range is further compatible with experimental values obtained in the course of the measurements described below. In the case of GaAs, however, it may not be valid for electronic diffusion, due to the very high D in this material. For this situation, the full time-domain theoretic treatment of the semiconductor with finite thickness developed by Luke and Cheng (1987) must be implemented. Another candidate for the full theoretic treatment would be ultrahigh-quality Si, such as magnetic CZ wafers, which often have z on the order of a few hundred psec.

IV. Instrumental Foundations of Photothermal Radiometric DeepLevel Transient Spectroscopy: The Lock-in Rate-Window Method Among the instrumental boxcar rate-window (RW) techniques (Lang, 1979), the lock-in amplifier (L1A)-based method offers the highest signal-tonoise ratio (SNR) (Chen and Mandelis, 1992) and overall the second highest SNR (Mandelis, 1994) after the optimum (matched) filter correlation method introduced by Miller, Ramirez, and Robinson (1975). Even though the latter exhibits maximum sensitivity to signal decay lifetime, nevertheless, the LIA-based method is easiest to use and modern high-dynamic-range lock-in instruments allow for extremely broad tunability from microseconds to several seconds, thus being capable of capturing very wide ranges of (opto)-electronicprocesses in semiconductors. The well-documented serious limitation of the LIA as an instrument extremely sensitive to bridge or loading electrical voltage transients in E-DLTS applications (Miller, Ramirez, and Robinson, 1975; Miller and Patterson, 1977), is not a problem in PTR-DLTS, since the excitation is purely optical and no electrical pulses are involved in this technique to overload the entrance circuitry of the LIA. In terms of photothermal applications, the extremely narrow band detection capability of the LIA over conventional dual-gate boxcar averagers wields

6 IMPURITIES AND DEFECTS IN SEMICONDUCTORS

189

SNR advantages, because all thermally generated signals suffer from inherently low dynamic range. Therefore, SNR enhancing schemes are of far greater importance with photothermal signals than with purely optical (or electrical) ones. The RW-LIA method combines the superior SNR of a tuned electronic filter, familiar with frequency-domain PTR detection (Sheard, Somekh, and Hiller, 1990), with the simple and straightforward interpretation of the time-domain PTR signal in terms of electron-hole or defect-impurity recombination time delays (Cho and Davis, 1989). In the RW-LIA method, the in-phase (IP) or quadrature (Q) lock-in signal output is measured. This is proportional to the fundamental Fourier component b, or a,, respectively, of the PTR signal S,,(t;z), Eq. (12), over the pulse repetition period To (Day et al., 1979; Chen and Mandelis, 1992): Si,(t;z) sin(w,t) dt

(15)

where w, = (2n/T0). A similar expression is valid for a , , with cos(w,t) replacing the sin(o,t) in the integrand of Eq. (15). Equation (15) has an extremum if the measurement is performed at a constant temperature and the pulse repetition period or the pulse duration is varied. The I P component of the LIA output is obtained by weighing the fundamental Fourier coefficient b, by the square-wave LIA reference function eR(t) of duration To, phase-tuned so as to align the origin of all times at t = 0. This corresponds to setting the positive (or negative) edge of the reference square wave in-phase with the rising edge of the optical pulse. The square wave reference function is valid in the limit when the RC-time constant of the LIA low-pass filter is very long compared with the longest lifetime T of the semiconductor de-excitation dynamics, as well as to To (Mandelis, 1994). The result of this operation gives the demodulated IP LIA signal: b , sin(o,t)eR(t)dt 2

= - b,b, 7l

(16)

To)

As the repetition period or the pulse duration is scanned, the condition for the existence of the RW extremum becomes

A. MANDELIS, A. BUDIMANAND M. VARGAS

190

V. Experiment and Discussion The experimental setup for PTR-DLTS is shown in Fig. 2. A time-gated laser beam obtained by modulating a cw argon ion (or an IR) laser using an acousto-optic modulator (AOM) is slightly focused on the sample to the size of approximately 1mm diameter. This beam size was used to keep the data analysis within the one-dimensional regime of the theoretical treatment of Part IV, and it does not reflect the resolution limits of PTR-DLTS. The actual resolution of the technique is set either by the wavelength of the IR emission (on the order of a few microns), or, more likely, by the size of the IR sensor element. The IR blackbody radiation emitted from the semiconductor surface is collected by two off-axis paraboloidal mirrors and is detected using a liquid nitrogen-cooled photoconductive mercury-cadmium-telluride (MCT) detector, the size of which was 50 x 50pm2 in the

,I + :1

, ,

computer interfaces

'

detector

,

I

I

knife edge

I I

mirror

indicator /

cryostat

FIG.2. Experimental system of photothermal radiometric deep-level transient spectroscopy (PTR-DLTS). Ar' argon-ion; AOM, acousto-optic modulator; HgCdTe, mercury-cadmiumtellurium; pre amp., preamplifier.

6

IMPURITIES AND

DEFECTS IN SEMICONDUCTORS

191

setup of Fig. 2. The size of this active element practically limits the spatial resolution of the technique to a maximum of 25 x 25pm2, with today's MCT technologies. The typical frequency bandwidth of the detector-preamplifier system is dc - 1.5 MHz, capable of capturing PTR decay events on the order of -0.7psec and longer, that is, within the recombination lifetime range of major semiconductors, such as Si and GaAs. Faster decays may also be monitored using photovoltaic MCT detectors. The baseline drift of the detector associated with the dc-level is minimized by choosing low-frequency preamplifier cut-off filters in the 0.1 Hz range, which exclude the dc level. The transient signal from the MCT preamplifier is directed to a LIA or to a boxcar averager. For completeness, the setup has the capability of scanning both the sample temperature and the optical pulse duty cycle. The PTR-DLTS apparatus further allows the acquisition of optical conductance and capacitance O-DLTS spectra with the use of a metallic needle probe, a voltage-pulse generator (the same as the one driving the AOM) and a Boonton Electronics Corp. (Randolph, N.J.), Model 7200 capactance meter, the output of which is connected to a second LIA interfaced with the computer. There are two distinct DLTS modes of operation with the instrumentation of Fig. 2: (i) constant-temperature boxcar or LIA duty-cycle scan and (ii) constant-duty-cycle temperature scan. 1. CONSTANT-TEMPERATURE PHOTOTHERMAL RADIOMETRIC

DEEP-LEVEL TRANSIENT SPECTROSCOPY OF SILICON In this mode a set of n- and p-type Si samples were studied. Unless otherwise noted, the isothermal scans are implied to have been performed at ambient temperature. The samples had been intentionally contaminated with deep-level forming impurities (chromium, iron, and gold) during Czochralski crystal growth. They were subsequently RCA-cleaned and oxidized to minimize the effects of surface recombination, without further annealing (e.g., in a hydrogen atmosphere). The surface recombination velocity of these wafers is nominally 4197 cm/sec (Pierret, 1987). The rest of the properties of the Si wafers studied are summarized in Table I. In Table I, the wafer is indicated by N (for n-type) or P (for p-type) in the first column. The optical absorption coefficient of intrinsic Si is given by Dash and Newman (1955). For the argon ion laser radiation (L = 514 nm; 2.41 eV), the absorption coefficient is approximately lo4 cm-l, or the optical absorption length is 1 pm,which is much less than the sample thickness of 500 pm. Therefore, Si can be considered optically semi-infinite and totally ~ 3.33 x lO-'Osec, much less than the earliest experimental opaque ( T w observation time considered in PTR-DLTS). Photothermal radiometric transient signals obtained from the impurity-doped Si wafers exhibited

-

192

A. MANDELIS. A. BUDIMANAND M. VARGAS TABLE I PROPERTIES OF SILICON WAFERS USED IN THIS STUDY

Sample

N4

N5 N11 P4

P5

P11

Metal impurity

Resistivity Cncml

Oxygen concentration (PPW

Concentration of metal (atoms/cm3) x

Chromium Iron Gold Chromium Iron Gold

5-13 6 - 16 4-13 11 15 9 - 13 10 15

11.8 14.5 17.2 11.6 15.4 14.2

17 47 11 16 61 31

-

Cubic zirconium (100) 65-mm diameter.

purely exponential decay characteristics. This behavior is consistent with the theoretical result, Eq. (1l), which is valid for materials with high-quality (low recombination velocity) surfaces. A numerical dual-gate boxcar RW was set up in the computer and the transients were scanned after fixing the gate-delay times so that t , / t , = 2. Figure 3 shows the boxcar RW signal of a chromium-doped n-type wafer (N4 in Table I). Since only the pulse decay transient was used in the boxcar RW scan, the pulse duration z p does not

OJ 0.00

0.01

0.02

0.03 *I

0.04

0.05

0.

[msl

FIG.3. Boxcar rate-window signal of chromium-doped n-type silicon sample (N4). Total number of averaged transients is 2500. The dotted line is the experimental result and the solid line is the smoothed curve of the experimental data. tprpulse duration.

193

6 IMPURITIESAND DEFECTS IN SEMICONDUCTORS TABLE 11 EXPERIMENTAL RESULTS OF CARRIER BULKLIFETIMES(IN p e c ) MEASUREDUSING THE PHOTOTHERMAL RADIOMETRIC RATE-WINDOW METHOD -

N4

NS

Sample: N11

lob 30' SOb

2 1.6" 8.2 14.4 16.8

-

18.W

11.7" 9.7 12.9 11.7 11.W

5.1" 5.3 4.0 1.5 4.5'

TP >

-

P4

P5

P11

21.6" 22.9 21.1 22.2 8.2'

5.5" 7.0 6.1 6.3 0.6'

2.7" 3.1 2.4 0.9 1.1'

"Resultsmeasured by use of boxcar rate-window method. The values shown are best fits (in the least squares sense) to each experimental curve. bResultsmeasured by use of the lock-in amplifier rate-window method. The values shown are best fits (in the least squares sense) to each experimental curve. 'Measurements with laser microwave deeplevel transient spectroscopy technique.

affect the maximum in the RW signal. The maximum is related to the time constant of a purely exponential decay as follows (Lang, 1974):

Using the (t where the maximum RW signal occurs (= 15 psec), the carrier bulk lifetime z = 21.6psec was obtained from Eq. (18). Table I1 summarizes the experimental results using this method and further provides results from the LM-DLTS technique for comparison (Chen et al., 1993). The RW-LIA method was also used with the Si samples of Table I. Based on the fact that the PTR signal decays were exponential, the To-scanned LIA RW signals were fitted to Eq. (16). With Si,(t) given according to Eqs. (11) and (12), b,(z, To)was obtained from Eq. (15). The theoretical result is

where

6 = tan- l(w,z)

194

A.

MANDELIS, A.

BUDIMANAND M. VARGAS

6

FIG.4. Lock-in amplifier rate-window (LIA RW) signals of the iron-doped n-type silicon sample (N5) with four different laser pulse durations. The solid lines are theoretical simulations using Eq. (19) with T = 11 psec. LIA filter time constant is 1 sec.

The IP-LIA rate-window data were fitted to Eq. (19) for various pulse durations and scanned pulse repetition time. Figure 4 shows the resulting fits to the data from the iron-doped Si wafers (N5). The curves were normalized to the peak of curve (d), zp = 200psec. The only fitting parameter was the bulk lifetime of the semiconductor, which turned out to be approximately z = 11.0 f 1.1 psec for this sample (Table 11). In fitting the rest of the curves of Fig. 4,the best-fit value to curve (d) of z = 11 psec was used and no other adjustments were introduced in the theoretical simulations, so as to show the degree of tolerance of the PW-LIA method to small variations in the value of the lifetime. The IP-LIA RW signals of chromiumiron-, and gold-doped p-Si wafers are shown in Fig. 5. The superior SNR of the LIA rate-window determination of the maximum in the fundamental Fourier component of the PTR transient has a z resolution of 1 psec, limited only by the dynamic range of the data acquisition system. The measured carrier bulk lifetimes are summarized in Table 11. The results are consistent with each other and with those from the boxcar measurements, with some notable exceptions. The results from wafers N4, N5, and N11 also agree with those obtained from the LM-DLTS method. On further investigation of the transient signals from the wafers, which did not exhibit agreement between the PTR-DLTS and the LM-DLTS techniques, it was found that the origin of the discrepancies lay mainly in the nonexponential nature of the decay of

6 IMPURITIESAND DEFECTS IN SEMICONDUCTORS

8

0.8

iij

3 0.6

x

195

0

-

isa,

0.4-

dc

0.2-

1

0.c)

I

cl

-0.2 1

50

Au

p -Si I

I 100

200

150

250

300

To [PSI FIG. 5. Lock-in amplifier rate-window (LIA-RW) signals of chromium ((3)-, iron (Fe)-, and gold (A@-doped p-silicon (Si) wafers with pulse duration tp = 50psec. The solid lines are theoretical simulations using Eq. (19).

the PTR signals when the pulse duration was at least one order of magnitude longer than the bulk lifetime of (mostly gold-doped) wafers. The origin of this nonlinearity is unclear at this time; however, it is most likely related to neglecting the onset of the thermal-wave contribution to the theoretical model in Part 111. In any case, the time-gating feature of PTR-DLTS via selective ratewindowing can be used to completely separate out the free plasma contribution (Mandelis, Bleiss, and Shimura, 1993). As an example, Fig. 6 is the PTR transient from an n-type chromium-doped Si sample with preoxidized and etched, high-quality surface. This sample exhibits strongly overlapping carrier recombination and thermal conduction contributions. Two experiments were performed at each wavelength (514nm and 1.06pm), one with z p = 30 psec and another with op = 1 msec. The temperature increases were estimated to be less than 1 K in both cases. Figure 7 shows the RW-LIA signal (IP) as a function of the laser time-gated square pulse repetition period To,resulting from the short-duration pulse. The theoretical fits were made using Eq. (19), which assumes purely exponential plasma decay. This type of decay was verified for all high-quality surfaces of this Si batch, consistent with the theory. Two different lifetimes were obtained, one for each laser wavelength, as shown in Fig. 7. The difference can be explained by the widely different optical penetration depths of the excitation pulses at

A. MANDELIS, A. BUDIMAN AND M. VARGAS

196 in

m

5l

iij cI 0.6-

Cr in n-Si

.-

X=514nm ~,=30ms

z i L 0.4,

Time [ms]

-

0.8 -

iij

0.6-

B

cI

0

Cr in n-Si h = 1.06 pm ~~=30m

.s 13

8

0.4-

0.2 -

0.0 S

-

i

-

i

o

1

2

3

4

s

6

Time [ms] FIG.6. Photothermal radiometric (PTR) transient response of a chromium (Cr)-doped n-type silicon (Si) wafer (1.7 x lO”Cr atoms/cm2; 5 to 13Llcm) to (a) an Ar ion laser acousto-optically time-gated pulse; and (b) a similar neodymium:yttrium-aluminum-garnet (Nd:YAG) laser pulse. Pulse duration T~ = 30 msec.

514 nm and 1.06pm. The former pulse probes a region very close to the Si surface, in which near-surface defects can provide an additional free-carrier recombination channel, thus shortening the effective lifetime (Ling and Ajmera, 1991). However, the deeply penetrating 1.06 pm pulse is expected to give a better measurement of the true value of the bulk recombination

6 IMPURITIESAND DEFECTS IN SEMICONDUCTORS

197

1.o

-m

6

0.8

iij

g

2

0.6

5il

Cr in n - S i

c

2 P

0.4

c 4

3%

0.2

0.0

1.o

-6 m

0.8

;ij

g

5

0.6

Cr in n -Si

0

2 0.4 CI

.Fi;

1

0.2

x= 1.06p.m z = 24.5 p

0.0

FIG.7. Lock-in amplifier rate window (LIA-RW) photothermal radiometric (PTR) response to a repetitive square laser pulse of duration T~ = 30 psec. Squares indicate data; lines indicate least squares fits of theory, Eq. (19) to data. (a) The 514nm photon excitation of chromium (Cr) doped n-type silicon (Si) wafer is T = 16.0 0.8 psec. (b) The 1.06 pm photon excitation is z = 24.5 5 0.9 psec.

lifetime. In this case near-surface recombination is much less significant in its contribution to the effective lifetime, which is longer and characteristic of bulk processes. The technique of RW-scanned PTR-DLTS is very efficient in decoupling easily and totally the thermal from plasma evolution effects, a task which is,

198

A. MANDELIS, A. BUDIMAN AND M. VARGAS

at best, quite cumbersome with any other combinations of photothermal and photoacoustic measurement methodologies (Mandelis, 1987). In the foregoing examples, when the laser pulse duration was increased to 1 ms, the PTR responses from both the argon and neodymium: yttrium-aluminumgarnet (Nd:YAG) lasers exhibited maxima of the I P signals at approximately 2msec from the onset of the pulse. This is in good agreement with domination of the IR PTR signal by a thermal transient, which in the 500-pm-thick freestanding Si wafer requires a thermal round-trip time (Chen and Mandelis, 1992)

t,,

- L271

-

4%

1.96msec

where 01, is the thermal diffusivity of Si (equal to 0.94cm2/sec; Mandelis, 1987). Similar effects also have been observed with more recent studies using GaAs wafers. It already has been mentioned that the completely decoupled free-carrier-plasma and thermal transients with effective RW pulse timegating cannot be obtained easily with frequency-scanned PTR detection because of the time-multiplexed nature of this methodology. Furthermore, they cannot be obtained with pulsed laser excitation either, due to poor (that is, nonideal square) pulse line shapes and limitations in the extent of variability of the repetition rate, pulse duration, or both, of today’s laser technologies.

PHOTOTHERMAL RADIOMETRIC 2. CONSTANT DUTY-CYCLE DEEP-LEVEL TRANSIENT SPECTROSCOPY OF SEMI-INSULATING-GALLIUM ARSENIDE In this PTR-DLTS mode the pulse-duration to repetition-time ratio (duty cycle) is fixed, whereas the sample temperature is scanned over a range that is known to correspond to one, or more, thermal energy quanta kT- A E j = IE, - E T j I , where E , denotes the energy level at the relevant band edge of the semiconductor, and E T j represents a bandgap trap (impurity or defect) state j . This PTR-DLTS mode is similar to conventional types of DLTS in methodology (Lang, 1974). It has the advantage of trap-level selectivity over its RW-scanned counterpart at fixed temperature, through its ability to measure the trap activation energy A E j . Three samples were examined: a chromium-compensated semi-insulating (SI) sample; an n-GaAs chromium-compensated tellurium-doped (8 x 10’’ cmU3)wafer; and a gold-coated GaAs wafer, used as a purely thermal-wave

6 IMPURITIES AND DEFECTS IN SEMICONDUCTORS

199

PTR signal waveform generator for comparison purposes. Temperature scans were performed ranging from 310 to 400K, including the 0-DLTS peak at approximately 375K due to the chromium level in either por n-type GaAs (Martin et al., 1980). Photothermal radiometric RW scans of the GaAs:Cr were obtained, clearly indicating an optimal duty cycle, z,/T,, ranging from 40% to 60%. The PTR transient waveform from the GaAs:Cr wafer using the argon ion laser with pulse duration 25psec is shown in Fig. 8. Two regions with different slopes can be seen clearly for both the rising and falling parts of the transient. A combined electronicthermal-wave model was constructed based on the solution to Eq. (1) and its thermal-wave counterpart in a solid of finite thickness (Leung and Tam, 1984). The theoretical superposition-field best-fit to the data of Fig. 8 is shown in Fig. 9. No other combination of two purely electronic or purely thermal-wave superpositions or one of each could yield the characteristic "break" in the decay transient shown in Fig. 8, and at that particular instant in time after the end of the pulse. In the fit, the constants 7 and z, were used as adjustable parameters. The best fit was obtained for the values 3.10 and 1.0psec, respectively. This pair of values is unique, because the steep decay profile following the pulse cut-off fixes the value of zs; then the temporal position of the "knee" below the peak fixes the value of z, while the late slow decay slope depends on the value of the thermal diffusivity of the sample. The calculated value for z is

t [PSI FIG. 8. Photothermal radiometric (PTR) transient of GaAs:Cr sample with argon ion laser excitation. Optical pulse tp = 25 psec and To = 50 psec. Temperature is 310 K.

200

A.

MANDELIS,

A. BUDIMANAND M. VARGAS

18.5

ca W

3

17.5

8

iij

f

16.5

15.5

FIG. 9. Theoretical simulation of the GaAs:Cr transient of Fig. 8 using superposition field of Eq. (1) and a thermal transient (Leung, W. P., and Tam, A. C. (1984). Techniques of Flash Radiometry. J . Appl. Phys. 56, 153-161). The gallium arsenide (GaAs) thermal diffusivity is 0.24cm2/sec. The thickness is 400 pm. The visible optical absorption coefficient is a = 1 x IO'cm-'. The infrared absorption coefficient is aIR= 2.4cm-I. The best-fit electronic parameter values are 7 = 3.0 psec and 5, = 1.0 psec. PRT, photothermal radiometry.

25

20 n

> E

0

15

C

cn

6

10

5

0

FIG. 10. Photothermal radiometric (PTR) transient of GaAs:Te sample with argon ion laser excitation. = 25 psec and To = 50 psec.

6 IMPURITIESAND DEFECTS IN SEMICONDUCTORS

201

in excellent agreement with the theoretical carrier lifetime range of 1 to 3 psec calculated for semi-insulating GaAs: Cr (Papastamatiou and Papaioannou, 1990). The PTR transient from the GaAs:Te sample is shown in Fig. 10 for comparison. This sample exhibits the “knee” between the early and late portions of the waveform much more abruptly than does the chromium-doped wafer, with an early slope indicative of a much shorter lifetime than the GaAs:Cr response. It was not possible to analyze the early decay kinetics using RW-LIA detection, owing to the low-frequency cut-off of the LIA ( 100 kHz). Therefore, only the long decay was accessible to the LIA-interfaced instrumentation. A theoretical fit of Eq. (10) to the early portion of the PTR transient from the GaAs:Te sample yielded the best-fit pair of values (z = 17.2 psec, z, = 0.12 psec) (Fig. 11). The clear departure of the late portion of the curve from the purely free-carrier decay dynamics of Eq. (10) is prominent in Fig. 11. Based on the fact that GaAs:Te is not known to produce deep electronic levels, the evolution of the long signal decay was traced to purely thermal response of the lattice by comparing the transient to, and finding it very similar with, thermal evolution waveforms from the gold-coated GaAs wafer as well as a piece of metal in the T, = 2.5 msec range. Our experience with the use of purely electronic responses to fit experimental transients, such as Figs. 8 and 10, shows that the presence of a superposed thermal transient greatly overestimates the value of z, whereas the value of z, remains within a factor of two. Therefore,

-

20

0

FIG. 11. Least-squares theoretical fit of Eq. (10) to the early portion of the data in Fig. 10. 17.2 0.4psec and ts = 0.12 0.03 psec. PRT, photothermal radiometry.

t=

A. MANDELIS, A. BUDIMANAND M. VARGAS

202 4.01,

I

,

1

,

,

J

4.0

rn rn

5

3.3

300

3:

T

[KI

FIG. 12. Photothermal radiometric deep-level transient spectrum (PTR-DLTS) of (a) chromium-compensated semi-insulating gallium arsenide (GaAs) and (b) of chromium-compensated and tellurium-doped n-GaAs. t p= 15 psec and To = 30 psec. a.u., arbitrary units.

the foregoing estimate of the z value for GaAs:Te must be treated with extreme caution. The early-time response of the GaAs:Te wafer is undoubtedly of electronic nature and related to surface recombination processes. The transient behavior of this sample was subsequently used to study how the purely thermal (lattice) response affects the characteristics of PTR-DLTS at later (msec) times. The PTR-DLTS spectra of the GaAs:Cr and the GaAs:Te samples were obtained under identical experimental conditions, with z p = 15 psec and repetition period 30psec (Fig. 12). In GaAs:Cr a peak consistent with the chromium level in GaAs appeared at approximately 370 K (Fig. 12(a)), the exact position of the peak depending on the actual To chosen; significant peak shifts to lower temperatures were observed when the pulse repetition period was varied between 30 and 90psec (Fig. 13). The position of the rather flat GaAs:Te peak is at approximately 320K (Fig. 12(b)), the peak separation between the two samples, attesting to the high spectral resolution of PTR-DLTS. As discussed above, the DLTS spectrum from the GaAs:Te sample, excluding the inaccessible earliest experimental time range, is expected to be purely of thermal origin. Indeed the GaAs:Te peak shifted in the opposite direction from that of the GaAs:Cr sample with varying To.As a consequence of the foregoing, and other similar, observations, the shift of the PTR-DLTS peaks to lower temperatures with increasing To has been

6 IMPURITIES AND DEFECTS IN SEMICONDUCTORS

203

n

j 1.00

0

Y

-

.-UcJ3

0.95

p 0.90 1 n I L11

0.85

0.80

FIG. 13. Photothermal radiometric deep-level transient spectra (PTR-DLTS) of GaAs:Cr sample with argon ion laser excitation. Optical pulse repetition period varied between 30 and 90 psec, as shown, with 50% duty cycle. am, arbitrary units.

associated with purely free-carrier-plasma phenomena, whereas the shift of the peaks to higher temperatures is taken to be purely thermal in nature, or due to a superposition between electronic and thermal transport phenomena. (See also the discussion after Eq. (22)). In Fig. 12, the GaAs:Cr peak is broader than are those obtained by Martin et al. (1980) using capacitance 0-DLTS, which is partially due to the fact that a LIA, rather than a dual-gate boxcar integrator, was used for these measurements (Chen and Mandelis, 1992). Fujisaki, Takano, and Ishiba (1986) have obtained a DLTS peak from liquid-encapsulated Czochralski (LEC) semi-insulating GaAs using an all-optical microwave detection scheme akin to PTR-DLTS. Their line shape is quite broad and similar to Fig. 12, perhaps a feature of the optical plasma probing mechanism. It is important to emphasize that, even though the apparent broadening of the PTR-DLTS peaks may potentially result in compromised resolution of neighboring features in some samples, nevertheless the ability of the PTR-DLTS technique to generate DLTS peaks in samples that cannot yield such data through other techniques (e.g., very high resistivity SI-GaAs) and the noncontacting remote detection are major advantages: No DLTS peak could be found within the To < 100 psec range (which characterizes the spectra of Fig. 12) in the n-GaAs wafer coated with an approximately 1000-A-thick gold layer. Further investigation showed that there is a peak for this sample in the 1 msec range, which

204

A, MANDELIS, A. BUDIMANAND M. VAUGAS

FIG.14. Photothermal radiometric deep-level transient spectra (PTR-DLTS) of GaAs:Cr sample with argon ion laser excitation. Optical pulse period varied between 20 and 60 msec as shown with 50% duty cycle. Note that the direction of peak-shift with To is opposite that of Fig. 13. am, arbitrary units.

is purely due to thermal conduction in the gold thin film, following optical absorption and the subsequent nonradiative de-excitation (Mandelis, Bleiss, and Shimura, 1993), similar to observations with Si wafers described in $1 of Part V. Figure 14 shows the PTR-DLTS peak shift of the GaAs:Cr sample of Fig. 13 to higher temperatures with increasing To in the msec time range, where the response is consistent with purely thermal transport evolution. An additional major advantage of PTR-DLTS over DLTS methods requiring electrical contacts, such as 0-DLTS, is in the strength of the signal: the latter methods exhibit a much lower signal response level from intrinsic and other high-resistivity materials, such as SI-GaAs. Neither a capacitance nor a conductance 0-DLTS signal could be obtained for the GaAs:Cr wafer for To < 100 psec. Only for To 2 500 p e c were we able to obtain high enough SNR in the conductance (but not in the capacitance) channel, so as to record an 0-DLTS spectrum, as shown in Fig. 15, in agreement with earlier reports (Martin et al., 1980). The signal level was more than 20 times lower than the PTR-DLTS. Nevertheless, the peak is well-resolved in the 370 K range, in good positional agreement with the spectrum (a) of Fig. 12. Generally, it was observed that GaAs PTR-DLTS spectra yield optoelectronic information in the To < 100psec range. The

6 IMPURITIES AND DEFECTS IN SEMICONDUCTORS

T

205

[KI

FIG. 15. Conductance optical deep-level transient spectrum (0-DLTS) of GaAs:Cr. T~ = 250 psec and To = 500 psec. The applied bias voltage is 0 V. a.u., arbitrary units.

To < 100pec transient signal waveforms (Figs. 8 and 10) were quite different from thermal conduction transients (Leung and Tam, 1984), giving very clear evidence that the DLTS peaks and early-time peak-shift trends in GaAs: Cr wafers at To < 100 psec are due to electronic, rather than thermal, phenomena. This was confirmed with the superposition simulation shown in Fig. 9. The temperature-dependence of the lifetimes under low injection conditions was modeled by use of the Shockley-Read-Hall (SRH) theoretical formalism (Shockley and Read, 1952). Assuming a trap density much smaller than the equilibrium intrinsic carrier density, the SRH formula reduces to (Lang and Logan, 1975) z(T,) = const. x T,-’exp(AE/kT,)

(22)

where T, is the temperature of the DLTS maximum (peak) and AE = ET - E, (for p-type materials), or AE = E, - E T (for n-type materials). Here the subscripts V; C, and T stand for valence band, conduction band, and trap-impurity energy level, respectively. The PTR-DLTS maxima for several pulse repetition periods from the GaAs: Cr and GaAs:Te wafers were plotted in an Arrhenius plot in Fig. 16. Consistent with the direction of the PTR-DLTS peak-shifts, the signs of the two slopes were opposite. For the GaAs:Cr sample, an activation energy of AE = 0.73 eV was extracted from

206

A. MANDELIS, A. BUDIMANAND M.

VARGAS

3.0

2.5 2.0 n

N

E

‘;T r -

1.5

v

1.o

0.5 0.0

2.2

2.4

2.6

2.8

3.0

3.2

1000/T FIG.16. Arrhenius plot of the photothermal radiometric deep-level transient spectroscopic (PTR-DLTS) temperature-dependent lifetimes from the peak positions of several spectra of GaAs:Cr (see Fig. 13) and similar GaAs:Te samples obtained with 50%duty cycle and To = 30, 50, 70, and 90 p e c .

the slope of the curve. This value for the chromium-level activation energy in GaAs is in excellent agreement with the 0.6 to 0.88eV value range obtained using E-DLTS (Lang and Logan, 1975) and O-DLTS (Martin et al., 1980). Furthermore, both those conventional DLTS techniques measure carrier lifetime temperature dependencies that decrease with increasing temperature, 17: and this is also the case with PTR-DLTS. This lifetime behavior can be considered to be characteristic of thermal activation (ejection) processes from chromium traps as thermal energy from the lattice becomes available to trapped carriers, thus decreasing the trap-residence time of the carrier system with increased temperature, in excellent agreement with the magnitude of the PTR-DLTS lifetimes calculated by Papastamatiou and Papaioannou (1990). Unfortunately, the signal and SNR levels of concurrent O-DLTS spectra were insufficient to perform meaningful z(T) measurements from the peak values of curves similar to Fig. 15. For the GaAs:Te sample an activation energy of A E = -0.36eV was extracted from the slope of the curve. The negative sign is consistent with the earlier observation that the PTR-DLTS spectrum may be purely thermal, resulting in the lifetime increasing with increasing I: This temperature-dependence can be explained easily by one of two plausible mechanisms:

6 IMPURITIESAND DEFECTS IN SEMICONDUCTORS

207

1. A purely thermal response, in which the increased temperature lowers the thermal diffusivity of the GaAs lattice (Touloukian et al., 1973). This amounts to an effectively increased thermal transport-relaxation time constant and results in a negative slope, such as the one observed in Fig. 16. 2. A purely electronic response, in which the increased density of intrinsic free carriers at higher temperatures fills up existing trapping (mostly surface) sites, increasing the lifetime of the optically injected carriers.

Sheard (1987) has observed similar lifetime behavior in the frequencydomain PTR of heated Si wafers, which was further accompanied by a reduction in the surface recombination velocity. Increased lifetimes with temperature in impurity-doped Si also have been observed with LM-DLTS (Kirino et al., 1990; Shimura, Okui, and Kusama, 1990) whereas SPV-DLTS of n-GaAs exhibited thermal-activation behavior (Lagowski, Edelman, and Morawski, 1992a). Under the present configuration using an argon ion laser pump with optical penetration depth on the order of l p m , and in the absence of deep bulk levels to influence the photoexcited carrier recombination, the surface conditions of the GaAs:Te wafer are expected to dominate the PTR-DLTS signal, yielding an activation energy AE characteristic of surface-state and near-surface-state trapping processes. A similar mechanism has been reported with PTR-RW detection from deep-level impurity (chromium-) doped n-type Si (Mandelis, Bleiss, and Shimura, 1993). Changing the wafer surface treatment conditions (e.g., by a chemical etch), or comparing PTR-DLTS results obtained with shallow and deep-penetrating photon sources, or both, the relative bulk-surface contributions to the signal can, in principle, be deconvoluted to a large extent (see $1 of Part V) and the observed transient waveform of the GaAs:Te sample can be better understood.

VI. Potential for Ion-Implantation Diagnostics and Conclusions PTR-DLTS, a novel, sensitive, noncontact, all-optical, diagnostic DLTS of semiconductors based on IR radiometric detection has been described. Its physical principles and the first few applications to Si and GaAs wafers with surface or bulk deep impurity levels, or both, have been discussed. The instrumental setup and experimental characteristics of PTR-DLTS regarding the free-carrier plasma or the lattice thermal transport origin, or both, of the signal have been presented. Rate-window detection using a lock-in amplifier or a boxcar integrator scheme has been evaluated with Si wafers.

208

A. MANDELIS, A. BUDIMANAND M. VARGAS

Deep impurity levels originating with chromium, iron, and gold concentrations in the range of 10'3cm-3 clearly have been detectable by the isothermal detection mode of the present technique: Figure 5 indicates a ~ lower, , for chromium in p-type Si. detectivity floor of at least 10" ~ m - or Furthermore, in the constant-duty-cycle temperature-scan mode, PTRDLTS has been able to yield very high SNR DLTS spectra of SI-GaAs samples at fast repetition rates commensurate with relaxation lifetimes in these materials, an impossible task by other types of DLTS techniques, including O-DLTS. Based on the foregoing evidence, there are excellent prospects for the novel technique to be developed to a remote, fully noncontacting, process quality control diagnostic for (i) real-time in situ characterization of native and process-induced deep-level defects in semiconductor wafers, and (ii) for fast, high-spatial-resolution scanning imaging and mapping of specific defects with well-characterized PTR-DLTS peaks at particular substrate temperatures (e.g., Cr, in GaAs). Specifically for the requirements of ion implantation diagnostics, both constant temperature and constant duty cycle modes of PTR-DLTS can be readily applied at the industrial wafer level, owing to the remote signal acquisition nature of the technique. Ion-implant-generated defect levels in semiconductors are shallow (Lang, 1979) and therefore, the sample under investigation would have to be in contact with a cooling element during a PTR-DLTS temperature scan below ambient values. A potential disadvantage of this operating mode could be the expected decreased strength of the PTR signal with decreasing temperature (Chen and Mandelis, 1992). On the other hand, isothermal RW PTR-DLTS scans at room-temperature and above are most likely to yield direct information on the effects of ion implantation on the decay time constant(s) of the signals, with a unique dependence of the lifetime on the type, energy, and dose of implant. This type of experimental mode has the additional advantage of compatibility with industrial implanter set-ups for rapid, on-line characterizations. Several aspects of the application of PTRDLTS to ion-implanted wafers are currently under investigation in the Photothermal and Optoelectronic Diagnostics Laboratories of the University of Toronto.

ACKNOWLEDGMENTS The authors acknowledge the support of the Natural Sciences and Engineering Research Council of Canada through a Collaborative Research Grant, which made this work possible. Arief Budiman thanks the Office of Assessment and Application of Technology, Indonesia, for financial support, Miguel Vargas acknowledges the support of CONACyT, Mexico.

6 IMPURITIESAND DEFECTS IN SEMICONDUCTORS

209

REFERENCES Auret, F. D., and Nel, M. (1988). Single Scan Defect Identification by Deep Level Transient Spectroscopy Using a Two-Phase Lock-in Amplifier (IQ-DLTS). J . Appl. Phys. 63, 973-976. Auret, F. D.,and Nel, M. (1991). Detection of Near-Surface Defects in Semiconductors by Deep-Level Transient Spectroscopy. Measurement Sci. Technol. 2, 623-627. Bakes, H. P. (1976). On the Validity of Kirchhoff's Law of Heat Radiation for a Body in a Non-Equilibrium Environment. In Progress in Optics (E. Wolf, ed.) North-Holland, Amsterdam, Chapt. I. Chantre, A., Vincent, G., and Bois, D. (1981). Deep-Level Optical Spectroscopy in GaAs. Phys. Rev. B 23, 5335-5359. Chen, K., and Mandelis, A. (1992). Scanning Photothermal Rate-Window Spectrometry. Methodologies and Applications to the Thermal Diffusivity Measurement of Ultrahigh Thermal Conductors: CVD Diamonds. Phys. Rev. B 46, 13526-13539. Chen, Z.H., Bleiss, R., Mandelis, A., Buczkowski, A., and Shimura, F. (1993). Photothermal Rate-Window Spectrometry for Noncontact Bulk Lifetime Measurements in Semiconductors. J . Appl. Phys. 73, 5043-5048. Cho, K., and Davis, C. C. (1989). Time-Resolved Infrared Radiometry of Laser-Heated Silicon. IEEE J . Quantum Electronics QE-25, 1112- 1117. Dash, W. C., and Newman, R. (1955). Intrinsic Optical Absorption in Single-Crystal Germanium and Silicon at 77 K and 300K. Phys. Rev. 99, 1151-1155. Day, D. S., Tsai, M. Y., Streetman, B. G., and Lang, D. V. (1979). Deep-Level-Transient Spectroscopy: System Efiects and Data Analysis. J . Appl. Phys. 50, 5093-5098. Desnica, U. V., Petrovic, B. G., Skowronski, M., and Cretella, M. C. (1991). On Quantitative Mapping of EL2 Concentration in Semi-Insulating GaAs Wafers. J . de Physique 111 France 1, 1481-1487. Dobrilla, P., and Blakemore, J. S. (1985). Experimental Requirements for Quantitative Mapping of Midgap Flaw Concentration in Semi-Insulating GaAs Wafers by Measurement of Near-Infrared Transmittance. J . Appl. Phys. 58, 208-218. Edwards, D. F., and Maker, P. D. (1962). Quantitative Measurement of Semiconductor Homogeneity From Plasma Edge. J . Appl. Phys. 33,2466-2468. Fujisaki, Y.,Takano, Y., and Ishiba, T. (1986). Nondestructive Characterization of Deep Levels in Semi-Insulating GaAs Wafers Using Microwave Impedance Measurements. Japan. J . Appl. Phys., 25, L874-L877. Kachare, A. H., Spitzer, W. G., Euler, F. K., and Kahan, A. (1974). Infrared Reflection of Ion-Implanted GaAs. J . Appl. Phys. 45, 2938-2946. Kirino, Y., Buczkowski, A., Radzimski, Z. J., Rozgonyi, G . A., and Shimura, F. (1990), Noncontact Energy Level Analysis of Metallic Impurities in Silicon Crystals. Appl. Phys. Lett. 57, 2832-2834. Lagowski, J., Balestra, C., and Gatos, H. C. (1972). Determination of Surface State Parameters from Surface Photovoltage Transients: CdS. Surface Sci. 29, 203-212. Lagowski, J., Edelman, P., and Morawski, A. (1992a). Non-contact Deep Level Transient Spectroscopy (DLTS) Based on Surface Photovoltage. Semiconductor Sci. Technol. 7, A2 11-A214. Lagowski, J., Morawski, A,, and Edelman, P. (1992b). Non-Contact, No Wafer Preparation Deep Level Transient Spectroscopy Based on Surface Photovoltage. Japan. 1. Appl. Phys. 31 (Part 2), L1185-L1187. Lang, D. V. (1974). Deep-Level Transient Spectroscopy: A New Method to Characterize Traps in Semiconductors. J . Appl. Phys. 45, 3023-3032.

210

A. MANDELIS, A. BUDIMANAND M. VARGAS

Lang, D. V. (1979. Thermally Stimulated Relaxation in Solids (P. Braunlich, ed.) Topics in Applied Physics, Vol. 37. Springer-Verlag, New York, 93- 133. Lang, D. V., and Logan, R. A. (1975). A Study of Deep Levels in GaAs by Capacitance Spectroscopy. J . Electronic Materials 4, 1053- 1066. Leung, W. P., and Tam, A. C. (1984). Techniques of Flash Radiometry. J . Appl. Pkys. 56, 153-161. Ling, Z. G., and Ajmera, P. K. (1991). Measurement of Bulk Lifetime and Surface Recombination Velocity by Infrared Absorption Due to Pulsed Optical Exitation. J . Appl. Phys. 69, 5 19-521. Little, I., Crean, G . M., and Sheard, S. J. (1990). Modelling of the Photothermal Radiometric Response of a Layered Dielectric-on-Semiconductor Structure. Materials Sci. Engng. B 5, 89-93. Luke, K. L., and Cheng, L.-J. (1987). Analysis of the Interaction of a Laser Pulse with a Silicon Wafer: Determination of Bulk Lifetime and Surface Recombination Velocity. J . Appl. Phys. 61, 2282-2293. McKelvey, J. P. (1966). Solid State and Semiconductor Physics. Krieger, Malabar, FL, Chapt. 10.

Mandelis, A. (Editor) (1987). Photoacoustic and Thermal- Wave Phenomena in Semiconductors. North-Holland, New York. Mandelis, A. (1994). Signal-to-Noise Ratios in Lock-in Amplifier Synchronous Detection: A Generalized Communications Systems Approach with Application to Frequency-, Time-, and Hybrid (Rate-Window) Photothermal Measurements. Rev. Sci. Instrum. 65, 33093323. Mandelis, A. (1995). Unpublished. Mandelis, A., Bleiss, R., and Shimura, F. (1993). Highly Resolved Separation of Carrier- and Thermal-Wave Contributions to Photothermal Signals from Cr-Doped Silicon Using Rate-Window Infrared Radiometry. J . Appl. Phys. 74, 3431-3434. Mandelis, A., and Chen, Z. (1992). Lock-in Rate Window Thermomodulation (Thermal Wave) and Photomodulation Spectrometry: Technique, Instrumentation and Measurement Methodologies. Rev. Sci. Instrum. 63, 2977-2988. Mandelis, A,, Budiman, R. A,, Vargas, M., and Wolff, D. (1995). Noncontact Photothermal Infrared Radiometric Deep-Level Transient Spectroscopy of GaAs Wafers. Appl. Phys. Lett. 67, 1582-1584. Mandelis, A. and Wagner, R. E. (1996). Quantitative Deconvolution of Photomodulated Thermoreflectance Signals from Si and Ge Semiconducting Samples. Jpn. J . Appl. Phys. 35, 1786-1797. Martin, G . M., Mitonneau, A., Pons, D., Mircea, A,, and Woodard, D. W. (1980). Detailed Electrical Characterisation of the Deep Cr Acceptor in GaAs. J . Phys. C: Solid State Phys. 13, 3855-3882. Miller, G. L., Ramirez, J. V., and Robinson, D. A. H. (1975). A Correlation Method for Semiconductor Transient Signal Measurements. J . Appl. Phys. 46, 2638-2644. Miller, M. D., and Patterson, D. R. (1977). Transient Capacitance Deep Level Spectrometry Instrumentation. Rev. Sci. Instrum. 48, 237-239. Papastamatiou, M. J., and Papaioannou, G . J. (1990). Recombination Mechanism and Carrier Lifetimes of Semi-Insulating GaAs:Cr. J . Appl. Phys. 68, 1094- 1098. Pierret, R. F. (1987). Modular Series on Solid State Devices (R. F. Pierret and G . W. Neudeck, eds.) Advanced Semiconductor Fundamentals, Vol. VI, Addison-Wesley, Reading, 188-192. Sheard, S. J. (1987). Photothermal Radiometric Microscopy. Ph.D. thesis, University College, London.

6 IMPURITIESAND DEFECTS IN SEMICONDUCTORS

21 1

Sheard, S. J., Somekh, M. G., and Hiller, T. (1990). Non-Contacting Determination of Carrier Lifetime and Surface Recombination Velocity Using Photothermal Radiometry. Materials Sci. Engng. B 5, 101- 105. Shimura, F., Okui, T., and Kusama, T. (1990). Noncontact Minority-Carrier Lifetime Measurement at Elevated Temperatures for Metal-Doped Czochralski Silicon Crystals. J . Appl. Phys. 67, 7168-7171. Shockley, W., and Read, W. T. (1952). Statistics of the Recombination of Holes and Electrons. Phys. Rev. 87, 835-842. Smith, R. A. (1978). Semiconductors. 2nd ed. Cambridge University Press, Cambridge, 118-1 19. Sopori, B. L. (1985). Measurement of High Boron Concentrations in Silicon by Infrared Spectroscopy. Appl. Phys. Lett. 47, 39-41. Spitzer, W. G., and Fan, H. Y.(1957). Determination of Optical Constants and Carrier Effective Mass of Semiconductors. Phys. Rev. 106, 882-890. Spitzer, W. G.,and Whelan, J. M. (1959). Infrared Absorption and Electron Effective Mass in n-Type Gallium Arsenide. Phys. Rev. 114, 59-63. Touloukian, Y. S., Powell, R. W., Ho, C. Y., and Nicolaou, M. C. (1973). Thermal Dzfusivity. IFI/Plenum, New York. Ulmer, E. A., and Frankl, D. R. (1968). Infrared Emission from Free Carriers in Germanium. Proceedings IXth International Conference on Physics of Semiconductors, Nauka, 170- 174. Wolf, H. F. (1969). Silicon Semiconductor Data. Pergamon Press, 17-96.

This Page Intentionally Left Blank

SEMICONDUCTORS AND SEMIMETALS. VOL. 46

CHAPTER 7

Ion Implantation into Quantum-Well Structures R. Kalish SOLID

STATE INSTITUTE AND PHYSICS DEPARTMEKT

TECHNION-ISRAGL INSTITUTEOF TECHNOLOGY HAFA, ISRW

S. Charbonneau INSTITUTE FOR

MICROSIRUCTUFXL SCIENCES

NATIONAL RPSEARCH COUNCIL

OTTAWA, CANADA

1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. GENERAL BACKGROUND. . . . . . . . . . . . . . . . . . . . . . . . . 1. Ion-Implantation-Related Damage . . . . . . . . . . . . . . . . . . . 2. Annealing Ion-Implantation-Related Damage . . . . . . . . . . . . . . . 3. Evaluation of Structural Modijicaiions . . . . . . . . . . . . . . . . . 4. Evaluation By Optical Techniques . . . . . . . . . . . . . . . . . . . 111. ION-BEAM-INDUCED MODIFICATIONS OF QW STRUCTURES . . . . . . . . . . 1. Point Defect and Impurity-Induced Heterostructure Interdifusion . . . . . . 2. Threshold Dose for Intermixing of Quantum Wells: Size of Interface Area Affected by Individual Ion Tracks . . . . . . . . . . . . . . . . . . . 3. Intermixing of Interfaces Far Beyond ihe Ion Range: The Ion Channeling Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Realization of Zero- and One-Dimensional Structures by the Use of Focused Ion-Beams (FIBS). . . . . . . . . . . . . . . . . . . . . . . . . . 5. Defect D@usion in Ion-Implanted AlGaAs and InP Systems . . . . . . . . 6. Implantation Temperature and Dose-Rate Dependence . . . . . . . . . . IV. FUTURE TRENDS AND APPLICATIONS . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

213 216 217 220 221 224 230 231 232

237 238 240 241 248 251

I. Introduction Since the 1970s the major growth area within semiconductor physics has been the study of layered structures, grown by molecular beam epitaxy (MBE), metal-organic chemical vapor deposition (MOCVD), or chemical beam epitaxy (CBE). These structures consist of semiconductor layers with 213 Copyright 0 1997 by Academic Press All rights of reproduction in any form reserved. 0080-8784/97 $25

214

R. KALISHAND S. CHARBONNEAU

the interface geometry, doping level, and chemical composition defined with atomic-scale precision. Superlattices (SLs) and multiple quantum wells (QWs) are artificially structured materials made by interleaving layers of semiconductors with different electronic properties. In this way it is possible to control the electronic properties of the SL and QW by redesigning matter on a quantum mechanical scale. The sharpness of the interface between the different semiconductors plays a major role in the physical properties of the quantum structure. The most commonly studied interface is that between gallium arsenide (GaAs) and Al,Gal -,As (aluminum gallium arsenide), where x varies typically between 0.1 and 0.4. As the two semiconductors are closely matched in lattice spacing, high-quality hetero-epitaxial layers can be grown. The band structure of the two differ so that the conduction band minimum energy of AI,Gal-,As is approximately x eV above that of GaAs, whereas the valence band maximum energy is approximately 0.35 x eV below that of GaAs. Quantum well structures are realized by sandwiching a GaAs layer between two layers of Al,Gal -,As (Kelly and Nicholas, 1985). The progress in epitaxial growth techniques has led to a new understanding of semiconductor physics at a fundamental level. Furthermore, it has extremely important device applications in that it allows the design of semiconductors that do not exist in nature. In particular, it gives control over important material parameters such as the bandgap (through confinement effects), the carrier concentration, the carrier mobility, and the carrier lifetime. The close lattice matching of the GaAs and the AlGaAs system has played a key role in the development of 111-V multilayer devices,just as the high integrity of the silicon-silicon dioxide (Si-SiO,) interface has been crucial to Si technology. In fact, since the mid-1980, use of 111-V material systems has revolutionized the semiconductor industry, especially in the fields of optoelectronics and photonics (De La Rue and Marsh, 1993). Among other 111-V semiconductors, the indium gallium arsenide-indium phosphide (InGaAs-InP) and InGaAs-GaAs material systems have matured in the past few years to the extent that smaller bandgaps and increased lattice mismatch can be used to tailor desired device performances more precisely. Different 11-VI material systems offer narrow-gap, wide-gap, and magnetic options for optical applications, and the field of Si-Sil -xGex (silicon-silicon germanium) multilayers, given their associated lattice mismatch, is another emerging field that may lead to Si-based novel optoelectronic and microelectronic devices. The fabrication of large arrays of very small but identical features by various etching techniques is another challenge. If one reduces the electron states from being two-dimensional, as in layered QW structures, to oneor zero-dimensional, as in layered quantum-wire or quantum-dot structures,

7 ION IMPLANTATIONINTO QUANTUM-WELL STRUCTURES

215

improved laser performances such as lower current thresholds, higher quantum efficiencies, and greater temperature stabilities are predicted (Kelly, 1990). All these advantages are easily lost by unacceptable fluctuations in the feature size of the individual wires or dots. Diffusion of defects over a few atomic layers can radically alter the electrical properties of such small structures. One of the yet unresolved problems connected with growth of QW structures is the structural disorder on the atomic scale that occurs at the growth surfaces that create the interfaces of the hetero-structures. This disorder determines, to a large extent, the quality of the grown structures, and thus is decisive for exploitation of two-dimensional devices. Our present knowledge of structural, chemical, and electronic properties of interfaces between semiconductors is still far from complete. For example, many properties of interfaces in QW structures and of devices consisting of multiple interfaces depend on seemingly minor details of the technology used to produce them, as well as on fundamental physical properties of the material used. Interfaces can be accessible for direct inspection by the existing surface characterization methods if the probing species used (electrons, light ions, or photons) may propagate throughout the analyzed sample to the detecting unit without being scattered along the way. Recently, considerable progress has been made in developing investigative techniques suitable for postgrowth analysis of QW hetero-structure interfaces. This relates in particular to luminescence techniques and transmission electron microscopy (TEM), which are reviewed briefly in the next section. Selective postgrowth modification of Q W hetero-structures has been a goal actively pursued worldwide, mostly for its potential application in optoelectronic integration. Compositional disordering of 111-V compound superlattice structures has received the most attention, primarily because of its simplicity. It can be induced by several techniques: (i) impurity-induced disordering (Chen and Steckl, 1995; Tan and Gysele, 1987; Xia et al., 1992), where the impurities are introduced into the structure by either diffusion or ion implantation; (ii) defect-induced disordering (Ralston et al., 1989; Koteles et al., 1992; Shi et al., 1994; Deppe et al., 1986), where, for example, a striped pattern of SiO, is deposited on the surface of the sample; (iii) laser beam-induced disordering (Kirillov et al., 1984), and (iv) ion-implantationenhanced interdiffusion (Koteles et al., 1989, 1992; Zucker et al., 1992; Poole et al., 1994). In this chapter we concentrate on the disordering of hetero-structures due to ion implantation and subsequent interdiffusion. The QW intermixing process relies on the fact that there is an abrupt, large concentration gradient of atomic species across a QW-barrier inter-

216

R.

KALISHAND

S. C~ARBONNEAU

face. At elevated sample temperatures a mixing of Q W and barrier materials can occur, with this process being dramatically enhanced by the presence of defects or dopant atoms in the interface region. Therefore, the sole purpose of the implantation process is to produce vacancies, interstitials, and other defects or to introduce the atomic species required to promote enhanced diffusion. The number of defects generated depends on many factors such as the mass of the ions, their energy, the dose or fluence (i.e., ions/cm2), the ion flux (current density), and the temperature at which the implantation is performed. In subsequent sections we attempt to summarize the field of ion-induced QW disordering and the techniques most frequently used in quantifying this process. There is a continuing stream of new information concerning different aspects of ion-induced QW disordering, which not only provides additional data but also often modifies the interpretation of a particular concept or of experimental results. Therefore, rather than trying to include all papers published on this subject, only representative references are given to permit the reader to become acquainted with the current status and research trends concerning the subject. We apologize to those whose work has not been mentioned.

11. General Background

The intermixing of QW structures can be caused by several different processes, as previously mentioned, including the following: thermal interdiffusion across the interface; interdiffusion assisted by the presence of specific impurities; kinematic intermixing due to momentum transfer to atoms in the interface region by ions that are shot through the structure, and diffusion assisted by defects. Generally, the results of such intermixing can be determined by either direct or indirect measurements. Direct “observations” of changes in atomic arrangements across a hetero-epitaxial interface are obtained from high-resolution TEM (with chemical image enhancement) (Ourmazd, 1993; Bode et al., 1991), from secondary ion mass spectroscopy (SIMS) measurements that yield information on atomic composition as a function of depth (Leier et al., 1990), and to a much lesser degree of sensitivity, by Rutherford backscattering spectrometry (RBS). Indirect yet very accurate information on changes to QW structures as a result of intermixing can be obtained from various forms of optical measurements. Particularly useful are a variety of luminescence techniques (both photoluminescence (PL) and cathodoluminescence (CL)), which include direct observations of the PL spectra and of integrated peak intensity,

7 ION IMPLANTATIONINTO QUANTUM-WELL STRUCTURES

217

photoluminescence excitation (PLE) experiments or time-resolved PL measurements. A detailed review of these techniques can be found in Chapter 3 of this volume. In the next sections, we briefly summarize the main experimental and theoretic points relevant to ion-beam-related intermixing of hetero-epitaxial interfaces. More detailed information can be found elsewhere (Marsh, 1993; Cohen, 1993; Bradley et a/., f1993). 1. ION-IMPLANTATION-RELATED DAMAGE Ion implantation is a violent process in which energetic ions (the implants) are forced into the target material. During their slowing down in the solid, large amounts of damage are inflicted on the sample until the implants come to rest. It is important to bear in mind that, unless the implantation-related damage is annealed out, a process that is accompanied by excessive diffusion (i.e., intermixing of layers), most measureable changes in the implanted layer are due to damage and not directly to the presence of the implanted species. The volume density of the energy deposited in the stopping medium by the ion during its slowing down is the important parameter that determines the damage inflicted on the material by each implanted ion. Two dose regimes are particularly important for the intermixing of QW structures due to ion-beam-induced damage and subsequent annealing. The first “low-dose” regime is defined as that in which only a small fraction of the interface area has experienced the damaging effects of an individual ion traversing it. The second “high-dose” regime is defined by the high areal density of impinging ions such that the total interface area, on average, has been affected by the implantation process. The “critical dose” that defines the border between these two regimes depends, as will be shown subsequently, on the implantation and annealing conditions and on defects diffusivities throughout the structure. Information about the implant distribution in the target and the related damage inflicted can be obtained from computer simulations, of which TRIM (The TRansport of Ions in Matter; Biersack and Haggmark, 1980) is the most commonly used. The TRIM program is a Monte Carlo code that statistically follows the collisions individual ions undergo while in motion in the target material. The input parameters to the program are the specific experimental conditions (ion type and energy, target material, composition, and density) and the displacement energy E d , which is the energy required to displace a target atom far enough from its lattice site so that it will not fall back into the vacancy it has left behind. This parameter is usually not

218

R. KALISHAND S. CHARBONNEAU

well known. An uncertainty in Ed is directly reflected in uncertainties in the calculated numbers of vacancies and interstitials created during the implantation process. Typical damage cascades, as calculated by the TRIM computer code, due to a single 8-MeV bismuth (Bi) ion and a single 100-keV Ga ion, shot into GaAs are shown in Figs. l(a) and l(b), respectively. (In these simulations a value of 20eV was used for the displacement energy.) The large difference between these two cases both in the total ion range and in the range-straggling (1.07 f 0.32pm for Bi and 0.044 L 0.020pm for Ga) as well as in the number of displaced target atoms (about 75,000 for Bi and 1900 for Ga) should be noted. Furthermore, it is evident from Figs. l(a) and

FIG. 1. Collision cascade following penetration into gallium arsenide of (a) a single 8-MeV bismuth ion; (b) a single 100-keV Ga ion, as calculated by the Monte Carlo computer code TRIM (The TRansport of Ions in Matter). (M. Shaanan, unpublished).

7 ION IMPLANTATIONINTO QUANTUM-WELL STRUCTURES

219

FIG. 1. (Continued)

l(b) that the damage cascade has “branches” in which point defects are clustered in high densities, whereas other regions along the ion track contain only very few, well-isolated, defects. Once a large enough number of ions has hit the target surface, the effects due to the individual ion-impacts are washed out and only average values of defect densities and ion ranges are of importance. Defect (vacancy) distributions as obtained from the TRIM simulation, averaged over several thousand Bi (8 MeV) or G a (0.1 MeV) ions implanted into GaAs, are shown in Fig. 2. The large difference in the thickness of the layer affected by the implantation process is again evident. Whereas 100-keV Ga ions will heavily damage a layer approximately lOOOA deep, the damaging effects of 8-MeV Bi ions will extend, rather homogeneously, well over l p m in depth. Since

R. KALISHAND S . CHARBONNEAU

220

CJ X

'"'4

0.0

5.0

10.0

15.0

DEPTH I R )

2 -0

%lo'

FIG.2. Defect distribution (total target displacements) for 8-MeV Bi (xi) and 100-keV Ga ions in GaAs as obtained by TRIM (The TRansport of lons in Matter). The size of a typical quantum structure is marked for comparison.

most quantum structures are confined to the near-surface region of the material (about 1 pm), MeV ion implantations will homogeneously affect the whole structure, with no doping effects whatsoever in the active layer, whereas sub-MeV implantations may affect only parts of the structures, and the implanted ions may come to rest in the midst of the QW or SL structure. Here, the discussion mainly concentrates on the intermixing of interfaces due to ion-implantation-related damage rather than due to chemical doping effects. 2. ANNEALINGION-IMPLANTATION-RELATED DAMAGE Most applications of semiconductors require as perfect as possible crystallinity. Ion-implantation-related damage is thus undesirable and must be removed by thermal annealing. This annealing, which is based on the thermal diffusion of vacancies and interstitials, is also accompanied by atomic diffusion, and hence by the exchange of atoms across interfaces, that is, interdiffusion. Furthermore, compound semiconductors (e.g., binaries and ternaries) may have the tendency to undergo stoichiometric changes due to differences in diffusivities and volatility of some constituents. These pro-

7 ION IMPLANTATION INTO QUANTUM-WELL STRUCTURES

221

cesses obviously are undesirable and need to be suppressed. The schemes commonly employed to anneal the materials of which most QW structures are composed (e.g., GaAs-GaAlAs, and In-GaAs-InP) make use of rapid thermal annealing (RTA), in which the specimen is rapidly heated in an inert atmosphere, with the surface protected against effusion by either encapsulation or proximity caps. Typical annealing conditions for 111-V systems are rapid heating to 700 to 900°C for from 0.1 to 5 min. Such annealing conditions seem to be adequate in restoring crystallinity, because full PL intensity is restored following implantation and annealing; that is, defect centers giving rise to nonradiative recombinations are annihilated.

3. EVALUATION OF STRUCTURAL MODIFICATIONS Two questions regarding the state of QW structures following ion implantation and annealing that need to be addressed are: (1) Did the crystallinity of the material recover (that is, are all defects annealed out)? (2) To what extent did the interface sharpness get blurred out due to interdiffusion. Subsequently, we briefly describe and critically evaluate the most commonly used techniques for structural evaluation of interfaces. a.

Transmission Electron Microscopy and Chemical Mapping

High-resolution TEM (HRTEM) makes possible the imaging of structures with atomic resolution. It is also commonly used to image extended defects such as dislocations, stacking faults, and inclusions. As usually operated, HRTEM thus is not capable of observing antisite defects, that is, the interchange of group 111 atoms across a hetero-structure interface. Ourmazd recently has shown (1993) that by properly choosing the angle of the recorded reflection of the diffracted electrons, chemical imaging can be achieved. This mode of TEM permits the observation of compositional changes in the atomic occupancy of an ordered subset of lattice sites. Thus it is the ideal tool to observe the structure and sharpness of interfaces between hetero-epitaxial layers. Figure 3 shows a typical chemical image of the GaAs- AlAs interface together with the compositional profiles across the top (AIAs-GaAs) and bottom (GaAs-A1As) interfaces. As can be seen, both interfaces are equally sharp and are abrupt to within one lattice constant. It is obvious that chemical mapping is a most suitable technique for the assessment of modifications to interfaces due to ion implantation and annealing. Its only drawback lies in the complexity of specimen preparation

222

R. KALISHAND S.CHARBONNEAU

FIG. 3. Chemical lattice image and composition profiles for two aluminum arsenidegallium arsenide (Ah-GaAs) interfaces. Each data point represents the composition of a 1-pm segment of an atomic plane parallel to the interface. A1 Conc., aluminum concentration. (With permission from Bode, M., Ourmazd, A., Cunningham, J. and Hong, M. (1991). Interaction of Energetic Ions with Inhomogeneous Solids. Phys. Rev. Lett. 67(7), 843-846).

and data analysis, which are the reasons this powerful technique is not widely used to evaluate interface qualities. However, HRTEM as commonly done can be successfully used to identify the nature and spatial distribution of residual defects after annealing. Accurate quantitative measurement of the interdiffusion length, deduced from the contrast in the electron image due to the difference in the atomic scattering factors between GaAs and GaAlAs layers, remains difficult, especially in the presence of residual defects, qualitative information can however be readily obtained by this technique (Bitnell and Tobbs, 1989). Thus this technique is a powerful complementary technique used to investigate in detail the behavior of defects before and after annealing.

6. Random and Channeling Incidence Rutherford Backscattering Spectroscopy Rutherford backscattering spectroscopy (RBS) is a quick, direct, easy, non-destructive experimental method for determining the composition of

7

ION IMPLANTATIONINTO QUANTUM-WELL STRUCTURES

223

thin layers, provided that the atomic mass of the constituent atoms are sufficiently different to allow their resolution in the RBS spectra. Hence, GaAs is not suitable for elemental RBS evaluation, whereas for example InP is most favorable for mass sensitive RBS measurements. Since the depth sensitivity of RBS depends primarily on the detector resolution and on the scattering geometry employed, the method usually is not sensitive enough to detect intermixing of only a few atomic layers around a hetero-epitaxial layer. However, when combined with channeling, RBS allows to determine the perfection of the crystallinity, and thus is a good tool for studying the buildup of implantation damage with increasing dose and annealing (Akano et a!., 1993). Furthermore, when channeling experiments are performed under nonnormal incidence conditions, strain in QW and SL structures can be measured and quantified (Chu, 1988). Chapter 8 of volume 45 describes this technique in more detail.

c.

Secondary Ion Mass Spectrometry and Auger Electron Spectroscopy

The techniques of SIMS and Auger electron spectroscopy (AES) are based on detection of sputtered atoms (in SIMS) or of typical signatures of atoms on the sample surface (in AES). Hence they are both surface-sensitive techniques and are complementary to the optical technique discussed subsequently. The depth resolution in both is obtained by gradual, controlled erosion of the specimen by sputtering and can reach an accuracy of a few nanometers for very thin layers or for low mixing rates (Anderson, 1979). However, for high implantation doses, enough mixing is generated to the extent that quantitative measurements of the interdiffusion length for high implantation can be obtained. This mode of digging into the material is destructive and may induce some intermixing of atomic layers by the sputtering process itself. Hence it does not have the high atomic sensitivity that HRTEM offers. Nevertheless, SIMS and, to a lesser extent, AES are widely accepted techniques for the evaluation of intermixing of QW structures. Figure 4 (Leier et ~ l . ,1990) illustrates the use of SIMS for the assessment of interdiffusion of an AlGaAs-GaAs superlattice subjected to different implantation and annealing treatments. It is worth emphasizing that SIMS and PL are complementary techniques that allow characterization of the disordering for low doses (low mixing rates) to high doses (high mixing rates). Furthermore, SIMS also can be used before the annealing step to separate the effect of the bombardment itself from the diffusion of the effects during annealing. However, this technique does not give any indication of the nature and spatial distribution of residual defects after annealing.

R. KALISHAND S. CHAREIONNEAU

224

I

I

1.5.1 O'5cm-2

I

85OoC, 0.5h 230keV Ar

1 o5

1

104

0

lo5 .-& m z 1o4 1 o5 1'0 1 o5 10'

U

0

200 400 depth (nm)

FIG.4. Secondary ion mass spectroscopy (SIMS) signal as a function of depth of differently processed AIGaAs-GaAs superlattices. From bottom to top: as-grown sample; annealed sample (SSOT, 30min); sulfur (S)-implanted sample (1.5 x 10LScm-2,100keV); silicon (Si)implanted (1.5 x 1015cm-Zat 100keV) and annealed sample (850°C for 30min); S-implanted (1.5 x 1015cm-2 at 100keV) and annealed sample (SSOT for 30 min); and argon (Ar)implanted (1.5 x 1 0 ' 5 ~ m - 2at 230keV) and annealed sample (850°C, 30 min). arb., arbitrary; impl., implanted. (With permission from Leier, H., Forchel, A., Hyrcher, G., Hommel, J. Bayer, S., and Rothfritz, H. (1990). Mass and Dose Dependence of Ion-Ionplantation-Induced Intermixing of GaAs/GaAlAs Quantum-Well Structures. J. Appl. Phys. 67(4), 1805- 1813.)

4. EVALUATION BY

OPTICAL

TECHNIQUES

Optical techniques provide a nondestructive tool for the analysis of the structure of materials and of intrinsic impurity- or defect-related electronic transitions in semiconductors. Photoluminescence (PL) is the most common technique for examining structural disorder, on an atomic scale, at the surfaces creating the interfaces of the QW hetero-structures. Photoluminescence is luminescence by excited photons (usually delivered by a laser) and is distinct from cathodoluminescence (CL), in which excitation is achieved by electron impact (usually in an electron microscope). Both these excitation

7 ION IMPLANTATION INTO QUANTUM-WELL STRUCTURES

225

techniques promote electrons from the valence to the conduction band of the semiconductor, resulting in the creation of electron-hole pairs that may recombine and emit characteristic luminescent radiation (Herman, Bimberg, and Christen, 1991). Another frequently used technique is Raman scattering. These three optical characterization techniques are briefly reviewed subsequently.

a. Luminescence Spectroscopies Luminescence techniques belong to the most sensitive, nondestructive methods of analyzing both intrinsic and extrinsic semiconductor properties. Despite this, the application of these techniques for analyzing interfaces and epitaxial growth processes has only started recently. The QW first studied by optical techniques was the rectangular- or square-shaped potential well structure shown schematically in Fig. 5(a) (Miller and Kleinman, 1985; Allard et al., 1992). It generally consists of a

I

I-

- - -- - - - -- -,

I

FIG. 5. Schematic showing the band structure of a quantum well (a) before and (b) after compositional disordering. EJx) is the bandgap energy as a function of concentration x. (With permission from Allard, L. B., Aers, G. C., Charbonneau, S., Jackman, T. E., Williams, R. L. Templeton, 1. M., and Buchanan, M. (1992). Fabrication of Nanostructures in Strained InGaAs/GaAs Quantum Wells by Focused-Ion-Beam Implantation. J. Appl. Phys. 72(2), 422-428.)

226

R. KALISHAND S. CHARBONNEAU

thin GaAs layer sandwiched between larger bandgap Al,Ga, -,As slabs. These two kinds of layers are usually referred to as well material and barrier material, respectively. The investigations performed on such structures by absorption spectroscopy have evidenced experimentally, for the first time, a fundamental quantum-mechanical rule, known from physics textbooks since the 1920s. It concerns the bound states of a particle confined in a onedimensional, quantum well of thickness L,, which is comparable to the de Broglie wavelength of the particle. In this case a discrete spectrum of energy levels is created for the particle in the potential well. It is important, however, that the energies of these levels depend strongly on the well thickness L, and on the well material. In the simplest case of an infinitely deep rectangular potential well, the solution of the Schrodinger equation leads to the energy eigenvalues of a particle in a box, that is, En = (n2112/ 2m*) (n/L,)’, n = I, 2,3... . In the limit of L, + co , a continuum of states results and the particle will no longer be in the quantum limit. A typical PL emission spectrum of GaAs QWs of varying thicknesses is shown in Fig. 6 (curve (a)). Each QW emits radiation at a characteristic energy that is inversely proportional to the square of the QW thickness. The previous discussion is related to a QW with infinitely sharp interfaces. We now focus our attention on the effect of interdiffusion on the potential profile of QW structures. As will be discussed, both the presence of impurities (e.g., zinc and silicon), implantation damage, or both can enhance the interdiffusion, thus rounding up the potential profile at the interface. The simplest model for interdiffusion describes an isotropic diffusion of a single species across the hetero-structure. It applies, for example, to the diffusion of the group I11 elements in the lattice-matched GaAs-Al,Ga, -,As system and in the strained In,Gal -,As-GaAs hetero-structure, with a diffusion coefficient independent of the A1 concentration and the In concentration, respectively (Allard et al., 1992). A simple measure of the degree of QW intermixing is the interdiffusing length Ai (defined as one quater of the distance across the QW interface over which the concentration decreases from 90% to 10% of the original value) obtained by modeling the ternary concentration after implantation and annealing as

for aluminum (upper signs) or indium (lower signs), where x is the original concentration and L, is the initial width of the square well (Fig. 5(a)). This model is used to calculate local bandgaps and effective masses from which the PL transition energy is calculated, and Ai is adjusted to fit the observed energy shift.

7 ION IMPLANTATION INTO QUANTUMWELLSTRUCTURES Cloddlfq GOo.tA1aaAm

227

Gobrlnml thkh-8 10.0

5.0 9 W 2 7.0 OW3

5500

; 4500 c

.-C a

-6 d

3500

>

k W

2500

c

z

2

1500

so0

-500

B 1.5 1.6 1.7

ENERGY IeV) FIG.6. Photoluminescence (PL) (at 1.4K) of quantum well (QW) structures following 8.0-MeV bismuth ion implantations and rapid thermal annealing at 900°C for 4 min. (a) Unimplanted and annealed (b) 2.5 x 10'0/cm2; (c) 7.5 x 10'o/cmz; and (d) 2.5 x 10"/cm2. The dashed and dotted lines are meant to guide the eye, showing the shifts of the different PL lines. The inset shows the multiple QW structures. arb., arbitrary; GaAs, gallium arsenide; Al, aluminum. (With permission from Kalish, R., Kramer, L.-Y., Law, K.-K., Merz, J. L. Feldman, L. C., Jacobson, D. C., and Weir, B. E. (1992). Local Intermixing of GaAs/GaAIAs Quantum Structures by Individual Ion Implant Tracks. Appl. Phys. Lett. 61(21), 2589-2591.

The interdiffusion process leads to a band profile such as that shown in Fig. 5(b). The net effect of the interdiffusion on the electron and hole quantized states in the well is an effective downshift with respect to the bottom of the bands. The transition energy, which is observed in PL experiments, between the first heavy-hole state and the first electron state E:" and E;, respectively

R. KALISH AND S. CHARB~NNEAU

228

(a)

”

> -10-

P a,

C W

’.,.A€; ’\

-16 -

%

x = 0.05

-20

...

I

I

3E

A

Y

0.0

0.6

1.0

1.6

Ai (nm) FIG. 7. Calculated transition energy shifts BE,, for a 3.0nm wide indium gallium arsenide (InGaAs) well plotted as a function of the interdiffusion length Ai for (a) 5 % and (b) 20% indium concentrations. Also shown are the heavy-hole and electron eigenenergy shifts that are dominated by the charge in the bandgap to produce the net positive AE,,. (With permission from Allard, L. B., Aers, G. C., Charbonneau, S., Jackman, T. E., Williams, R. L. Templeton, I. M., and Buchanan, M. (1992). Fabrication of Nanostructures in Strained InGaAs/GaAs Quantum Wells by Focused-Ion-Beam Implantation. J. Appl. Phys. 72(2), 422-428.

(measured with respect to their respective band minima) is

Err = Eth + E;

+ E;”(Z = 0) - E,,

where E F ( z = 0) is the heavy-hole bandgap and E,, the exciton binding energy. In Figs. 7(a) and 7(b), the calculated transition energy (blue) shift AErr is shown for an initial well width L, = 3.0nm, as a function of interdiffusion length Ai for initial indium fraction of 5% (i.e., Ino.o,Gao.95As-GaAs) and

7 ION IMPLANTATIONINTO QUANTUM- WELLSTRUCTURES

229

20% (i.e., Ino.20Gao.80As-GaAs). Obviously the larger the concentration (i.e., the deeper the well), the greater the possible shift due to interdiffusion. Also shown in Figs. 7(a) and 7(b) are the heavy-hole and electron eigenenergy shifts that are dominated by the change in the bandgap to produce a net positive (blue shift) AEtr. Therefore, the shift of the peak energy AE or the shift of the wavelength AI of the PL between an intermixed QW and its as-grown state allows us to monitor and estimate the compositional disordering Ai resulting from implantation and annealing. This disordering process, as measured by the PL shift, is greatly influenced by the dose of ions implanted (number of defects created), the width of the QW, and the magnitude of the potential difference between the well and the barrier material. A number of studies have focused on the dose dependence of the intermixing process. Kalish et al. (1992) used high-energy (%MeV) Bi ions to disorder five GaAs-AlGaAs QWs of different widths (Fig. 6). In general, the as-implanted QWs (before annealing) show no measurable luminescence having an integrated PL intensity about four orders of magnitude lower than the as-grown material. This quenching of the PL signal is due to the large number of residual defects (e.g., vacancies and interstitials) left by the implanted ion damage cascade in the QWs. These defects act as nonradiative recombination centers, which are very efficient in trapping photoexited carriers. The absence of PL shift before annealing is an indication that the implant doses used are below those required for collisional (ion beam) mixing. On annealing, the situation changes dramatically as can be seen in Figs. 6(b) to 6(d). The energy shifts observed occur due to the change in the geometry of the well, as discussed previously. Wide wells, which correspond to the lowest energy PL line, shift a small amount because the intermixed region is a small fraction of the total well width. In narrow wells the energy levels are located near the top of the well, hence for these only a relatively small shift is possible. Quantum wells of intermediate width display the largest PL shift, as indicated in Fig. 6. Another very sensitive luminescence technique used to study defect generation through implantation is time-resolved luminescence. This technique represents the most important and most powerful optical tool for gaining information on exciton lifetimes. Different capture processes can be distinguished by measurements of the transient behavior of the various peaks of a luminescence spectrum. As discussed previously, defects generally act as nonradiative decay centers that lead to a decrease in the observed decay times. (Allard et al., 1992). It should be emphasized that each luminescence technique possesses advantages and disadvantages. For example, CL, in which excitation is achieved by finely focused electron beams, is limited in lateral resolution by the diffusion of excitons in QWs (e.g., -1pm at 4.2K), whereas the

R. KALISHAND

230

s. CHARBONNEAU

resolution in PL is limited by how narrowly one can focus the probing laser beam on the sample (diffraction limited). The probing depth in CL easily can be varied by changing the electron energy. The time resolution obtainable in PL is about three orders of magnitude higher than that in CL due to the ability of obtaining ultrashort laser pulses (Charbonneau, 1989). Tunable laser excitation (resonant PL) can be used to investigate various radiative transitions, providing more information on the binding (localizing) centers.

b. Raman Scattering Spectroscopy Another optical technique commonly used in the characterization of QW structures is Raman scattering. In the Raman effect a photon is scattered inelastically by the crystal, with the creation or annihilation of a phonon or magnon. The Raman effect is made possible by the strain-dependence of the electronic polarizability. This technique is extremely sensitive to the damage created by ion implantation and was successfully used to characterize impurity-enhanced interdiffusion of GaAs- AlGaAs QWs after RTA (Choo et al., 1992; Kumar et al., 1993) or to study the interdiffusion of InGaAs-InP SL by thermal annealing (Yu et al., 1991). The interdiffusion profiles of group 111 and V atoms were obtained by measuring the shift of frequencies and the variation of intensities of longitudinal optical phonon peaks. From the development of three of the complementary techniques reviewed, (namely, PL, SIMS, and TEM) critical points of ion-beam-induced mixing of semiconductor hetero-structures such as measurement of the mixing rate, characterization of the homogeneity of the mixing, recovery of the electronic properties after annealing, defect behavior, and comprehension of intermixing mechanisms can be efficiently investigated.

111. Ion-Beam-Induced Modifications of QW Structures

As mentioned in 01 of Part 11, ion implantation creates damage in the target material. Hence following the implantation process a certain layer of the material, the thickness of which depends mainly on implant mass and energy, contains large numbers of vacancies and interstitials as well as the implanted ions themselves. Whereas the common use of ion implantation in semiconductor technology is for doping the material n- or p-type or for isolation, this does not apply to doping QW structures. The reason is that the feature size of most structures is much smaller than the width of the ion distribution ( ARJ. Hence selective doping of specific layers by ion implantation is impossible and needs to be done during epitaxial growth.

-

7 ION IMPLANTATION INTO QUANTUM-WELL STRUCTURES

231

Nevertheless, the presence of vacancies, interstitials, and dopants in the QW structure, in particular in the vicinity of the interfaces, has very pronounced effects on interface interdiffusion. In the next section, we briefly review the well-known effects that excess group 111 vacancies or interstitials as well as dopants have on diffusion of group I11 constituents in 111-V semiconductors. We then discuss the intermixing effects of interfaces due to ionimplantation-related defects alone for low-dose and high-dose implantation and for three different depth regimes: (1) the ions pass through the QW structure, (2) the ions stop within the stucture, and (3) the ions stop well ahead of the structure.

AND IMPURITY-INDUCED HETEROSTRUCTURE 1. POINT-DEFECT INTERDIFFUSION

The excellent review by Deppe and Holonyak (1988) covers the topic of hetero-structure interdiffusion very well, hence here we only briefly mention the points most relevant to the present discussion. Both point defects and the presence of specific impurities (e.g., silicon, sulfur, or zinc) have a pronounced effect on the interdiffusion of 111-V hetero-structures (Deppe, 1986, 1987). The quantity of group I11 vacancies or interstitials can be controlled by selecting annealing conditions that either favor or inhibit group 111 vacancy generation. For example, encapsulation with siliconnitride (Si,N,) or annealing under As underpressure, inhibits group I11 vacancy formation, whereas encapsulation with silicon-dioxide (SiO,) or annealing under As overpressure results in group I11 vacancy formation. All these have been shown to enhance GaAs-AlGaAs interface interdiffusion (Vieu et al., 1991, 1992). The presence of dopants, such as Si or zinc (Zn) also has a pronounced effect on structural modifications of 111-V semiconductors insofar as they greatly enhance the diffusion of group 111 elements over that of group V constituents. Figure 4 (Leier et al., 1990) clearly demonstrates the relative importance of the presence of defects and dopants to the intermixing of a GaAs-AlGaAs superlattice. Figure 4 shows SIMS profiles of the A1 signal for the as-grown, annealed (unimplanted), asimplanted (with 100keV 1.5 x l O I 5 S/cm2) and implanted (with 1.5 x 1015cm-Z lOOkeVS, Si, and 230keV Ar ions) and annealed cases. It is clearly seen that 1. Annealing alone has hardly any effect on the SL, hence purely thermal interdiffusion of hetero-structures is negligible. 2. Implantation alone has little effect on the SL structure, hence kinematic mixing alone hardly affects interface intermixing.

232

R. KALISHAND S. CHARBONNEAU

3. Damage due to implantation with no chemical effect (Ar) does cause substantial intermixing following annealing. 4. Implantation of dopants (in particular, Si) has a dramatic effect on washing out the SL to the extent of complete intermixing of the GaAs- AlGaAs layers. This list clearly demonstrates the great potential that lies in the possibility of locally modifying QW structures by selective area ion implantation. This can be achieved either by implantations through properly designed masks or by the use of focussed ion beams (FIB) for the implantation of ultrafine features. A key question is what is the minimum feature size achievable by these techniques (following annealing). We now address this question.

WELLS: 2. THRESHOLD DOSEFOR INTERMIXING OF QUANTUM SIZEOF INTERFACE AREAAFFECTED BY INDIVIDUAL ION TRACKS The size limit for the creation of low-dimensional (one- or zero-dimensional) quantum structures achievable by implantation of two-dimensional structures through photolithographic masks or by using finely focused ion beams is determined by the cross-sectional damage area affected by an individual ion impact. On annealing, this area may increase beyond the originally created damage track due to defect-assisted diffusion (Kalish et al., 1992). The lateral size of the damaged region caused by individual ion impact has been directly observed by HRTEM and has been indirectly deduced from RBS channeling experiments. The TEM work has shown that for the case of 80 keV gold (Au) ions implanted at a very low dose (10" cm-') into GaAs, amorphous regions approximately 3 nm in diameter can be associated with damage due to impingement of a gold atom onto the surface (Bode et al., 1991). RBS-channeling experiments on GaAs implanted with 600-keV gold and 8-MeV Bi ions at ever-increasing doses (Kalish et al., 1993) indicate that effective damage diameters of 4 and 2nm must be associated with the damage due to the gold and Bi ions, respectively. These diameters relate to damaged regions before annealing and find support from computer simulations (TRIM). Information about the interface mixing caused by the impact of single ions was obtained by the chemical imaging technique (Bode et al., 1991). Figure 8 shows the chemical lattice images and the deduced composition profiles obtained from different regions in an AlAs-GaAs interface, im-

7 IONIMPLANTATION INTO QUANTUM-WELL STRUCTURES

233

FIG. 8. Chemical lattice image of an aluminum arsenide on gallium arsenide interface, implanted with 5 x 10” ions/cm2. The chemical profiles, obtained by averaging over 50 %( segments of planes parallel to the interface, show substantial variations in the GaAs- AlAs transition width. (With permission from Bode, M., Ourmazd, A,, Cunningham, J. and Hong, M. (1991). Interaction of Energetic Ions with Inhomogeneous Solids. Phys. Rev. Lett. 67(7), 843-846).

planted at 77 K with 320 keV 5 x 10’ Ga+/cm2. As can be seen from Fig. 8, the width of the interface is sometimes (in those regions where an ion has passed) substantially broadened, changing from about 0.25 nm for the as-grown interface sharpness (Fig. 3) to about 0.62 nm. We therefore can conclude that at low temperatures the passage of an energetic heavy ion through a hetero-structure leaves behind, on average, a damaged region that extends laterally over a few nanometers (from TEM and RBS measurements) and that it causes an interface mixing of a couple of lattice constants (from chemical imaging). Postimplantation annealing of QW structures removes the implantation related damage; at the same time, it may cause excessive interdiffusion of the hetero-structure. A way to determine the lateral dimensions of ion-impact affected regions following annealing is to measure the PL from QWs that have been subjected t o implantations (through the structures) over a wide range of doses (lo7 to 10’3cm2). This dose range covers the transition from isolated intermixed regions, created by individual ion impacts, to doses at which the intermixed regions overlap to yield a uniformly mixed layer. The PL spectra of such an experiment (Kalish et al., 1992) in which a structure

234

R. KALISHAND S. CHARBONNEAU

containing five GaAs-AlGaAs wells of different widths has been implanted by 8-MeV Bi ions at different doses followed by RTA have been shown in Fig. 6. Interestingly, for doses below a critical dose D, of 7.5 x 1010cm-2, the implantation has no noticeable effect on the PL spectra. Only for doses exceeding D, can a gradually increasing blue shift be noticed. The physics underlying the abrupt changes observed in the PL spectra at a particular critical dose is that the PL of a GaAs quantum well altered by local intermixing depends on the relative size of the affected and nonaffected areas and on their spatial distribution. For a low concentration of small, isolated, intermixed regions, the PL spectra should be those of the unaffected material as the excitons may diffuse inside the well and eventually decay in regions of minimum energy, that is, in the areas unaffected by the ioninduced mixing. However, with increasing dose, a threshold is reached at which the proximity of intermixed areas starts to be on the order of the size of the exciton (- 10 nm). Some excitons thus will decay in the intermixed QWs, giving rise to blue-shifted PL originating from implantation mixed regions. The emission measured in this case thus will include PL from both affected and nonaffected areas. For even higher ion doses, the complete area will have been intermixed due to one or more ion impacts. The PL spectrum now will be completely blue-shifted and the lines broadened, reflecting the probability that some areas of the sample have been mixed by more than one ion. From this stage on, additional ion implantation will cause further ion-beam-induced mixing, with a monotonic increase in the PL blue-shift with ion dose (Charbonneau et al., 1995). This description is consistent with the data shown in Figs. 6 and 9. The critical dose deduced from the PL data has been used to extract a lateral mixing radius rm around each ion track (following annealing) by assuming that each ion impact affects an area nr; (Kalish et al., 1992). At the critical dose the whole area has been intermixed, that is, D;ltr; = 1. By taking D, = 7.5 x 1010cm-2, an intermixing radius of 20nm has been obtained for the case of %MeV Bi ions traversing the GaAlAs-GaAs interface. Similar experiments were performed on different unstrained GaAs-AlGaAs QW structures for implantations at room temperature and at elevated (350°C) temperatures (Kalish et al., 1993) and on unstrained GaAs-AlGaAs and strained InGaAs-GaAs QW structures for indium concentrations ranging from 5% to 20% implanted at room temperature (Kalish et al., 1992; Allard et al., 1994). The radii of the intermixed regions as deduced from the critical doses for different GaAs-AlGaAs structures vary, for implantations at room temperature, from 18nm (Kalish et al., 1992) to 23nm (Allard et al., 1994). For the hot implantation case this radius increased to 30nm (Kalish et al., 1993). For the strained InGaAs-GaAs

I ION IMPLANTATION INTO QUANTUM-WELL STRUCTURES

= *-

'9 20

235

0

0

- - - m *

0

5.0nm

0

0

0

k

40+1 20

20-0nm

107

10'

IMPLANT

lo"

DOSE ( C W r )

10'3

FIG. 9. Photoluminescence (PL) line shifts as a function of implantation dose for the different quantum wells studied. The thickness of the QW is specified in each panel. Implantation conditions are 8.0-MeV of bismuth at room temperature and annealed at 900°C for 4 min. (With permission from Kalish, R.,Kramer, L.-Y., Law, K.-K., Merz, J. L. Feldman, L. C., Jacobson, D. C., and Weir, B. E. (1992). Local Intermixing of GaAs-GaAIAs Quantum Structures by Individual Ion Implant Tracks. Appl. Phys. Left. 61(21), 2589-2591.)

QWs (Allard et al., 1994), the PL lines were found to shift already at the lowest implantation dose employed ( lo9Bi/cm2), yielding a lower limit for the mixing radius of 180nm! In a similar study (Myers et al., 1988), the intermixed radii were shown to increase with In concentration in the well. It should be noted that these intermixing radii are huge, as compared with the radii of the ion-beam-affected spots before annealing! The measured PL line shifts can be converted to the interdiffusion lengths Ai by the use of Eq.

236

R. KALISHAND S. CHARBONNEAU

-

0

10'

2.0

.

10'"

10"

5

iol'

1,

10'~

I

A

E 5 4-

I

-0.5

10'

10'" 10'' 10" Dose (Bi'/cm2)

10'~

FIG. 10. (a) Interdiffusion lengths Ai fitted to the photoluminescence (PL) energy shifts for the GaAs-A1,~,Ga0,,As samples as a function of bismuth ion dose. In order of distance from 9.5 nm (O), the sample surface, the data for the different well widths are 3.5 nm (o),5.6 nm (o), and 18.5 (0). Data for another sample are for well widths 8.0nm (A)(4.5 nm (A), and 2.8nm (A). The dashed lines follow the extreme values at each dose. (b) Interdiffusion lengthsAi fitted to the PL energy shifts for 3.0-nm-wide QWs in the samples as a function of bismuth ion dose. In order of distance from the sample surface, the data are x = 0.071 (a)0.105 , (A), 0.147 (A) and 0.180 (A).Data for another sample are x = 0.041 (o),0.074 (o),0.111 (O),and 0.159 (0). The dashed lines follow the extreme values at each dose. (With permission from Allard, L. B., Aers, G . C., Piva, P. G., Poole, P. J., Buchanan, M. Templeton, I. M., Jackman, T. E., and Charbonneau, S. (1994). Threshold Dose for Ion-Induced Intermixing in InGaAs/GaAs Quantum Wells. Appl. Phys. Lett. 64(18), 2412-2414.)

(1). When this is done for the data of Allard et al. (1994) for both strained and unstrained structures (Figs. lO(a) and (b), respectively), intermixing lengths on the order of 1 nm are obtained for both strained and unstrained systems, however, with the intermixing starting at much lower doses for the strained InGaAs system.

7 ION IMPLANTATION INTO QUANTUM-WELL STRUCTURES

237

Intermixing, both in the plane of the interface and perpendicular to it, must be explained in terms of defect-assisted diffusion. However, it should be noted that this diffusion is peculiar in that it is extremely unisotropic. Whereas the intermixing along the interface plane extends laterally over tens of nm, its effects across the interface are felt over distances 10 to 100 times smaller (i.e., a few nm at most).

3.

INTERMIXING OF INTERFACES FAR

BEYONDTHE ION RANGE:

THEION CHANNELING EFFECT During their study of the effects of FIB implantations into a set of GaAs-GaAlAs QWs located at different depths from the surface, Laruelle et al. (1990) observed interface mixing at depths well in excess of the ion range. The explanation offered for this observation was that accidentally channeled ions that penetrated deep into the target material were responsible for this intermixing. This explanation finds supports in the finding that very low doses ( w 10' cmW2)are sufficient to cause complete intermixing of GaAs-AlGaAs hetero-structure interfaces (Kalish et al. 1992). Since mixing experiments have typically been carried out at approximately 1013/cm2,only a very small, yet quite reasonable, fraction of the implanted ions need be channeled for effective mixing beyond the amorphous range. Thus the previous results on the low doses required for complete hetero-structure intermixing support the notion of channeling as the cause of this anomalous ion-beam-induced deep penetration phenomenon. An experiment explicitly designed to test the importance of ion channeling in the intermixing of QWs located beyond the projected (amorphous) ion range is reported in the work of Jackman et al. (1991). In that work, four InGaAs-GaAs quantum wells of equal thickness, yet with different In content (hence exhibiting different PL emission lines), located at 50, 100, 250, and 450nm below the surface, were subjected to 5 x 1013cm-Z 250-keV Au implantations under different conditions: into the (100) channel, along a carefully chosen random direction and through a thin (20nm), amorphous, Si3N, beam-scattering cap layer. The expected random ion penetration parameters for this implantation are R, f ARp = 51 L- 18nm. Hence if channeling was of importance for interface intermixing, the QWs lying well beyond the random ion range should exhibit a blue shift only for the implantations in which a sufficient number of gold ions (i.e., in the channeling tail) reaches them. This indeed was found by experiment. Excessive gold diffusion through the structures, as was observed by SIMS to accompany the annealing of the implanted specimens, has been proven (Jackman et al., 1991) not to be responsible for the observed P L line shifts.

238

R. KALISHAND

s. CHARBONNEAU

4. REALIZATIONOF ZERO-AND ONE-DIMENSIONAL STRUCTURES BY THE USEOF FOCUSED ION-BEAMS (FIBs)

One of the major applications of FIBs is to write fine features by FIB implantation into QW structures, thus obtaining lower-dimensional quantum systems. This is achieved by locally affecting two-dimensional structures (i.e., QWs or SLs) either by damaging them (for isolation applications) or by selectively intermixing their interfaces, and thus following annealing, modifying their electronic energy levels. In this way quantum well wires (QWWs) and quantum well boxes (QWBs) can be realized from twodimensional quantum well layers. Two kinds of FIBs are commonly used: nondopant ion beams and dopant beams. Among the nondopant beams, Ga ions are the most commonly used because G a is a nonalien atom to most III-V hetero-structures studied, and because of technologic aspects related to ion-source-beam optics. Obviously, the effect of such beams is to induce interfacial mixing due to damage alone. Among the dopant ion beams used, Si is a widely used ion, because it strongly affects the diffusion of group I11 elements, as shown previously (see Fig. 4). The beams commonly used have an energy of several hundred keV, and can be focused down to submicron spots. As a result, these beams carry a very high current density and deliver to the implanted spot extremely high power densities. A beam of lOpA, when focused onto a spot lOOnm in diameter, will carry a current density of 0.1 A/cm2. If the ions have an energy of 100keV, then the power density delivered to the implanted spot will be lo4 W/cm2. This is a very high power density that may cause substantial local heating to the irradiated spot even for short dwell times. Several papers have been published over the last few years describing the realization of quantum wires and quantum boxes by FIBs followed by RTA. The optical features of the structures thus created have been evaluated by the use of PL and CL techniques. As described above, the effect of ion implantation and subsequent annealing on hetero-structures is to intermix the layers and thus to cause a blue shift. Hence single ion implants or irradiation of dots by the use of FIBs always will lead to regions of higher-energy emissions, that is, to antiquantum dots. Quantum dots or quantum wires thus can be realized only by bracketing the desired area by ion-beam-affected blue-shifted regions, hence leaving the dot or wire unaffected. The properties of the structures obtained by such implantations are best observed by the use of CL, due to the possibility of combining the high lateral resolution of electron beams (produced in an electron microscope) with optical emission spectroscopy originating from the spot excited by the electrons.

I

ION IMPLANTATION INTO QUANTUM-WELL STRUCTURES

239

FIG. 11. Schematic representation of the cathodoluminescence (CL) spectrum at T = 14K and the corresponding micrographs of two wires and a box. The quantum well (1.1-nm wide) is 100nm below the surface and the line separation is 200 nm. The inset is a schematic view of the implantation pattern used. arb., arbitrary. (With permission from Laruelle, F., Bagchi, A,, Tsuchiya, M.,Merz, J. and Petroff, P. M.(1990). Focused Ion Beam Channeling Effects and Ultimate Sizes of GaAlAs/GaAs Nanostructures. Appl. Phys. Lett. 56(16), 1561- 1563.)

240

R. KALISHAND S. CHARBONNEAU

This application of FIB and its evaluation by CL have been developed and widely used by Laruelle et al. (1990). High-energy (150-keV) Ga ions, focused down to a spot size of 80nm were used to write line or grid patterns in GaAs-AlGaAs QW structures, thus creating quantum wires or quantum boxes. Figure 11 shows the low-temperature two-dimensional CL maps obtained at different wavelengths of the luminescence spectrum. By using this technique, quantum wires were bracketed between two implanted lines, and quantum boxes between two pairs of perpendicular lines. Very fine features can be obtained by taking advantage of the lateral spreading of the ion-beam-affected areas due to the excessive lateral diffusion and the lateral straggling of the focused ion beams. As mentioned previously, low-dimensional quantum structures also can be realized by wide-area implantations through small features created by high-resolution electron-beam lithography. The first such structures were reported by Cibert et al. (1986) who implanted a single GaAs-GaAlAs QW with 210-keV Ga ions to a dose of 5 x 1013cm-2 through arrays of metal wires and dots. Cathodoluminescence analysis of the structures obtained following RTA and mask lift-off clearly showed that QWWs and QWBs could be obtained. The cross-sectional CL data also demonstrated the extensive lateral spreading of the intermixed regions to extend below the masked off regions, thus leading to quantum features smaller than the features originally blocked off from the beam by the mask. Direct optical observation of the energy levels due to an array of simple quantum wires has been reported by Hirayama et al. (1986, 1988). In that work, a single GaAs-GaAlAs QW was locally intermixed by writing a set of finely spaced (200nm) lines by a focused (100nm) Ga ion beam (200pA, 100keV, 1.6 x 1016cm-2). After annealing (SOOOC 60min with a GaAs proximity cap), PL and PLE measurements have revealed features in the spectra that could be attributed to transitions between the electron and hole levels in a QWW with interdiffused boundaries, thus demonstrating the realization of one-dimensional structures and their quantum properties.

5. DEFECTDIFFUSION M ION-IMPLANTEDAlGaAs AND InP SYSTEMS As pointed out in 111.3, the intermixing of QW layers at depths significantly beyond the expected mean ion range requires ion channeling to assist the transport of the ions to the QW region. Experiments combining PL with SIMS in the InGaAs-GaAs and GaAs-AIGaAs systems indicate that the ion damage must be delivered close to the QW region, either via direct implantation or through channeling effects, to have a significant

7

ION IMPLANTATION INTO QUANTUM-WELL STRUCTURES

241

mixing effect (Jackman et al., 1991; Kieslich et al., 1992). Taken together, these results suggest that defects created in the ion implantation process do not diffuse very rapidly through undoped or p-doped GaAs or AlGaAs material. This is consistent with the results of Kahen et al. (1989) who obtained a low value for the vacancy diffusion constant in A10,,,Gao,62As. In this section, we review the importance of the location of the initial distribution of ion-induced damage with respect to the structure before annealing (e.g., in front of the QW, at the QW and underneath the QW) and its importance to the intermixing process in two material systems (Gillin et al., 1993; Hamoudi et al., 1995).

a. Experimental Evidence of Defect Difision In the work of Poole et al. (1995), two InGaAs QW structures were studied. They consisted essentially of (a) InGaAs-GaAs QWs embedded in Alo.70Gao.30As cladding layers; and (b) InGaAs-InGaAsP QWs embedded in InP cladding layers. In both samples, the QWs were at approximately the same depth from the surface (about 1.85pm) and the cladding layers, through which the defects created by implantation must diffuse, were of comparable width. Ion implantation was performed, with the sample normal tilted 7 degrees to the beam (to deliberately avoid channeling), at energies ranging from 300keV to 8.6MeV. For both structures, the ion species were chosen to avoid doping the structure: As' for the InGaAs-GaAs-AlGaAs sample and phosphorus ions (P') for the InGaAs-InGaAsP-InP sample. After implantation, the samples were annealed at 850°C and 700"C,respectively, for 30sec in a nitrogen atmosphere. In recent studies (Poole et al., 1995), these temperatures were found to produce significant intermixing of the QW in implanted regions of the samples without causing significant intermixing in the unimplanted regions. The PL energy shift of the InGaAs QWs was used to monitor the compositional disordering, while the line width and intensity of the PL emission were used t o evaluate the relative number of nonradiative recombination sites left by the implantation and not removed by the annealing. Figure 12 shows the P L peak from the InGaAs QWs obtained from as-grown samples together with spectra obtained after ion implantation and 30sec RTA. In the AlGaAs based structure (Fig. 12(a)), the P L peak changes significantly as a function of the ion implantation energy. For ion energies below 2MeV (corresponding to an expected ion range less than 1 pm, i.e., short of the QW structure) there is very little broadening and no shift of the QW peak. For implant energies between 2 and 4 MeV, for which

R. KALISHAND

242

s. CHARBONNEAU

I+0.5 MeV

k 8.6MeV

0.88 0.90

0.92

0.94 1.251.30 1.35 1.40 1.45 1.50

Wavelength (pm)

Wavelength (pm)

Photoluminescence (PL) emission spectra from (a) single strained In, z,Gao 79 As-GaAs quantum well (QW) in the InGaAs-GaAs-AIGaAs laser structure; and from (b) five unstrained In,~,,Ga,,,,As-ln,,,~Ga,~~~ A S ~ ~ ~ ,QWs P ~ ,in~ the , InGaAs-InGaAsP-lnP laser structure. Spectra are given for typical as-grown samples (dashed lines) together with the results obtained after implantation with (a) arsenic ions and (b) phosphorous ions at the indicated energies, followed by a 30-sec annealing at 850 and 700"C, respectively. In (a) the QW peak intensity has been normalized to the PL peak from the doped gallium arsenide cap (not shown here). In (b) the QW peaks have been normalized to have the same intensities. arb., arbitrary. (With permission from Poole, P. J., Charbonneau, S. Aers, G. C., Jackman, T. E., Buchanan, M., Dion, M., Goldberg, R. D., and Mitchell, I. V. (1995). Defect diffusion in ion implanted AlGaAs and InP. Consequences for Quantum Well Intermixing. J. Appl. Phys. 78(4), 23672371.)

the majority of defects are located in the vicinity of the active region, the PL peak is found to be very broad and weak (note the magnification factor in Fig. 12), but does not shift upward in energy. Above 4-MeV ion energy, that is, the range of the ions is beyond the QW, the peak exhibits an upward (blue) energy shift of about 30meV and is again intense and quite narrow. This change occurs over a narrow range of implantation energies between 4 and 5 MeV. (The energy shifts AE and peak widths are summarized in Fig. 14(a) and 14(b)). A quite different picture is observed in the PL spectra from the InP-based material, shown in Fig. 12(b). The most obvious difference between these spectra and those of Fig. 12(a) is that substantial intermixing already occurs at very low implantation energies and shifts close to the maximum value are

ION IMPLANTATIONINTO QUANTUM-WELL STRUCTURES

7

243

20

c .-0

15

7

'E c

-

*

Y

10

Ion energy (MeV)

5 I I

a n ~. 00

Vi

0

I

-

I

0.0

1.0

2.0

3.0

I

4.0

Depth ( p m ) FIG. 13. Ion range (open circles) and vacancy (solid circles) distributions for 4-MeV arsenic implantation in the InGaAs-GaAs- AlGaAs structure, calculated using TRIM 91 (The TRansport of Ions in Matter) Computer Code. Note the discontinuities in the vacancy production at the depths corresponding to the cap and active region interfaces. The inset shows the depth corresponding to the peak in the implanted ion depth distribution and peak vacancy generation as a function of incident ion energy. The horizontal dotted line indicates the depth ( 1 . 8 8 ~ of ) the InGaAs quantum well (QW). (With permission from Poole, P. J., Charbonneau, S. Aers, G. C., Jackman, T. E., Buchanan, M., Dion, M., Goldberg, R. D., and Mitchell, I. V. (1995). Defect diffusion in ion implanted AlGaAs and InP: Consequences for Quantum Well Intermixing. J. Appl. Phys. 78(4), 2367-2371.

already obtained with ions having as little as 2MeV energy, that is, with projected ranges well short of the QWs. Furthermore, there is no energy region in which the line width is dramatically larger or the intensity significantly lower than at other energies. This implies that defect complexes in the QW region that might act as nonradiative recombination sites are either not created as easily in this InP-based system or are more efficiently removed during the annealing. In order to understand the different results obtained in these two material systems, that is, the importance of the initial distribution of ion-induced damage, TRIM 91 simulations were used. Figure 13 shows the results of such a simulation for both the implanted ion distribution and the total vacancy distribution (created by ions and recoils) for 4-MeV As ions implanted into the InGaAs-GaAs- AlGaAs structure. The inset shows for each implantation energy the depth at which the ion and total-vacancy distributions reach their maximum values.

244

R.KALISHAND S. CHARBONNEAU

I

C

0

Ti.-

*&) Y

ion energy (MeV) FIG. 14. Photoluminescence (PL) and TRIM 91 (The TRansport of Ions in Matter) results for the InGaAs-InGaAsP-AICaAs structure. (a) The energy shift AE in peak PL emission versus ion implantation energy after 30,60, and 12Osec annealing at 700°C. (b) Corresponding full width at half maximum of the PL peak. (c) Local implanted ion deposition and vacancy generation (number/nm/incident ion) at the depth of the middle quantum well (QW) (1.83 p) and total number of created vacancies in the sample (number/incident ion) calculated with the TRIM 91 program. (With permission from Poole, P. J., Charbonneau, S. Aers, G . C., Jackman, T. E., Buchanan, M., Dion, M., Goldberg, R. D., and Mitchell, I. V. (1995). Defect diffusion in ion implanted AlGaAs and InP. Consequences for Quantum Well Intermixing. J. Appl. Phys. 78(4), 2367-2371.)

Figure 14(c) shows the calculated concentrations of ions and vacancies created at the depth of the Q Win the InGaAs-GaAs-AlGaAs structure. In addition, the total number of vacancies created in the structure per incident ion is displayed. There is very good correlation between the energy at which the PL peak is most broadened (Fig. 14(b)) and reduced in intensity (Fig. 12(a)) and the energy at which the simulation predicts the peak in the ion

7

ION IMPLANTATION INTO QUANTUMWELLSTRUCTURES

245

distribution (end-of-range) to occur in the QW region. Furthermore, the magnitude of the blue shift after 30 sec of annealing correlates well with the concentration of vacancies generated in the QW region, including a reduction in shift at the highest implant energy, where most vacancies are created significantly beyond the QW. The main point to be emphasized is that after 30 sec of annealing, intermixing occurs for the AlGaAs-based structure only at energies above 4 MeV, where the simulation predicts a sharp increase in vacancy creation at the QW. Figure 15 gives the same comparison of simulation and experiment for the InP-based structure. The damage produced per ion is lower in this case, mainly due to the lower scattering cross section and lower mass of the phosphorus ion relative to that of As. Figure 15(a) clearly shows that, after 30 sec of annealing, QW intermixing occurs at energies for which no ions or vacancies have been deposited in the Q W region. However, the correlation with the total number of vacancies created anywhere in the sample is very good. It also can be seen that the PL line width does not change as dramatically as in the previous sample and, for implantation energies up to 2 MeV, does not increase significantly from that observed in unimplanted material. Considering first the InGaAs-GaAs- AlGaAs structure, the previous data, together with earlier work (Kahen et al., 1989) support the following picture of the intermixing process, which is assumed to be mediated by defects that must diffuse primarily through AlGaAs during anneal: 1. Significant shifts in the PL peak, after the first annealing, occur only at energies for which the simulation predicts the creation of large numbers of vacancies in the QW region. Thus for efficient Q W intermixing by ion implantation to occur in InGaAs-GaAs QWs embedded in AlGaAs, the defects must be created in the vicinity of the QW. This suggests that the defect diffusion length in the Alo.71Gao.29Ascladding layers is short during the RTA. This is in qualitative agreement with a very short neutral vacancy diffusion length (-0.1 pm) found in a somewhat different Alo.3sGao.62Asalloy (Kahen et al., 1989). Of course, there is a possibility that the defect diffusion is inhibited once the interface between the AlGaAs cladding and the GaAs barrier region is reached by the defects. 2. The maximum broadening and intensity reduction of the PL peak occur at lower implantation energies than those resulting in large peak shifts and correlates well with the energy at which most of the implanted ions are deposited in the well region. It can be speculated that this is due to the resulting defect complexes acting as nonradiative recombination centers, as previously reported (Allard et al., 1992;

246

R. KAL~SH AND S. CHARBONNEAU

i

i

3

4

6

8

ion energy (MeV) FIG. 15. Photoluminescence (PL) and TRIM 91 (The TRansport of Ions in Matter) results for the InGaAs-InGaAsP-InP structure. (a) energy shift AE in peak PL emission versus ion implantation energy after 30, 60, and 120 set annealing at 700°C. (b) Corresponding full width at half maximum of the PL peak. (c) Local implanted ion deposition and vacancy generation (number/nm/incident ion) at the depth ofthe middle quantum well (QW) (1.83 p ) and total number of created vacancies in the sample (number/incident ion) calculated with the TRIM 91 program. (With permission from Poole, P. J., Charbonneau, S. Aers, G . C., Jackman, T. E., Buchanan, M., Dion, M., Goldberg, R. D., and Mitchell, I. V. (1995). Defect diffusion in ion implanted AlGaAs and InP. Consequences for Quantum Well Intermixing. J. Appl. Phys. 78(4), 2367-2371.)

Laruelle et al., 1990). Such complexes may be hard to dissolve during annealing. In the InGaAs-InGaAsP-InP structure, where defect diffusion occurs mostly through InP, the situation appears to be quite different. The strong correlation between the PL energy shift and the calculated total vacancy creation in the sample per incident ion suggests that the defects are

I

ION IMPLANTATIONINTO QUANTUM-WELL STRUCTURES

247

extremely mobile in InP and that, wherever they are created in the structure, they can diffuse to the QW region and promote intermixing. Figures 14 and 15 also show the dependence of the peak shift and line width on annealing time for the two structures. In the case of the InGaAsGaAs-AlGaAs structure (Fig. 14), the degree of intermixing increases slowly with annealing time, suggesting the slow transport of defects to the QW region. This is particularly noticeable at 3 MeV, where vacancies are created just short of the QWs. At 4 MeV, intermixing clearly is retarded by more extensive damage. At energies for which ions are deposited in the QW region, the linewidth decreases with annealing time, confirming that the heavy QW damage is only slowly removed by the annealing process. For the InGaAs-InGaAsP-InP structure (Fig. 15) the situation is rather different. After annealing for 60 sec or longer, significant PL peak shifts are noticeable at all implantation energies, suggesting that defects can diffuse all the way from the cap region to the Q W interfaces to promote intermixing. Finally, because of the similarities between the calculated initial depth distributions of ions, vacancies, or interstitials, further work is required to identify uniquely the dominant defect species responsible for the intermixing of quantum wells in these systems, and the role that the effect of strain might play in the interdiffusion process. 6. IMPLANTATION TEMPERATURE AND DOSE-RATE DEPENDENCE

The temperature dependence of ion beam mixing in many materials is reasonably well understood in terms of standard treatments of radiationenhanced diffusion. At low temperatures, mixing is mainly due to atomic transport within the displacement cascades and thus is nearly independent of temperature. At higher temperatures, when point defects become mobile, the mixing increases exponentially with temperature. For 111-V compound semiconductors, it was found that hetero-structure mixing increases with temperature up to a critical temperature, at which point it precipitately drops (Forbes et al., 1995; Anderson et al., 1988). This temperature has been proposed as the temperature at which an instantaneous amorphous-to-crystalline transition takes place. The magnitude of the low-temperature mixing parameters in these compound semiconductors is very large. It was found to correlate inversely with the melting temperature, decreasing in the order of InP > GaAs > AlAs while melting temperatures increase in the reverse order (Forbes et al., 1995; Anderson et al., 1988). Recently, the disordering of GaAs- AlGaAs and InGaAs-GaAs QWS (Charbonneau et al., 1995) by means of ion bombardment at elevated temperatures was reported. For both systems it was concluded that the effect

248

R. KALISHAND S. CHARBONNEAU

of implanting at high temperatures is to enhance the diffusion and recombination of interstitials and vacancies. The net result of high-temperature implantation is a steady-state interstitial-vacancy population, which enhances the mixing of the QW with a minimum of residual damage left after implantation and RTA. Dose-rate effects on amorphization or on interdiffusion are generally thought to be the result of both temporal and spatial overlap of defects within the collision cascades. Studies on dose-rate effects on damage accumulation in implanted 111-V materials have shown that the generation rate of interstitial-vacancy pairs increases with increasing ion flux (dose rate) (Haynes and Holland, 1991a and 1991b). Recently it has been shown (Charbonneau et al., 1995) that the intermixing of hetero-structures is also dose-rate dependent. The explanation of this phenomenon is most likely similar to that posed for the amorphization of bulk 111-V materials.

IV. Future Trends and Applications The realization of optoelectronic devices in low-dimensional QW structures is most tempting because such devices are expected to exhibit superior performance. Whereas the problems related to the growth of two-dimensional QWs by various epitaxial growth techniques are well under control, the implementation of zero-dimensional (quantum dots) or one-dimensional (quantum wires) systems is still in its infancy. As pointed out previously, it is now possible to take advantage of the fact that ion implantation (followed by proper annealing) shifts the energy levels in the affected regions to bracket areas within the two-dimensional QW structure. Hence small squares (QWBs) or narrow lines (QWWs) can be fabricated. The dimensions of these features are dictated by the possibility of selectively implanting small regions, either through openings in masks or by direct writing by the use of FIBS. The fact that after annealing the implantation-affected areas spread out due to enhanced interdiffusion leads to the broadening of the dimensions of the modified regions, thus further shrinking the active areas bracketed by them. This defect-enhancedhetero-structure interdiffusion thus aids the realization of ultrafine features in QWs. This very promising possibility of the miniaturization of low-dimensional QW structures unfortunately cannot be utilized effectively, as yet, due to insufficient defect removal. The presence of residual damage in the ion-beam affected areas, which cannot be completely removed by the annealing process, seems to degrade the optical performance of the devices due to the presence of nonradiative recombination centers that survive the annealing

7 ION IMPLANTATION NTO QUANTUM-WELL STRUCTURES

249

process. Effective ways to remove these residual defects have to be found in order to allow the practical application of low-dimensional quantum structures confined by ion-implantation-related interdiffusion of hetero-structures. Another important field in which ion-beam-assisted diffusion of QWs has proven to be useful is that of monolithic integration in optoelectronics. Since the original proposals by Miller and Kleinman (1989, monolithic integration of several optoelectronic devices in photonic-integrated circuits and optoelectronic integrated circuits has been a goal actively pursued worldwide. The primary requirement for the monolithic integration of optoelectronic devices is the mutual compatibility of optical bandgap energies among the various components on a given chip. For example, integrated wavelengthdivision multiplexing components for optical communications systems require the integration of lasers, modulators, and waveguiding regions. The bandgap energy requirements of a circuit consisting of a QW laser, a modulator, and a wave guide (shown schematically in Fig. 16) are that the

Energy FIG. 16. Schematic illustration of the relative bandgap positions of laser, modulator, and waveguide needed for successful photonics integration. (With permission from Charbonneau, S., Poole, P. J., Piva, P. G., Aers, G. C., Koteles, E. S., Fallahi, M., He, J.-J., McCaffrey, J. P., Buchanan, M., Dion, M., Goldberg, R. D., and Mitchell, 1. V. (1995). Quantum well intermixing for optoelectronic integration using high energy ion implantation. J. Appl. Phys. 78(6), 3697-3705.).

R. KALISHAND S. CHARBONNEAU

250

bandgap energy of the injection laser be lower than that of the modulator, which in turn must be lower than that of the waveguide; hence, local modification of the bandgap is required. Several approaches to integration based on QW active layers have emerged in recent years. In most cases the variation of these bandgap energies on a wafer usually involves complicated etch and regrowth processes that are possible in principle but difficult in practice. The selective interdiffusion of QW structures, and hence the possibility of locally tuning the bandgap, appears to be the most promising emerging technology for the fabrication and integration of optoelectronic devices. Selective interdiffusion offers a planar technology that can be used to integrate, laterally, regions of different bandgaps. Since the bandgap of the intermixed alloy is larger than that of the original QW structure, ion-implantation-related selective-area interdiffusion may provide a route to form low-loss optical waveguides, bandgap-shifted quantum-confined modulators, lasers, and detectors, using only one epitaxial step. In addition, because the bandgap is increased in the interdiffusion regions, the refractive

1420

1440

1460

1480

1500

Wavelength (nm) FIG. 17. Emission spectra of the bandgap-shifted broad-area lasers ( 6 0 p x SOOps). The emission wavelength, external quantum efficiency, and threshold current are: (a) 1497 nm, 30.7%,410mA;(b) 1485nm, 31.5%, 410mA; (c) 1449nm, 25.8%,400mA; (d) 1433nm, 24.6%, 420mA. arb., arbitrary. (With permission from Charbonneau, S., Poole, P. J., Feng, Y., Aers, G . C., Dion, M., and Davies, M. Band-gap tuning of InGaAs/InGaAs/InP laser using high energy ion implantation. Appl. Phys. Letter 67(20), 2954-2956,)

7 ION IMPLANTATION INTO QUANTUM-WELL STRUCTURES

251

index can be modified. In structures containing only a few QWs, this change in the refractive index will have only a small effect on the optical propagation constant; however, in multiple QW structures, the optical overlap between the intermixed well and the optical wave can be large enough to allow useful changes in the refractive index to provide optical confinement gratings or even laser reflectors. Several other potential applications of Q W intermixing techniques in integrated optoelectronics can be identified. These include bandgap-tuned modulators, bandgap-tuned lasers, low-loss waveguides for interconnecting components on an optoelectronic integrated circuit, integrated extended cavities for line-narrowed lasers, single-frequency distributed Bragg reflector lasers and mode-locked lasers, nonabsorbing mirrors, and either gain or phase gratings for distributed feedback lasers. In a recent publication, Charbonneau et al. (1995) demonstrated that the technique of ion-induced Q W intermixing can be successfully used to tune the emission wavelength of an InGaAs-InGaAsP-InP multiple Q W laser operating at 1.55pm.Figure 17 shows the lasing spectra from four differently implanted regions (on the same sample), which have undergone different degrees of intermixing (blue-shifted by as much as 63 nm). The high quality of the bandgap-shifted lasers that is retained despite implantation and annealing and as demonstrated by the constant threshold current obtained on the blue-shifted layers, is remarkable. The spatial selectively inherent in this technique makes it useful for the monolithic integration of optoelectronic devices having different functionalities (Alferness et al., 1992). Finally, the fact that SL or QW structures can be locally damaged and intermixed by selective area ion implantations, even without annealing, may find applications in the isolation of various devices adjacent to each other on the same chip.

REFERENCES Akano, U. G., Mitchell, I. V., and Shepherd, F. R. (1993). Influence of Dose Rate and emperature on the Accumulation of Si-Implantation Damage in Indium Phosphide. Appl. Phys. Lelt. 62(14), 1670-1672. Akano, U. G., Mitchell, I. V., Shepherd, F. R., Miner, C. J., and Rousina, R. (1993). Implant-Damage Isolation of InP and InGaAsP. J. Vuc. Sci Technol. A 11(4), 1016-1021. Alferness, R. C., Koren, U., Buhl, L. L., Miller, B. I., Young, M. G., Koch, T. L., Raybon, G., and Burrus, C. A. (1992). Broadly Tunable InGaAsP/lnP Laser Based on a Vertical Coupler Filter with 57-nm Tuning Range. Appl. Phys. Letr. 60(26), 3209-321 1. Allard, L. B., Aers, G . C., Charbonneau, S.,Jackman, T. E., Templeton, I. E., and Buchanan, M. (1992). Focused-Ion-Beam Implantation in Strained InGaAs-GaAs Quantum Wells. Canad. J . Phys. 70, 1023-1026.

252

R. KALISHAND S. CHARBONNEAU

Allard, L. B., Aers, G . C., Charbonneau, S., Jackman, T. E., Williams, R. L., Templeton, I. M., and Buchanan, M. (1992). Fabrication of Nanostructures in Strained InGaAs/GaAs Quantum Wells by Focused-Ion-Beam Implantation. J. Appl. Phys. 72(2), 422-428. Allard, L. B., Aers, G . C., Piva, P. G., Poole, P. J., Buchanan, M., Templeton, I. M., Jackman, T. E., and Charbonneau, S. (1994). Threshold Dose for Ion-Induced Intermixing in InGaAs/GaAs Quantum Wells. Appl. Phys. Leff. 64(18), 2412-2414. Anderson, H. H. (1979). The Depth Resolution of Sputtering Profiling. Appl. Phys. 18, 131. Anderson, K. K., Donnelly, J. P., Wang, C. A., Woodhouse, J. D., and Haus, H. A. (1988). Compositional Disordering of GaAs/AIGaAs Multiple Quantum Wells Using Ion Bombardment at Elevated Temperatures. Appl. Phys. Lett. 53(17), 1632-1634. Biersack, J. P., and Haggmark, L. G. (1980). A Monte Carlo Computer Program for the Transport of Energetic Ions in Amorphous Targets. Nucl. Instrum. Merh. 174, 257. Bitnell, E. G., and Tobbs, W. (1989). Philos. Mag. A60, 39. Bode, M., Ourmazd, A., Cunningham, J., and Hong, M. (1991). Interaction of Energetic Ions with Inhomogeneous Solids. Phys. Rev. Lett. 67(7), 843-846. Bradley, I. V., Gillin, W. P., Homewood, K. P., and Webb, R. P. (1993). The Effects of Ion Implantation on the Interdiffusion Coefficients in In,Ga, -,As/GaAs Quantum Well Structures. J. Appl. Phys. 73(4), 1686-1692. Charbonneau, S. (1989). Picosecond Photoluminescence Spectroscopy in Highly Excited Semiconductors. Optical Engrg. 28(10), 1101-1 107. Charbonneau, S., Poole, P. J., Feng, Y.,Aers, G. C., Dion, M., Davies, M., Goldberg, R. D., and Mitchell, I. V. (1995). Bandgap Tuning of InGaAs/InGaAsP/InP Laser using High Energy Ion Implantation. Appl. Phys. Left. 67(20), 2954-2956. Charbonneau, S., Poole, P. J., Piva, P. G., Aers, G . C., Koteles, E. S., Fallahi, M., He, J.-J., McCaffrey, J. P., Buchanan, M., Dion, M., Goldberg, R. D., and Mitchell, I. V. Quantum Well Intermixing for Optoelectronic Integration using High Energy Ion Implantation. J. Appl. Phys. 78(6), 3697-3705. Chen, P., and Steckl, A. J. (1995). Selective Compositional Mixing in GaAs/AIGaAs Superlattice Induced by Low Dose Si Focused Ion Beam Implantation. J. Appl. Phys. 77, 5616-5624. Choo, A. G., Gupta, V., Jackson, H. E., Boyd, J. T., Steckl, A. J., Chen, P., Weiss, B. L., and Burnham, R. D. (1992). Raman Characterization of AlGaAs Superlattice Channel Waveguide Structure Formed by CIB and FIB Implantation. Materials Res. SOC.Symp. Proc. 240,691-695. Chu, Wei-Kan. (1988). Resonance Planar Channeling Effect in Superlattices. In Nuclear Physics Applications on Materials Science (E. Recknagel and J. C. Soares, eds.). NATO AS1 Series, Vol. 144, Kulvar Academic Publishers, 117- 132 and references therein. Cibert, J., Petroff, P. M., Dolan, G . J., Pearton, S. J., Gossard, A. C., and English, J. H. (1986). Optically Detected Carrier Confinement to One and Zero Dimension in GaAs Quantum Well Wires and Boxes. Appl. Phys. Left.49(19), 1275-1277. Cohen, R. M. (1993). Interdiffusion in Alloys of the GaInAsP System. J. Appl. Phys. 73(10), 4903-4915. De La Rue, R. M., and Marsh, J. H. (1993). Integration Technologies for 111-V Semiconductor Optoelectronics Based on Quantum Well Waveguides. Critical Rev. 45, 259-288. Deppe, D. G., Guido, L. J., Holonyak, N. Jr., Hsieh, K. C., Burnham, R. D., Thornton, R. L., and Paoli, T. L. (1986). Stripe-geometry Quantum Well Heterostructure Al,Ga, _,AsGaAs Lasers Defined by Defect Diffusion. Appl. Phys. Left. 49(9), 510-512.

7

ION IMPLANTATIONINTO QUANTUM-WELL STRUCTURES

253

Deppe, D. G., and Holonyak, N., Jr. (1988). Atom Diffusion and Impurity-Induced Layer Disordering in Quantum Well 111-V Semiconductor Heterostructures. J. Appl. Phys. 64(12), R93-R113. Deppe, D. G., Holonyak, N., Jr., Kish, F. A,, and Baker, J. E. (1987). Background Doping Dependence of Silicon Diffusion in p-Type GaAs. Appl. Phys. Lett. 50(15), 998-1000. Forbes, D. V., Coleman, J. J., Klatt, J. L., and Averback, R. S. (1995). Temperature Dependence of Ion-beam Mixing in 111-V Semiconductors. J. Appl. Phys. 77(7), 3543-3545. Gillin, W. P., Rao, S. S., Bradley, I. V., Homewood, K. P., Smith, A. D., and Briggs, A. T. R. (1993). Vacancy Controlled Interdiffusion of the Group V Sublattice in Strained InGaAs/ InGaAsP Quantum Wells. Appl. Phys. Left. 63(6), 797-799. Hamoudi, A,, Ougazzaden, A,, Krauz, Ph., Rao, E. V. K., Juhel, M., and Thibierge, H. (1995). Cation Interdiffusion in InGaAsP/InGaAsP Multiple Quantum Wells with Constant P/As Ratio. Appl. Phys. Lett. 60(6), 718-720. Haynes, T. E., and Holland, 0. W. (1991b). Dose Rate Effects on Damage Formation in Ion-Implanted Gallium Arsenide. Nucl. Instrum. Meth. Phys. Res. B5910, 1028- 1031. Haynes, T. E., and Holland, 0. W. (1991b). Dose Rate Effects on Damage Formation in Ion-Implanted Gallium Arsenide. Nucl. Instrum. Meth. Phys. Res. B59/H60,1028- 103 1. Herman, M. A,, Bimberg, D., and Christen, J. (1991). Heterointerfaces in Quantum Wells and Epitaxial Growth Processes: Evaluation by Luminescence Techniques. J. Appl. Phys. 70, RlLR52. Hirayama, Y., Suzuki, Y.,and Okamoto, H. (1986). Compositional Disordering and Very-Fine Lateral Definition of GaAs-AIGaAs Superlattices by Focused Ga Ion Beams. Surface Sci. 174, 98- 104. Hirayama, Y., Tarucha, S., Suzuki, Y., and Okamoto, H. (1988). Fabrication of a GaAs Quantum-Well-Wire Structure by Ga Focused-Ion-Beam Implantation and its Optical Properties. Phys. Rev. B 37(5), 2174-2777. Jackman, T. E., Charbonneau, S., Allard, L. B., Williams, R. L., Templeton, I. M., Buchanan, M., Vos, M., Mitchell, I. V., and Jackman, J. A. (1991). Compositional Disordering of Strained InGaAs/GaAs Quantum Wells by Au Implantation: Channeling Effects. Appl. Phys. Lett. 59, 27-29. Kahen, K. B., Peterson, D. L., Rajeswaran, G., and Lawrence, D. J. (1989). Properties of Ga Vacancies in AlGaAs Materials. Appl. Phys. Lett. 55(7), 651-653. Kalish, R., Feldman, L. C., Jacobson, D. C., Weir, B. E., Merz, J. L., Kramer, L.-Y., Doughty, K., Stone, S., and Lau, K.-K. (1993). Implantation Induced Changes in Quantum Well Structures. Nucl. Instrum. Meth. Phys. Res. B80I81, 129-733. Kalish, R., Kramer, L.-Y., Law, K.-K., Merz, J. L., Feldman, L. C., Jacobson, D. C., and Weir, B. E. (1992). Local Intermixing of GaAs/GaAlAs Quantum Structures by Individual Ion Implant Tracks. Appl. Phys. Lett. 61 (21), 2589-2591. Kelly, M. J. (1990). Low-dimensional Devices: Future Prospects. Semicond. Sci. Technol. 5, 1209-1214. Kelly, M. J., and Nicholas, R. J. (1985). The Physics of Quantum Well Structures. Rep. Progr. Phys. 48, 1699-1741. Kieslich, A,, Straka, J., and Forchel, A. (1992). Optical Study of Ar' Implantation-Induced Damage in GaAs/GaAlAs Heterostructures. J. Appl. Phys. 72(12), 6014-6016. Kirillov, D., Merz, J. L., Dapkus, P. D., and Coleman, J. J. (1984). Laser Beam Heating and Transformation of a GaAs-AIAs Multiple-quantum-well Structure. J. Appl. Phys. 55(4), 1105-1109.

254

R. KALISHAND S. CHARBONNEAU

Koteles, E. S., Elman, B., Holmstrom, R. P., Melman, P., Chi, J. Y., Xin Wen, Powers, J., Owens, D., Charbonneau, S., and Thewalt, M. L. W. (1989). Modification of the Shapes of GaAs/AIGaAs Quantum Wells Using Rapid Thermal Annealing. Superlattices Microstructures 5(3), 321-325. Koteles, E. S., Masum Choudhury, A. N. M., Levy, A., Elman, B., Melman, P., Koza, M. A., and Bhat, R. (1992). Quantum Well Shape Modification in Quaternary Quantum Wells. Materials Res. SOC.Symp. Proc. 240, 171-176. Kumar, M., De Brabander, G. N., Chen, P., Boyd, J. T., Steckl, A. J., Choo, A. G., Jackson, H. E., Burnham, R. D., and Smith, S. C. (1993). Optical Channel Waveguides in AlGaAs Multiple Quantum Well Structures Formed by Focused Ion Beam Induced Compositional Mixing. Materials Res. SOC.Symp. Proc. 281, 313-318. Laruelle, F., Bagchi, A., Tsuchiya, M., Merz, J., and Petroff, P. M. (1990). Focused ion Beam Channeling Effects and Ultimate Sizes of GaAIAs/GaAs Nanostructures. Appl. Phys. Lett. 56(16), 1561- 1563. Laruelle, F., Hu, Y. P., Simes, R., Robinson, W., Merz, J., and Petroff, P. M. (1990). Optical Study of GaAs/GaAIAs Quantum Structures Processed by High Energy Focused ion Beam Implantation. Surface Sci. 228, 306-309. Leier, H., Forchel, A., Horcher, G., Hommel, J., Bayer, S., and Rothfritz, H. (1990). Mass and Dose Dependence of Ion-implantation-induced intermixing of GaAs/GaAlAs Quantumwell Structures. J. Appl. Phys. 67(4), 1805-1813. Marsh, J. H. (1993). Quantum Well Intermixing. Semiconductor Sci. Technol. 8, 1136- 1155. Millar, R. C., and Kleinman, D. A. (1985). Excitons in GaAs Quantum Wells. J. Luminescence 30, 520-540. Myers, D. R., Dawson, L. R., Biefeld, R. M., Arnold, G. W., Hills, C. R.,and Doyle, B. L. (1988). Ion-implantation Damage and Annealing Effects in (InCa)As/GaAs Strained Layer Semiconductor Systems. Superlattices Microstructers 4(4/5), 585-589. Ourmazd, A. (1993). Mapping the Composition of Materials at the Atomic Level Materials Sci. Reports 9(6), 201-250. Poole, P. J., Charbonneau, S., Aers, G . C., Jackman, T. E., Buchanan, M., Dion, M., Goldberg, R. D., and Mitchell, I. V. (1995). Defect Diffusion in Ion Implanted AlGaAs and I n P Consequences for Quantum Well intermixing. J. Appl. Phys. 78(4), 2367-2371. Poole, P. J., Piva, P. G., Buchanan, M., Aers, G. C., Roth, A. P., Dion, M., Wasilewski, 2. R., Koteles, E. S., Charbonneau, S., and Beauvais, J . (1994). The Enhancement of Quantum Well Intermixing Through Repeated Ion Implantation. Semicond. Sci. Technol. 9, 21342137. Ralston, J. D., Schaff, W. J., Bow, D. P., and Eastman, L. F. (1989). Room-temperature Exciton Electroabsorption in Partially intermixed GaAs/AlGaAs Quantum Well Waveguides. Appl. Phys. Lett. 54(6), 534-536. Shi, S., Li Kam Wa, P., Miller, A., Pamulapatia, J., Cooke, P., and Mitra Dutta (1994). The Controlled Disordering of Quantum Wells using Surface Oxidation. Semicond. Sci. Technol. 9, 1564-1566. Tan, T. Y., and Gysele, W.(1987). Destruction Mechanism of 111-VCompound Well Structures Due to Impurity Diffusion. J. Appl. Phys. 61(5), 1841-1845. Tan, H. H., Jagadish, C., Williams, J. S., Zou, J., Cockayne, D. J. H., and Sikorski, A. (1995). Ion Damage Buildup and Amorphization Processes in AI,Ga, -,As. J. Appl. Phys. 77(1), 87-94. Vieu, C., Schneider, M., Launois, H., and Descouts, B. (1992). Damage Generation and Annealing in Ga' implanted GaAs/(Ga, AI)As Quantum Wells. J. Appl. Phys. 71(10), 4833-4842.

7

ION IMPLANTATIONINTO QUANTUM-WELL STRUCTURES

255

Vieu, C., Schneider, M., Planel, R., Launois, H.. Descouts, B., and Gao, Y. (1991). Mixing of GaAs/(Ga, AI)As Interfaces by G a + Implantation. J. Appl. Phys. 70(3), 1433-1443. Xia, W., Pappert, S. A,, Zhu, B., Clawson, A. R., Yu, P. K. L., Lau, S. S., Poker, D. B., White, C. W., and Schwarz, S. A. (1992). Ion Mixing of IIJ-V Compound Semicoductor Layered Structures. J. Appl. Phys. 71(6), 2602-2610. Yu, S. J., Asahi, H., Emura, S., Gonda, S., and Nakashima, K. (1991). Raman Scattering Study of Thermal Interdiffusion in InGaAs/lnP Superlattice Structures. J. Appl. Phys. 70(1), 204-208. Zucker, J. E., Tell, B., Jones, K. L., Divino, M. D., Brown-Goebeler, K. F., Joyner, C. H., Miller, B. I., and Young, M. G. (1992). Large Blueshifting of InGaAs/InP Quantum-Well Band Gaps by Ion Implantation. Appl. Phyx. Lett. 60(24), 3036-3038.

This Page Intentionally Left Blank

SEMICONDUCTORS AND SEMIMETALS, VOL. 46

CHAPTER 8

Ion Implantation and Thermal Annealing of 111-V Compound Semiconducting Systems: Some Problems of 111-V Narrow Gap Semiconductors Alexandre M. Myasnikov SEMICONDUCTORS PHYSICS ACADEMYOF SCIENCES OF RUSSIA INSTITUTE Of

NOVosmlRSK,

RUSSIA

Nikolay N. Gerasimenko JON INSTINTE foR SEMICONDUCTOR RESEARCH ZELENOORAD, Moscow. RUSSIA

1. INTRODUCTION . . , . . . , . . . . . , . . . . . . . . . . . . 11. MATERIALS AND IMPURITIES . . . . , . . . , . . . . . . . . . . . 1. Materials . . . . . , . . . . . . . . , . . . . . . . . . . . 2. Impurities . . . . . . . . . . . . . . . . . . . . . . . . . . 111. ION IMPLANTATION. . . . . . . . . . . . . . . . . . . . . . . . 1. Range Statistics . . . . . . . . . . . . . . . , . . . . . . . . 2. Ion Implantation Damage . . . . , . . . . . . , . . . . . . . . 3. ion implantation Damage in indium Arsenide . . . . . . . . . . . 4. Ion Implantation Damage in Indium Antimonide and Gallium Antimonide 5. Swellingofindium Antimonideund G a h m Antimonide . . . . . .

. . 1. Damage Annealing . .

. . . . . . 2. CappingLayers . . . . , . 3. Capless Annealing . . . . .

. . . . . . . . . . . . . . . . . . , . . . . . . . . . . . . . . . . . . . . . 4. Redistribution of impurities During Annealing . . . CONCLUSION . . . . . . . . . . . . . . . . . . References . . , . . . . . . . . . . . . . . . .

IV. ANNEALING.

v.

. . .

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

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

. . . . . . . .

. . . .

,

.

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

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

251 258 258 260 261 261 264 266 271 212 219 279 219 281 281 290 291

I. Introduction As is known, after silicon (Si), gallium arsenide (GaAs) is the most important semiconductor material for the manufacture of electronic devices. The unique properties of GaAs and also of the other 111-V semiconductors 257 Copyright 0 1997 by Academic Press All rights of reproduction in any form reserved. 0080-8784/97 $25

258

A. M. MYASNIKOV AND N. N. GERASIMENKO

make them superior materials for many applications, particularly in the areas of optoelectronics and microwave devices. The advantages of 111-V semiconductors are due primarily to the characteristics of their band structure. In comparison with Si, the bandgap of 111-V semiconductors can be either wider (GaAs) or narrower (GaSb, indium arsenide (InAs), and indium antimonide (InSb)). However, unlike Si, electronic band-to-band transitions of 111-V semiconductors are direct. For this reason, 111-V semiconductors are suitable for devices such as light-emitting diodes, lasers, and high-efficiency solar cells. 111-V semiconductors also have a high electron mobility that arises from the steep curvature of the conduction band, thus making them attractive for applications requiring low power and high speed. Contrary to GaAs ion implantations studies, less research has been performed on 111-V narrow gap semiconductors. However, recently great interest has been shown in these materials that have been the object of several scientific conferences and meetings. This chapter is a review of the achievements in ion implantation and annealing of 111-V semiconductors and describes the main problems of narrow gap semiconductor compounds such as GaSb, InAs, and InSb.

11. Materials and Impurities 1. MATERIALS

The properties of 111-V narrow gap semiconductors such as InAs, GaSb, and InSb in comparison with GaAs are summarized in Table I. GaAs and other 111-V semiconductor compounds provide processing and growth challenges not encountered in elemental semiconductors. At high temperatures, the vapor pressures of elements from column I11 (Ga and In) and from column V (As and Sb) over 111-V compounds are not equal, with As and Sb being the more volatile species. Thus in GaAs, both As and Ga begin to evolve at 450°C and the compound dissociates incongruently at 637°C (Robinson, 1992), far below its melting temperature and most processing temperatures. In other 111-V compound semiconductors, we have the same behavior of elements as of those from column I11 and V. In the process of heating, the atoms from the elements in column V begin to vaporize at temperatures far below the melting points of compounds and atoms from column I11 to remain in the wafer. As the temperature is increased, the pressure of the elements from column V also increases steadily. Because of the preferential

8 ION IMPLANTATION AND THERMAL ANNEALING

259

TABLE I PROPERTIES OF

FOUR SEMICONDUCTOR MATERIALS AT 300 K.

Property Crystal structure Lattice constant, A Atomic density, x loz2cm-* Atomic weight Density, g/cm3 Melting point, "C Free energy of formation, kJ/Kg.atom Dielectric constant Energy gap, eV Effective mass, m, Electrons Heavy holes Light holes Mobility, cmZ/Vs Electrons Holes Displacement energy, eV Column I11 atom Column V atom

GaAs

InAs

GaSb

InSb

ZB

ZB

5.653 4.42 144.79 5.32 1238 35.29 11.1 1.43

6.058 3.60 189.91 5.66 943 19.22 11.7 0.36

ZB 6.095 1.76 191.65 5.6 712 28.39 14.0 0.7

6.4789 2.94 236.77 5.78 525 12.89 17.7 0.18

0.07 1 0.5 0.12

0.023 0.3 0.025

0.047 0.39 -

ZB

0.012 0.5 0.015

8500 400

23000 200

5000 1000

100000 1700

8.8 10.1

6.7 8.5

-

5.8 6.8

-

GaAs, gallium arsenide; InAs, indium arsenide; GaSb, gallium antimonide; InSb, indium antimonide: ZB, zinc-blende. Data from Baransky P. I., Klochkov, V. P., and Potykevich, 1. V. (1975) and Ryssel, H., and Ruge, I. (1986).

loss of V atoms, the vacancy concentration in the 111 sublattice is directly proportional to the vapor pressure of V atoms; the vacancy concentration in the V sublattice is inversely proportional to the pressure of the V atoms. For this reason, wafers of 111-V compound semiconductors must be either capped or annealed in an ambient of V atoms during high-temperature processing to maintain stoichiometry. At 1 atm pressure, solid 111-V semiconductors decompose directly into the vapor phase before melting. To prevent this, ingots of 111-V semiconductors must be grown under high pressure. The most common growth technique is liquid encapsulated Czochralski in which a boron oxide liquid encapsulates the ingot. The ambient pressure, impurity content, 111-V semiconductor stoichiometry, and crystal diameter must all be closely controlled during growth to obtain device quality material. In recent years, careful attention to these variables has led to considerable improvement in the quality of boules. However, nonuniformities in dislocation density, impurity content, and other properties still exist. Moreover, the background impurities in

260

A. M. MYASNIKOV A N D N. N. GERASiMENKO

111-V compounds are well within the intrinsic carrier concentration at room temperature, meaning that the Fermi level is always controlled by extrinsic factors. These problems can lead to a degradation of desirable properties, such as electron mobility, and also to nonuniform behavior of produced devices. Other problems associated with 111-V semiconductor compounds include the lack of a stable native oxide and compensation of p-type dopants. The lack of a stable native oxide means that dielectric materials must be deposited instead of grown, complicating the process. Activation of p-type dopants is a major problem of 111-V narrow gap semiconductors because defects compensate implanted p-type impurities. GaAs, InAs, GaSb, and InSb also present some unique material problems that arise simply from the fact that these are compound semiconductors. There are six primary point defects in 111-V compounds, four more than in Si. These are vacancies and interstitials of atoms of column 111, vacancies and interstitials of atoms of column V, and two antisite defects. These defects also can form complexes, such as 111-V vacancy pairs. The point-defect chemistry of 111-V compounds thus is much more complicated than that of Si, and provides considerable challenges when attempting to model phenomena that involve point-defect interactions. Thus 111-V semiconductor compounds are valuable materials for optoelectronic and high-frequency devices. Serious problems remain to be resolved, however, before they can evolve from their “niche” and make a broader contribution in all areas of electronic devices. To achieve this, the cost-effectiveness of the processing must be improved, and the material and electrical properties must be controlled in a uniform and reproducible manner.

2. IMPURITIES In 111-V compound semiconductors doping with impurity atoms was used for the formation of mobile carriers. Table I1 presents behaviors of some elements in I11 and V sublattices of 111-V compounds. In the 111 sublattice the elements from column I1 are acceptors and doping of the 111-V compound with beryllium (Be), magnesium (Mg), zinc (Zn), and cadmium (Cd) forms p-type material. The elements from column VJ located in sites of the V sublattice are donors and these dopants form n-type material. The elements of column IV such as carbon (C),Si, germanium (Ge), and tin(Sn), when they are in the 111 sublattice, produce donors and form n-type conduction. When these elements are located in the V sublattice, p-type conduction material is created.

8 ION IMPLANTATION AND THERMAL ANNEALING

261

TABLE 11 BEHAVIORS OF SOME ELEMENTS FROM COLUMNS 11, IV, AND VI I N 111 AND V SUBLATTICES OF III-V COMPOUNDS

Acceptor I1

111

Donor Acceptor JV

V

Donor VI

'N

4Be 0.975

0.88

6C 0.77

0.7

0.678

12Mg

1 3 ~ 1

I4Si

'5P

1.3

1.26

1.17

1.1

16S 1.04

30Zn

3'Ga

34Se

1.26

'*Ge 1.22

33As

1.31

1.18

1.14

48Cd 1.48

491n

'OSn

"Sb

"Te

1.44

1.4

1.36

1.32

5B

beryllium; 'B, boron; 6C, carbon; 'N, nitrogen; '0, oxygen; "Mg, magnesium; 'Al, aluminium; %, silicon; "P, phosphorus; I6S,sulfur; "Zn, zinc; "Ga, gallium; "Ge, germanium; "As, arsenicum; '4Se, selenium; 48Cd, cadmium; "'In, indium; 50Sn, tin; 51Sb,antimony; 5ZTe,tellurium. The covalent radius is given in A(ngstrem) under each elements. Reprinted from Baransky, P. I., Klochkov, V. P., Potykevich, I. V. (1975). Semiconductor Electronics. Handbook on Properties of Materials. Naukova Dunka, Kiev; with permission.

111. Ion Implantation 1. RANGESTATISTICS Of central importance to the discussion of ion implantation processing of semiconductors is consideration of the impurity range distributions that can be obtained by the implantation process. The penetration of ions into a solid is governed by an elastic interaction with the target electron system. The depth distribution of the implanted ion is a function of the ion energy and mass, and of the atomic number of both the ion and the target atoms. According to the classic theory of Lindhard, Scharff, Schiartt (1963), (LSS), the range of an impurity implanted into amorphous targets is given by a symmetric Gaussian distribution (see theoretic Chapters 2 and 3 of Volume 45). Improvements in the range statistics have mostly been achieved through refinements in ion stopping power (e.g., in Ziegler, Biersack, and Littmark (1985)). Unfortunately, most simulations of implantation assume an amorphous target. This is so because determination of the electronic potential as a function of position in a crystalline lattice is very difficult. For this reason, channeling effects, which can cause considerable deviation from Gaussian distribution, cannot yet be simulated using a first principles approach.

262

A. M. MYASNIKOV AND N. N. GERASIMENKO

Channeled profiles can be simulated using empiric methods, which use higher moment distribution, such as Pearson IV. Unfortunately these methods do not provide details on implant damage. Detailed discussion on the theory of how ions are stopped in solids is given in the literature (e.g., LSS theory, and Ryssel and Ruge (1986)). Ions lose energy in solids in two ways, by nuclear interactions and electronic interactions. Nuclear interactions can be described as elastic binary collisions that cause angular deflections of both the incident ion and the target atom. These interactions can be modeled quite well using classical mechanics. Conversely, electronic interactions are inelastic collective energy exchanges caused by electron excitation and ionization. These interactions are more difficult to model. The type of stopping that an ion will undergo depends on the energy of the ion. At high energies, electronic stopping dominates; at low energies, nuclear stopping dominates. The critical energy E, is defined as the crossover between the two processes and is a function of both the ion and target masses. High-energy light ions primary lose energy to electronic interactions as they first move through the crystal. When their energy has been reduced below E,, they begin to undergo nuclear collisions. Low-energy heavy ions undergo nuclear collisions from the time they first enter the crystal. Since the majority of damage is caused by nuclear interactions, light ions tend to localize their damage just before R,, where most nuclear collisions occur. Heavy ions distribute their damage more uniformly from the surface to R,. In the widely used TRIM (The TRansport of Ions in Matter) computer program, (Ziegler, 1994) the penetration of ions into solids is calculated using a Monte Carlo method. This program has been developed gradually and in the recent version the ions may have energies from 10 eV to 2 GeV. The program may accept complex targets made up of compound materials, with up to three layers made up of different materials. It will calculate both the final distribution of the ions and also all kinetic phenomena associated with the energy loss of the ion: target damage, ionization, and phonon production. All target atom cascades in the target are followed in detail, and the redistribution of these target atoms is determined. This program results from the original work of Biersack and Haggmark (1980) on range algorithms Ziegler, Biersack, and Littmark (1985). We used this program to calculate range statistics for III-V narrow gap semiconductors such as InAs and InSb Khryaschev et al. (1994). After simulation a comparison of Be and Mg profiles produced by ion implantation and calculated by TRIM was carried out. In that work, InAs and InSb were irradiated with Be' with a dose of 1 x lo" cm-' and with Mg' ions with doses of 1 x 1014 to 10 x 10'4cm-2. Depth distribution of Be' and Mg' was investigated by secondary ion mass spectrometry (SIMS). TRIM

8 ION IMPLANTATIONAND THERMAL ANNEALING

263

TABLE 111 A COMPARISON OF THE STATISTICAL COEFFICIENTS OF EXPERIMENTAL AND CALCULATED PROFILES Ion/Material Be/InAs Mg/lnAs Be/InSb Mg/InSb

R,, Pm 0.5154 0.5311 0.2726 0.2597 0.6060 0.5589 0.3321" 0.2729

wn

Y

a

Method

0.2123 0.1604 0.1375 0.1115 0.2569 0.1790 0.1822* 0.1247

-0.149 -0.683 0.641 -0.04 1 0.011 -0.569 0.356" 0.075

2.71 3.10 4.51 2.49 2.59 2.87 3.28" 2.39

SIMS TRIM SIMS TRIM SIMS TRIM SIMS TRIM

AR,,

Be, beryllium; InAs, indium arsenide; Mg, magnesium; InSb, indium antimonide; SIMS, secondary ion mass spectroscopy; TRIM, The TRansport of Ions in Matter. "At a dose of cm-z

simulations of ion-implanted profiles were carried out for lo4 ions. Profiles were calculated at densities of InAs and of InSb as 5.7 and 5.8 g/cm3, respectively. The statistical coefficients such as the mean projected range R,, straggling A R p , skewness y, and kurtosis p for the SIMS depth profiles of the as-implanted samples in comparison with TRIM simulations are given in Table 111. The implanted profiles of Be' and Mg' in InAs have been found to be in relatively good agreement with theoretic simulations. The experimental deviation of AR, was 25% and 19% in the case of Be' and Mg' ions, respectively. Thus Be' and Mg' distributions in InAs could be approximated by Pearson IV functions for SIMS and TRIM profiles. Table I11 also presents the statistical coefficients for Be' and Mg' distributions in InSb. However, comparison of these is not correct because of the different shapes of the profiles. The distinction between experimental and calculated Mg ' profiles in InSb could be caused by swelling (see 55 of Part 111) of the irradiated layer and by the change in its stoichiometry. An increase in the dose from 1 x l O I 4 to 5 x 1014cm-2 results in a growth in concentration at the main maximum of 0.354pm and at the additional maximum of 0.185pm and appearance of a profile bend at 0.474 ,urn. Further increase in the dose up to 1 x 1015cm-2 resulted in a growth in concentration once more at a depth of 0.474 pm, a main maximum, and at a depth of 0.643 pm, appearance of the profile bend. Using SIMS, Pearton et al. (1988) studied Mg and Si profiles in InAs and GaSb at an energy of 100 keV and with doses ranging from 5 x l O I 3 to

264

A. M. MYASNIKOV AND N. N. GERASIMENKO

10''

-

l

.

'

*

'

0

l

*

1

*

2

x, p m

FIG. 1. Magnesium (Mg) profiles in indium arsenide (InAs) before and after annealing at 700°C for 10 sec. N , impurity concentration; D, dose; E , energy; x,depth. (Reprinted from Pearton, S. J., von Neida, A. E., Brown, J. M., Short, K. T., Oster, L. J., and Chakrabarti, U. K. (1988). Ion Implantation Damage and Annealing in InAs, GaSb, and Gap. J. Appl. Phys. 64,629-636;with permission.)

1 x lOI5 crnp2.These profiles for unannealed and rapid thermal annealing (RTA) wafers are shown in Figs. 1 to 4.

2. IONIMPLANTATION DAMAGE Ion implantation damage in semiconductors is either simple point defects, such as vacancies and interstitials, or complexes of point defects of lattice and impurities. These kinds of damage are the cause of many of the levels within the forbidden gap and exert an effect on the electrical properties of implanted layers. In 111-V narrow gap semiconductors, the forbidden gaps at 300K are from 0.18 to 0.7 eV, This means that up to the middle of the forbidden gap all energy levels are shallow and may produce carriers in the conduction and valence zones. Thus ion implantation damage in 111-V narrow gap semiconductors could be either donor or acceptor damage. Moreover, in 111-V narrow gap semiconductors the process of defect formation is complicated (unlike in Si), due to the existence of I11 and V

8 ION IMPLANTATION AND

THERMAL ANNEALING

265

Si : InAs D = 10~~cm-~ E = 100 keV as-implanted --- 7OO0C, 10"

loza 19

-

10

101(

x, p m FIG. 2. Silicon (Si) profiles in InAs before and after annealing at 700°C for 10 sec. N, impurity concentration; D, dose; E, energy. (Reprinted from Pearton, S . J., von Neida, A. E., Brown, J. M., Short, K. T., Oster, L. .Iand ., Chakrabarti, U. K. (1988). Ion Implantation Damage and Annealing in InAs, GaSb, and Gap. J . Appl. Phys. 64,629-636; with permission.)

1o'O

Mg :GrSb D= Em'* E = 100 keV

10''

-as-im

lrnted

--- 650 '8,

10"

1

10"

0

1

x. p m

FIG. 3. Magnesium (Mg) profiles in gallium antimonide (GaSb) before and after annealing at 650°C for 10 sec. N, impurity concentration; D,dose; E, energy. (Reprinted from Pearton, S. J., von Neida, A. E., Brown, J. M., Short, K. T., Oster, L. J., and Chakrabarti, U. K. (1988). Ion Implantation Damage and Annealing in InAs, GaSb, and Gap. J. Appl. Phys. 64, 629-636; with permission.)

266

A. M. MYASNIKOV AND N. N. GERASIMENKO

loao

'a

10''

Si :GaSb D= ern-' E=100 kfV -as-im ante --6OO"< 30" * * * 600"C, 120"

0

d 10"

0

1

x, p.Lm

2

Frc. 4. Silicon (Si) profiles in gallium antimonide (GaSb) before and after annealing at 650°C for 30 and 120 sec. N, impurity concentration; NS, sheet electron concentration; T, temperature; D, dose; E, energy. (Reprinted from Pearton, S. J., von Neida, A. E., Brown, J. M., Short, K. T., Oster, L. J., and Chakrabarti, U. K. (1988). Ion Implantation Damage and Annealing in InAs, GaSb, and Gap. J . Appl. Phys. 64, 629-636; with permission.)

sublattices and also low displacement energies (see Table I). In InAs and InSb the displacement energies of I11 and V atoms are one third the Si displacement energy. The low threshold of ion implantation damage causes the emergence of many defects. The TRIM simulation of Be and Mg ion implantation in InAs and InSb with 200keV energy has shown that the amount of vacancies per ion creates 1300 vacancies per ion for Be ion implantation of InAs, and 7000 vacancies per ion for Mg ion implantation of InSb (Khryashchev, 1994). The behavior of radiation damage in 111-V narrow gap semiconductors during and after ion implantation is especially interesting in relation to the condition of the implanted impurity. The properties of ion-implanted layers in these semiconductors are very different and are discussed separately for each material herein.

3. IONIMPLANTATION DAMAGE IN INDIUM ARSENIDE The main peculiarity of p-InAs is the conversion of the conduction type due to the effect of ion implantation damage. Thus Gerasimenko et al. (1988) studied the electrical properties of n-layers induced by argon (Ar) ion

8 ION IMPLANTATION AND THERMAL ANNEALING

267

implantation of p-InAs with E = 250 keV at Tmp= 20°C with doses ranging = 350°C with doses ranging from from 1 x 1013 to 1 x 10'6cm-2 and qmp 1 x 1014 to 3 x 1015 cm-' by using the van der Pauw technique. Conductivity type and sheet carrier concentration n, were measured at 77 K. For samples implanted at 20°C, n, has a weak dependence on the implanted dose. In fact, n, varied from 4 x 1014 to 1 x 1015cm-2 by changing the dose from 1 x 1013to 1 x 10l6 ern-'. Curve 1 of Fig. 5 shows the dependence of sheet electron concentration n, versus annealing temperature T at a dose of 5 x IOl4 cm-'. The same temperature dependence was observed for all other doses. Samples remained as n-type for all temperatures studied. The sheet carrier concentration n, reached a maximum value of 2 x I O I 4 for 5 x 10'4cm-2 at a temperature of 350°C for all doses. The layers formed by implantation at 350°C had approximately the same value of n,. The results of our investigation showed that the donor concentration ( N , > 10" ~ m - remained ~ ) high up to an annealing temperature of 650°C. Our data were in accordance with the results of Akimchenko et al. (1980) who used high doses and low energies of Cd ions in order to reveal the acceptor properties of Cd in n-1nAs and to promote a high level of volume impurity concentration for compensation of donor centers induced by implantation. Conversion of a p-type semiconductor to an n-type semiconductor by ion implantation is typical for narrow gap material such as lnSb (Bogatyrev and Kachurin, 1977a). Usually these phenomena are explained by either preferential formation of donor defects in implanted layers or by reactions of defects with amphoteric impurities resulting in a change in the sublattice occupation (Blaut-Blachev, 1980). Here an additional model for conduction conversion was proposed as the essential feature, of which a comparable concentration of donor and acceptor states induced by irradiation in the forbidden gap takes exists. In this case it is known that spatial fluctuations of the potential arise and create deep (about the forbidden gap) tails in the state density for electron and holes. When the effective masses of the charge carriers are comparable and the deep tails of the state density are localized (e.g., in amorphous Si), then a semiconductor becomes an insulator. If m, << mh, then holes are localized at deep states. Due to their small effective masses, electrons are more weakly bound than are holes and can migrate under the action of external fields, inducing n-type conduction in a compensated material. Local disordered nonstoichiometric regions may also be the reason for fluctuations of the potential. An argument in favor of p-n conversion being due to large fluctuations of the potential and not to the preferential generation of donor defects, is the

268

A. M. MYASNIKOV AND N. N. GERASIMENKO

100

300 500 T, "C

FIG. 5. The sheet carrier concentration in a layer of indium arsenide (InAs) irradiated with argon (1) and sulfur (2) ions as a function of annealing temperature. Ion implantation with a dose of 5 x l O I 4 cm-', an energy of 250 keV, and a temperature of 77K. n,, sheet carrier concentration.

absence of clearly expressed stages on the curve of the isochronous annealing (see curve 1 of Fig. 5). An additional argument in favor of the proposed model comes from the results of optical measurements close to the fundamental absorption edge of InAs at hv < E,. These measurements show (Akimchenko et al., 1979) the existence of tails in the state density, which is typical of a disordered crystal, while a shift of the absorption edge into the short-wave band according to the Moss-Burstein effect is observed for heavily doped semiconductors.Thus the contact between the disordered layer and the crystals of InAs may be interpreted as a semimetal-semiconductor barrier but not as a p-n junction. Furthermore, it was shown that n-type conversion due to radiation damage was also possible at sulfur ion implantation and took trouble to select the contribution of doping impurity in the formation of an n-type conduction layer. Gerasimenko et a!. (1991) compared the electrical properties of n-type conduction layers obtained in InAs by S + and Ar+ (which have almost the same mass) ion implantation. The comparison was carried out in order to select the contribution of radiation defects and doping impurity. The conditions of sample preparation and layer measurements were analogous to those in previous works Gerasimenko et al. (1989). The main differencewas that InAs wafers were divided into two groups for implantation: (a) implantation with Ar' ions and (b) implantation with S + ions. The results of the investigation showed that S f ion implantation, as well as that of Ar', led to conversion of the p-type semiconductor to an n-type semiconductor. The sheet carrier concentration in these implanted layers had no dependence on the chemical origin of ions and ranged from 5 x 1OI3

8 ION IMPLANTATION A N D THERMAL ANNEALING

269

to 5 x 1014cm-2for all doses. At an annealing temperature greater than 200"C, a transformation of a set of defects induced by irradiation and a change in the sheet carrier concentration (curve 2 of Fig. 5 ) took place. At an annealing temperature of about 300°C there was a stage that was characterized by an increase in sheet carrier concentration. This stage was for both implanted ions. Some defects were likely to dissociate at 400°C annealing and the chemical donors (sulfur) were shown at this stage. The sheet carrier concentration in InAs layers doped by sulfur was higher than the one in layers irradiated by Ar ions that contained only the radiation defects. The difference between the sheet carrier concentrations reached a value of about 1.4x 10'4cm-2 (at a dose of 5 x 1014cm-2) after 350°C annealing and did not change significantly at higher temperatures. It is logical to correlate this difference with the dopant activation. An increase in the dose of up to 1 x 10I6cmL2(by more than one order of magnitude) resulted in a slight increase in the difference between the sheet carrier concentrations in layers of 3 x 1014 to 4 x 1014cm-2. The sulfur efficiency has been reduced from 0.3 to 0.04 at a dose of 5 x 1014 to 1 x 1016cm-2 (Table IV). It should be noted that the contribution of the electrically active sulfur was separated only at doses over 5 x 10'4cm-2. Therefore, we cannot declare that sulfur efficiency tends to unity for lower doses, even though the high limit of the sulfur solubility in InAs (> 1019~ m - (Schillmann, ~ ) 1956) suggests that this might be so. A comparison of the profile of the charge carrier concentration and the profile of implanted sulfur in InAs carried out by Koltsov et al. (1983) showed that sulfur efficiency did not exceed 0.1 at a dose of 5 x 1014cm-2. In this case the efficiency was estimated without subtracting the contribution from the radiation defects. Our data show that the highest contribution of the electrically active sulfur occurs at a dose of 5 x 10'4cm-2. At lower doses conduction was caused by radiation disorder of the lattice, but at the higher

TABLE IV THE

DIFFERENCE N, BETWEENTHE SHEET CARRIER CONCENTRATIONS IN LAYERSIMPLANTED WITH SULFURAND ARGONAND SULFUREFFICIENCY K FOR DEFERENT DOSES D.

D,cm-2

N,, cm-*

K = NJD

5 3

4 x 1014 3 1015 3.8 1014

0.3 0.1 0.04

1014

1015

loL6

T"".= 400°C.

270

A. M. MYASNIKOV AND N. N. GERASIMENKO

doses the sulfur efficiency decreased and a larger fraction of sulfur formed defect complexes. Conversion of conduction in p-type InAs induced by hydrogen plasma treatment and proton implantation was also found by Polyakov et al. (1993). It was shown that both treatments can produce an n-type layer at the surface of p-InAs. For the hydrogen plasma treatment the effect was explained by hydrogen donors forming complexes with the Be and Zn acceptors and rendering them electrically inactive, and thus uncompensated. In proton-implanted samples the p-n conversion is due to a creation of donor-type lattice defect. Like Polyakov et al. (1993), Koltsov et al. (1995) observed conversion of p-InAs at proton implantation. These authors found the effect of proton bombardment on the galvanomagnetic properties of p-InAs. It was shown that implantation of protons leads to the appearance of donor type defects in the thin surface layer and allows control of the properties of the inversion layer in p-InAs at low temperatures. Using Rutherford backscattering spectroscopy (RBS) and transmission electron microscopy (TEM), Pearton et al. (1988) investigated the characteristics of ion-implantation-induced damage in InAs and its removal by Rapid Thermal Annealing (RTA). These authors found that there is relatively poor regrowth of this material if it is amorphized during implantation, leaving significant densities of dislocation loops and microtwins. For implant doses below the amorphization threshold, RTA produces good recovery of the lattice disorder, while backscattering yields results similar to unimplanted materials. An interesting study of defects induced by ion implantation in InAs was carried out by Wendler et al. (1995). They investigated high-energy implantation of 2-MeV As’ ions into (100) InAs using RBS. The ion dose varied from 2 x lOI3 to 5 x 10l6cm-’. The relative defect concentration npd, calculated from the backscattering spectra with the help of the computer code DICADA2, (Wendler et al., 1995) was compared with the distribution of the electronic and the nuclear energy density obtained from TRIM simulations. Comparing the measured defect profiles npd with the distribution of the electronic and the nuclear deposited energy densities, it was found that at low ion doses ( < 3 x 1013cm- ’) npdwas in good agreement with the profiles of the primary nuclear energy deposition. In the region of low electronic energy deposition the defect concentration increases continuously with increasing dose. For high doses of ions (> 1014cm-2) the defect concentration in the near-surface region, where a large amount of energy per unit volume is deposited into electronic processes, is low in comparison with the nuclear energy deposition. This led to the conclusion that the defect mobility and the defect annealing during room temperature implantation are strongly influenced by the total energy density deposited into electronic processes.

8 ION IMPLANTATION AND THERMAL ANNEALING

271

4. IONIMPLANTATION DAMAGE IN INDIUM ANTIMONIDEAND GALLIUM ANTIMONIDE Unlike GaSb, behavior of ion implantation damage in InSb has been extensively investigated and the defect properties can be separated into those induced from low-dose implantation, on the one hand, and high-dose implantation, on the other. The former will be dealt with in this section, whereas the latter will be discussed in 95 of Part 111. Guseva et al. (1976) presented the results of the investigation of conduction, concentration, and mobility of carriers in ion-implanted layers formed by bombardment of InSb with S + ions at 40 keV energy at 280°C. It was shown that the conversion of conduction type takes place at irradiation doses higher than 3 x 1014cm-'. In ion-implanted layers there is a chemical doping effect with sulfur concentrations at about the solubility limit of sulfur in InSb. Bogatyrev and Kachurin (1977a) studied the electrical properties of InSb bombarded with ions of different masses with energies of 60 to 300 keV and annealed between 20 and 400°C. It was shown that the n-layers, which become stable from 100 to 150"C, were only formed after proton irradiation at room temperature. Irradiation of InSb under the same conditions with middle mass ions and He ions led to the creation of p-type layers with high hole concentration. Korshunov et al. (1978) carried out a comparison of results of bombardment of InSb, irradiated with the ions of inert gases (argon (Ar) and xenon(Xe)), of donor (S and tellurium (Te)), and acceptor (Zn, Cd) impurities at energies of 32 to 80 keV. It was found that the radiation defects formed during bombardment at room temperature did not lead to the conversion type of p-InSb doped with manganese (Mn) and iron (Fe), unlike p-InSb doped with germanium. It was established that p-InSb irradiated with Ar ions has the same stages and activation energy of isochronous annealing as does p-InSb irradiated with neutrons. In another work Korshunov, Mirkin, and Tikhonov (1979) investigated the isothermal annealing (with hot implantation) of p- and n-type InAs irradiated with mid-range energy ions by layer etching with Hall and resistivity measurements. It was established that the change of initial impurity composition (e.g., the interchange of Ge by Mn in p-InSb) allowed an increase in temperature of hot implantation from 280°C to 320°C and an increase in electron mobility in inversion layers. Blaut-Blachev et al. (1980) studied the p-n conversion of the thermal treatment of monocrystals of InSb irradiated with neon (Ne), Ar, and krypton (Kr) ions. The initial material was doped with In, Zn, and Ge. It was found that irradiation of InSb in a temperature range of 250 to 300°C

272

A. M. MYASNIKOV AND N. N. GERASIMENKO

or at room temperature with postimplantation annealing at 250 to 300°C resulted in the formation of n-type layers only on samples doped with Ge at concentrations over 1 x 10'4cm-3. It also was concluded that after ion implantation with thermal treatment in InSb there were structure reorganizations that changed the electrical site of the amphoteric impurity Ge, and that this process was the cause of p-n conversion. Further investigation performed by Lezheyko, Lyubopytova, and Obodnikov (1982) showed that n-layers formed by H: ion implantation in p-InSb are of a different nature. The first type of layer occurred immediately after ion implantation and was related to the defect complexes of hydrogen, whereas the second type of layer arose during annealing at 300 to 350°C due to ion-implantationinduced defects. All these results show that after ion implantation of InSb there is damage that needs defect annealing in order to transfer the impurity atoms in the lattice sites and to recover the lattice disorder. However, Trokhin et at. (1988) found that Be atoms were located in sites of the In sublattice immediately after ion implantation of n-InSb and conversion of conduction type was only found after annealing of the radiation damage, thus compensating the electrical activity of ion-implanted Be. The rearrangement of Be atoms in the In sublattice was promoted by the large amount of In vacancies induced during ion implantation, by the high mobility of nonequilibrium point defects in InSb, and by the small ion radius of Be atoms.

5. SWELLING OF INDIUM ANTIMONIDE AND GALLIUM

ANTIMONIDE For ion implantation of semiconductor compounds there appeared problems that were unknown in the case of ion implantation of simple semiconductors such as Ge and Si. During ion implantation of semiconductor compounds, stoichiometric disturbances may occur (Christel and Gibbons, 1981). In terms of free energy formation of compounds, the dissociation of compound having the least free energy is the most probable. Free energy formation of compounds varies (see Table I) from 12.89 kJ/Kg.atom (InSb), 19.22 kJ/Kg.atom (GaSb), 28.39 kJ/Kg.atom (InAs), to 35.29 kJ/Kg.atom (GaAs). In reality it was found that ion implantation in InSb (Danilov, Popov, and Tulovchikov, 1978), GaSb (Callic et al., 1991), and in some cases GaAs (Whan and Arnold, 1970) was caused by swelling of the implanted layer. The process of swelling was clearly visible after ion implantation as a step between the implanted and unimplanted regions and as powder on the implanted surface. Danilov and Tulovchikov (1980) studied swelling of InSb

8 ION IMPLANTATION AND THERMAL ANNEALING

273

at Mg, Zn, Ar, and Ne ion implantation and established that the step height between implanted and nonimplanted regions is a function of ion energy and dose. In a range of doses D from 1 x l O I 4 to 2 x 10’’ cm-2 and in a range of energies E from 40 to 200 keV, there is a dependence of step height

where A is the coefficient that depends on the ion mass, increasing with an increase in the mass, and n is approximately 1. Detailed investigations of ion-implanted layers in InSb showed that the layers were very perturbed and did not anneal up to the melting point of InSb (Destefanis and Gailliard, 1980). The bombarded area has a spongelike appearance that penetrates below the initial level of the surface. The structure of the spongy area shown by the transmission electron microscopy (TEM) corresponds to an ion-implanted InSb sample and shows voids of several hundred angstroms. The void formation cannot be understood by the classic model of nuclear collision displacement but should include a cooperative mechanism: The energy loss of one ion induces a thermal spike, which may lead to a local overpressure. When this overpressure is higher than the elastic limit, the material yields plastically, leaving a void of volume that can be calculated. For InSb, it was found that there is good agreement between the experimental and theoretic values of the cavity volumes. By using TEM, Maksimov, Pitirimova, and Pavlov (1982) found that formation of porous structures is the most typical feature of the defect formation process in InSb during the implantation of ions of average mass at room temperature. Irradiation is likely to be accompanied by a partial loss of more volatile atoms of Sb, by the accumulation of redundant In atoms. Supersaturation of the activated volume with vacancies generates pores, which are displaced in the electron photomicrographs. The redundant In atoms are carried to the surface of the pores and canals and can react with the residual gases of the acceleration chamber, thus forming a film of InO, which is registered in the electronograms. Thus ion implantation is accompanied by the destruction of irradiated materials and by formation of a two-phase mixture of InSb and In, which is then oxidized. Reduction in the dose and mass of impurities leads to a smaller concentration of redundant vacancies. The porous structure is not formed, and in the volume vacancy loops are accumulated. The redundant In atoms seem to be responsible for the loops of interstitial type. Accumulation of vacancies and pores in the activated volume leads to its swelling and such a bend of the heterogeneous system at which the activated layer substrate has the irradiated surface on the convex side. The presence of I n 0 on the surface of pores, dividing the InSb into blocks may, at least

214

A. M. MYASNIKOV AND N. N. GERASIMENKO

partially, hinder the crystalline structure recovery during annealing. An increase in the target temperature contributes to the increase of the diffusion mobility of vacancies and, probably, of the redundant In atoms. As a result, the flow of In atoms into the implantation surface or their penetration into the depth of the crystal becomes possible. Local supersaturation with vacancies and deviation from stoichiometry are not observed, pores are not formed, and the activated volume remains single crystalline. It is, however, natural that maintaining a single crystal leads to stronger elastic strains, which earlier were compensated for by the formation of a two-phase pore volume. An increase in the diffusion mobility of point defects with temperture apparently causes development of a dislocation structure in the depth, substantially exceeding the average projection range. In 1984, Alberts used proton and a-particle channeling to study the radiation damage caused by the implantation of Mg' ions at 160keV energy with doses of 5 x 1013to 10l6cm-2. It was established that at lower doses (up to 1 x 1014cm-2) no well-defined damage peak is observable and the damage produced during the implantation process is shown only as an enhanced dechanneling in the aligned spectra, indicating that the atoms were displaced from their crystal lattice sites to distances exceeding the Thomas-Fermi screening radius. At doses exceeding 1 x 10'4cm-2, a welldefined damage peak appears, extending deeper into the crystal at higher doses. The damage peak almost reached the random level after a dose of 1 x 10'5cm-2, starting right at the surface. Dechanneling decreases with energy of a-particles according to the indication that interstitial atoms are mainly responsible for the dechanneling, which is supported by the tendency of InSb to form polycrystals because of the existence of a great number of In and Sb in a nonstoichiometric state. The isochronous annealing behavior of room temperature implantation shows that the annealing takes place in two phases, the first starting at 200°C and the second at about 300°C. Complete recrystallization of damaged crystals does not occur even at temperatures just below the melting point. This was the case for all crystals implanted at beam intensities of 0.05, 0.1, 0.5, and 1 ,uA/cm2 at a dose of 1 x 1015cm-2. The annealing behavior for 1 x 10l5cm-2 was shown to be a function of the implantation temperature. No annealing was reached up to 400°C followed by a very sharp annealing stage. This indicates that the polycrystalline phase only starts at 400°C and that there might be an increase in the mobility of vacancies, preventing the creation of complex and stable defect structures. At 500°C the crystals were not restored to their initial condition before implantation. The radiation damage was measured as a function of the implanted dose at room temperature. With increasing dose, the number of stable defect

8

ION

IMPLANTATIONAND THERMAL ANNEALING

275

structures increases due to an increase in the number of interstitial crystal nuclei in the paths of the mobile point defects. At doses between 1 x 1014 and 5 x 1014cm-2, an accumulated high level of point defects does not appear to be possible, resulting in an equilibrium between the rate at which point defects are created and the mobility of vacancies forming more complex and stable defects. At higher doses the individual disordered zones merge to form an amorphous layer. Also by using cc-particle channeling Stoyanova (1988) investigated the kinetics of accumulation of radiation damage in differently oriented InSb implanted with Mg ions. The conditions of implantation were the following: (a) implantation energies 40 to 150 keV; (b) implantation doses 3 x 1013 to 1 x l O I 5 cm-’; and (c) current densities 6 nA/cm’ to 0.5 pA/cm’. It was found that at a dose of 3 x 1014cm-’ the level of defects in crystals in the (100) plane was the maximum, whereas in the (110) plane it was the minimum. For the crystals with (100) and (111) orientations the simple structure defects were predominant (e.g., interstitial atoms, clusters of interstitial atoms, associates of simple defects with impurity atoms, and lattice defects). In (1 10) orientation crystals there were the predominate extended defects as dislocation type. It was established that there are three types of dependencies of the relative defectivity x on the current density of the ion j :

-

In the range of low currents (j< 20nA/cm2) there was dependence on x j o.8. The increase of defectivity that appeared with the increase in the rate of point-defect generation was caused by the reduction of its recombination time. As with the increase of j , the concentration of point defects was increased at time interval z and its recombination was reduced and the probability of formation of steady defect complexes was increased during the irradiation process. The second type of dependence (x Const) was in a range of current density from 15 to 240 nA/cm’. In this j range with increasing ion current density j , the recombination of primary defects was negligibly small and did not affect the processes of secondary formation of defects. The main role in the formation of defects was played by the process of formation of steady complexes and the amount of those reaching maximum value at a given dose and energy was constant in a wide range of current densities.

-

276

A. M. MYASNIKOV AND N. N. GERASIMENKO

The third type of dependence (x -jO.')was at doses higher than 2 x 1014 cm-2 and was the behavior of changing of defectivity in the sponge layer. In a range of doses from 3 x 1013to 2 x 1014CII-' the dose dependence of x was a function of the dose in 0.6 degrees at j < 15 nA/cm2 and in one degree in the case in which 15 nA/cm2 < j < 240 nA/cm2. With the beginning of swelling in InSb the dose dependence of x was changed and transformed into a weak dependence on dose (x- DO.'). The energy dependence of x shows a change of degree from 1.5 at D = 6 x 1013cm-2 to 3 at D = 3 x 10'4cm-2. Thus the dependence of x on energy E, dose D, and current density j, may be presented as

where 1 is 0.8, 0, and 0.1 at small j, at middle j , and swelling, respectively; m is 0.6, 1, and 0.1 at small j , middle j, and swelling, respectively. The n degree has a dose dependence and changes from 1.5 to 3. In a study of GaSb, Pearton et al. (1988) took into consideration the fact that the behavior of GaSb during ion implantation and annealing has the same peculiarity as does InSb. The regrowth process after amorphization of the near-surface region was unsuccessful. The improvement with annealing was not sufficient and the surface region had either inclusion of atoms lighter than Sb or loss of Sb. These authors connected this phenomenon with the inclusion of oxygen from the atmosphere into the heavily disordered region near the surface and swelling of GaSb. In 1991, Callec et al. (1991)investigated Ne', Mg', Si', S', Ar', and selenium (Se') ion implantation of GaSb and observed anomalous elevations up to 6 pm of the ion-implanted GaSb surface. This swelling phenomenon was very similar to that of InSb described previously and was related to the formation of a porous layer. The step height between implanted and nonimplanted regions was dependent on the mass, energy, and dose of the implanted ions. These dependencies are shown in Figs. 6 and 7. Although a large amount of oxygen was measured in the porous layers, it is not likely that it is responsible for the swelling. Callec et al. (1993) studied the production and annealing of damage in GaSb implanted at room temperature with 150 keV Ar ions over a wide range of doses. This study was performed using RBS in combination with the channeling technique and TEM. Above the critical dose of 4 x 1013cm-2, the introduction of radiation damage induced a swelling of the implanted region. This phenomenon was related to the formation of a layer containing voids and microtwins and, at higher doses, porous polycrystalline GaSb. For doses below the swelling threshold, the RTA process produced good recovery of defects. Otherwise, for doses higher than the

8 ION IMPLANTATION AND THERMAL ANNEALING

277

FIG.6. Variation of the step height as a function of the dose for argon implantation at various energies. GaSb, gallium antimonide. (Reprinted from Callec, R., Favannec, P. N., Salvi, M., L'Haridon, H., and Gauneau, M. (1991). Anomalous Behavior of Ion-Implanted GaSb. A p p l . Phys. Lett. 59, 1872-1874; with permisson.)

FIG. 7. Variation of the step height as a function of the ion mass for 300 keV implantation of 3 x 1014 and 1 x 1015 cm-'. GaSb, gallium antimonide; Ce, cessium; Se, selenium; Ar, argon; s, sulfur; Si, silicon; Mg, magnesium; Ne, neon. (Reprinted from Callec, R., Favannec, P. N., Salvi, M., L'Haridon, H., and Gauneau, M. (1991). Anomalous Behavior of Ion-Implanted GaSb. A p p l . Phys. Lett. 59, 1872-1874; with permisson.)

A. M. MYASNIKOV AND N. N. GERASIMENKO

278

10'' 1

Critical dose, em-'

150 keV --> GaSb

10"

10''

Te

I so

10 " 0

50

100

1

200

Ion Mass, a.m.u. FIG. 8. TRIM (The TRansport of Ions in Matter) simulated variation of the critical dose as a function of the ion mass for 150 keV implantation. GaSb, gallium antimonide; a.m.u., arbitrary atom units. (Reprinted from Callec, R., Poudoulec, A,, Salvi, M., L'Haridon, H., Favannec, P. N., and Gauneau, M. (1993). Ion Implantation Damage and Annealing in GaSb. Nucl. Instrum. Meth. B80181,532-537; with permission.)

critical dose, the annealing process was less efficient. When the swelling was low, there was regrowth of the porous layer after annealing at 600°C but large voids remained; when the swelling was high, the regrowth procedure was unsuccessful. Calculation using TRIM has shown that the critical dose corresponds to a simulated number of displacements of atoms in the target, reaching 1 x 1022cm-3. The results of the simulations performed at an energy of 150keV are shown in Fig. 8. Gauneau et al. (1993) pointed out that oxygen alone does not play a major role in the process of swelling. The mechanism of the phenomenon is related to the damage created by the ion bombardment rather than to the presence of oxygen, which would modify the stoichiometry of GaSb after implantation. Although analogous study of InSb was not carried out, it seems that the role of oxygen in the phenomenon of InSb swelling is the same. However, Lin et al. (1995) investigated the production and annealing of implantation damage in GaSb implanted with 2-MeV nitrogen (N) ions and

8 ION IMPLANTATION AND THERMAL ANNEALING

279

doses from 1013 to lo1' cm-2 by using RBS and channeling, spectroscopic ellipsometry, and scanning electron microscopy (SEM). These authors established the idea that radiation damage increases as the implantation dose increases. No swelling phenomenon was observed on either of the implanted samples. The damaged layer induced by 2-MeV N ion implantation was buried and transformed into an amorphous layer at doses higher than 3 x 10'4cm-2. The damage resulting after implantation depends on the dose rate for a given ion dose. A good recovery of the damaged layer is obtained when the annealing temperature is 600°C for 25 sec by RTA. These unusual results of GaSb ion implantation could be related to the high energy and low mass of N ions.

IV. Annealing

1. DAMAGE ANNEALING Annealing of ion implantation damage and dopant activation is generally done with a thermal treatment as well as with laser beams (Bogatyrev and Kachurin, 1977b). Self-annealing can occur during ion implantation if the process is done at elevated temperatures or if the current density is high. Thermal annealing of 111-V narrow gap semiconductors takes place at temperatures from 300 to 800°C. Both furnace annealing and RTA are used. Annealing of 111-V narrow gap semiconductors occurs at lower temperatures than Si and GaAs due to lower melting temperatures (see Table I) and the high vapor pressure of column V atoms. The exact temperature depends on the impurity implanted, the amount of diffusion that can be tolerated, and the electrical properties required, such as mobility and concentration of carriers.

2. CAPPING LAYERS As pointed out in $1 of Part 11, in 111-V compound semiconductors at high temperatures there are problems due to the differences between the vapor pressures of elements from column I11 (Ga and In) and from column V (As and Sb). The encapsulant should not stress the substrate as a result of intrinsic stress as a result of differences in the thermal expansion coefficient. The ideal film for encapsulating 111-V compounds for annealing should have the following properties (Duncan and Westphal, 1985):

280

A. M. MYASNIKOV AND N. N. GERASIMENKO

a. It should be capable of deposition at temperatures below that of the incongruent evaporation of the substrate. b. Interdiffusion or chemical reaction between substrate or implanted atoms and encapsulant should not take place during encapsulant deposition or annealing. c. It should be stable and adherent at room temperature and elevated temperatures. d. The encapsulant should not stress the substrate as a result of stress intrinsic in the form or as a result of differences in the thermal expansion coefficient. The technique that has been generally used for thin-film dielectric encapsulation of GaSb, InAs, and InSb is chemical vapor deposition (CVD). Of importance to the quality of an encapsulant is surface preparation. The details of surface preparation are particularly important for III-V compounds due to the formation of native oxide, the thickness and composition of which are strongly dependent on the ambient oxidizing conditions. Currently, there exist no universally accepted encapsulants for III-V narrow gap semiconductors. In the case of InSb encapsulation, insulating layers such as sandwich structures were used. Thus layers of SiO, (1500 A pyrolytic SiO, deposition at 340°C) were used by Foyt, Lindley, and Donnelley (1970). Wei et al. (1980) also used pyrolytic SiO, deposited in a CVD reactor at temperatures ranging from 100 to 200°C. Thom et al. (1980) evaluated several dielectric materials used as the gate insulator in InSb MOS (Metal Qxide Semiconductors) structures, including silicon monoxide (SiO), Al,O,, SiO,N,, In,O,, anodically grown native oxide of InSb, TiO,, and SiO,. From all the layers, SiO, deposited by low-temperature CVD at 220°C were utilized in the InSb arrays. A sandwich structure of Si0,-Si,N, was successfully applied by Hurwitz and Donnelly (1975), as early as 1975 for the fabrication of planar InSb photodiodes. An anode oxide-Al,O, layer was employed by Fujisada and Kawada (1985) for the formation of ion-implanted InSb p-n junctions. In 1980, InSb was annealed in a vacuum at encapsulation of the surface with a sandwich structure of anode oxide and SiO, (Blaut-Blachev et al., 1980). For InAs encapsulation, insulating layers such as sandwich structures were also applied but the temperature of deposition was higher. Akimchenko et al. (1979) used SiO, when before annealing of implanted S + or Mg' a capping layer was deposited at 350°C. Different capping layers were deposited at 200°C in a CVD reactor by Gerasimenko et al. (1992): A SiO, layer for annealing of the Mg-implanted layer; a sandwich structure of anode oxide-SiO, (Gerasimenko et al., 1989)

8 ION IMPLANTATIONAND THERMAL ANNEALING

281

for production of n’p junctions by S f ion implantation or of SiO, + Si,N, (Gerasimenko et al., 1988) for the formation of n’p junctions by Ar ion implantation. A sandwich structure of anode oxide ’and Si,N, was also employed by Astakhov et al. (1992) for the formation of planar InAs diodes.

3. CAPLESS ANNEALING In parallel with annealing using encapsulants there are works in which capless annealing was carried out. Thus a study of behavior of column IV atoms ion implanted (and annealed in a vacuum at 600°C) in InSb was accomplished by Guseva et al. (1974). Similarly the activation of impurities was carried out in a vacuum by Korshunov et al. (1978). In addition in some works the “hot” implantation was used with ion implantation at high temperature (up to 350°C) where the recovery of lattice damage occurred simultaneously. For InSb the hot implantation was conducted at 100 to 350°C by Bogatyrev, Kachurin, and Smirnov (1978), at 280 to 320°C by Korshunov et al. (1979), and at 280°C by Guseva et al. (1976) 350°C hot implantation of InAs was also used by Gerasimenko et al. (1988) for the production of an n-layer induced with Ar ions. The benefits of RTA have resulted in a great deal of research activity in the activation of both n-type and p-type dopants in ion-implanted III-V semiconductors (Williams and Pearton, 1985). The technique appears to provide electrical properties at least as good as those achievable by conventional furnace heating and has other more specific advantages over furnace heating, which are derived from the shorter heating time of RTA. With the advance of RTA, it is now possible to carry out annealing without using capped layers and without degradation of III-V narrow gap semiconductors. Thus, Pearton et al. (1988) annealed InAs and GaSb ion-implanted layers between 400 and 700°C for 10 to 120sec. This annealing was found to be able to recover implantation damages. In addition, capless annealing of GaSb by using RTA was carried out by Callec et al. (1993) at 400 to 600°C for 15 sec and by Lin et al. (1995) at the same temperatures for 25 sec. 4. REDISTRIBUTION OF IMPURITIES DURINGANNEALING

Of fundamental importance is the redistribution of impurity during annealing. We investigated (Gerasimenko et al., 1992, 1996) the behavior of Beryllium (Be) and Mg ion implanted and annealed in InAs and InSb in the process of annealing. Thus Mg in InAs was investigated by Gerasimenko et al. (1992). InAs wafers were implanted by Mg at 200 keV with a dose of 1 x 10’’ ern-,. The

A. M.MYASNIKOV AND N. N. GERASIMENKO

282

Mg:InAs D=lO’%m-* E=200 keV fi

m99..

3S0°C, 30’

“1

‘El0

for SOO°C, 30’

-calculated

x, p m FIG.9. Magnesium (Mg) profiles formed in indium arsenide (InAs) by ion implantation and annealing for 30 min. T,,,,, temperature; t,,,,, time. D, dose; E, energy; N, impurity concentration; x, depth.

samples were then annealed at temperatures ranging from 250 to 600°C in a nitrogen ambient for 30 min. Depth and area distributions of Mg were investigated by SIMS. The experimental data and the calculated curve at an annealing temperature of 500°C for a period of 30 min are shown in Fig. 9. The Mg profile was calculated with a Mg diffusivity

D = Do exp

(- E ) AE

where T is the temperature, k is the Boltzmann constant, AE = 1.17 eV, and Do = 1.98 x cm2/sec (Schillmann, 1956). Our measurements showed that the Mg profile had no changes up to 350°C annealing. At annealing temperatures higher than 350°C there was a redistribution of the Mg profile. The Mg profiles had surface peaks and long “tails” at a depth of some micrometers. The concentration in the surface peak and in the tail region increased with increasing temperature. The comparison of calculated and experimental profiles demonstrated a considerable discrepancy in the depth of Mg penetration in InAs.

8 ION IMPLANTATION AND THERMAL ANNEALING

283

w2

'E 0 d

lo'*

10"

10

0

10

20

x, p m FIG. 10. Magnesium (Mg) profiles formed in indium arsenide (InAs) by ion implantation time. D, dose; E, energy; N, impurity and annealing for 30 min. Tan,., temperature; t, concentration; x, depth.

The Mg profiles in samples annealed at temperatures of 400 and 600°C and measured at high sputtering speed about 3 nm/sec are also shown in Fig. 10. These data enable us to estimate the Mg diffusivity in the tail region, assuming that diffusion occurred by two fluxes. This assumption was made because the experimental profiles had exponential tails [N(x) exp( -ax)], while theoretic profiles should show a dependence N ( x ) exp( - ax') due to diffusion by one flux. Thus a simple assumption was the diffusion of Mg by two independent fluxes (each with its proper parameters), and the concentration in the tail region could be

--

where L , = (4D,t)1'2, L, = (4D,t)'12, D , = Do, exp(-DE,/kT), D , = Do, exp( - DE,/k T). By adjusting the parameters N , , N , , L,, and L, at 400 and 600°C, N(x) was fitted to the experimental data, and values of L , (400"C), 4 (40O0C),L , (600"C), and L, (600°C) were found. These values were used to calculate the diffusivity of Mg at the tail region, giving AE, = 0.13eV, D o , = cm2/sec. cm2/sec, BE, = 0.18eV, Do, = 4.0 x 1.0 x To avoid the effect of mass interference in the Mg yield from singly charged polyatomic ions and multiply charged monoatomic ions of impuri-

284

A. M. MYASNIKOV AND N. N. GERASIMENKO

I , ,

24

,

.

!

25

'

I

I

26

'

MIISS,a.m.u. FIG.11. Mass spectra of secondary ions measured at depths of 0.1 (1) and 25mm (2). a.m.u., arbitrary atom units.

ties or the substrate, implantation with 24Mg+,2 5 Mg', and 26Mg' ions at natural abundance was carried out. The natural abundance is 79% for 24Mg', 10% for "Mg', and 11% for 26Mg+ isotopes. Spectra of 24Mg+, 25Mg+,and 26Mg+secondary ions measured at the samples surface and at a depth of 25pm are shown in Fig. 11. As seen in Fig. 11, the natural abundance is constant at the beginning and at the end of profile. Thus we must conclude that it is the Mg that diffuses and the other impurities of samples had no effect on our measurements. By the area distributions measured at a depth of 2pm in the tail region, it was shown that there was an area inhomogenity of Mg. The number of local regions enriched with Mg per unit of area is 103cm-2. This value is on the order of the dislocation density in InAs. Thus the assumption can be made that in the tail region the dislocations were decorated by Mg. As a test of this assumption, selective dislocation etching was used to compare the Mg pattern in an area distribution. It is found that the local regions enriched with Mg coincided with the dislocation etch pits, but part of the dislocation etch pits was without Mg. Thus Mg was located at dislocations and its penetration through the sample was connected with the fast volume diffusion by the interstitial mechanism. Depending on the origin the dislocation inclusions may or may not be decorated by Mg. Magnesium penetration through the sample along dislocations with exits in the surface region could not be excluded. Then dislocations without exits in the surface region would be without Mg.

8 ION IMPLANTATIONAND THERMAL ANNEALING

205

Near the surface the Mg profiles of samples annealed at temperatures of 400 to 6 0 0 T had surface peaks, and this effect could be connected with the Mg decoration of structural defects near the surface and in the range of maximal energy losses for implanted ions. By defect decorating the Mg was kept out of diffusion flux resulting from upgoing diffusion. This Mg constriction also was demonstrated at RTA of ion-implanted samples of InAs (Pearton et al., 1988) (see Fig. 1). Gerasimenko et al. (1996) studied the behavior of Be in InAs and InSb and Mg in InSb. InAs and InSb wafers were ion implanted at an energy of 200keV. Doses of implanted ions were 1 x 101'cm-' for Be and 5 x 10'4cm-2 for Mg. Temperatures of annealing were below 0.85 x Tmelt, where Tmeltwas the melting temperature of 525°C for InSb and 943°C for InAs. The samples were then annealed at temperatures ranging from 300 to 450°C for InSb and from 300 to 800°C for InAs in a nitrogen ambient. Figures 12 to 14 show SIMS profiles of Mg ion-implanted and annealed at 350, 400 and 45OoC, respectively. As shown in Fig. 12, at 350°C annealing there was no profile broadening even for 240 min. The migration of Mg to the surface was observed only from the region near the maximum of concentration. An increase in annealing temperature up to 400°C induced (as shown in Figs. 12 and 13) essential redistribution of Mg: concentration was decreased

10

1 10 n

'El

10

Mg:InSb

D = 5x10" cm-' F = 7nn kpv

T-. = 350°C as implanted - - _ - tarn.= 30' tmn.= 60'

..... ..... tarn.= ..... t-.

120'

= 240'

i 10

10

x, ct m FIG. 12. Magnesium (Mg) profiles in indium antimonide (InSb). 350°C annealing. D, dose; E, energy; N , impurity concentration;x, depth-Ten,.,temperature; t,,,,, time.

A. M. MYASNIKOV AND N. N. GERASIMENKO

286

Mg:InSb 10

D = 5x10" cm-2 E = 200 keV Tan=. = 40yoC - as imD anted

=.

n

'8

10

lo'.:,,

'

'

*

.

'

'

'

1'

"

tun, = 120' t-. = 240'

, , ' 1 ' , .

,

, 2

x, pm FIG. 13. Magnesium (Mg) profiles in indium antimonide (InSb). 400°C annealing. D,dose: E, energy; N, impurity concentration;x, depth; T,,,,,temperature; t,,,,, time..

in the region at mean projected range R , = 0.36 pm and went on to increase near the surface. Annealing at 400°C for 60 min changed the shape of profile, which had three maxima: the main one near the surface up to a concentration of about 1 x 1020cm-3, and the two others at 0.32 pm and 0.48 pm with equal concentrations of about 8 x 10'8cm-3. Further increase in annealing time resulted in a shift of the maximum from 0.32pm to the surface and a decrease of its concentration even relative to maximum at 0.48pm. With an increase in annealing time, the impurity profile became deeper in comparison with the as-implanted profile, and broadening amounted to 0.2pm for 240min at a concentration level of 1 x 1017cm-3. As seen from Fig. 14, annealing at 450°C did not change the general trend of Mg behavior observed at 400°C; however, broadening amounted to 0.3 pm for 240 min at a concentration level of 1 x 10" cm-3. This unusual behavior of Mg was connected to the presence of swelling defect surface layers on InSb formed by Mg ion implantation, as demonstrated by Destefanis and Gailliard (1980). As was found, the defective surface layer was a mixture of amorphous and polycrystalline phases of InSb depleted of Sb and looked like a sponge, and monocrystalline structure could not be restored by annealing. Presence of this structural damage in InSb irradiated with Mg ions was also confirmed by TEM (Maksimov, Pitirimova, and Pavlov, 1992). In planar p-n junctions this defect surface

8 ION IMPLANTATION AND THERMAL ANNEALING

l o mh

R

I

4

10

MgrInSb

D

-

287

5 ~ 1 0 ’o ~m -2

E = 200 keV

T-.

= 450°C

as implanted _ _ _ - tmn.= 30’

..... = ..... t-. tmn.=

60’

120’

I E

d

10

10

U

1

x, prn

FIG. 14. Magnesium (Mg) profiles in indium antimonide (InSb). 450°C annealing. D, dose; E, energy; N , impurity concentration;x, depth; T,,,,, temperature; ran.,, time.

layer touching at the perimeter of the p-region on the substrate induced the additional current leakage (Hurwitz and Donnelly, 1975). In our case, the presence of a swelling surface layer on InSb formed by Mg ion implantation was also pointed out by a 30% decrease in Sb yield during profile measuring up to a depth of 0.25pm. During annealing at 400 to 450”C,the diffusivity of Mg in the surface layer of the defect was higher than in the monocrystalline region. This effect resulted in the distortion of the profile and the Mg migration resulted in the swelling layer to the Si0,-InSb interface, which acted as a sink. The lower value of diffusivity of Mg in monocrystalline InSb was evidenced by an increase in the depth profile L = 0.2pm at 400°C annealing and L = 0.3pm at 450°C annealing for t = 240 min. The estimation of diffusivity for Mg in monocrystalline InSb by the equation L = (Dt)’” gave values of diffusivities D = 5 x lo-’’ cm2/sec at 400°C and D = 10-14cm2/sec at 450°C.These values are very similar to the predicted values using the Mg diffusivity parameters Baransky, Klochkov, and Potykevich (1975) of AE = 1.17eV and Do= 4 x cm2/sec in InSb. Figure 15 shows Be profiles in InSb as-implanted and after 40°C annealing for 30 min. These two profiles almost coincide, and further increase in annealing time does not change the Be profile. Both profiles have two maxima: the sharp maximum at a depth of about 0.2pm, and the smoother one at a depth of about 0.65pm. The sharp

A. M. MYASNIKOV AND N. N. GERASIMENKO

288

10 xo

10

= 1OI6 om-' E = 200 keV

Be:InSb D

as implaated T,,,.=450 C, t,,,.=30'

19

Ql

'310

0

1

2

3

x, p m FIG. 15. Magnesium (Mg) profiles in indium antimonide (InSb). D, dose; E, energy; N , impurity concentration;x, depth; T,,,., temperature; tan,., time.

maximum was not connected with an error of ion implantation because the Be profile in InAs formed simultaneously with InSb. It is likely that formation of the additional maximum on the Be profile in InSb was connected with the swelling surface layer as at Mg ion implantation in InSb. Indirect confirmation of existence of this swelling layer in the case of Be ion implantation in InSb at an energy of 200 keV was the removal of the surface layer with a thickness of 0.4pm for the leakage suppression of planar p-n junctions (Hurwitz and Donnelly, 1975). In our case, when Be was implanted in InSb, absence of changes in profile could be connected with the 90% location of Be in In lattice sites that already exist during ion implantation and almost 100% activation of Be after 350°C annealing for 30 minutes (Trokhin et al., 1988). It was found by the channeling technique and the reaction of 'Be(a, n, y) "C that during annealing Be did not change its location and remained in In lattice sites. The behavior of Be in InAs is completely opposite from the behavior of Be in InSb. Figure 16 shows Be profiles in InAs for different annealing temperatures. It was observed that Be redistribution took place starting from 400°C as the lowest temperature of annealing. As seen in Fig. 16, the increase in annealing temperature from 400 to 700°C resulted in a severe distortion of the original profile: pronounced migration of Be to the Si0,-InAs interface as well as to the bulk of the sample. Beryllium

8 ION IMPLANTATIONAND THERMAL ANNEALING

10

"[

Be:InAs

289

D = 10'' cm-' E = 200 keV

10 n

g 10

-8OOOC

z" 10

10 FIG. 16. Magnesium (Mg) profiles in indium antimonide (InSb). D, dose; E, energy; N , impurity concentration; x, depth; T,,,,,temperature; t,,,,, time.

concentration in the region at mean projected range R, = 0.57 pm decreased with the increase in annealing temperature up to 700°C. However, an additional increase in annealing temperature to 750°C the maximum practically disappeared at R,, and the Be concentration in this region rose from 1018cm-3after 700°C annealing, to 3.6 x 1OI8cm-3 after 750°C annealing, and to 1.2 x 1019cm-3 after 800°C annealing. In addition, straggling of this maximum became narrower with an increase in temperature. The general trend of the formation of renewable maximum was proved as follows: If after 650°C annealing when the maximum in the region at mean projected range R, = 0.57 pm practically disappeared when an additional annealing at 800°C was carried out, then the Be concentration at a depth of 0.57pm increased from 1.1 x 1018cm-3 to 8.2 x 10I8~ m - ~ . This behavior of Be, when the concentration decreased in the region at mean projected range R, = 0.57 pm at relatively low temperatures, showed that annealing may be connected with the different locations of Be after ion implantation in InAs and InSb. If Be in InSb after ion implantation was already located in the In lattice sites, then Be in InAs could be in the interstitial site. In addition, in the process of ion implantation defects were formed in the region with maximal Be concentration. These defects did not interact with Be and did not anneal at temperatures lower than 700°C. Therefore, Be at temperatures up to 700°C behaved without peculiarities

290

A. M. MYASNIKOV AND N. N. GERASIMENKO

and migration of Be went on to the Si0,-InAs interface as well as to the bulk of the sample. With temperatures over 700°C these defects were formed of the gettering layer, which interacted with Be and caused its redistribution.

V. Conclusion The physical phenomena induced by ion implantation of impurities in 111-V narrow gap compound semiconductors such as InSb, InAs, and GaSb and postimplantation annealing, first of all, are connected with activation and depth redistribution of impurity atoms. During these processes point defects that in many cases are electrically active were in competition with impurities. The formation of n-layers on p-InAs and the conversion of n- and p-InSb distinguish the examples of this competition. During study of the diffusion of impurities implanted in InAs and InSb a number of general peculiarities were found: 1. In the case of Be and Mg in InSb, the ion-implanted impurities interact actively with point defects during irradiation before annealing and form defect-impurity complexes. On heating, these complexes either suppress the process of diffusion (as does Be in InSb) or allows broadening of the concentration profiles that is insignificant (as does Mg in InSb). 2. At the same time, in the case of Be and Mg in InAs there is the effect of anomalous diffusion. Such diffusion consists of either penetration of Mg along dislocations of up to tens of micrometers or the effect of self-gettering when Be returns to the region of the mean project range in the process of annealing. 3. Annealing of layers on InSb formed by ion implantation of amphoteric impurities results in a change in the occupation of I11 and V sublattices and conversion of the conduction type. This transition may be confused with defect activity. The effects of InSb and GaSb high-dose implantation, when surface swelling layers are produced, are also interesting. The swelling layers have spongelike structures that do not recover up to the melting point temperature. The most up-to-date information obtained and given in this chapter could not be synonymously understood because no detailed knowledge exists of structure defects in 111-V narrow gap semiconductors. A study of structure defects (e.g.. with the aid of HRTEM) may provide answers to many questions in the near future.

8 ION IMPLANTATIONAND THERMAL ANNEALING

291

ACKNOWLEDGMENTS This work was supported by Dr. V. I. Obodnikov, and without his help it would not have been possible. The authors are indebted to their colleagues at the Institute of Semiconductors Physics, especially Drs. G. A. Kachurin, G. S. Khryashchev, G. L. Kuryshev, A. A. Nesterov, and L. N. Safronov for their contributions to the study of the 111-V narrow gap semiconductors. Thanks are also expressed to Dr. N. A. Valisheva for his help in sample preparation and to Dr. B. A. Zaytsev, Mrs. S. I. Alysheva, and Mrs. L. N. Romashko for the preparation of this manuscript.

REFERENCES Akimchenko, I. P., Panshina, E. G., Tikhonova, 0. V., and Frimer, E. A. (1979). Photoelectical Properties of Indium Arsenide Implanted with S and Mg at 350 keV Energy. Sou. Phys. Semiconductors 13, 2210-2215. Akimchenko, 1. P., Panshina, E. G., Tikhonova, 0. V., and Frimer, E. A. (1980). Spectra of Photoelectical Force of lnAs Implanted with Cd ions. Kratkie Soobshcheniya Fizike 7,3-7. Alberts, H. W. (1984). A Channeling Study on Mg Implanted InSb Single Crystals. Materials Res. SOC.Symp. Proc. 27, 335-340. Astakhov, V. P., Danilov, Yu. A. Dudkin, V. F., Lesnikov, V. P., Sidordova, G . Yu., Suslov, L. A., Taubkin, I. I., and Eskin, Yu. M. (1992). Planar Diodes on InAs Material. Pis'ma Zhurnal Tekhnicheskoy Fiziki 18, 1-5. Baransky, P. I., Klochkov, V. P., Potykevich, 1. V. (1975). Semiconductor Electronics. Handbook on Properties of Materials. Naukova Dumka, Kiev. Biersack, J. P., and Haggmark, L. G. (1980). A Monte Carlo Computer Program for the Transport of Energetic Ions in Amorphous Targets. Nucl. Instrum. Meth. 174, 257-269. Blaut-Blachev, A. N., Gerasimenko, N. N., Lezheyko, L. V., Lyubopytova, E. V., and Obodnikov V. I. (1980). On the nature of p-n Conversion of InSb Crystals Irradiated with Ions. Soviet Phys. Semiconductors 14, 306-310. Bogatyrev, V. A,, and Kachurin, G. A. (1977a). Defect Annealing in Indium Antimonide after Ion Bombardment. Sou. Phys. Semiconductors 11, 1360-1363. Bogatyrev, V. A,, and Kachurin, G. A. (1977b). Formation of Low Resistance N-Layers on P-lnSb by Pulse Laser Irradiation. Sou. Phys. Semiconductors 11, 100- 102. Bogatyrev, V. A,, Kachurin, G . A., and Smirnov, L. S. (1978). Ion Implantation into Indium Antimonide at Elevated Temperatures. Sou. Phys. Semiconductors 12, 102-105. Callec, R., Favannec, P. N., Salvi, M., L'Haridon, H., and Gauneau, M. (1991). Anomalous Behavior of Ion-Implanted GaSb. Appl. Phys. Lett. 59, 1872- 1874. Callec, R., Poudoulec, A,, Salvi, M., L'Haridon, H., Favannec, P. N., and Gauneau, M. (1993). Ion Implantation Damage and Annealing in GaSb. Nucl. Instrum. Meth. B80/81,532-537. Christel, L. A,, and Gibbons, J. F. (1981). Stoichiometric Disturbances in Ion Implanted Compound Semiconductors. J. Appl. Phys. 52, 5050-5055. Danilov, Yu. A,, Popov, Yu. S., and Tulovchikov, V. S. (1978). Effect of Anomalous Radiation Disturbance of Indium Antimonide at Ion Implantation. In Interaction of Atomic Particles with Solids, Part 2, Minsk, USSR, 132-134.

292

A. M. MYASNIKOV AND N. N. GERASIMENKO

Danilov, Yu. A,, and Tulovchikov, V. S. (1980). Anomalous Radiation Disturbance of Indium Antimonide at Ion Implantation. Soviet Phys. Semiconductors 14, 197-200. Destefanis, G. L., and Gailliard, J. P. (1980). Very Efficient Void Formation in Ion Implanted InSb. Appl. Phys. Lett. 36, 40-42. Duncan, W. M., and Westphal, G. H. (1985), GaAs Material Preparation and Characterization. In VLSI Electronics Microstructure Science, Vol. 11, Academic Press, Orlando, California, Chapt. 2,41-86. Foyt, A. G., Lindley, W. T., and Donnelly, J. P. (1970). n-p Junction Photodetectors in InSb Fabricated by Proton Bombardment. Appl. Phys. Lett. 16, 335-337. Fujisada, H., and Kawada, M. (1985). Temperature Dependence of Reverse Current in Be Ion Implanted InSb p + n Junctions. Japan. .I Appl. . Phys. 24, L76-L78. Gauneau, M., Chaplain, R., Rupert, A,, Toudic, Y., Riviere, D., Callec, R., and Andre, E. (1993). Secondary Ion Mass Spectrometry Generates Swelling of GaSb Depth Resolution and Secondary Ion Yields. 1.Appl. Phys. 73,2051-2056. Gerasimenko, N. N., Khryashchev, G. S., Kuryshev, G. L., Myasnikov A. M., and Obodnikov, V. I. (1996). SIMS Study on Redistribution on Implanted Impurities in InSb and InAs during Post-Implantation Annealing. Nucl. Instrum. Meth., in press. Gerasimenko, N. N., Kuryshev, G. L., Myasnikov, A. M., Obodnikov, V. I., Safronov, L. N., and Khryaschev, G. S. (1991). The Electrical Properties of Planar n+-p Junctions in InAs Produced by S + Ion Implantation. Nucl. Instrum. Meth. B58, 187-190. Gerasimenko, N. N., Myasnikov, A. M., Nesterov, A. A., Obodnikov, V. I., Safronov, L. N., and Khryaschev, G. S. (1988). Conversion of Conduction Type in P-InAs Layers Irradiated with Ar' Ions. Sou. Phys. Semiconductors 22, 753-755. Gerasimenko, N. N., Myasnikov, A. M., Nesterov, A. A., Obodnikov, V. I., Safronov, L. N., and Khryaschev, G. S. (1989). Conversion of Type Conductivity in P-InAs Layers Induced by Ar+ Ion Implantation. Nucl. Instrum. Meth. B39,480-482. Gerasimenko, N. N., Myasnikov, A. M., Obodnikov, V. I., Plotnikov, V. L., Safronov, L. N., and Khryaschev, G. S. (1989). Doping of Indium Arsenide with Sulfur Ions. In Impurity and Dejects in Narrow Gap Semiconductors. Proceedings of 1 lth All-Union Seminar on Problems of Physics and Chemistry of Semiconductors, Part 2, Pavlodar, USSR, 43-47. Gerasimenko, N. N., Myasnikov, A. M., Obodnikov, V. I., and Safronov, L. N. (1992). Redistribution of Magnesium in InAs during Postimplantation Annealing. Sou. Phys. Semiconductors 26,926-927. Gibbons, J. F. Johnson, W. S., and Mylroie, S. E. (1975). Projected Range Statistics. Dowden, Hutchinson and Ross, Stroudsbourg, Pennsylvania. Guseva, M. I., Mansurova, A. N., Tikhonov, V. G., and Khorvat, S . N. (1976). A Study of Hall Effect in Sulfur Ion Implanted Indium Antimonide. Sou. Phys. Semiconductors 10, 14691471. Guseva, M. I., Zotova, N. V., Koval', A. V., and Nasledov, D. N. (1974). A Study of the Behavior of Column IV Elements in Indium Arsenide Doped by Ion Implantation. Sou. Phys. Semiconductors 8, 59-63. Hurwitz, C. E., and Donnelly, J. P. (1975). Planar InSb Photodiodes Fabricated by Be and Mg Ion Implantation. Solid State Electronics 18, 753-756. Khryashchev, G. S., Kuryshev, G. L., Myasnikov, A. M., and Obodnikov, V. I. (1994). Concentration Profiles of Be and Mg in InSb and InAs. Abstracts of 20th International Conference on Ion Implantation Technology. Catania, Italy. Koltsov, G. V., Krutenyuk, Yu. V., Bologov, S . V., Boriskina, L. V., Simonov, A. N., and Frimer, E. A. (1983). Distribution of Electrically Active Centers in Indium Arsenide after Ion Implantation and Annealing. Pouerkhnost' (Surface) 8, 118-122.

8 ION IMPLANTATIONAND THERMAL ANNEALING

293

Koltsov, G . I., Krutenyuk, Yu. V., and Skipetrov, E. P. (1995). Modification of Properties of the Inversion Layer in P-InAs(Zn) by Proton Implantation. Nucl. Instrum. Meth. B96,835-837. Korshunov, A. B., Kuznetsov, G. M., Makarov, A. G., Olenin, V. V., and Postnikov, I. V. (1978). Investigation of Isochronous Annealing of P-InSb Irradiated with Middle Energy Ions. Sou. Phys. Semiconductors 12, 938-943. Korshunov, A. B., Mirkin, L. I., and Tikhonov, V. G. (1979). Investigation of Isothermal Annealing of Indium Antimonide Irradiated with Middle Energy Ions. Sou. Phys. Semiconductors 13, 645-648. Lezheyko, L. V., Lyubopytova, E. V., and Obodnikov, V. I. (1982). Conversion of Conduction Type at Isochronous Annealing InSb Irradiated with Hydrogen Ions. Sou. Phys. Semiconductors. 16, 1638-1640. Lin, C., Qian, Y., He, Z., Zheng, Y., Jeynes, C., Wilson, R. J., Hemment, P. L. F., and Zou, S. (1995). Damage Behavior of MCV N + Ion Implanted GaSb. Nucl. Instrum. Meth. B96, 872-875. Lindhard, J., Scharff, M., and Schiatt, H. E. (1963). Range Concepts and Heavy Ion Ranges. Det Kongelige Danske Videnskabernes Selskab, Matematisk-Fysiske Meddelelser 33, N14. mova, E. A., and Pavlov, P. V. (1982). Formation of Defects in the Process of Ion Implantation into A3B5. Radiation Efects 66, 95-100. Pearton, S. J., von Neida, A. E., Brown, J. M., Short, K. T., Oster, L. J., and Chakrabarti, U. K. (1988). Ion Implantation Damage and Annealing in InAs, GaSb, and Gap. J. Appl. Phys. 64, 629-636. Polyakov, A. Y., Ye, M., Pearton, S. J., Wilson, R. G., Milnes, A. G., Stam, M., and Erickson, J. (1993). The Influence of Hydrogen Plasma Treatment and Proton Implantation on the Electrical Properties of InAs. J. Appl. Phys. 73, 2882-2887. Robinson H. G . (1992). Difusion ofp-Type Dopants in Gallium Arsenide. Ph.D. thesis, Stanford University, California. Ryssel H., and Ruge I. (1986). Ion Implantation. John Wiley & Sons, Chichester, UK. Schillmann, E. (1956). Ueber Einbau und Wirkung von Fremdstoffen in Indiumarsenid. Z . Naturforschung lla, 463-472. Stoyanova, I. G., Skakun, N. A., Svetashev, P. A., Trokhin, A. S., and Shestakov, A. V. (1988). Kinetics of the Accumulation of Radiation Damage in Indium Antimonide at Magnesium Ion Implantation. Poverkhnost’ (Surface) 3, 129- 134. Thom, R. D., Koch, T. L., Langan, J. D., and Parrish, W. J. (1980). A Fully Monolithic InSb Infrared CCD Array. IEEE Trans. Electron Devices ED-27, 160-169. Trokhin, P. A,, Skakun, N. A,, Stoyanova, I. G., Deev, A. S., Kamushkin, G. V., Oleynik, V. A,, I. G., and Svetashev, P. A. (1988). Localization of Beryllium Atoms in Crystalline Lattice of Indium Antimonide at Ion Implantation. Pouerkhnost’ (Surface) 8, 144- 146. Wei, C.-Y., Wang, K. L., Taft, E. A., Swab, J. M., Gibbons, M. D., Davern, W. E., and Brown, D. M. (1980). Technology Development for InSb Infrared Imagers. IEEE Trans. Electron Devices ED-27, 170- 175. Wendler, E., Wilson, R. J., Jeynes, C., Wesch, W., Gaertner, K., Gwilliam, R. M., and Sealy, B. J. (1995). 2 MeV As’ Implantation in InAs. Nucf. Instrum. Meth. 896, 298-301. Whan, R. E., and Arnold, G. W. (1970). Lattice Expansion and Strain in Ion Bombarded GaAs and Si. Appl. Phys. Lett. 17, 78-380. Williams, J. S., and Pearton, S. J. (1985). Rapid Annealing of GaAs and Related Compounds. Materials Res. SOC. Symposia Proc. 35, 427. Ziegler, J. F. (1994). TRIM’94 Manual. Yorktown, New York. Ziegler, J. F., Biersack, J. P., and Littmark, U. (1985). Stopping Power and Ranges of Ions in Solids. Pergamon Press, New York.

This Page Intentionally Left Blank

Index atomic diffusion, 220 atomic hydrogen, 86 average network parameter, 152

A absorptance spectrum, 156 absorption, 75 coefficient, 156 acceptor damage, 264 activation, 40 energy, 129, 130, 135, 147 AES: Auger electron spectroscopy, 223 ambibolar diffusivity, 117 amorphisation, 127, 160 amorphized layer, 165 amorphous silicon, 116 semiconductor, 67, 151 crystalline interface, 49 crystalline transition, 69 matrix, 172 phase, 285 angular frequency, 117 anisotropic effect, 3 annealing, 87, 88, 90, 92, 94, 115, 128, 217, 220,221, 223, 230 temperature, 40, 49, 63, 67, 129, 133, 166, 282 kinetics, 53, 116, 121 recovery mechanism, 129 damage, 152 time, 287 annihilation, 12, 174 mechanism, 122 of defects, 152 process, 39 anomalous diffusion, 290 Anti-Stokes, 78 antiquantum dot, 238 antisite defect, 260

B band tail state, 163, 167, 174 edge, 154, 167, 173 filling, 13, 142, 146 structure, 258 bandgap, 214 energy, 117 re-normafization, 95 shifted broad-area laser, 250 tuned, 251 beam-scattering, 237 bilayer form, 49 binding centers, 230 binding energy of excition, 76 bivacancy, 67 black absorber, 157 blue-shift, 234, 238, 251 bound excition, 76 bragg reflector laser, 251 Brillouin zone, 10, 165 C

calorimetric measurement, f 73 capacitance, 152 capless annealing, 257, 281 capping layer, 257, 279 carbon tetrachloride, 157 carrier mobility, 86, 151, 214 lifetime, 214 scattering, 68

295

INDEX Carrier (continued) concentration, 61, 63, 64 cathod-luminescence, 87 channeled profile, 262 channeling, 288 effect, 261 chemical imaging, 221, 232 mapping, 221 profile, 233 doping, 87, 271 donor, 269 CL: Cathodoluminescence, 216, 224, 238 classic dispersion theory, 41 cluster, 131 concentration of activated carriers, 55 communication system, 249 compensator, 5 complete recrystallization, 64 comp 1ex reflection coefficient, 3, 4 field coefficient, 3 dielectric function, 1, 9 reflectance ratio, 4 dielectric constant, 41 of point defects, 264 conduction band, 11,258 conduction zone, 264 conductive effective mobility, 53 mass, 56, 63 conductivity, 61, 152 contactless technique, 39 continuous random network, 152 critical concentration of defects, 161 dose, 49, 147, 234, 278 energy, 262 amorphization dose, 54, 86, 94 crystal with diamond structure, 78 crystallografic axis, 48 Cuprous oxide, 97 CVD: chemical vapor deposition, 280

D damage cascade, 219 nucleation, 152 annealing, 257, 279 damping, 13

danglin bond, 151 deep level traps, 136 defects, 86 defect, 151 diamagnetic, 153 engineering, 151 assisted, 168 diffusion, 241 impurity complex, 290 mobility, 270 formation, 273 assisted diffusion, 237 induced disordering, 215 degree of activation, 40 density of state, 11 depth resolution, 223 distribution, 262 di-interstitial, 152 diamagnetic, 162 dichroisrn, 159 dielectric constant, 77 diffracted electrons, 221 diffusion mobility, 274 digging, 223 diode lasers, 79, dislocation, 166, 284 density, 128, 259 structure, 274 disorder, 127, 139 disordered semiconductor, 67, 151 dissociation, 272 divacancy, 133, 152, 156, 159, 160, 162, 166, 167, 174 absorption, 161 concentration, 153 donor damage, 264 dopant beam, 238 activation, 279 dose rate dependent, 147, 248 double negatively charged, 160 beam method, 5 Drude effect, 137, 142, 146 dye lasers, 79

E effect of annealing, 166

INDEX effective mass, 63, 64, 67, 68, 226, 267 medium theory, 1, 9, 67 radius, 118 conductive mass, 67 electrical activation, 42, 65, 69, 126, 151 measurements, 65 conductivity, 152 property, 266, 268, 271, 281 activity, 86, 98 electrically active, 63, 131 inactive, 270 electromagnetic radiation, 153 theory, 3 wave, 3 electron charge, 56 difraction, 158 mobility, 258, 260 paramagnetic resonance, 97 beam lithography, 238 electronic transition, 170 diffusivity, 117, 136 property, 230 transport properties, 151 band-to-band transition, 258 device, 257 potential, 261 interaction, 262 electronograms, 273 electroreflectance measurement, 40 ellipsometer, 5 ellipsometric analysis, 1 angle, 7 ellipsometry, 1, 2, 4 elliptical polarizer, 6 emission spectra, 250 encapsulant, 280 Czochralski, 259 energy formation of compound, 272 conservation, 75 epitaxial growth, 248 EPR: electron paramagnetic resonance, 152, 159, 162

297

evaporation, 280 excite-and-probe, 82, 83 excitons, 76 extinction coefficient, 61 extrinsic semiconductor. 225

F Fano-type interaction, 101 Fermi energy position, 162 film sputtered, 167 front medium, 156 five-vacancy, 152 FIB: Focussed Ion Beam, 232, 248 fluorescence, 8 1 forbidden gap, 267 four-vacancy, 152 free-hole plasma, 41 free-carrier, 53, 63, 117 concentration, 57, 63, 65, 67 contribution, 146 density, 117 absorption, 55, 69 impurities concentration, 64 Fresnel equation, 8 fringe amplitude, 49 pattern, 49 FTIR spectrophotometer, 53, 79, 82 FTIR: Fourier transform infrared, 49, 53 fundamental absorption edge, 41, 268 furnace annealing, 279 heating, 281

G GaAs implanted with Sn, 101 gallium arsenide, 1 galvanometric property, 270 gap state, 154 GaSb implanted with Ga, 101, 102 germanium, 1 global amorphization, 135

H Hall mobility, 132 effect, 64 harmonic resonance. 41

298

INDEX

hetero-structure, 215, 233, 237, 248 hetero-epitaxial, 214, 216 high-fluence, 41 high-efficiencysolar cell, 258 high-frequency device, 260 high-temperature implantation, 248 higher-energy emission, 238 homogeneous, 49 HRTEM: High-resolution TEM, 221, 222, 223, 232 hydrogen plasma, 270 hydrogenated, 86 amorphous carbon, 154

I impedance bridge, 6 implantation, 221 energy, 49, 57, 126 induced defect, 174 angle, 126 dose, 41, 49, 129, 126, 158 implanted sputtered film, 174 layer thickness, 49 impurity, 40 impurity, 151, 257 range distribution, 261 induced heterostructure interdiffusion, 231 induced disordering, 215 inactive site, 131 incident power, 119 index of refraction, 137 indirect bandgap, 164 indirect optical transition, 164 indium phosphide, 1 induced damaged, 39 infrared spectroscopy, 40 infrasil quartz, 157 inhomogeneous material, 65 insulator, 267 interdiffusion, 216, 220, 221, 226, 280 length, 235, 236 process, 247, 248 intermixing, 237 interface state, 154 interference, 40,49 effect, 41, 46 fringe, 41, 49, 69 induced oscillations, 156

phenomena, 40 intermixing length, 236 intermixed radii, 234 interstitial, 131, 220, 247 mechanism, 284 defect, 264 vacancy pair, 248 intrinsic stress, 279 ion induced damage, 151 flux, 216, 248 penetration, 220, 237 track, 232 implantation damage, 257, 263 implanted structure, 290 beam-affected spot, 235 beam-assisted diffusion, 249 beam-induced mixture, 234 beam-induced damage, 217 implantation-enhanced interdiffusion, 215 source-beam optic, 238 isochronal annealing, 48, 126, 166, 168, 271, 274 isothermal annealing, 271 111-V narrow gap semiconductor, 257 111-V compound semiconducting system, 257 111-V hetero-structure, 238 Indirect absorption, 75 Indium Antimonide (InSb), 75 I n P Fe, 90 Integrated luminescence, 92 IR spectroscopy, 42 IR resonance, 41 J

junction depth, 63

K Kramer-Kronig, 5, 10, 13

L laser, 258 reflector, 250 beam-induced disordering, 215 annealing, 84, 94 lattice damage, 122

299

INDEX site, 272 deformation, 77 lifetime, 146 light velocity in vacuum, 56 light emitting diode, 258 line dislocations, 133 LMA: Law of Mass Action, 129 LO phonons, 78 local bandgaps, 226 long-range order, 10, 151 loop dislocations, 133 low-loss waveguide, 251 LPCVD: low-pressure chemical vapor deposition, 11, 12 LSS theory: Lindhard-Scharff-Schiott theory, 261 luminescence, 152 spectrum, 238

M MBE: Molecular Beam Epitaxy, 213 melting, 259 temperature, 258 metal line, 116 microelectronics devices, 214 microwave device, 258 miniaturization minimum reflectivity, 64 mobility, 61 MOCVD metal-organic chemical vapor deposition, 213 mode-locked laser, 79, 251 modulation frequency, 117 modulator, 6, 249, 251 momentum-space wavefunction, 67 momentum transfer, 216 monochromator, 157 monolithic integration, 249, 251 Monte Carlo method, 262 multilayer, 9 optical model 1 multiphonon absorption, 168 multiwavelength, 8 N

narrow gap semiconductor, 258, 264, 266, 279 Nd: YAG laser, 82

near surface region, 220, 270 bandgap excitation, 92 negative annealing, 92, 128, 134 nitrogen atmosphere, 241 non-contact, 115 nonabsorbing mirror, 251 nondegenerate statistics, 61 nondestructive, 41, 115, 146 technique, 39 analysis, 1 nondimensional, 61 nondopant ion beam, 238 nonequilibrium point defect, 272 nonradiative recombination, 87, 220 center, 248 nonstoichiometric, 267 state, 274 normal incidense reflection, 40 normal incident reflectance, 9 nuclear collision, 273 interaction. 262 0

optical absorption, 40, 152, 154, 159 characterization, 40 property, 151, 172 spectroscopy, 39 transmission, 42, 68 gap, 154, 169, 170 reflectance, 42, 68 transition, 163 coefficient, 61 optoelectronic, 214 integrated circuit, 249 device, 248, 260 oxide layer, 47

P PA: photoacoustic, 153 parabolic band, 11 PDS: Photothermal Deflection Spectroscopy, 151, 153, 174 permittivity, 61 phase grating, 251 transition, 128 phonon peak intensity (PPI), 109

INDEX phosphorus implanted Si, 90, 91, 92, 93, 94 photoconductivity, 152 photolithographic, 232 photoluminescence, 74, 75 photometric ellipsometer, 6 photomicrographs, 273 photomodulated, 115 photon energy, 117 flux, 117 photonics, 214 integrated circuit, 249 photothermal, 115 equation, 115 PL: Photoluminescence, 216,223, 224, 226, 235, 246 planar p-n junction, 286 plasma wavelength, 63, 64, 65 effect, 49, 140 diffusion length, 117, 118 wave, 117, 142 wavelength minimum, 55 coefficient, 117, 137 contribution, 138 density, 117 PMTR: Photomodulated Thermoreflectance, 115-146 point defect, 12, 127, 128, 152, 231, 264 complexe, 167 polarimeter-type-ellipsometer, 2 polarimetry, 3, 4 polarized light, 4 polarizer, 5 polyatomic ion, 283 polysilicon, 104 positively charged state, 159 position sensor, 155 postimplantation annealing, 233, 272, 290 Probability of radiative transition, 75 probe beam, 117, 119, 155 progressive etching, 90 projected range, 263 proton bombardment, 270 pseudo-dielectric function, 9 photoelectronic device, 258 pumb beam, 117, 119

Q Q-swhithed lasers, 79, 98

quantum well, 213, 232 structure, 232 box, 238 mechanical rule, 226 dot, 238 wire, 238 quarter-wave plate, 5 QWB quantum well box, 238,248 QWs: Quantum Wells, 214, 215, 224 QWW quantum well wire, 238, 240,248

R radiation induced damage, 151 damage, 266 defect, 268 transition, 152, 230 RAE: rotating-analyzer ellipsometer, 6 Raman, 74,77 spectroscopy, 11, 172 scattering, 230 technique, 80-81 conservation of momentum, 77 of phosphorus implanted Si, 106- 109 backscattering, 80 tensor, 77 conservation of energy, 77 frequency shift, 84 scattering theory, 77 electric field, 170 range statistics, 257, 261 Rayleigh scattering, 77 RBS: Rutherford backscattering, 222, 279 spectrometry, 9, 88, 270 reaction kinetic equation, 129, 131 recombination, 275 lifetime, 117, 136 velocity, 136 recovery of lattice disorder, 270 recrystallization, 126, 131 reflection spectra, 57 minimum, 49, 53, 65 spectroscopy, 39 coefficient, 57 reflectivity, 115 minimum, 60, 65 reflectometry, 3, 4 refractive index, 8, 40, 41

INDEX gradient, 155 regrowth, 278 relative strength of Stokes to anti-Stokes, 78 relaxation time, 61 relationship, 130 equation, 134 residual defect, 221, 223, 349 residual damage, 248 resistivity, 61 resonance, 41 rotating-element, 2 RPE: rotating-polarizer ellipsometer, 6 RTA: rapid thermal annealing, 221, 270, 279, 281 S

SE: spectroscopic ellipsometry, 3 Secondary ion mass spectroscopy (SIMS), 88 self quenching, 160 self-annealing, 122, 124, 279 self implanted, 11 SEM: scanning electron microscopy, 279 semi-infinite, 117 Sharp phonon modes, 97 sheet resistance, 42 sheet carrier concentration, 267, 268, 269 short range order, 10 Si Charge-coupled device (CCD), 8 1 single beam spectrometer, 41 single beam thermowave technique, 144 Single-crystal indium phosphide (InP), 88 SIMNI: separation by implantation of nitrogen, 1 SIMOX: separation by implantation on oxygen, 1 SIMS Secondary Ion Mass Spectroscopy, 216, 223, 231, 237, 238, 262, 263, 282,285 SLs: Superlattices, 214 SNR: signal to noise ratio, 145, 168 SOI: silicon-on-insolator, 1 solubility limit, 271 SOS: silicon on sapphire, 154, 166, 167, 168 spreading resistance, 53, 63, 65 sputtered film, 158 structural relaxation, 152 stoichiometry, 259, 263, 278 stopping power, 261 straggling, 218, 263

301

strain-associated behavior, 173 structural disorder, 67, 170, 172 relaxation processes, 167, 172, 173, 174 defect, 285 subgap spectra, 173 subgap absorption, 152, 166 sublattice, 259, 260, 266, 267, 272 substitutional site, 131 substrate properties, 146 supersaturation, 273, 274 SUPREM: Stanford University Process Engineering Model, 49 surface recombination velocity, 117, 146 state, 154 peak, 282 resistance, 86 swelling, 263 synchronous periodic deflection, 155

T TEM: Tranmission Electron Microscopy, 215,216,232,270,273 temperature induced disorder, 170 gradient, 155 tetrahedral bond-angle distortion, 152 there-dimensional diffusion, 118 thermal diffusion length, 118, 147, 155 diffusivity, 117 activation, 117 disorder, 170 interdiffusion, 216 wave, 116 contribution, 138, 139 coefficient, 137 wave contribution, 142 conductivity, 132 treatment, 279 annealing, 98, 257 degradation, 89 wave photothermal reflectance, 116 thermoreflectance, 115 thin film, 1, 119, 154 Ti: Sapphire lasers, 79 TO phonon, 78 total vacancy creation, 243 transmission

302

INDEX

coefficient, 57, 58 spectrum, 55, 57 spectroscopy, 39 transmittance spectrum, 156 trapping, 162 TRIM: The transport of ions in matter, 216, 218, 232, 243, 246, 262, 263, 265, 270, 278 two frequency hetero-dyne modulation, 145 two phase model, 10 two layer model, 145

complex, 67 pair, 260 concentration, 259 valance band, 1 1 band tail, 163 zone, 264 Van der Paw Measurements, 53, 267 volatile species, 258 volume vacancy loop, 273

W U ultrafast laser pulse, 230 unimplanted sputtered film, 174 unisotropic, 237 Urbach parameter, 163, 167, 169, 170, 171, 264 V vacancy, 67, 133, 152, 220, 243 diffusion constant, 241 impurity complex, 152 distribution, 243

waveguide, 249 minimum reflectivity, 66

X XTEM: cross-section transmission electron microscopy, 9 Z

zero-dimensional structure, 238

Contents of Volumes in This Series

Volume 1 Physics of 111-V Compounds C. Hilsum, Some Key Features of 111-V Compounds Franco Bassani, Methods of Band Calculations Applicable to 111-V Compounds E . 0. Kane, The k-p Method 1/: L. Bonch-Brueuich, Effect of Heavy Doping on the Semiconductor Band Structure Donald Long, Energy Band Structures of Mixed Crystals of 111-V Compounds Laura M. Roth and Petros N. Argyres, Magnetic Quantum Effects S. M. Puri and T. H. Geballe, Thermomagnetic Effects in the Quantum Region W. M. Eecker, Band Characteristics near Principal Minima from Magnetoresistance E. H. Pulley, Freeze-Out Effects, Hot Electron Effects, and Submillimeter Photoconductivity in InSb H. Weiss, Magnetoresistance Betsy Ancker-Johnson, Plasma in Semiconductors and Semimetals

Volume 2 Physics of 111-V Compounds M. G. Holland, Thermal Conductivity

S. I. Novkova, Thermal Expansion U. Piesbergen, Heat Capacity and Debye Temperatures G. Giesecke, Lattice Constants J. R Drabble, Elastic Properties A. U. Mac Rae and G. W. Gobeli, Low Energy Electron Diffraction Studies Robert Lee Mieher, Nuclear Magnetic Resonance Bernard Goldstein, Electron Paramagnetic Resonance T. S. Moss, Photoconduction in 111-V Compounds E. Antoncik ad J. Tauc, Quantum Efficiency of the Internal Photoelectric Effect in InSb G. W. Gobeli and I. G. Allen, Photoelectric Threshold and Work Function P. S. Pershan, Nonlinear Optics in IIILV Compounds M. Gershenzon, Radiative Recombination in the 111-V Compounds Frank Stern, Stimulated Emission in Semiconductors

303

304

CONTENTS OF VOLUMES IN THISSERIES

Volume 3 Optical of Properties 111-V Compounds Marvin Hass, Lattice Reflection William G. Spitzer, Multiphonon Lattice Absorption D. L. Stierwalt and R F. Potter, Emittance Studies H. R Philipp and H.Ehrenveich, Ultraviolet Optical Properties Manuel Cardona, Optical Absorption above the Fundamental Edge Earnest J. Johnson, Absorption near the Fundamental Edge John 0. Dimmock, Introduction to the Theory of Exciton States in Semiconductors B. Lax and J. G. Mavroides, Interband Magnetooptical Effects H. Y. Fan, Effects of Free Carries on Optical Properties Edward D. Palik and George B. Wright, Free-Carrier Magnetooptical Effects Richard H. Bube, Photoelectronic Analysis B. 0. Seraphin and H. E. Bennetr, Optical Constants

Volume 4 Physics of 111-V Compounds N. A. Goryunova. A. S. Borschevskii, and D. N, Tretiakov, Hardness N. N. Sirota, Heats of Formation and Temperatures and Heats of Fusion of Compounds A"'BV Don L. Kendall, Diffusion A. G. Chynoweth, Charge Multiplication Phenomena Robert W. Keyes, The Effects of Hydrostatic Pressure on the Properties of 111-V Semiconductors L. W. Aukerman, Radiation Effects N. A. Goryunova, E P. Kesamanly, and D. N. Nasledov, Phenomena in Solid Solutions R T. Bate, Electrical Properties of Nonuniform Crystals

Volume 5 Infrared Detectors Henry Levinstein, Characterization of Infrared Detectors Paul W.Kruse, Indium Antimonide Photoconductive and Photoelectromagnetic Detectors M. B. Prince, Narrowband Self-Filtering Detectors Ivars Melngalis and T. C. Harman, Single-Crystal Lead-Tin Chalcogenides Donald Long and Joseph L. Schmidt, Mercury-Cadmium Telluride and Closely Related Alloys E. H. Putley, The Pyroelectric Detector Norman B. Stevens, Radiation Thermopiles R J. K e y s and T. M. Quist, Low Level Coherent and Incoherent Detection in the Infrared M. C. Teich, Coherent Detection in the Infrared E R. Arums, E. W. Surd, B. J. Peyton, and E P. Pace, Infrared Heterodyne Detection with Gigahertz IF Response H. S. Sommers, Jr., Macrowave-Based Photoconductive Detector Robert Sehr and Ruiner Zuleeg, Imaging and Display

Volume 6 Injection Phenomena Murray A. Lampert and Ronald B. Schilling, Current Injection in Solids: The Regional Approximation Method Richard Williams, Injection by Internal Photoemission Allen M. Barnett, Current Filament Formation

CONTENTS OF VOLUMES IN THISSERIES

305

R Baron and J. W. Mayer, Double Injection in Semiconductors W. Ruppel, The Photoconductor-Metal Contact

Volume 7 Application and Devices Part A John A. Copeland and Stephen Knight, Applications Utilizing Bulk Negative Resistance l? A. Padovani, The Voltage-Current Characteristics of Metal-Semiconductor Contacts P. L. Hower, W. W. Hooper. B. R. Cairns, R D. Fairman. and D. A. Tremere, The GaAs Field-Effect Transistor Marvin H. White, M O S Transistors G. R Antell, Gallium Arsenide Transistors T. L. Tansley, Heterojunction Properties

Part B T. Misawa, IMPATT Diodes H. C. Okean, Tunnel Diodes Robert B. Campbell and Hung-Chi Chang, Silicon Junction Carbide Devices R E. Enstrom. H. Kressel, and L. Krassner, High-Temperature Power Rectifiers of GaAs,-,P,

Volume 8 Transport and Optical Phenomena Richard J. Stirn, Band Structure and Galvanomagnetic Effects in 111-V Compounds with Indirect Band Gaps Roland W. Ure, Jr., Thermoelectric Effects in 111-V Compounds Herbert Piller, Faraday Rotation H. Barry Bebb and E. W. Williams, Photoluminescence I : Theory E. W. Williams and H. Burry Bebb, Photoluminescence 11: Gallium Arsenide

Volume 9 Modulation Techniques B. 0. Seraphin, Electroreflectance R L. Aggarwal, Modulated Interband Magnetooptics Daniel F. Blossey and Paul Handler, Electroabsorption Bruno Batz, Thermal and Wavelength Modulation Spectroscopy fvar Balslev, Piezopptical Effects D. E. Aspnes and N.Bottku, Electric-Field Effects on the Dielectric Function of Semiconductors and Insulators

Volume 10 Transport Phenomena R L. Rhode, Low-Field Electron Transport J. D. Wiley, Mobility of Holes in 111-V Compounds C. M. Worfe and G. E. Stillman, Apparent Mobility Enhancement in Inhomogeneous Crystals Robert L. Petersen, The Magnetophonon Effect

306

CONTENTS OF VOLUMES IN THISSERIES

Volume 11 Solar Cells Harold J. Hovel, Introduction; Carrier Collection, Spectral Response, and Photocurrent; Solar Cell Electrical Characteristics; Efficiency; Thickness; Other Solar Cell Devices; Radiation Effects: Temperature and Intensity; Solar Cell Technology

Volume 12 Infrared Detectors (11) W. L. Eiseman, J. D. Merriam, and R E Potter, Operational Characteristics of Infrared Photodetectors Peter R. Bratt, Impurity Germanium and Silicon Infrared Detectors E. H. Putley, InSb Submillimeter Photoconductive Detectors G. E. Stillman, C M. Wolfe, and J. 0. Dimmock. Far-Infrared Photoconductivity in High Purity GaAs G. E. Stillman and C. M. Wolfe, Avalanche Photodiodes P. L. Richards, The Josephson Junction as a Detector of Microwave and Far-Infrared Radiation E. H. Putley, The Pyroelectric Detector-An Update

Volume 13 Cadmium Telluride Kenneth Zanio, Materials Preparations; Physics; Defects; Applications

Volume 14 Lasers, Junctions, Transport N. Holonyak, Jr. and M. H. Lee, Photopumped 111-V Semiconductor Lasers Henry Kressel and Jerome K. Butler, Heterojunction Laser Diodes A Van der Ziel, Space-Charge-Limited Solid-state Diodes Peter J . Price, Monte Carlo Calculation of Electron Transport in Solids

Volume 15 Contacts, Junctions, Emitters B. L. Sharma, Ohmic Contacts to 111-V Compounds Semiconductors Allen Nussbawn, The Theory of Semiconducting Junctions John S. Escher, NEA Semiconductor Photoemitters

Volume 16 Defects, (HgCd)Se, (HgCd)Te Henry KresseI, The Effect of Crystal Defects on Optoelectronic Devices C. R Whitsert, J. G. Broerman. and C. J. Summers, Crystal Growth and Properties of Hg, _,Cd,Se alloys M . H. Weiler, Magnetooptical Properties of Hg, _,Cd,Te Alloys Paul W.Kruse and John G. Ready, Nonlinear Optical Effects in H&-,Cd,Te

Volume 17 CW Processing of Silicon and Other Semiconductors James F. Gibbons, Beam Processing of Silicon Arto Lietoila. Richard B. Gold. James F. Gibbons, and Lee A. Christel, Temperature Distribu-

CONTENTS OF VOLUMES IN THISSERIES

307

tions and Solid Phase Reaction Rates Produced by Scanning CW Beams Arto Leitoila and James E Gibbons, Applications of CW Beam Processing to Ion Implanted Crystalline Silicon N . M. Johnson, Electronic Defects in CW Transient Thermal Processed Silicon K. E Lee, T. J. Stultz, and James E Gibbons, Beam Recrystallized Polycrystalline Silicon: Properties, Applications, and Techniques T. Shibata. A. Wakita, T. W. Sigmon, and James E Gibbons, Metal-Silicon Reactions and Silicide Yves I. Nissim and James E Gibbons, CW Beam Processing of Gallium Arsenide

Volume 18 Mercury Cadmium Telluride Paul W. Kruse, The Emergence of (H&-xCd,)Te as a Modern Infrared Sensitive Material H. E. Hirsch, S. C. Liang. and A. G. White, Preparation of High-Purity Cadmium, Mercury, and Tellurium W. E H. Micklethwaite, The Crystal Growth of Cadmium Mercury Telluride Paul E. Petersen, Auger Recombination in Mercury Cadmium Telluride R M. Broudy and V. J. Mazurczyck, (HgCd)Te Photoconductive Detectors M. 5. Reine, A. K. Soad, and T. J. Tredwell, Photovoltaic Infrared Detectors M . A. Kinch, Metal-Insulator-Semiconductor Infrared Detectors

Volume 19 Deep Levels, GaAs, Alloys, Photochemistry G. E Neumark and K. Kosai, Deep Levels in Wide Band-Gap 111-V Semiconductors David C. Look, The Electrical and Photoelectronic Properties of Semi-Insulating GaAs R E Brebrick. Ching-Huu Su, and Pok-Kai Liao, Associated Solution Model for Ga-In-Sb and Hg-Cd-Te Yu. Yu. Gurevich and Yu. V. Pleskon, Photoelectrochemistry of Semiconductors

Volume 20 Semi-Insulating Ga As R. N. Thomas, H. M. Hobgood, G. W. Eldridge, D. L. Barrett, T. T. Braggins, L. B. Ta, and S. K. Wang, High-Purity LEC Growth and Direct Implantation of GaAs for Monolithic Microwave Circuits C. A. Stolte, Ion Implantation and Materials for GaAs Integrated Circuits C. G. Kirkpatrick, R. T. Chen, D. E. Holmes, P. M. Asbeck. K. R Elliott, R D. Fairman, and J. R Oliver, LEC GaAs for Integrated Circuit Applications J. S. Blakemore and S. Ruhimi, Models for Mid-Gap Centers in Gallium Arsenide

Volume 2 1 Hydrogenated Amorphous Silicon Part A Jacques I. Pankove, Introduction Masataka Hirose, Glow Discharge; Chemical Vapor Deposition Yoshiyuki Uchidu, di Glow Discharge T. D. Moustukas, Sputtering Isao Yumadu, Ionized-Cluster Beam Deposition Bruce A. Scott, Homogeneous Chemical Vapor Deposition

308

CONTENTS OF VOLUMESIN THISSERIES

Frank J. Kampas, Chemical Reactions in Plasma Deposition Paul A. Longeway, Plasma Kinetics Herbert A. Weakliem, Diagnostics of Silane Glow Discharges Using Probes and Mass Spectroscopy Lester Gluttman, Relation between the Atomic and the Electronic Structures A. Chenevas-Paule, Experiment Determination o f Structure S. Minomura, Pressure Effects on the Local Atomic Structure David Adler, Defects and Density of Localized States

Part B Jacques I. Pankove, Introduction G.D. Cody, The Optical Absorption Edge of a-Si: H Nabil M. Amer and Warren B. Jackson, Optical Properties of Defect States in a-Si: H P. J. Zanzucchi, The Vibrational Spectra of a-Sk H Yoshihiru Hamakawa, Electroreflectance and Electroabsorption Jefrey S. Lannin, Raman Scattering of Amorphous Si, Ge, and Their Alloys R A. Street, Luminescence in a-Si: H Richard S. Crandall, Photoconductivity J. Tauc, Time-Resolved Spectroscopy of Electronic Relaxation Processes P. E. Vanier, IR-Induced Quenching and Enhancement of Photoconductivity and Photoluminescence H. Schade, Irradiation-Induced Metastable Effects L.Ley, Photoelectron Emission Studies

Part C Jacques I. Pankove, Introduction J. David Cohen, Density of States from Junction Measurements in Hydrogenated Amorphous Silicon P. C. Taylor, Magnetic Resonance Measurements in a-Si: H K. Morigaki, Optically Detected Magnetic Resonance J. Dresner, Carrier Mobility in a-Si: H T. Tiedje, Information about band-Tail States from Time-of-Flight Experiments Arnold R Moore, Diffusion Length in Undoped a-Si: H W. Beyer and J. OverhoJ Doping Effects in a-Si: H H.Fritzche, Electronic Properties of Surfaces in a-Si: H C. R Wronski, The Staebler-Wronski Effect R J. Nemanich, Schottky Barriers on a-Si: H B. Abeles and T. Tiedje, Amorphous Semiconductor Superlattices

Part D Jacques I. Pankove, Introduction D. E. Carlson, Solar Cells G.A. Swartz, Closed-Form Solution of I-V Characteristic for a a-Si: H Solar Cells Isamu Shimizu, Electrophotography Sachio Ishioka, Image Pickup Tubes

CONTENTS OF VOLUMES IN THISSERIES

309

P. G. LeComber and W E. Spear, The Development of the a-Si: H Field-Effect Transistor and Its Possible Applications D. G. Ast, a-Si: H FET-Addressed LCD Panel S. Kuneko, Solid-State Image Sensor Masakiyo Matswnura, Charge-Coupled Devices M. A. Bosch, Optical Recording A. D’Amico and G. Fortunato, Ambient Sensors Hiroshi Kukimoto, Amorphous Light-Emitting Devices Robert J. Phelan, Jr., Fast Detectors and Modulators Jacques I. Pankove, Hybrid Structures P. G. LeComber, A. E. Owen, W.E. Spear, J. Hajto, and W. K. Choi, Electronic Switching in Amorphous Silicon Junction Devices

Volume 22 Lightwave Communications Technology Part A Kazuo Nakajima, The Liquid-Phase Epitaxial Growth of IngaAsp W. T. Tsang, Molecular Beam Epitaxy for 111-V Compound Semiconductors G. B. Stringfellow, Organometallic Vapor-Phase Epitaxial Growth of 111-V Semiconductors G. Beuchet, Halide and Chloride Transport Vapor-Phase Deposition of InGaAsP and GaAs Manijeh Razeghi, Low-Pressure Metallo-Organic Chemical Vapor Deposition of Ga,in,-xAsP,-y Alloys P. M. Petrofi Defects in 111-V Compound Semiconductors

Part B J. P. van der Ziel, Mode Locking of Semiconductor Lasers Kam Y. Lm and Ammon Yariv, High-Frequency Current Modulation of Semicond Actor Injection Lasers Charles H. Henry, Special Properties of Semiconductor Lasers Yasuharu Suematsu. Katsumi Kishino, Shigehisa Arai, and Fumio Koyama. Dynamic SingleMode Semiconductor Lasers with a Distributed Reflector W. T. Tsang, The Cleaved-Coupled-Cavity (C3) Laser

Part C R J. Nelson and N. K. Dutta, Review of InGaAsP InP Laser Structures and Comparison of Their Performance N. Chinone and M. Nakamura, Mode-Stabilized Semiconductor Lasers for 0.7-0.8- and 1.1- 1.6-pm Regions Yoshiji Horikoshi, Semiconductor Lasers with Wavelengths Exceeding 2 pm B. A. Dean and M. Dixon, The Functional Reliability of Semiconductor Lasers as Optical Transmitters R H. Saul, T. P. Lee, and C. A. Burus, Light-Emitting Device Design C. L. Zipfel, Light-Emitting Diode-Reliability Tien Pei Lee and Tingye Li, LED-Based Multimode Lightwave Systems Kinichiro Ogawa, Semiconductor Noise-Mode Partition Noise

310

CONTENTS OF VOLUMES IN THISSERIES

Part D Federico Capasso, The Physics of Avalanche Photodiodes T. P, Pearsall and M. A. Pollack, Compound Semiconductor Photodiodes Taka0 Kaneda, Silicon and Germanium Avalanche Photodiodes S. R Forrest, Sensitivity of Avalanche Photodetector Receivers for High-Bit-Rate LongWavelength Optical Communication Systems J. C. Campbell, Phototransistors for Lightwave Communications

Part E Shyh Wang, Principles and Characteristics of Integrable Active and Passive Optical Devices Shlomo Margalit and Amnon Yariv, Integrated Electronic and Photonic Devices Takaoki Mukai, Yoshihisa Yamamoto, and Tatsuya Kimura, Optical Amplification by Semiconductor Lasers

Volume 23 Pulsed Laser Processing of Semiconductors R. F. Wood, C. W. White, and R T. Young, Laser Processing of Semiconductors: An Overview C. W. White, Segregation, Solute Trapping, and Supersaturated Alloys G. E. Jellison, Jr., Optical and Electrical Properties of Pulsed Laser-Annealed Silicon R F. Wood and G.E. Jellison, Jr., Melting Model of Pulsed Laser Processing R F. Wood and E W . Young, Jr., Nonequilibrium Solidification Following Pulsed Laser Melting D. H. Lowndes and G. E. Jellison, Jr., Time-Resolved Measurement During Pulsed Laser Irradiation of Silicon D. M.Zebner, Surface Studies of Pulsed Laser Irradiated Semiconductors D. H. Lowndes, Pulsed Beam Processing of Gallium Arsenide R B. James, Pulsed CO, Laser Annealing of Semiconductors R T. Young and R E Wood, Applications of Pulsed Laser Processing

Volume 24 Applications of Multiquantum Wells, Selective Doping, and Superlattices C. Weisbuch, Fundamental Properties of 111-V Semiconductor Two-Dimensional Quantized Structures: The Basis for Optical and Electronic Device Applications H. Morkoc and H. Unlu, Factors Affecting the Performance of (Al,Ga)As/GaAs and (Al,Ga)As/InCaAs Modulation-Doped Field-Effect Transistors: Microwave and Digital Applications N. T. Linh, Two-Dimensional Electron Gas FETs: Microwave Applications M. Abe et al., Ultra-High-speed HEMT Integrated Circuits D. S. Chemla. D. A. B. Miller, and P. W. Smith, Nonlinear Optical Properties of Multiple Quantum Well Structures for Optical Signal Processing F. Capasso, Graded-Gap and Superlattice Devices by Band-Gap Engineering W. T. Tsang, Quantum Confinement Heterostructure Semiconductor Lasers G. C. Osbourn et al., Principles and Applications of Semiconductor Strained-Layer Superlattices

CONTENTS OF VOLUMES IN THISSERIES

311

Volume 25 Diluted Magnetic Semiconductors W. Giriat and J. K. Furdyna, Crystal Structure, Composition, and Materials Preparation of Diluted Magnetic Semiconductors W. M. Becker, Band Structure and Optical Properties of Wide-Gap Af'_,Mn,BiV Alloys at Zero Magnetic Field Saul Oseroff and Pieter H. Keesom, Magnetic Properties: Macroscopic Studies Giebultowicz and T. M. Holden, Neutron Scattering Studies of the Magnetic Structure and Dynamics of Diluted Magnetic Semiconductors J. Kossut, Band Structure and Quantum Transport Phenomena in Narrow-Gap Diluted Magnetic Semiconductors C. Riquaux, Magnetooptical Properties of Large-Gap Diluted Magnetic Semiconductors J. A. Gaj, Magnetooptical Properties of Large-Gap Diluted Magnetic Semiconductors J. Mycielski, Shallow Acceptors in Diluted Magnetic Semiconductors: Splitting, Boil-off, Giant Negative Magnetoresistance A. K. Ramadas and R. Rodriquez, Rarnan Scattering in Diluted Magnetic Semiconductors P. A . Wol& Theory of Bound Magnetic Polarons in Semimagnetic Semiconductors

Volume 26 I11-V Compound Semiconductors and Semiconductor Properties of Superionic Materials Zou Yuanxi, 111-V Compounds H. V. Winston, A. T. Hunter, H . Kimura. and R. E. Lee, 1nAs-Alloyed GaAs Substrates for Direct Implantation P. K Bhaitachary and S. Dhar, Deep Levels in IIILV Compound Semiconductors Grown by MBE Yu. Yu. Gurevich and A . K. Ivanov-Shits, Semiconductor Properties of Supersonic Materials

Volume 27 High Conducting Quasi-One-DimensionalOrganic Crystals E. M. Conwell, Introduction to Highly Conducting Quasi-One-Dimensional Organic Crystals I. A. Howard, A Reference Guide to the Conducting Quasi-One-Dimensional Organic Molecular Crystals J. P. Pouquei, Structural Instabilities E. M. Conwell, Transport Properties C. S. Jacobsen, Optical Properties J. C. Scott, Magnetic Properties L. Zuppiroli, Irradiation Effects: Perfect Crystals and Real CrystaIs

Volume 28 Measurement of High-speed Signals in Solid State Devices J. Frey and D.Ioannou, Materials and Devices for High-speed and Optoelectronic Applications H. Schumacher and E. Strid, Electronic Wafer Probing Techniques D. H. Auston, Picosecond Photoconductivity: High-speed Measurements of Devices and

Materials J. A. Valdmanis, Electro-Optic Measurement Techniques for Picosecond Materials, Devices,

and Integrated Circuits.

of Integrated Circuits and High-speed Devices G. Plows, Electron-Beam Probing A. M, Weiner and R. B. Marcus, Photoemissive Probing J. M. Wiesenfeld and R K. Jain, Direct Optical Probing

312

CONTENTS OF VOLUMES IN THISSERIES

Volume 29 Very High Speed Integrated Circuits: Gallium Arsenide LSI M. Kuzuhara and T. Nazaki, Active Layer Formation by Ion Implantation H. Hasimoro, Focused Ion Beam Implantation Technology T. Nozaki and A. Higashisaka, Device Fabrication Process Technology M. In0 and T. Takada, GaAs LSI Circuit Design M. Hirayama, M. Ohmori, and K. Yamasaki, GaAs LSI Fabrication and Performance

Volume 30 Very High Speed Integrated Circuits: Heterostructure H. Watanabe, T. Mizutani, and A. Usui, Fundamentals of Epitaxial Growth and Atomic Layer Epitaxy S. Hiyamizu, Characteristics of Two-Dimensional Electron Gas in 111-V Compound Heterostructures Grown by MBE T. Nakanisi, Metalorganic Vapor Phase Epitaxy for High-Quality Active Layers T. Nimura, High Electron Mobility Transistor and LSI Applications T. Sugera and T. Ishibashi, Hetero-Bipolar Transistor and LSI Application H. Matsueda, T. Tanaka, and M. Nakamura, Optoelectronic Integrated Circuits

Volume 3 1 Indium Phosphide: Crystal Growth and Characterization J. P. Farges, Growth of Discoloration-free InP M. J. McCoNum and G. E. Stillman, High Purity InP Grown by Hydride Vapor Phase Epitaxy T. fnada and T. Fukuda, Direct Synthesis and Growth of Indium Phosphide by the Liquid Phosphorous Encapsulated Czochralski Method 0. Oda, K. Katagiri, X Shinohara, S. Katsura. Y. Takahashi, X Kainosho, K. Kohiro, and R Hirano, InP Crystal Growth, Substrate Preparation and Evaluation K. Tada, M. Tatsumi, M. Morioka, T. Araki, and T. Kawase, InP Substrates: Production and Quality Control M. Razeghi, LP-MOCVD Growth, Characterization, and Application of InP Material T. A. Kennedy and P. J. Lin-Chung, Stoichiometric Defects in InP

Volme 32 Strained-Layer Superlattices: Physics T. P. Pearsall, Strained-Layer Superlattices Fred H . Pollack, Effects of Homogeneous Strain on the Electronic and Vibrational Levels in Semiconductors J. Y. Marzin, J. M. Gerard, P. Voisin, and J. A. Brum, Optical Studies of Strained 111-V Heterolayers R. People and S. A. Jackson, Structurally Induced States from Strain and Confinement M. Jaros, Microscopic Phenomena in Ordered Suprlattices

Volume 33 Strained-Layer Superlattices: Materials Science and Technology R Hull and J. C. Bean, Principles and Concepts of Strained-Layer Epitaxy William J. Schax Paul J. Tusker. Marc C. Foisy, and Lester I? Eastman, Device Applications of Strained-Layer Epitaxy

CONTENTS OF VOLUMES IN THISSERIES

313

S. T. Picraux, B. L. Doyle, and J. Y. Tsao, Structure and Characterization of Strained-Layer Superlattices E. Kasper and I? Schaffer, Group IV Compounds Dale L. Martin, Molecular Beam Epitaxy of IV-VI Compounds Heterojunction Robert L. Gunshor. Leslie A. Kolodziejski, Arto V. Nurmikko, and Nobuo Otsuka, Molecular Beam Epitaxy of 11-VI Semiconductor Microstructures

Volume 34 Hydrogen in Semiconductors J. I. Pankove and N. M. Johnson, Introduction to Hydrogen in Semiconductors C. H. Seager, Hydrogenation Methods J. I. Pankove, Hydrogenation o f Defects in Crystalline Silicon J. W. Corbett, P. Deizk, U. V. Desnica, and S. J . Pearton, Hydrogen Passivation of Damage Centers in Semiconductors S. J. Pearton, Neutralization of Deep Levels in Silicon J. I. Pankove, Neutralization of Shallow Acceptors in Silicon N. M. Johnson, Neutralization of Donor Dopants and Formation of Hydrogen-Induced Defects in n-Type Silicon M. Stavola and S. J. Pearton, Vibrational Spectroscopy o f Hydrogen-Related Defects in Silicon A. D. Marwick, Hydrogen in Semiconductors: Ion Beam Techniques C. Herring and N. M . Johnson, Hydrogen Migration and Solubility in Silicon E. E. Huller, Hydrogen-Related Phenomena in Crystalline Germanium J. Kakalios, Hydrogen Diffusion in Amorphous Silicon J. Chevalier, B. Clerjaud, and B. Pajor, Neutralization of Defects and Dopants in 111-V Semiconductors G. G. DeLeo and W. B. Fowler, Computational Studies of Hydrogen-Containing Complexes in Semiconductors R F: Kiejl and T. L. Esfle, Muonium in Semiconductors C. G. Van de Walle, Theory of Isolated Interstitial Hydrogen and Muonium in Crystalline Semiconductors

Volume 35 Nanostructured Systems Mark Reed, Introduction H. van Houten, C. W. J. Beenakker, and B. J. van Wees, Quantum Point Contacts G. Timp, When Does a Wire Become an Electron Waveguide? M. Biittiker, The Quantum Hall Effects in Open Conductors W. Hansen, J. P. Kotthaus, and U. Merkt, Electrons in Laterally Periodic Nanostructures

Volume 36 The Spectroscopy of Semiconductors D. Heiman, Spectroscopy of Semiconductors at Low Temperatures and High Magnetic Fields Art0 V. Nurmikko, Transient Spectroscopy by Ultrashort Laser Pulse Techniques A. K. Ramdas and S. Rodriguez, Piezospectroscopy of Semiconductors Orest J. Glembocki and Benjamin K Shanabrook, Photoreflectance Spectroscopy of Microstructures David G. Seiler. Christopher L. Littler, and Margaret H. Wiler, One- and Two-Photon Magneto-Optical Spectroscopy of InSb and Hg, _xCd,Te

314

CONTENTS OF VOLUMES IN THISSERIES

Volume 37 The Mechanical Properties of Semiconductors A,-B. Chen, Arden Sher and W. i7 Yost, Elastic Constants and Related Properties of Semiconductor Compounds and Their Alloys David R Clarke, Fracture of Silicon and Other Semiconductors Hans Siethox The Plasticity of Elemental and Compound Semiconductors Sivaraman Guruswamy, Katherine i7 Faber and John P. Hirth, Mechanical Behavior of Compound Semiconductors Subhanh Mahajan, Deformation Behavior of Compound Semiconductors John P. Hirrh, Injection of Dislocations into Strained Multilayer Structures Don Kendall, Charles B. Fleddermann, and Kevin J. Malloy, Critical Technologies for the Micromachining of Silicon Zkuo Matsuba and Kinji Mokuya, Processing and Semiconductor Thermoelastic Behavior

Volume 38 Imperfections in IIW Materials Udo Scherz and Matthias Schefler, Density-Functional Theory o f sp-Bonded Defects in III/V Semiconductors Maria Kaminska and Eicke R. Weber, El2 Defect in GaAs David C. Look, Defects Relevant for Compensation in Semi-Insulating GaAs R C. Newman, Local Vibrational Mode Spectroscopy of Defects in III/V Compounds Andrzej M. Hennel, Transition Metals in III/V Compounds Kevin J. MaUoy and Ken Khachaturyan, DX and Related Defects in Semiconductors V. Swaminathan and Andrew S. Jordan, Dislocations in III/V Compounds Krzysztof W.Nauka, Deep Level Defects in the Epitaxial III/V Materials

Volume 39 Minority Carriers in 111-V Semiconductors: Physics and Applications Niloy K Dutfa, Radiative Transitions in GaAs and Other 111-V Compounds Richard K. Ahrenkiel, Minority-Carrier Lifetime in 111-V Semiconductors Tomofumi Furuta, High Field Minority Electron Transport in p-GaAs Mark S. Lundsrrom, Minority-Carrier Transport in 111-V Semiconductors Richard A. Abram, Effects of Heavy Doping and High Excitation on the Band Structure of GaAs David Yevick and Witold Bardyszewski, An Introduction to Non-Equilibrium Many-Body Analyses of Optical Processes in 111-V Semiconductors

Volume 40 Epitaxial Microstructures E. E Schuberr, Delta-Doping of Semiconductors: Electronic, Optical, and Structural Properties of Materials and Devices A. Gossard, M. Sundaram, and P. Hopkins, Wide Graded Potential Wells P. Petros Direct Growth of Nanometer-Size Quantum Wire Superlattices E. K q o n , Lateral Patterning of Quantum Well Heterostructures by Growth of Nonplanar Substrates H. Temkin. D. Gershoni, and M. Punish, Optical Properties of Gal-,In,As/InP Quantum Wells

CONTENTS OF VOLUMESIN THISSERIES

315

Volume 41 High Speed Heterostructure Devices E Capasso, F. Betlram. S. Sen, A. Pahlevi, and A. Y. Cho, Quantum Electron Devices: Physics and Applications P. Solomon, D. J. Frank, S. L. Wright, and F. Canora, GaAs-Gate Semiconductor-InsulatorSemiconductor FET M. H. Hashemi and U.K. Mishra, Unipolar InP-Based Transistors R Kiehl, Complementary Heterostructure FET Integrated Circuits T. Ishibashi, GaAs-Based and InP-Based Heterostructure Bipolar Transistors H. C. Liu and T. C. L. G. Sollner, High-Frequency-Tunneling Devices H. Ohnishi, T. More, M. Takatsu, K. Imamura, and N. Yokoyama, Resonant-Tunneling Hot-Electron Transistors and Circuits

Volume 42 Oxygen in Silicon F. Shimura, Introduction to Oxygen in Silicon W. Lin, The lncorporation of Oxygen into Silicon Crystals T. J. Schafliter and D. K. Schroder, Characterization Techniques for Oxygen in Silicon W. M. Bullis, Oxygen Concentration Measurement S. M. Hu,Intrinsic Point Defects in Silicon B. Pajof, Some Atomic Configurations of Oxygen J. Michel and L. C. Kimerling, Electical Properties of Oxygen in Silicon R C. Newman and R Jones, Diffusion of Oxygen in Silicon 7: Y. Tan and W. J. Taylor, Mechanisms of Oxygen Precipitation: Some Quantitative Aspects M. Schrems, Simulation of Oxygen Precipitation K. Simino and I. Yonenugu, Oxygen Effect on Mechanical Properties W. Bergholz, Grown-in and Process-Induced Effects E Shimura, Intrinsic/Internal Gettering H. Tsuya, Oxygen Effect on Electronic Device Performance

Volume 43 Semiconductors for Room Temperature Nuclear Detector Applications R. 8. James and T. E. Schlesinger, Introduction and Overview L. S. Darken and C. E. Cox, High-Purity Germanium Detectors A. Burger, D. Nason. L. Van den Berg, and M. Schieber, Growth of Mercuric Iodide X J. Bao. T. E. Schlesinger, and R B. James, Electrical Properties of Mercuric Iodide X. J. Bao, R B. James, and T. E. Schlesinger, Optical Properties of Red Mercuric Iodide M. Huge-Ali and P. Siyeert, Growth Methods of CdTe Nuclear Detector Materials M. Huge-Ali and P Sifert, Characterization of CdTe Nuclear Detector Materials M. Huge-Ali and P. Styerr, CdTe Nuclear Detectors and Applications R B. James, T. E. Schlesinger, J. Lund, and M . Schieber, Cd,-,Zn,Te Spectrometers for Gamma and X-Ray Applications D. S. McCregor, J. E. Kammeraad, Gallium Arsenide Radiation Detectors and Spectrometers J. C. Lund, E Olschner, and A. Burger, Lead Iodide M. R. Squillante. and K. S. Shah, Other Materials: Status and Prospects V. M. Gerrish, Characterization and Quantification of Detector Performance J. S. Iwanczyk and B. E. Patr, Electronics for X-ray and Gamma Ray Spectrometers M. Schieber, R B. James. and T. E. Schlesinger, Summary and Remaining Issues for Room Temperature Radiation Spectrometers

316

CONTENTS OF VOLUMESIN THIS SERIES

Volume 44 11-IV BluelGreen Light Emitters: Device Physics and Epitaxial Growth J. Hun and R L. Gunshor, MBE Growth and Electrical Properties of Wide Bandgap ZnSe-based 11-VI Semiconductors Shizuo Fujita and Shigeo Fujita, Growth and Characterization of ZnSe-based 11-VI Semiconductors by MOVPE Easen Ho and Leslie A. Kolodziejski, Gaseous Source UHV Epitaxy Technologies for Wide Bandgap 11-VI Semiconductors Chris G. Van de Walle, Doping of Wide-Band-Gap 11-VI Compounds-Theory Roberto Cingolani, Optical Properties of Excitons in ZnSe-Based Quantum Well Heterostructures A. Ishibashi and A. V. Nurmikko, 11-VI Diode Lasers: A Current View of Device Performance and Issues Supratik Guha and John Petruzello, Defects and Degradation in Wide-Gap 11-VI-based Structures and Light Emitting Devices

Volume 45 Effect of Disorder and Defects in Ion-Implanted Semiconductors: Electrical and Physiochemical Characterization Heiner Ryssel, Ion Implantation into Semiconductors: Historical Perspectives You-Nian Wang and Teng-Cui Ma, Electronic Stopping Power for Energetic Ions in Solids Sachiko T. Nakagawa, Solid Effect on the Electronic Stopping of Crystalline Target and Application to Range Estimation G. Miiller. S. Kalbitzer and G. N. Greaves, Ion Beams in Amorphous Semiconductor Research Jumana Boussey-Said, Sheet and Spreading Resistance Analysis of Ion Implanted and Annealed Semiconductors M. L. Polignano and G. Queirolo, Studies of the Stripping Hall Effect in Ion-Implanted Silicon J. Stoemenos, Transmission Electron Microscopy Analyses Roberta Nipoti and Marco Servidori, Rutherford Backscattering Studies of Ion Implanted Semiconductors P. Zaumseil, X-ray Diffraction Techniques

Volume 46 Effect of Disorder and Defects in Ion-Implanted Semiconductors: Optical and Photothermal Characterization M. Fried, T. Lohner and J. Gyulai, Ellipsometric Analysis Antonios Seas and Constantinos Christofides,Transmission and Reflection Spectroscopy on Ion Implanted Semiconductors Andreas Othonos and Constantinos Christofides, Photoluminescence and Raman Scattering of Ion Implanted Semiconductors. Influence of Annealing Constantinos Christofides, Photomodulated Thermoreflectance Investigation of Implanted Wafers. Annealing Kinetics of Defects U. Zammit, Photothermal Deflection Spectroscopy Characterization of Ion-Implanted and Annealed Silicon Films Andreas Mandelis, Arief Budiman and Miguel Vargas, Photothermal Deep-Level Transient Spectroscopy of Impurities and Defects in Semiconductors R Kalish and S. Charbonneau, Ion Implantation into Quantum-Well Structures Alexandre M. Myasnikov and Nikolay N. Gerasimenko, Ion Implantation and Thermal Annealing of 111-V Compound Semiconducting Systems: Some Problems of 111-V Narrow Gap Semiconductors

This Page Intentionally Left Blank

I S B N 0-32-752346-3