Electrochemistry of Semiconductors and Electronics Processes and Devices - John Mchardy

ELECTROCHEMISTRY OF
SEMICONDUCTORS AND
ELECTRONICS
Processes and Devices
Edited by
John McHardy and Frank Ludwig

Hughes Aircraft Company
El Segundo, California
I.
np I
NOYES PUBLICATIONS
Park Ridge, New Jersey, U.S.A.
Copyright 0 1992 by Noyes Publications
No part of this book may be reproduced or utilized in
any form or by any means, electronic or mechanical,
including photocopying, recording or by any informa-
tion storage and retrieval system, without permission
in writing from the Publisher.
Library of Congress Catalog Card Number: 91-46659
ISBN 0-8155-1301-1
Printed in the United States
Published in the United States of America by

Noyes Publications
Mill Road, Park Ridge, New Jersey 07656
10 9 8 7 6 5 4 3 2 1
Library of Congress Cataloging-in-Publication Data
Electrochemistry of semiconductors and electronics : processes and

devices / edited by John McHardy and Frank Ludwig.
p. cm.
Includes bibliographical references and index.
ISBN 0-8155-1301-1
1. Semiconductors--Design and construction. 2. Electrochemistry.
I. Ludwig, Frank.
TK7871.85.EM2 1992
621.381’52-dc20 91-46659
CIP
MATERIALS SCIENCE AND PROCESS TECHNOLOGY SERIES
Editors
Rointan F. Bunshah, University of California, Los Angeles (Series Editor)
Gary E. McGuire, Microelectronics Center of North Carolina (Series Editor)
Stephen M. Rossnagel, IBM Thomas J. Watson Research Center
(Consulting Editor)
Electronic Materials and Process Technology

DEPOSITION TECHNOLOGIES FOR FILMS AND COATINGS: by Rointan F. Bunshah et al
CHEMICAL VAPOR DEPOSITION FOR MICROELECTRONICS: by Arthur Sherman
SEMICONDUCTOR MATERIALS AND PROCESS TECHNOLOGY HANDBOOK: edited by
Gary E. McGuire
HYBRID MICROCIRCUIT TECHNOLOGY HANDBOOK by James J. Licari and Leonard R.
Enlow
HANDBOOK OF THIN FILM DEPOSITION PROCESSES AND TECHNIQUES: edited by Klaus
K. Schuegraf
IONIZED-CLUSTER BEAM DEPOSITION AND EPITAXY: by Toshinori Takagi
DIFFUSION PHENOMENA IN THIN FILMS AND MICROELECTRONIC MATERIALS: edited by
Devendra Gupta and Paul S.Ho
HANDBOOK OF CONTAMINATION CONTROL IN MICROELECTRONICS: edited by Donald
L. Tolliver
HANDBOOK OF ION BEAM PROCESSING TECHNOLOGY: edited by Jerome J. Cuomo,
Stephen M. Rossnagel, and Harold R. Kaufman
CHARACTERIZATION OF SEMICONDUCTOR MATERIALS-Volume 1: edited by Gary E.
McGuire
HANDBOOK OF PLASMA PROCESSING TECHNOLOGY: edited by Stephen M. Rossnagel,
Jerome J. Cuomo, and William D. Westwood
HANDBOOK OF SEMICONDUCTOR SILICON TECHNOLOGY: edited by William C. O'Mara,
Robert €3. Herring, and Lee P. Hunt
HANDBOOK OF POLYMER COATINGS FOR ELECTRONICS: by James J. Licari and Laura
A. Hughes
HANDBOOK OF SPUTTER DEPOSITION TECHNOLOGY: by Kiyotaka Wasa and Shigeru
Hayakawa
HANDBOOK OF VLSl MICROLITHOGRAPHY: edited by William 6. Glendinning and John
N. Helbert
CHEMISTRY OF SUPERCONDUCTOR MATERIALS: edited by Terrell A. Vanderah
CHEMICAL VAPOR DEPOSITION OF TUNGSTEN AND TUNGSTEN SILICIDES: by John E.J.
Schmitz
ELECTROCHEMISTRY OF SEMICONDUCTORS AND ELECTRONICS: edited by John
McHardy and Frank Ludwig
HANDBOOK OF CHEMICAL VAPOR DEPOSITION: by Hugh 0. Pierson
(continued)
V
vi Series
Ceramic and Other Materials-Processing and Technology

SOL-GEL TECHNOLOGY FOR THIN FILMS, FIBERS, PREFORMS, ELECTRONICS AND
SPECIALTY SHAPES: edited by Lisa C. Klein
FIBER REINFORCED CERAMIC COMPOSITES: by K.S. Mazdiyasni
ADVANCED CERAMIC PROCESSING AND TECHNOLOGY-Volume 1: edited by Jon G.P.
Binner
FRICTION AND WEAR TRANSITIONS OF MATERIALS: by Peter J. Blau
SHOCK WAVES FOR INDUSTRIAL APPLICATIONS: edited by Lawrence E. Murr
SPECIAL MELTING AND PROCESSING TECHNOLOGIES: edited by G.K. Bhat
CORROSION OF GLASS, CERAMICS AND CERAMIC SUPERCONDUCTORS: edited by
David E. Clark and Bruce K. Zoitos
HANDBOOK OF INDUSTRIAL REFRACTORIES TECHNOLOGY: by Stephen C. Carniglia and
Gordon L. Barna
Related Titles
ADHESIVES TECHNOLOGY HANDBOOK: by Arthur H. Landrock
HANDBOOK OF THERMOSET PLASTICS: edited by Sidney H. Goodman
SURFACE PREPARATION TECHNIQUES FOR ADHESIVE BONDING: by Raymond F.
Wegman
FORMULATING PLASTICS AND ELASTOMERS BY COMPUTER: by Ralph D. Hermansen
HANDBOOK OF ADHESIVE BONDED STRUCTURAL REPAIR: by Raymond F. Wegman and
Thomas R. Tullos
~ ~ ~~
CONTRIBUTORS
Giulio Di Giacomo R. David Rauh

IBM EIC Laboratories, Inc.
Hopewell Junction, NY Norwood, MA
Robert C. DeMattei Keshra Sangwal

Center for Materials Research Institute of Chemistry
Stanford University Pedagogical University of
Stanford, CA Czestochowa
Czestochowa, Poland
Robert S. Feigelson
Center for Materials Research Robert T. Talasek
Stanford University Texas Instruments
Stanford, CA Dallas, TX
Ian D. Raistrick Micha Tomkiewicz

Los Alamos National Laboratory Department of Physics
Los Alamos, NM Brooklyn College
Brooklyn, NY
ix
To the best of our knowledge the information in this
publication is accurate; however, the Publisher does not
assume any responsibility or liability for the accuracy or
completeness of, or consequences arising from, such
information. Mention of trade names or commercial
products does not constitute endorsement or
recommendation for use by the Publisher.
Final determination of the suitability of any information or

product for use contemplated by any user, and the
manner of that use, is the sole responsibility of the user.
We recommend that anyone intending to rely on any
recommendation of materials or procedures for use in
electrochemistry involving semiconductors and/or
electronics mentioned in this publication should satisfy
himself as to such suitability, and that he can meet all
applicable safety and health standards. We strongly
recommend that users seek and adhere to the
manufacturer’s or supplier’s current instructions for
handling each material they use.
X
PREFACE
This book reflects the confluence of two trends. On the one hand,
Electrochemistry is reemerging as a vital scientific discipline after many
years of relative obscurity. Issues such as the space race, the energy
crisis, and the environmental movement have prompted rapid expansion
in electrochemical research and the subject is becoming an important
foundation of modern technology. On the other hand, the relentless drive
towards faster, more compact electronic devices continues to probe the
limits of materials science, setting ever higher goals for semiconductor
purity, crystal uniformity, and circuit density. The following chapters
discuss possible electrochemical avenues towards these goals. The aim
is to highlight opportunities in electronics technology to match current
advances in areas such as energy conversion, batteries, and analytical
chemistry.
In Chapter 1, R.C. De Mattei and R.S. Feigelson review electrochemical

methods for the deposition and doping of semiconductors. Potential
advantages of these methods over thermally driven processes include
electrical control over the deposition rate, relatively low deposition
temperatures. and applicability to a wide range of materials. Despite these
advantages, electrochemical methods have been overlooked as a route
to electronic semiconductors. The incentive for research described in this
chapter has come largely from photovoltaic applications.
The next three chapters deal with electrochemical aspects of

semiconductor processing. In Chapter 2, K. Sangwal reviews the
principles and applications of chemical etching. Although the technology
vii
viii Preface
is well established, the electrochemical viewpoint of this article provides

fresh insight for the development of improved processes.
In Chapter 3, R.T. Talasek reviews the anodic passivation of Il-VI

semiconductors such as the (Hg,Cd)Te alloys used in infrared imaging
detectors. This is one area of electronics in which electrochemical
methods have already become the industrial standard.
In Chapter 4, R.D. Rauh introduces the relatively new subject of

photoelectrochemical processing. The injection of photon energy at an
electrochemical interface adds an extra dimension to the processing
capability, be it for selective etching, patterned electrodeposition, or the
fabrication of optical elements. These concepts offer intriguing possibilities
for the future of both electronic and opto-electronic technologies.
In Chapter 5, Micha Tomkiewicz reviews photoelectrochemical methods

for characterizing the defect structure and doping levels of semiconductor
wafers. Application of these techniques on a real-time basis should
provide feedback which can be used to fine-tune the manufacturing
process, assuring consistently high quality wafers.
The quest for ever more compact circuitry requires progressive reductions
first in the width and spacing of conductor lines and second in the size of
individual circuit elements. As conductors become finer and more closely
spaced, the incidence of electrochemical migration phenomena become
increasingly critical. In Chapter 6, G. DiGiacomo reviews the principles
underlying these phenomena. Understanding gained from this review will
provide a basis for controlling or avoiding migration-related failures in
future circuit designs. The final chapter by I.D. Raistrick reviews the
subject of electrochemical capacitors: devices which promise to reduce the
size of capacitors and/or batteries utilized in electronic circuits.
August, 1991 John McHardy

El Segundo, California
CONTENTS
.
1 ELECTROCHEMICAL DEPOSITION OF
SEMICONDUCTORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Robert C. DeMattei and Robert S. Feigelson
.
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
.
2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
.
3 Elemental Semiconductors . . . . . . . . . . . . . . . . . . . 8
3.1 Silicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 8
Silicate-Based Melts . . . . . . . . . . . . . . . . . .9
Fluorosilicate-Based Melts . . . . . . . . . . . . . 14
Organic Electrolytes . . . . . . . . . . . . . . . . . . 15
.
4 Compound Semiconductors . . . . . . . . . . . . . . . . . 16
4.1 Il-VI Compounds . . . . . . . . . . . . . . . . . . . . . . 16
Aqueous Solvents . . . . . . . . . . . . . . . . . . . 16
Non-Aqueous Solvents . . . . . . . . . . . . . . . 24
Molten Salts . . . . . . . . . . . . . . . . . . . . . . . 30
Ternary Alloys and Compounds . . . . . . . . . . 30
4.2 Ill-V Compounds . . . . . . . . . . . . . . . . . . . . . . 34
Gallium Phosphide . . . . . . . . . . . . . . . . . . .34
Indium Phosphide . . . . . . . . . . . . . . . . . . .39
Gallium Arsenide . . . . . . . . . . . . . . . . . . . . 44
4.3 IV-IV Compounds . . . . . . . . . . . . . . . . . . . . . 44
Silicon Carbide . . . . . . . . . . . . . . . . . . . . . 44
.
5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
.
6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2 . CHEMICAL ETCHING: PRINCIPLES AND APPLICATIONS . . 53

Keshra Sangwal
xi
xii Contents
.
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
.
2 Mechanism of Dissolution . . . . . . . . . . . . . . . . . . 54
2.1 Driving Force for Dissolution: Some Basic
Concepts and Definitions . . . . . . . . . . . . . . . . 54
2.2 Dissolution Process Controlled by Surface
Reactions and Volume Diffusion . . . . . . . . . . . 56
2.3 Types of Dissolution . . . . . . . . . . . . . . . . . . . . 58
2.4 Dissolution Kinetics in Terms of Interfacial
Layer Potential . . . . . . . . . . . . . . . . . . . . . . . 59
Dissolution of Ionic Compounds in
Aqueous Solutions . . . . . . . . . . . . . . . . . 59
Dissolution of Ionic Crystals in Acidic
and Alkaline Media . . . . . . . . . . . . . . . . .60
Dissolution of Metals . . . . . . . . . . . . . . . . . 62
Dissolution of Semiconductors . . . . . . . . . . . 65
2.5 Dissolution Kinetics in Terms of Surface
Adsorption Layers . . . . . . . . . . . . . . . . . . . . . 72
Two-Dimensional Nucleation Models . . . . . . 73
Surface Diffusion Model ............... 75
.
3 Mechanism of Selective Etching . . . . . . . . . . . . . . 77
3.1 Models of Etch-Pit Formation ............. 77
3.2 The Slope of Dislocation Etch Pits . . . . . . . . . . 81
3.3 The Role of Impurities . . . . . . . . . . . . . . . . . . 82
.
4 Composition of Etching and Polishing
Solutions ............................... 88
4.1 Ionic Crystals . . . . . . . . . . . . . . . . . . . . . . . . 88
Water-Soluble Crystals . . . . . . . . . . . . . . . 88
Water-Insoluble Crystals . . . . . . . . . . . . . . 89
4.2 Molecular Crystals . . . . . . . . . . . . . . . . . . . . . 89
4.3 Metallic Crystals . . . . . . . . . . . . . . . . . . . . . . 89
4.4 Semiconductors . . . . . . . . . . . . . . . . . . . . . . . 91
.
5 Photoetching . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
.
6 Electrolytic Etching and Polishing . . . . . . . . . . . . 98
.
7 Gas-Phase Chemical Etching . . . . . . . . . . . . . . . 101
.
8 Morphology of Chemical Etch Pits . . . . . . . . . . . 105
.
9 Correspondence Between Etch Figures and
Dislocations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
.
10 Etching Profiles . . . . . . . . . . . . . . . . . . . . . . . . . 112
.
11 Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . 116
.
12 References ............................. 119
3 . ELECTROCHEMICAL PASSIVATION OF (Hg.Cd)Te . . . . .. 127

Robert T. Talasek
.
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Contents xiii
1.1 Types of Infrared Detectors . . . . . . . . . . . . . 127

MIS Devices . . . . . . . . . . . . . . . . . . . . . . 128
Photovoltaic Devices . . . . . . . . . . . . . . . . 133
1.2 Electrochemical Passivation . . . . . . . . . . . . . 137
Material Effects . . . . . . . . . . . . . . . . . . . . 137
Surface Preparation . . . . . . . . . . . . . . . . . 139
.
2 Anodic Oxidation . . . . . . . . . . . . . . . . . . . . . . . . 141
2.1 Oxide Composition-Phase Diagrams . . . . . . 142
2.2 Oxide Analyses . . . . . . . . . . . . . . . . . . . . . . 144
2.3 Chemical Processes . . . . . . . . . . . . . . . . . . 149
.
3 Alternative Passivation Processes . . . . . . . . . . . 165
3.1 Electrochemical . . . . . . . . . . . . . . . . . . . . . . 165
3.2 Non-Electrochemical Passivation . . . . . . . . . 168
.
4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
4 . PHOTOELECTROCHEMlCAL PROCESSING OF
SEMICONDUCTORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
R. David Rauh
.
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
.
2 Experimental Procedures . . . . . . . . . . . . . . . . . . 178
.
3 Photoelectrochemical Etching . . . . . . . . . . . . . . 182
3.1 General Background . . . . . . . . . . . . . . . . . . 182
3.2 Periodic Structures . . . . . . . . . . . . . . . . . . . 187
Holographic Gratings . . . . . . . . . . . . . . . . 187
Non-Holographic Periodic Structures . . . . . 192
3.3 Focused Laser and Related Techniques . . . . 197
3.4 Miscellaneous Applications of Photoelectro-
chemical Etching . . . . . . . . . . . . . . . . . . . . . 201
.
4 Photoelectrochemical Deposition . . . . . . . . . . . . 203
.
5 Photoelectrochemical Surface Processing . . . . . 209
.
6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
.
7 Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . 211
.
8 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
5 . PHOTOELECTROCHEMICAL CHARACTERIZATION . . . . . 217

Micha Tomkiewicz
.
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
.
2 Direct Response . . . . . . . . . . . . . . . . . . . . . . . . . 219
2.1 Current-Voltage . . . . . . . . . . . . . . . . . . . . . 220
2.2 Rotating Ring Disc Electrodes . . . . . . . . . . . . 220
2.3 Absorption, Reflection and Photo-
luminescence Spectroscopies . . . . . . . . . . . . 222
2.4 Ellipsometry . . . . . . . . . . . . . . . . . . . . . . . . 224
3. Electric Field Modulation of System's Response . 224
xiv Contents
. .
3.1 impedance . . , . . . . . . . . . . . . . , . . . . . . 225
3.2 Photocapacitance . . . . . . , . . . . . . . . . . . . 232
3.3 Optical Techniques . . . . . . . . . . . . . . . . . . . 232
Electroreflectance , . . . . . . , . . . . . . . , . , 234
Photoreflectance , . , , , , , , , . . , . , . . . , . 235
Surface Photovoltage . , , . . . , , , , , . , . . , 240
Electromodulated infrared Spectroscopy . . 240
Other Modulation Techniques , , , , . , . , , . 241
4. Time Resolved Techniques . , , , , , , , , , . , , . . , . 241
4.1 Current , , , , . . . , . . , , . . , . , . . . . . . . . . . . 242
.
4.2 Potential . , . , , . , , , . , , , . . , . . . , , , , , , 242
.
4.3 Photoluminescence . , , . , , . . , . . , , . . . .
I 243
4.4 Microwave Conductivity , . , , ..
, , , , , . , . . . 243
4.5 Surface Restricted Transient Grating , , . , . , , 243
5. Photothermal Methods , , . . , , . . , , . , , , , , , , , , 244
6. Topographical Studies , , , , , . , , , ... , , , , , . , , 245
7. Acknowledgement . . , , , . . , . , , , , . . , , . , . . . . 245
8. References 247
6. ELECTROCHEMICAL MIGRATION . , , . . , , . . . . . . , , , , , , 255

Giulio Di Giacomo
1. Introduction . . . . . . . . . , . . , . . . . . . . . . . . 255
2. Model . , . . . . . , . . . . . . , . . . . . . . . . . . . . , . . . . 260
2.1 Current Density Through an Electrolyte , . . . . 260
Water Availability as a Function of RH . , . . 261
. .
B.E.T. RH-Function . . , . , . , . . , , , . . . 262
B.E.T. Time-to-Failure Model for
.
Dendrites . . , , . . . . . , , . . . . . . , . . . . 263
RH Function Based on Pore Distribution . . 263
t, for Dendrites Based on Pore
.
Distribution , , , . . . . , , , . , . . . , , . . , . 264
2.2 Current Density Through a Polymer Coating . . 265
3. Experimental . . , . . . , , . , . . . . . . . . , . . . . . . , . . 266
3.1 Parameters for Dendrite Model . . . , , , . . , , 266.
3.2 Parameters for Leakage Model . . . . . , . . . 273
3.3 Water-Drop Migration , , , . . , , . . . . . . . 276
n
Ag, Pb, and Cu Films . . , , , . . . . . . . . . . 276

Ni Films . . , . , , , . . . . , , , . , . . . , , . . . . 278
Cu-15% Ag-2.5% P Wires . , . . . , , . . , . 281
4. Discussion . . . . . . . . . . , . . . . . . . . . . . . . . . . . . 282
4.1 Model Acceleration and Materials/Process
Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
4.2 Materials Characterization by Water-Drop . . 285
4.3 Effect of Active Impurities . , , , . , , . . , , , . , , 286
Contents xv
4.4 Mechanism and Time-to-Failure Results . . .. 287

4.5 Polymer Coating . . . , , . . . . . , . . . . . . . . .. 289
5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
.
6. Acknowledgement . , , . , . . , . . , . . . , , , . . . .. 292
7. References , , , . , , , . , . . . , , . , , , , . . , , , . . . , , 293
7. ELECTROCHEMICAL CAPACITORS , . . , , , , . . . , . . . , . . 297

Ian D. Raistrick
. .
1. Introduction . . . . , , . . , . , , . , , , , . , , . . . . , . 297
2. Mechanisms of Charge Storage at the Electro-
chemical Interface , . , , , , , , . , , . . , . , , , , . . , , , 300
2.1 Double-Layer Capacitors . , . , , . . , , , . . , , , 300
2.2 Electrosorption Capacitance , , , . . . , . . . , , . 302
2.3 Surface Redox Processes , , . . , . , , , , . . , . 306
2.4 Thin-Film Bulk Reaction . . , . , . . , , . . . . . . . 308
3. Dynamic Behavior of Electrochemical
.
Capacitors . . , , , . , , , . . . . . . . , , . . . . . . . . . . 31 1
3.1 General Considerations . . . . , , . , , , , . . , , , 31 1
3.2 Uniform Transmission Line Model of the
Response of Porous Electrodes . . , . . . . . . . 312
3.3 Rough Electrode Surfaces , . . , . . , , . . . , . 315
3.4 Nonuniform Pores , , , . . . , , . . . . , , , , , , , . 317
3.5 Mass Transport into a Thin Film . . , , , . . , , , 320
3.6 Large-Signal Response . . , . . , . , , , . . , , . . 321
4. Carbon Electrochemical Capacitors . . , , , . . . . . 322
4.1 Introduction , , , , . , , , , . . , , . . , . , , , , . , , . 322
4.2 Systems with Aqueous Electrolytes . , , . . . , . 324
4.3 Nonaqueous Electrolyte Systems . , . , . . , , . 326
4.4 Solid Electrolyte Systems . . . , , . . , , , . . . . . 330
5. Transition and Noble Metal Oxide Capacitors , , , 331
5.1 Introduction , , , , . , , . , . , . . , , , , , , , , . , , . 331
5.2 Thermally Prepared Oxide Films . . . , , . . , , . 332
5.3 Anodically Prepared Films , . , , . . , , , . . , , , 336
.
5.4 Other Oxides . , . , . . , , . . , , . . . . . . . . .
I . 339
5.5 Metal Oxide Capacitors Utilizing Solid
and Polymeric Electrolytes . , , . . . . , , . . . . . 340
6. Conducting Polymers . . . . , . . . , , , , . , , , , . . , . 340
6.1 Introduction , , , , . , . , , . . , , , , . . . . , . . , . , 340
6.2 Charge-Storage Mechanism , , , , , . , , , , , , . 341
6.3 Electrical Response , , , . . , , . , , , , , , . . , . . 342
7. Conclusions , . . , , , . . . , . , . . . . , , . . , . , . . . , . 342
8. Acknowledgements , , , . . . . ............. 347
9. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
INDEX ....................................... .... 356

1
ELECTROCHEMICAL DEPOSITION OF
SEMICONDUCTORS
Robert C. DeMattei
Robert S. Feigelson
1.0 INTRODUCTION
Prior to the invention of the transistor and the birth of the semiconductor
industry, the field of electrochemistry was already very advanced with
respect to both theoretical understanding and industrial applications. It
is therefore surprising that electrochemical preparative techniques did
not play a significant role in the developmentof semiconductor materials.
The reason for this is unclear, but during the nearly forty years that have
elapsed since then, there have only been a few scattered papers published
in this field. When you compare this miniscule effortwith the vast body of
published papers on the research, development, and manufacturing of
semiconductor materials by other methods, it is not surprising that
electrochemical methods have not yet made a serious impact on this
multibillion dollar industry.
Most semiconductor materialsfor opto-electronic applications must
be in the form of single crystals with exceptional crystalline perfection and
purity. Typically, large boules are sliced into wafers, and devices are
prepared by either diffusing dopants into them and/or by depositing on
them compounds of either similar composition (homoepitaxy)or different
composition (heteroepitaxy). Some semiconductors in polycrystalline
film or bulk form have also been found useful in a few applications, the
most important being low cost solar cells. This latter application has
stimulated much of the recent work on the electrolytic deposition of
semiconductor materials.
Electrochemical preparative methods can be conveniently divided
into two categories: 1) low temperature techniques (usually aqueous
1
2 Electrochemistry of Semiconductors and Electronics
solutions, but organic electrolytes are sometimes used), and 2) high

temperature techniques (molten salt solutions). By far, the greater effort
has gone into low temperature processes because these systems are
simple to construct, operate and control, and because aqueous solution
chemistry is much better understood then complex molten salts. The
large metal plating industry (Cr, Au, Ag, Cu, etc.) is based on aqueous
electrochemical techniques.
The choice of solvent or electrolyte depends to a large extent on the
ability to put appropriate ions in solution. Low temperature solvents are
not readily available for many refractory compounds and semiconductor
materials of interest and, although aqueous techniques are preferablefor
the reasons stated above, they are often unsuitable. As a result, molten
salt electrolysis has found utility for the synthesis and deposition of
elemental materials such as AI, Si and a wide variety of binary and ternary
compounds such as borides, carbides, silicides, phosphides, arsenides,
and sulfides, and the semiconductors SIC, GaAs,and GaP and InP(1)(2)(3).
Molten salt electrolysis has proven to be a commercially important means
for refining aluminum from bauxite ore (the Hall process) and for alkali
metal separation.
While small single crystals of many compounds have been produced
electrolytically from molten salts as well as aqueous solutions, scaling up
to large size has generally been difficult. The subject of using molten salt
electrolysis for crystal growth was reviewed by Feigelson (3).
One of the unique features of electrodeposition is that it is an electrically
driven process capable of precise conlrol. This offers a potential advantage
over most other processing techniqueswhich are thermally driven. Other
attractive advantages include: 1) growth temperaturesare well below the
melting point so that the point defect concentration is low, 2) the solvents
have afluxing action on the cathode surface dissolving oxide impurities,
3) purification occurs during electrodeposition because of differences in
deposition potential between major and minor components in solution
(however,doping with certain elements is possible and can be controlled
through changes in concentration), 4) a wide range of compounds and
elements can be electrodeposited, and 5) electrolysis is convenient for
epitaxial deposition since growth occurs uniformly over the sample area.
The ability to produce thin uniform films on both simple and complex
shapes has been one of the traditional strengths of electrochemical
methods, and it is not surprising that the majority of the semiconductor
electrodeposition studies have concentrated on thin film deposition.
Semiconductor materials can be divided into two broad categories:
elements and compounds. The latter category may be further subdivided
by referenceto the column in the periodic table from which the constituent
elements come, and whether the compound is a binary or higher order.
Electrochemical Deposition of Semiconductors 3
This article reviews the history and most recent results of electrodeposition
of various semiconductors, including: 1) Si; 2) the Ill-V compounds,
GaAs, GaP and InP; 3) the Il-VIcompounds, CdSand CdTe; 4) Sic; and 5)
the important ternary compound CulnSe,.
2.0 THEORY
The electrodeposition of semiconductor compounds, like any other

chemical process, is governed by thermodynamic considerations. In the
case of electrodeposition, the reactions are thermodynamically unfavorable;
that is, the overall free energy change (AG) for the reaction is positive and
the electrical energy supplies the needed energy to drive the reaction.
Consider, the case of an ion M+"being reduced to M:
The change in free energy is given by (4)(5):
Eq. (2) A G = AGO + RTIn(a,/a,,,)
where R is the gas constant, T is the absolute temperature and ai is the

activity of species i. Activity is used instead of concentration in Eq. 2 to
account for the interaction of ions in solution, or for the difference in
reactivity of an atom in a molecule vs. that of an atom in the elemental state
where the activitywould be 1. In the solution case, activity is related to the
concentration by the activity coefficient,
A more complete discussion of activity and activity coefficients can be

found in references 4 and 5. For the sake of practicality, concentrations
will be used in the following discussion. Thus, Eq. 2 becomes:
Eq. (4) AG = AGO + RTIn( 1 / [ M + " ] )
It can be shown that (5)
where n is the number of moles of electrons involved in the reaction, F is

Faraday'sconstant and E is the potential. Equation 4 may now be written:
RT 1
Eq. ( 6 ) E = E,-- In-
nF [M+"]
where E, is the standard electrode potentialfor reaction (Eq. 1) referenced
to the standard electrode with [ M+,] = 1 mole/liter. Tables of standard
electrode potentials exist for aqueous solutions and some non-aqueous
systems.
A single electrode reaction such as given in Eq. 1 can not stand alone
since there must be a compensating reaction involving an oxidation
process. The overall reaction can be represented by:
Eq. (7) bAfa + a B b = bA + aB

and the electrode reactions by:
+ ae
-
Eq. (8a) A+a - 4 A (reduction at cathode)
Eq. (8b) B-b B + b e- (oxidation at anode)
The cell potential is given by
The reactions described in Eqs. 7 thru 9 are typical of those involved

in the deposition of an elemental semiconductor such as silicon or
germanium. The situation is somewhat more complicatedfor the formation
of a compound semiconductor such as GaAs or CdTe. In this case, two
materialsmust be codeposited at the cathode, and one of the species, the
nonmetal (As or Te above), is normally considered an anion. This
component of the semiconductor must be introduced into the solution in
a form such that it can be reduced at the cathode. This is usually
accomplished by using a starting compound that incorporatesthe desired
non-metal as part of an oxygen-containing ionic species (ASO;~ or
T e 0 i 2 for example). In general terms, the reactions involved in this
deposition would be:
Eq. (1Oa) M+, + m e- 4 M
Eq. (lob) NOin+(2y-n)e-- N+YO-~
Eq. (1Oc) 0-24 0 . 5 0 2+ 2e

yielding an overall reaction:
Eq. (1Od) 2Mfmt 2NOy-"t (m - n)02-2MN t (y t (m - n)/2) 0,

The cell potential is then given by:
RT 1
Eq. (1 1) E = Eo' -( 2 y t m - n)F In
[M'm]2[NO;n]2[0-2](m-n)
where Eiis the sum of the ELsfor reactions 10 a, b,c. From a practical point
of view, it is important to ensure that reactions 10a and 10b occur
simultaneously. This will occur if the potentials for the two reactions are
equal. The two cell reactions are:
Eq. (12a) 2M'"' t m0-24 2M t 0.5m0,
Eq. (12b) 2NOy-" 4 2N t (y - n/2) O2 t no'*
with cell potentials given by:
RT 1
Eq. (1 3a) EM=EM-- In
O 2mF [M+m]2[0-2]m
and
RT (02)"
Eq. (13b) EN=E"'- In -
O 2[2y-n]F (NOy-")'
where EoMand EoNare the sum of the standard potential for reactions 1Oa
plus 1Oc, and 1Ob plus 1Oc, respectively. Since the desired condition for
codeposition is EM= EN, Eqs. 13a and 13b can be combined to yield an
expression for determining the solution composition for codeposition:
Equation 14 is useful only if the Eis are known for the various species in
the solvent system being used. Often the investigator does not have this
information. The solution to this problem is the use of voltammetry. In this
technique, the voltage across an electrochemical cell is slowly increased
and the current is monitored. Ideally, there is no current flow until the
deposition potential is exceeded, as shown in Fig. 1 . In most practical

cases, some extrapolation of both the baseline and rising portion of a
current vs. voltage plot (I-V plot) is necessary to determinethe deposition
potential (Fig. 2 ) . A series of I-V plots with differing solution concentrations
will give the variation of deposition potential with concentration. Repeating
this procedure for each element in the semiconductor will give the range
of solution conditions under which codeposition of the elements is possible.
The foregoing discussion is a brief introduction to those parameters
which influence the thermodynamic aspects of electrodeposition. Kortum
and Bockris (6), and Bockris and Reddy (7) present a much more complete
discussion of the natureof ions in solution and the processes occurring at
electrified interfaces.
Thermodynamics is concerned with the equilibrium aspect of
deposition. Once the potential between theelectrodes is raised abovethe
deposition potential, the system is in a non-equilibrium condition and
kinetics must be considered. While it is possible to determine some
aspectsof the deposition process such as the electron transfer processes
occurring at the electrodes and the rate-controlling step for deposition,
from a practical point of view, it is more important to determine those
conditions which will yield a smooth deposit and what the expected
growth rate will be under those conditions.
In most cases, the rate of electrodeposition is limited by the onset of
dendritic growth on the electrode. This will occur if some critical current
density (i,) is exceeded. Despic and Popov(8) developed an equation to
determine this current density:
where io is the exchange current density, Q~ and aaare the cathodic and
anodic transfer coefficients, respectively, and q is the overpotential (the
difference between the potential measured between an electrode and a
referenceelectrodewith and without currentflowing) (7). Thevaluesof io,
ac and aa can be determined ( 7 ) . The value of iL,l, the limiting current
density to a flat plate, may be calculated from the expression (9):
If the values are available in the literature or if an investigator wishes to

determine them for his system, it is possible to calculate the critical
current densityfor dendritic growth and thus carry out the deposition at its
highest rate. The alternative is to accept the suggestion of KrolI (10) that
for most systems iL,l, which is less then or equal to ic (Eq. 15), equals 40
Figure 1. Theoretical current vs. voltage (I-V) plot showing deposition

potential (Ed).
400-
0
n
0
I
G
u
200 -
-a
-E
+
Figure 2. I-V (voltammogram) plot of 1 m/o K,SiF, in Flinak at 75OoC

showing extrapolated deposition potential (Ed).
mA/cm2.
In order to determine the rate of growth of the depositing layer, the
amount of material deposited per unit of time must be determined. Faraday’s
law of electrolysis gives the weight of material deposited by a given
amount of charge (4) as
where M is the molecular weight of the material and c is the deposition

efficiency. The volume is given by:
Eq. (18) v = (MEq) / (pnF) = Ay
where p is the density of the material, A is the area of deposition and y is

the deposit thickness. The rate of growth (dy/dt) is given by:
M€
-d y-- - dq=--M E I = - ME i
Eq’(19) dt p n F A dt pnF A pnF
The term M/pnF is a constant for any given deposition process and i is in
amps/cm2. As an example, consider the deposition of cadmium sulfide
(CdS) from cadmium ions (Cdt2) and sulfate ions ( S 0 i 2 ) . From Eq. 11,
the number of electrons transferred in Eq. 19 is 8, the density is 4.82 gm/
cm3 and the molecular weight is 144.46 gm, which yields a value for the
growth rate constant of 3 . 8 8 ~O1 5 cm3/(amp-sec). Using a value of 40
mA/cm2 (0.040 amp/cm2) as a practical current density, the maximum
growth rate for CdS (with E = 1) is 5 . 5 9 ~ 1 cm/hr
0~ or 55.9 pm/hr. In
practice, the actual value is somewhat less since e is usually less than 1.
3.0 ELEMENTAL SEMICONDUCTORS
The most important elemental semiconductor materialfor industrial

applications is silicon. Because of its commercial importance, the
electrodeposition of silicon has been studied to a greater extent than all of
the other semiconductor materials combined.
3.1 Silicon
Introduction. The first attempts to produce silicon electrolytically

date from the mid 1800’s. St. Claire De Ville (1 1) claimed that he produced
silicon as the result of the electrolysis of impure molten NaAICI,. Since the
material did not oxidize at white heat, the claim was probably not true.
Monnier(12) reports that DeVille did deposit silicon as platinum silicide on
a platinum electrode from a meltof NaF/KF containing SiO, at a later date.
In 1854, Gore (13) claimed to have produced silicon by the electrolysis of
an aqueous solution of potassium monosilicate. This was never confirmed
and silicon has never been deposited from any aqueous system. Ullik
(14), in 1865, was probably the first to deposit elemental silicon when he
electrolized a solution containing K,SiF, in KF. Iron- and aluminum-
silicon alloys were produced from solutions containing SiO, and iron or
aluminum oxide in NaCl t NaAIF, by Minet (15). Warren (16) produced a
silicon amalgam from SiF, in alcohol using a mercury cathode. All of this
work before 1900 established that both SiO, and fluorosilicates could be
used as source materials for silicon electrodeposition. This work also
showed that alkali halides as well as organic solvents were suitable
solvent materials for the process.
More systematic studies of silicon electrodeposition began in the
1930’s with Dodero’s (17)( 18) investigation of the electrolysis of molten
silicates at temperaturesof 800 to 125OoC. The very high potentials used
in these studies would be expected to liberate not only silicon but also
alkali and alkaline earth metals. There is no conclusive proof in Dodero’s
work that silicon was the primary cathode product or the result of a
reduction of the silicon containing compounds by alkali or alkaline earth
metals that had been produced by electrolysis. His best result was 72%
silicon produced from a melt composition of 5 SiO, - 1Na,O - 0.2NaF
electrolyzed at 1 15OoC.
The melts used for the electrodeposition of silicon can be broadly
classified by the silicon-containingspecies used: silicates and fluorosilicates.
Each will be discussed separately.
Silicate-Based Melts. Silicate or SiO, melts have been studied by
several investigators in an effort to develop a commercial process for
electrowinning silicon. The molten solutions most often studied contained
SiO, in cryolite.
Cryolite, Na, AIF,, was a logical choice as a solvent for use with SiO,
because of its ready availability and its successful use in the Hall process
for electrowinningaluminum. The SiO,/Na,AIF, systemwas studied both
in a laboratory environment and in pilot plant trials, first by Monnier, et al.
(12)( 19)(20), whose interest was in producing pure silicon, and later by
Grjotheim, et al. (21-24), whose primary interest was in AI-Si alloys. This
high temperature solution chemistry is not simple, leading to mixtures of
aluminum silicates and sodium aluminosilicates (12).
Monnier and his co-workers were able to obtain 99.9 to 99.99% pure
-
silicon from SiO, cryolite solutions in a two step process. The first step
was the deposition of silicon to form a molten copper-silicon alloy at the
cathode. The anode in thiscellwas graphiteand the measureddeposition

potentials at zero current could be calculated from thermodynamic data.
The second step involved using the cooper-silicon alloy as an anode and
electrorefining the silicon.
Monnier (12)( 19) was also responsible for the only reported pilot
plant study of the electrodeposition of silicon from Si0,-cryolite. The
study began in 1957 and the pilot plant was built and operated between
1960 and 1966. Two versions were built. One used auxiliary carbon
heating electrodes and operated at a maximum deposition current of 300
amp. The second furnace operated at currentsof up to 3000 amp and was
self heated as in the Hall cell for aluminum. Figure 3 is a diagram of
Monnier's cell. The current densities ran as high as 800 mA/cm and the
deposits were in the form of 1 - 3 mm crystals. After removal from the
solidified melt and zone refining, the silicon was reported to be of
semiconductor quality.
The limitation of this process is that silicon is deposited as a solid
which limits the rate of deposition as discussed earlier in this chapter and
also by Huggins and Elwell (26). Monnier's approach of depositing into a
liquid alloy cathode is one solution to this problem, but it does require a
second step toremove the alloying metal. The Hall process for aluminum
gets around this problem by depositing the metal above its melting point.
A second benefit arisesfrom the high currents used; after initial start up,
the electrolysis currents used provide enough Joule heating to keep the
system molten. A process similar to the Hall process was developed for
silicon by DeMattei, Elwell and Feigelson (27)(28) in 1981.
The main problem in using the Hall process for silicon electrodeposition
is that silicon melts at a much higher temperaturethan aluminum (1412OC
compared to 660OC). Cryolite can not be used at this temperature due to
volatization problems, so a binary or ternary melt containing SiO, had to
be developed that would be stable above this temperature. Johnson(29)
indicated that calcium and magnesium based silicate melts looked favorable,
while other alkaline earth and alkali metal silicates were less desirable.
DeMattei, Elwell and Feigelson (27) were the first to successfully
demonstrate a process for the electrodeposition of Si above its melting
temperature. The actual melt composition they preferredwas the eutectic
-
composition in the BaO-SiO, system (53% 47% by weight). About 15%
bariumfluoride was added to reduce viscosity. These melts were electrolyzed
at about 145OoC in the furnace shown in Fig. 4 using graphite crucibles
and graphite electrodes. Potentials in the range of 1 to 8 volts were used
together with currents of 0.1 to 2.0 amps for an electrode area of about 2
cm2. The electrodeposited silicon formed into spherical droplets (Fig. 5)
and, because of their lower density, floatedto the top of the melt. Faradaic
efficiency ranged from 20% to a high of 40% which is less than desiredfor
ElectrochemicaI Deposition of Semiconductors 11
ANODE
''
/ '\
+' -
CATHODE \+ ~ HEATER ELECTRODE
LID
CRYOLITE-
RICH
LlOUlD
SILICA-RICH
LIQUID
CONTAINER --I--
Figure 3. Cell for silicon production in which the electrolyte is contained

in an unmelted solid of the same composition (12)( 19).
12 Electrochemistryof Semiconductors and Electronics
4 ROTATION /TRANS L AT 1O N
VACUUM FEEDTHROUGH
I_(-
VACUUM Wi SUPPLY
G
S
U ' POWER FEEDTHROUGH
(WATER COOLED)
SHE/ HED
THEF 4OCOUPLE
Figure 4. Furnace used for electrodeposition of silicon and silicon carbide.

Figure5. Silicon lump (1.6 gm) produced by electrodepositionabove the

melting point (27).
commercial applications. The silicon produced by this method had a

typical purity of 99.98%with the main impurities being titanium (60 ppm),
and aluminum and iron (20 ppm). This is close to the purity required to
produce 10%efficient solar cell with one step of purification (30). The heat
generated by Joule heating in a commercial sized plant was judged
sufficient to maintain the temperature at 145OoC or above. Olson and
Kibbler (31) used tin in place of the aluminumor copper used in Monnier’s
(12) earlier work. This eliminated the shallow and deep-leveltraps which
these elements can induce in silicon. Their silicon was deposited from a
melt containing SiO, dissolved in Na81F6/LiF. Solidification of the Sn-Si
solution produced 1 - 10 mrn crystals of silicon which contained 10 ppm
of transition metals and exhibiteda resistivityof 0.05 - 0.1 n cm. The major
drawback of this approach (like Monnier’scase) was the need to separate
the silicon from the solidified tin.
Fluorosilicate-BasedMelts. Compounds containing fluorosilicate
ions (SiFi2)are a relatively inexpensive by-productof the fertilizer industry.
They may also be produced by the reaction of SiO, with alkali and alkaline-
earth fluorides at 1000 - 1 1OO°C (32). The process for deposition of Si
fromfluorosilicates is analogousto that usedfor titanium (33)(34)(35)and
suffers from the same limitations which makes it attractive only for special
applicationswhich may include films for solar cells, surface coatings, and
(in the form of powder or sponge) as a replacementfor metallurgicalgrade
silicon.
A study of silicon film deposition using K,SiF, was begun at Stanford
Universityin the Center for Materials Research in the early 1970’s. Of the
several fluoride systems studied, only LiF-KF and LiF-KF-NaF melts
proved suitable. In 1977, Cohen (36) reported that epitaxial Si layers
could be deposited from a LiF-KF eutectic containing K,SiF,. He also
produced continuous films using a dissolving silicon anode.
Similar melts were studied for the electrowinning of Si for use in
further processes. Initial work was done using graphite anodes and either
LiF-KF-NaF or LiF-KF eutecticsat 75OoC. In initial experiments by Rao, et
ai. (37)(38), silver was used as a cathode. Later experiments (39)(40)
used inexpensive graphite substrates. The concentration of K,SiF, was
-
maintained between 4 20 m/o to prevent the formation of powdery
deposits (38). Typical deposition conditions were: 1) constant current
experiments at 10 to 25 mA-cm”, and 2) constant potential experiments
at -0.74+ 0.04 volts versus Pt. Columnar growths (250 p m grain size) of
3 - 4 mm thickness were produced in 2 to 4 days. The electrodeposited
silicon was typically 99.99% pure although the best samples had impurity
levels less than 10 ppm. Current efficiencies could be as high as 80%.
Olson and Carleton (41) extended the use of these melt systems by
using a copper-silicon anode to simulate electrorefining metallurgical
grade silicon. In the same vein, Sharma and Mukherjee (42)investigated

the semi-continuous production of 99.99% pure silicon powder from
97.5% pure metallugical grade silicon. Bouteillon, et al. (43)(44),
demonstrated that improvements in both morphology and purity (to less
than 1 ppm impurity) could be obtained using pulsed electrolysis as
currently applied in copper refining.
Organic Electrolytes. One of the major drawbacks of any molten
salt process for electrodeposition is the energy needed to maintain the
system in its molten state. The energy that is added either by external
heaters or by Joule heating must add to the cost of the final product.
Austin (45)wasthefirstto reporttheelectrodepositionof amorphous
silicon from organic solvents in the temperature range of 20 to iOO°C.
Amorphous Si (a-Si) has been used commerciallyfor a number of years in
low cost solar cell applications (particularly calculators and watches).
The solvents of choice were aprotic materials such as propylene carbonate
or tetrahydrofuran using silane or a silicon halide as a source of silicon.
The deposits which contained impurities of less than 10 ppm showed a
resistivity of about 20 cm.
Attempts by Bucker and Amick (46) to reproduce Austin's work
produced films which were unstable in air. Heat treatment at about 35OOC
was recommended to remove hydrogen from the films before they were
exposed to air. Their recommended solvents were tetrahydrofuran/
benzene, tetrahydrofuran/toluene, dioxolane/benzene and dioxolane/
toluene, with silicon tetrachloride or trichlorosilane as solutes. The films
contained both chlorine and chromium in trace amounts.
Krogerand co-workers (47)(48)didaseriesof detailedstudieson the
deposition of amorphous silicon including the influence of dopants. Lee
and Kroger (47) investigated the deposition of fluorinated a-Si which
should have higher stability than hydrogen-containing films. Since fluorine
would normally be deposited at the anode, they used solutions of SiF, in
ethyl alcohol, dimethyl sulfoxideandacetonitrile. Potassiumfluoride was
added to produce K,SiF,, which contains a cathodic fluorine-containing
ion, and HF was added to increase conductivity. Deposition onto either
nickel or stainless steel cathodic substrates was carried out in a Teflon
vessel under an argon atmosphere using a platinum anode. The film
resistivity was about 10l2n cm. Phosphorus doping could change the
film from p- to n- type. The lowest resistivity observed was 10" a cm.
Rama Mohan and Kroger (48) investigated the deposition of a-Si
using tetraethyl orthosilicate or silicic acid in ethylene glycol or formamide-
ethylene glycol containing HF. Undopedfluorine containingfilmswere p-
type. Doping with phosphorus from triethyl phosphate reduced the
-
resistivity to 1O5 n cm.
A more complete discussion of silicon electrodeposition can be
found in Elwell and Rao (49).
Next Page
Previous Page
4.0 COMPOUND SEMICONDUCTORS
Compound semiconductors can be chosen to match their

optoelectronic properties to a particular application. This is particularly
true of ternary and higher order alloys and compounds in which the
stoichiometry as well as dopants can be used to vary the semiconductor
properties. These compounds are classed by the chemical groups to
which their constituent elements belong. Thus there are 11-VI, Ill-Vand IV-
IV type semiconductors. In the following discussion, ternary and higher
order compounds will be classed with the group of compounds which
contain the same non-metallic element.
4.1 Il-VI Compounds
Aqueous Solvents. The interest in the electrodeposition of Il-VI

semiconductors arosefrom the use of these compounds in solar cells and
photoelectrochemical energy conversion and storage. What appear to
be the initial twoarticles in this areawere published almost simultaneously
in 1971. Hodes, et al. (50),reported the growth of a 1 cm2polycrystalline
layer of CdSe on titanium. The cadmium and selenium were codeposited
from a solution of CdSO, and SeO,. Using 4.5 coulombs per electrode-
side, about 7.5 mmol of CdSe was deposited. No concentrations,voltages,
currents or current densitieswere reported. Two months later, Miller and
Heller (51) produced layers of CdS and Bi,S, by the anodization of
cadmium and bismuth in polysulfide solutions. CdS was produced using
-
a 1F Na,S solution and Bi,S3 using 1F Na,S 0.05F S. Again no voltages,
currents or current densities were reported.
The mechanism for the formation of sulfide films on cadmium was
studied by Peter (52). The anodizationwas performed on polished 99.999
(5N) polycrystalline cadmium rods in a solution of 0.1 M Na,S and 1 .O M
NaHCO,. A referenceelectrode consisted of a pool of mercury in contact
with red mercuric sulfide. The potential of this electrodewas estimatedto
be -0.74 volts on the hydrogen scale. A voltammogram (current versus
potential referenced to the reference electrode) showed three features
(Fig. 6).
A peak beginning at -0.5 volts corresponded to the formation of the
first monolayer of the sulfide. The plateau region and the rise following it
were investigated by a series of electrochemical experiments and
comparison to modelswhichwill not be detailed here. The results showed
that in the plateauregion the film grew to a thickness of about 5 nm by high
field ion migration. The donor concentration in this region was as high as
102’m-3. Above approximately 1.O volt, the film grew to approximately
500 nm by a diffusion controlled process. This portion of the film was
1 I I I
0 0 5 I O 15
V vs Hg / HgS
Figure 6. Linear sweep voltammogram. Polycrystalline cadmium electrode

(area 0.125 cm2) in 1 M NaHCO, t 0.1 M N%S. Sweep rate 100 mVs-’.
a) monolayer region, b) plateau region, c) transpassive region (52).
porous or polycrystalline. The process ended when the layer began to

crack.
Panicker,et al. (53), investigatedthe cathodic codeposition of cadmium
and tellurium to form CdTe layers. This paper analyzed the deposition
reactions in light of the existence range of CdTe and the equilibrium
potentials developed in this range. Under all conditions cadmium is the
potential-determining species and the deposition potential is given by:
Eq. (20) v, = -0.403 t -

0.295 log aCd2+AV
at the CdTe/Cd phase boundary and by:
Eq. (21) V, = 0.143 t 0.295 aCd2+

-AV
at the CdTe/Te phase boundary. AV is the overpotential and is given by
where q is the discharge overpotential, i the current density and R the

resistance per square centimeter of electrode between the cathode contact
and the saturated calomel electrode (SCE) used to measure the cathode
voltage. Depositions were done from aqueous solutions of cadmium
sulfate (1mole/liter) at a pH of 2.5 to 3. In some cases the Cd solution was
purified by pre-electrolysis. Tellurium oxide was added to the cadmium
solution. Concentrations of 1O 5 to 1O 3 moles/liter were used for unsaturated
solutions and excess TeO, was used for saturated solutions. In the
unsaturatedsolutions, twoanodeswere used, onewasTe (to maintainthe
Te concentration) and the other an inert material. The ratio of current
density (iTe/ iinert)was
2. After a number of experiments, Panicker and his
co-workerswereabletoestablish that the films depositedfrom CdSO,/Te
solutions having a rest potential (see ref. 53) between 0.2 volts and 0.6
volts vs. SCE were CdTe. When these films were deposited at room
temperature,they were amorphous. Between 35 and 90°C, the films were
semicrystallinewith a grain size of 500 to 1000 angstroms, and annealing
at 35OOC in argon increased the grain size to > 5000 angstroms. The
depositedfilms were n-type when -E,,,, > 0.3 volts and p-type when -E,,,,
e 0.3 volts. Films could also be doped with In (donor) or Cu (acceptor)
and were n- or p-type respectively for all deposition conditions.
Similar studies were done for cadmium sulfide films by Power, et al.
(54). The films were cathodically deposited according to the reaction:
Cd2+ t [SIt 2e- = CdS

The source of [SIwas thiosulfate (S,0,2-). A voltammogram of a 0.1 M
thiosulfate/0.002 M cadmium sulfate solution (pH 6.7)on a rotating

platinum cathode demonstrated that the reaction producing CdS did not
proceed to any appreciable extent (Fig. 7).The dominant reactions were
the deposition and stripping of cadmium. Lowering the pH to 2.8 by
addition of dilute sulfuric acid causedthe decomposition of the thiosulfate
to give colloidal sulfur and altered the voltammogram (Fig. 8 ) . The
cathodic (deposition) peak is increased (probably due to hydrogen
deposition) and the anodic peak is diminished and shifted to a more
positive potential. The authors attributed this anodic peak to the stripping
of excess cadmium from the CdSfilm. Reduction of the excess cadmium
ion in solution eliminated this peak. Repeated cycles under these conditions
gave a deep yellow, uniform, translucent film. These films were shown to
be photoactive in sulfide, thiosulfate, sulfite and sulfate solutions. The
shape of the I-V curves of the cathodically deposited films under illumination
in 0.1 M sodium sulfide solutions is different in both shape and efficiency
from anodically deposited films. The donor density of the cathodic film
was shown to be considerably higher by capacitance measurements.
Microprobe analysis of the cathodically deposited films showed roughly
equal amounts of Cd and S with no detectable contaminants.
CdTe film electrodeposition and photochemical behavior studies
were continued by Takahashi, et al. (55). These films were deposited on
either nickel or titanium electrodes from solutions of cadmium sulfate,
tellurium oxide and sulfuric acid. The I-V plot of a solution which was 1 M,
1 mM and 0.05 M in the respective reactants showed a current plateau
between -0.30 and -0.65 volts vs. a Ag/AgCI standard electrode. The I-V
curves were strongly affected by the TeO, concentration and a linear
relation was found between the limiting current and the concentration of
tellurium oxide (Fig. 9). The concentration of the cadmium ion had little
effect on the I-V curves. The thickness of the deposited films could be
correlated to the total charge passed through the cell (Fig. 10). X-ray
diffraction data confirmed that the films were CdTe (56). At the more
positive deposition potentials, a peak for Te was observed in the x-ray
patternswhich became stronger as the potentials became more positive.
The authors developedan exact expressionfor the total current based on
the reactions:
Eq. (23) HTeO,' t 3H+ + 4e- = Te t 2H,O
Eq. (24) Cd++ + 2e- t Te = CdTe

Under the conditions of the experiment,the exact expression reduced to:
-
i = 6DFC,,,02t / d
Figure 7. I-V plot of a solution containing 0.1 M Na,S,O, and 2x10-3 M

CdSO,, pH 6.7, at a Pt rotating disc electrode. Rotation speed 500 rpm.
Potential ramps speed 10 mVs". Temperature 25.OoC (54).
a
E
\
W
Figure 8. I-V plot of a solution containing 0.1 M Na,S,O, and 2x10.3 M

CdSO,, pH 2.8, at a Pt rotating disc electrode. Rotation speed 500 rpm.
Potential ramp speed 10 mVs". Temperature 25.OoC (54).
Figure 9. I-V plot of Ni electrode in a sulfuric acid solution (pH 1.4)

containing 1 M CdSO, and 1 mM TeO,. Insert: TeO, concentration
dependence of the current at -0.35 volts (vs. Ag/AgCI) (55).
Charge IC
Figure 10. The thickness of the electrochemically deposited films of Ni

as a function of charge passed. The deposition was carried out at -0.35
volts (vs. Ag/AgCI) in a sulfuric acid solution (pH 1.4) containing 1 M
CdSO, and 1 mM TeO,. Solid line shows the theoretical value calculated
by assuming six-electron process (55).
22 Electroc hemist ry of Semiconductors and Electronics
where D is the diffusion coefficient, F Faraday's constant, cHTsolt, the

concentration of HTeO,' and d the diffusion boundary thickness. Using
a density of 6.2 gm/cc for CdTe and the fact that the deposition is a six
electron process as indicated by the above equation, the authors calculated
a deposition efficiency of 75%.
The authorsalso studied both the diffuse reflection spectrum and the
photochemical behavior of the deposited films. The eff iciency of cathodic
photocurrent productionwas dependent on the deposition potential. The
efficiency peaked at a deposition potential of -0.40 volts. The authors
attributed this to the presence of excess Te at potentials more positive
than -0.40 volts and to the n-type character of the CdTe at potentials more
negativethan -0.40 volts. The effects of heat treatment and etching were
also studied. Heat treatment improved the photocurrent due to the
increased grain size which eliminated grain boundaries. Etching had the
effect of removing the surface layer of Te and improving photocurrents.
One of the potential advantages of electrochemical growth techniques
is in controlling impurities. Pre-electrolysis of the solution can reduce or
eliminate impurities. The proper selection of deposition conditions and
potential can prevent the codeposition of impurities, and complexing
agents can be added to render the impurities inactive. An example of the
latter approach isseen in theworkof PandeyandRoozon CdSefilms (57).
The authors grew CdSe films from solutions of 0.3 M cadmium sulfate and
0.009 M selenium oxide. The cadmium sulfate contained Cu (100 ppm),
Fe (46 ppm), Pb (62 ppm), Zn (39 ppm), AI (95 ppm), Ca (320 ppm) and
alkaline earths (0.1%). Ethylene diamine tetraacetic acid (EDTA) (probably
the disodium salt) was used as the complexing agent. A typical film
deposited at 0.67 volts (referenced to a saturated calomel electrode)
showed impurities of Zn, Pb, Fe and Cu when depth profiled by Auger
analysis (Fig. 11). The typical concentration was an order of magnitude
less than that of either the Cd or Se. Additions of EDTA removed these
impurities due to the formation of stable complexes. However the addition
of EDTA alters the Cd:Se ratio from 1:1.6 to 1:3. The photoresponseof the
CdSe films formed from EDTA containing solutions was improved. Careful
heat treatment and etching further improved the film performance. A
CdSe film deposited from the basic starting solutions showed an open
circuit voltage (OCV) of 200 mV and a short circuit current (SCC) of 1.7
mA/cm2, but the efficiency (eff) and fill factor (ff) were very low. The
addition of EDTA raised the values to: OCV 400 mV, SCC 4.0 mA/cm2,eff
0.9% and ff 0.28. The best annealed and etched film had values of OCV
500 mV, SCC 10.0 mA/cm2, eff 3.6% and ff 0.38. These compared
favorablywith the best films produced by co-evaporationof high purity Cd
and Se (58).
It has also proved possible to deposit CdTe films without the use of
0
I 10 20 30
Sputtering t i m e Iminl
I I I
0 10 20 30
S p u t t e r i n g t i m e Iminl
Figure 11. a) Auger depth profile of a typical CdSe sample deposited in

the absence of EDTA at 0.67 volts (57). b) Auger depth profile of a typical
CdSe sample deposited at 0.67 volts in the presence of 15x1O-* mol I-’
EDTA (57).
an external potential source. This method, electroless deposition, has

been studied by Bhattacharya and co-workers (59). The solutions used
were basicallythe same as had been usedfor conventional electrodeposition;
0.01 M cadmium acetate and 0.01 M tellurium oxide with the pH adjusted
to 2 with 10% sulfuric acid. Either titanium or NesatronTMglass (PPG
Industries) were used as substrates. The necessary deposition potential
was generated by short circuiting the substrate to an easily oxidizable
redox component. Aluminum, for example, undergoing the reaction
Eq. (26) AI = AI3+ t 3e'
has an electrode potential of 1.66 volts. Films were easily grown on

substrates shorted to both aluminum and cadmium foils. Optical absorption
measurements gave a band gap of 1.4 eV. EDAX measurements showed
that the films, as deposited, had aCd:Te ratioof 25:60. Annealing the films
for 30 min. at 59OoCchanged the ratio to 55:45. X-ray data provided more
evidence for compositional changes on annealing. Films, as deposited,
contained CdTe in both the hexagonal and cubic forms as well as hexagonal
Te. After annealing,only the hexagonalform of CdTe remained:there was
no free Te. The films exhibited photoelectrochemical behavior.
Non Aqueous Solvents. Several Il-VI compounds have also been
electrodeposited from non-aqueous solvents. The first report was by
Baranski and Fawcett (60) in 1980. Their approach was to deposit
cationic species electrochemically from a solution containing elemental
chalcogenide, dimethylsulfoxide (DMSO), dirnethylformamide (DMF)
and ethylene glycol (EG). A typical CdS deposition utilized a solution of
6 gm/l of sulfur and 10 gm/l of cadmium chloride. This was electrolyzed
at 11O°C with a current density of 5 2.5 mA/cm2. The quality of the CdS
deposit was independent of both the sulfur and cadmium chloride
concentration used and was not affected by the addition of 10% water.
The deposit composition was solution temperature dependent, however,
becoming highly non-stoichiometric below 90°C. X-ray diffraction data
showed that the crystallites in the film were all oriented with their [ 1 11]
planes parallel to the electrode surface. The resistivity of these films was
about 1O6 ncm which could be lowered by addition of sodium iodide to the
solution.
The sulfides of lead, bismuth, nickel, cobalt and thallium were also
produced bychangingthesalt in solution. CdTewasobtainedfrom aDMF
solution saturated with tellurium containing 10 gm/l cadmium chloride
and 10 gm/l potassium iodide.
The photochemical properties of the films were determined in a cell
using 1 Msodiumsulfideand 1 M sodium hydroxideastheelectrolyte. The
sulfides of cadmium and bismuthwere n-type while that of thallium was p-
type. Nickel and cobalt sulfides were metallic conductors and showed
catalytic properties for electrode processes involving sulfur compounds.
A second paper by Baranski and co-workers (61) extended the
previous work (60) on CdS. Again a DMSO solution was used. The
reactant concentrations were 0.055 M cadmium chloride and 0.19 M
sulfur. Cadmium/cadmium chloride was used as a reference electrode.
The films were grown on platinum substrates. Measurements of the
deposition potential versus time at constant current densities showed a
sharp rise characteristic of the formation of a double layer (Fig. 12). This
was followed by a nucleation step which lasted 100 to200 sec. depending
on the current density. The curvesthen became linear until the film “broke
down”atabout5volts. This “breakdown”wasdue toacrackingofthefilm
which the authors attribute to a secondary piezoelectric effect. The
cracking was eliminated by reducing the current density throughout the
deposition.
X-ray diffraction analysisconfirmed the results of the previous paper
and showed only reflections from the < 11 1 > planes. The spacing was
measured at 3.34angstroms in good agreement with the theoretical value
of 3.36 angstroms. Rutherford backscattering showed that the S:Cd ratio
was 0.9+ 0.1 :1 and that the ratiowas uniform throughout the depth of the
film. There was some evidence of chloride impurities.
Plots of the average thickness of the film versus time at constant
current density were compared to the thickness expected from Faraday’s
law and the deposition efficiency was calculated to be 81%. A similar plot
of depth versus current density at constant time indicated a more rapid
deposition at higher current densitieswhich was attributed to increased
deposition of cadmium.
Addition of thallium ions to the deposition solution changed the
conductivity of the films as evidenced by a lowering of the necessary
deposition potential as the thallium concentration was increased. It was
thus possible to tailor the electrical characteristics of the film by the
selective addition of various ions.
The deposition of CdTefrom propylene carbonate (PC) was studied
by Darkowski and Cocivera (62). This work was unusual in that unlike
previous non-aqueous deposition which used elemental chalcogenides
dissolved in a non-aqueous medium, this investigation used an
organometallic source - tri(n-buty1)phosphine telluride [ (C,H,),PTe]
(phosphine telluride). The solution concentrations employed were 2 - 10
mM cadmium perchlorate or cadmium trifluoromethane sulfonate, 120
mM lithium perchlorate or sodium trifluoromethane sulfonate, and 7 - 26
mM phosphine telluride. A silver/silver chloride reference electrode was
used throughout. Cyclic voltammogramswere made for solutions of the
cadmium alone, the phosphine telluride alone and for the combined
Figure 12. Plots of voltage against deposition time during the deposition
of CdS films at several current densities as indicated (61).
solution (Fig. 13). The deposition potential of the cadmium was considerably
more cathodic than that of the phosphine telluride. The phosphine
telluride peak was shifted slightly in the anodic direction in the combined
solution and the currents were larger. This seemed to indicate a weak
interaction between the cadmium ions and the phosphine telluride.
CdTefilms were deposited at potentials between -650 and -1 600 mV
from solutions containing 3 mM Cd(ll) and 12 and 18 mM phosphine
telluride. The stoichiometry of these films was determined by dissolving
the films and using polarography (see ref. 62 for details). TheTe:Cd ratio
in these films reached 1 (0.95 & 0.05) at -1 200 mV and remained at that
value up to -1600 mV. The ratio and deposition current depended
somewhat on the concentrations of the reactants.
The current density decreased with time during constant potential
deposition. The rate of decrease was dependent on the concentration of
the Cd(ll) relative to the phosphine telluride. The rate of decrease was
shown to be slower at higher relative concentrations indicative of a
complicated deposition mechanism, perhaps involving Cd-tri(n-butyl)
phosphine telluride complexes.
The photoelectrochemical activityof these filmswas measured in the
same cell and solutions in which they were grown. They produced a
cathodic photocurrent indicating that the films were p-type. Scanning
above the shut-off potential produced negligible photoanodic current
assuring that the photocurrent was not due to increased photoconductivity
(Fig. 14). The best films were producedfrom a solution of 3 mM Cd(ll), 12
mM phosphine telluride, and 100 mM lithium perchlorate electrolyzed at
-1 200 mV versus Ag/AgCI.
Fatas, et ai. (63), have also done a study on the electrodeposition of
CdS from nonaqueous solutions, in this case on stainless steel and tin
oxide. Theystudiedtwosolvents: DMSO andPC. Thesolutionswere0.19
M sulfur and 0.055 M cadmium chloride in DMSO, and PC saturated with
sulfur, cadmium chloride and potassium chloride. The depositions were
carriedout at a constant current at 12OOC. Allvoltageswere referencedto
a cadmium/cadmium chloride electrode.
The plots of deposition potential versus time for the two different
solutions were markedly different (Fig. 15). The PC solution maintained
essentially a constant potential after the initial transientwhile the potential
for the DMSO rose continuously. The initial transient potential was also
higher in DMSO which was attributable to the stronger solvation action of
this solvent.
Film thickness versus current density was also measured (Fig. 16).
At the same current density, the PC films were thicker than those deposited
from DMSO. Part of the explanation for this behavior was the possibility
of cracks forming in the films deposited from DMSO due to the higher
-I r-
--_i L-_-.
- 5 - h -1. 4 -1.2 -1.0 -a a -0. 6 14
FQTEHTIAL (VoILd
Figure 13. I-V plots in propylene carbonate under N, at 1OOOC and a scan
rate of a) 100 mV/sfor 0.003 M Cd(ll) and 0.10 M LiCIO,, b) 0.01 8 M tri-
n-butylphosphine telluride and 0.10 M LiCIO,, and c) 0.0003 M Cd(ll),
0.018 M tri-n-butylphosphine telluride, and 0.10 M LiCIO, (62).
"i
-0.5
-1.0-
-1.5-
Figure 14. Photoresponseat 1OOOCof electrodeposited CdTefilm in PEC

containing 0.003 M Cd(ll), 0.018 M tri-n-butylphosphine telluride, and
0.10 M LICIO, in propylene carbonate under nitrogen with no stirring (62).
- 1.2-
-0 1 -
h
-.
U
0
-u0.8 -
ul
-0.3 3
E
-0.2w
Figure 15. Plots of voltage against deposition time during the deposition
of CdS films, a) in DMSO, b) in PC (63).
J (rnAIcrn21
Figure 16. Plots of thickness of CdS film deposited on stainless steel

against current density. Deposition time, solution composition and
temperature were constant during the deposition, a) in DMSO, b) in PC
(63).
potential required. This seemed to occur at thicknesses greater than 0.8

pm. It may also have involvedthe decomposition of the solvent. Cracking
did not appear when PC was the solvent. By comparing the thicknesses
to that expectedfrom Faraday’slaw, theeff iciencies were found to be 90%
for PC and 70% for DMSO (additional evidence for the decomposition of
the DMSO).
The photoelectrochemical properties of these films were measured
using an electrolyte that consisted of 0.1 M sodium hydroxide, 0.1 M
sodium sulfideand0.01 M sulfur. Thefilmsshowed typical photochemical
behavior (Fig. 17).
X-ray diffraction data for films deposited between 100 and 140°C
exhibited hexagonal symmetry and were c oriented. At temperatures
above 14OoC,the films gave patterns very similar to powder patterns (Fig.
18). The grain size and the influence of deposition conditions on that size
were obtained from the x-ray data.
The films, as grown, had resistivities between 2 . 4 ~O5 1 and 9 . 0 ~O5
1
ncm. Heat treatment reduced this value to 500 n cm.
Molten Salts. ZnSe has been grown electrochemically on silicon
and germanium substrates by Yamamoto and Yamaguchi (64). Two
melts were studied: 1) 0.02 zinc chloride, 0.02 selenium chloride, 0.12
potassium chloride and 0.18 lithium chloride (in mole ratio), and 2) 0.003
zinc oxide, 0.003 sodium selenite, 0.4 potassium chloride and 0.5lithium
chloride. In both cases, the potassium and lithium chlorides were the
solvent. The melts were operated under dry argon between 430 and
550OC.
Melt 1 not only produced ZnSe, but also zinc oxide. Moreover the
results were not reproducible and the melt attacked germanium. Melt 2
produced high quality, reproducible films of ZnSe. These films showed
good orientation to the substrate ( < 111 > Ge) and exhibited red
photoluminescence at room temperature.
The optimum current density for growth lies between 0.5 and 3 mA/
cm2. Below 0.5 mA/cm2 no deposit was obtained, while above 3.0 mA/
cm2the deposit was polycrystalline. The ideal range of growth temperatures
was 450 to 5OOoC.
Ternary Alloys and Compounds. Baranski and co-workers (65), in
an extension of their previous work, grew solid solution films of cadmium
sulfur selenide (CdS,Se,). They used a solution of 0.055 M cadmium
chloride and 0.19 M sulfur with varying amounts of selenium in DMSO.
Good quality films were electrodeposited and the color varied from
yellow-gold for CdS to dark reddish-purple for films deposited from
saturated solution of Se (approximately 0.01M Se). The morphology of
these alloy films was crystalline, but highly cracked at low Se concentrations
and smooth and glass-like at high Se concentrations. These structural
100-
Figure 1 7. Current-voltage properties for n-type CdS electrode exposed

totheO.l MNaOH, 0.1 M Na2S*9H20and0.01MSelectrolyte. Thesweep
rate is 0.1 Vs-I (63).
Figure 18. X-ray diffraction spectra for CdS films obtained at different
temperatures during deposition, a) 7OoC, b) 14OoC,and c) 16OoC(63).
changes were also reflected in the x-ray patterns.

Plots of the S/Cd ratio versus the Se/Cd ratio in the films showed a
trend of decreasing S/Cd for increasing Se/Cd. This indicates that the
deposition mechanismsfor the two chalcogens are similar. A plot of Se/
Cd versus Se concentration in the film revealedthat it is possible to vary
the amount of selenium in the film over awide range. The ratio reached a
maximum value of 0.75.
The photoelectrochemical response of these films was measured.
As was to be expected, it and the band gap varied with film composition.
Some ternary compounds are also technologically important. The I-
lll-V12 chalcopyrite compounds are included here because the anion is
from groupVl and the average cation valence is 2. CulnSe,is an important
candidate material for highly efficient thin film solar cells.
The electrodeposition of copper indium selenide and telluride was
discussed by Bhattacharyaand Rajeshwar (66). The deposition solutions
were made in a multistep process. A 0.5 M solution (A) of indium chloride
was prepared. The copper solution (B) was prepared from cuprous
chloride dissolved in 30 ml of triethanolamine,40 ml of 30% ammoniaand
150 ml of water. The copper concentration was 0.5 M. SolutionsA and B
were mixed in equal amounts, diluted 10 times and adjusted to pH 1 with
HCI. Thissolution (C)wasagedfor 24 hrs. Theactual deposition bathwas
made from 20 ml of either 0.1 M selenium oxide or 0.1 M tellurium oxide
and 60 ml of solution C. The triethanolamine was added to the copper
solution as a complexing agent to shift the deposition potential. Depositions
were carried out at -1 .O volt versus a saturated calomel electrode. The
initial current density was 12 mA/cm2.
Beforedeposition, the behavior of the copper-indiumand chalcogen
solutions were investigated separately. The I-V plot of the copper-indium
solution showed both deposition and stripping peaks indicating a reversible
behavior (Fig. 19). The addition of the triethanolamine decreased the
separation of deposition potential between copper and indium to 0.37
volts instead of 0.68 volts. The I-V plots of the chalcogen oxide solutions
demonstrated irreversible behavior (Fig. 20).
X-ray diffraction patterns of the films showed peaks characteristic of
the chalcopyrite copper indium selenide or copper indium telluride
compounds. Auger spectra and depth profiles showed the expected
signal plus some carbon and surface oxidation. The stoichiometry varied
with depth.
A heterojunction of copper indium selenide on n-type cadmium
sulfide indicated that the films, as deposited, were p-type. These films
yielded an OCV of 100 mV and SCC of 2 mA/cm2. The main problem with
these films was poor grain morphology.
-0 5 70.3 10.1 -0 1 -03 -0.5

Potential, V vs SCE
Figure 19. I-V plot recorded using a Pt micro-electrode (potential scan

rate, 0.10Vs”)for asolutionof CuCI, and InCI, (5x1O-* Meach)in 0.1 M HCI
(66).
07
-0.2 0.5
2 4 6 0
PH
f L . 2 0 V/S
v/s
v/s
1 I I I 1 I
+o a +o 4 0 -0 4
Potential, V v s SCE
Figure 20. I-V plot on a Pt micro-electrode (parameter shown is the

- ~solution of SeO, in H,SO, (pH 7.8).
potential scan rate) for a 5 ~ 1 OM
Inset: dependence on pH of CV peak potentialsfor reduction and oxidation,
,, and E
E ,, respectively (66).
There is a review article by Hodes and Cahen (67) which outlines

numerousapproachesfor depositing copper indium sulfide and selenide.
It also outlines a technique whereby the copper and indium can be
codepositedand converted to the chalcogenide by annealing in hydrogen
sulfide or selenide.
4.2 Ill-V Compounds
Gallium Phosphide. Cuomo and Gambino ( 6 8 ) ,in 1968, were the

first to electrochemically synthesize a Ill-V compound (Gap). Gallium
phosphate is an important electroluminescent material. By molten salt
electrolysis, they prepared up to 100 mm thick epitaxial layers of GaP on
< 111 > , < 100 > and < 110 > silicon substrates (cathodes) at current
densitiesof up to 50 mA/cm2. Afused salt bath contained (in molar ratio)
2.0 NaPO,, 0.5 NaF and 0.1 225 - 0.25 G%O,. Itwas held at 8OOOC during
electrodeposition. They were able to dope their layers both p- or n- type
during growth, using small concentrations of ZnO and N%SeO, respectively.
An electroluminescentp-n junction was fabricated using electrochemical
techniques.
A schematic diagram of Cuomo and Gambino’s furnace is given in
Fig. 21. The quartz chamber was heated by a resistance furnace. The
graphite crucible, in addition to containing the melt, acted as the anode.
Two types of cathodes were used; 1/4 inch diameter graphite rods
(spectrographic grade) immersed in the melt to a depth of 1 inch were
used for synthesis experiments, while two oriented silicon single crystal
wafers mounted back to back on graphite rods were used for epitaxial
growth. Both constant current and constant voltage experiments were
studied. They used reagent grade chemicals except for the Ga,O, which
was 99’.999% pure. The gallium phosphide layers were readily removed
from the substrate by dissolving away the silicon substrate in HF:HNO,.
The GaP was coated with apiezon wax during the etching process. The
silicon wafers were placed back-to-back to prevent GaP depositing on
both surfaces of a single wafer and preventing Si removal.
The liquidus temperature of the melt composition (75OOC) set the
lower limit for the bath and deposition temperature while extensive
volatization at 1OOO°C fixed the upper limit. Voltages below 0.6 volts did
not produce solid products, while voltages of around 0.9 produced the
best results. The growth rates were typically 10 - 20 mm/hr initially, but
these decreased due to the relatively high resistivity of the GaP formed.
Doped layers with lower resistivities could be grown to thicker dimensions.
Typical deposits consisted of an epitaxial layer covered with a dendritic
polycrystalline overgrowth. The GaP films usually cracked along cleavage
planes. This problem was minimized by carefully cooling the layer after
growth. Depositswerefound to contain the following impurities (in atomic

percent); Si (O.OOOl), Mg (O.OOOl),AI (0.05), Te (0.01), and Cu (0.001).
Cuomo and Gambino (68) also showed that alkali halide salts could also
be used as solventsfor GaP electrodeposition, but the resultswere not as
good.
DeMattei, et al. (69), in alimited study in 1978, studied the conditions
necessary for the stable growth of GaP epitaxial layers on Si and GaP
substrates. They identifiedthe variables critical to controllingthe morphology
and uniformityof electrodeposited GaP layers. They used the same fused
salt composition and range of deposition temperatures (750 - 900OC) as
Cuomo and Gambino. The electrochemical cell reactions for GaP at the
cathode were probably:
+ 3P0,-3
-
Eq. (27) 8e- t Ga'+3t 4PO;+ GaP , or
Eq. (28) 8e- + Gaf3 t PO,' GaP + 30"

and at the anode:
A schematic of the apparatus used by DeMattei, et al. (69), is shown

in Fig. 22. As in Cuomo and Gambino's experiments, they used their
graphite crucible as the anode. Three different cathode materials were
studied: 1 ) Graphite, 2) P-doped, n-type (100) Si (0.3 0-cm) wafers
polished on one side, and 3) S-doped, n-type (1 11) as-cut wafers of GaP
(0.05n -cm). The (100) Si orientation has a close lattice match to GaP (as,
= 5.431, a ~ = ,5.449).
~ The minimum deposition potential for Si substrates
was about 0.5 volts, somewhat lower than observed by Cuomo and
Gambino (68), but the deposits grown at 40 mA/cm2(Fig. 23) had surface
craters due to excess P forming gas bubbles on the substrate surface.
Dendrites grew beneath the bubbles due to restricted solute transport in
these regions. These craters also appeared on graphite and GaP substrates
under certain conditions. Excess P evolution was also seen by Cuomo
and Gambinoat low potentials and most of theirexperimentswere carried
out at cell potentials of about 1 volt. More uniform, epitaxial crystalline
layers were produced at lower current densities (20 mA/cm2). Current
density is related to growth rate (G) and there is a maximum allowable
value of G above which the quality of single crystal films deteriorate for all
epitaxial growth methods used. The morphology of the crystallites in the
electrodeposited layers were comparable to that produced by early
Figure 21. Schematic of the apparatus for the preparation of GaP by

fused salt electrolysis (68).
VIEWING PORT VALVE FOR CATHODE

REMOVAL
-\I He OUT
VACUUM PORT
ANCDE
3 I I l l
II
He IN (BUBBLER)
FLANGE
-+
c AT HODE
Figure 22. Diagram for apparatus used for growth of GaP layers (69).
Figure 23. High current density deposit of GaP on silicon done early in
investigation. Shows layer with dendritic over-growth around "craters"
(40 mA/cm2) (69).
MOCVD deposition techniques (70).

In order to gain a better understandingof the deposition process and
to improve control of the epitaxial growth process for GaP on Si, DeMattei,
et al. (69), did a series of nucleation experiments at a constant cell
potential of 0.6 volts (slightly above the minimum deposition potential).
While analysis did not reveala consistent pattern, Si0,wasfound together
with the deposited GaP layer. The SiO, may have formed because the
electrodeposition potential was not applied until 5 minutes after the
substrate was dipped into the melt (part of the substrate cleaning process).
This conjecturewas confirmed by a later experiment in which it was found
that an applied potential was not necessary for the formation of GaP and
SiO, on a Si substrate. The results suggest that the silicon substratescan
react with the melt by the following reaction:
Eq. (31) 8Si t 2Ga20, t 4PO,--+ 4GaP t 8Si0, t 2 0 2
This reactivity of silicon with the melt and the poor thermal expansion
mismatch between Si and Gap, which was probably responsible for the
cracking observed by Cuomo and Gambino (68), led DeMattei, et al. (69),
to study the homoepitaxy of GaP on GaP substrates. The minimum
deposition potentials for GaP on (111) GaP at 8OOOC and at 900°C are
shown in Table 1, along with data for the electrodeposition of GaP on Si
and graphite substrates at 900OC.
Table 1
Minimum Deposition Potentials for GaP

on Various Cathodic Materials
Cathode Material WO€ 9oooc
Graphite 0.68 volts
Si (100) 0.55 volts
GaP (1 11) 1.16volts 0.86 volts

All deposition experiments were carried out at potentials 0.1 volts

greater than the minimum depositon voltages. Since substrate surfaces
were etched somewhat at 900°C, most depositions were carried out at
8OOOC. Current densities over 10 mA/cm2 gave layers with craters, as
seen in Fig. 23. At 48 mA/cm2, large dendritic craters on thin epitaxial
layers wereformed. In Fig. 24, a plot of current (I)versus overpotential is
given. DeMattei,et ai. (69), relatedthe first linear region to the growth rate
being controlled either by volume diffusion of solute ions or by interface
mechanisms, and the second linear region at higher current densities to
possible interface attachment mechanisms. The limit of this linear region
( q = 1.4 volts) corresponds to the maximum current density for stable
growth.
Both the preliminary work of Cuomo and Gambino (68), and DeMattei,
et ai. (69), have shown that electrodeposition of GaP is possible. More
carefully controlled systematic experiments should lead to much improved
epitaxial deposits. It is not likely, however, considering the major advances
of vapor deposition techniques, that a strong needfor this technologywill
be forthcoming.
Indium Phosphide. Indium phosphide, an important material for
high frequency electronic devices, is also of some interest for CdS/lnP
heterojunction solar cell applications. The only reported work on the
electrodeposition of InP was by Elwell, et ai. (71), in 1981. They explored
a wide range of solute/solvent systems and developed, for the first time,
the conditions necessary for the electrodeposition in InP on [OOOl]CdS
and [ 11 1 ] and [ 1001 InP substrates.
The use of molten salt electrolysis for InP synthesis and growth is
complicated by the high volatility of the most convenientsources of In ions
(In,O, and InF,) above 650°C, and also by the decomposition of InP at
elevated temperatures in an inert atmosphere. These factors limited the
temperature regime for stable deposition of InP to 600 - 65OOC.
After trying a number of low melting solvents as listed in Table 2,
Elwell, et ai. (71), were able to successfully deposit InP from a NaPOJ
KPO,/NaF/KF eutectic mixture named “Pofonak”. Current densities in
the range of 1 - 3 mA/cm2 were used.
InP probably forms electrochemically by the following cathode and
anode reactions:
Eq. (32) In3+ t 4PO; t 8 e 4 InP + 3PO:
Eq. (33) 4P0,3‘ 4 4P0,- + 20,T + 8e
The anode used in these experimentswas normally graphite, which

reacts with the 0, evolved to form CO or CO,. The minimum deposition
80 CATHODIC REGION
70
h
60
w
k
50
I
c
g3 40
c3
V
G 30
E
V
H
20
0 0.5 1.o 1.5

q (VOLTS)
( SU P ERSATURAT I 0 N )
Figure 24. Plot of current (I)vs. overpotential ( v ) for deposition of GaP

on GaP (69).
potentialwas 0.85f0.05volts and was not sensitive to In,O, concentration.

The chemicals used were reagent grade (98 - 99%) and major melt
impurities were 0.01 % Si, 0.001 % AI and Ca.
Whilecathode materialsmadeof graphite, Ni, Ge, Au, Mo, Nb, Ptand
Ta were tried, Ni was the only metal suitable for the deposition of
polycrystalline InP, due to poor adhesion on the other materials. Fig. 25
showsa deposit on Nifoil. TheCdS substratescamefromvarioussources
and had resistivities in the range of 3 - 10 n cm at 600OC. Commercial
crystals had to be annealed in a Cd vapor to reduce their initially high
resistivities (1 04-107ncm). These substrates had to be carefully cleaned
to prevent dendritic growth. An example of a InP layer on CdS is shown
in Fig. 26 in cross-section. Films 2 - 5 pm in thickness were deposited at
rates of 1 - 2pm/hr at potentials of 0.9 to 1.Ovolts. Deposition efficiencies
were around 50%.
Table 2
Solvent/Solute Composition Studied for InP Electrodeposition (17).
ComDosition Comment
NaPO,/NaF(ln,O, or InF,) Low solubility of In,O, and InF

at 60OoC
High viscosity
No InP deposits
KCI/LiCI eutectic Chlorides volatile

t NaPO,/ln,O, or InF,
LiF/NaF/KF eutectic t InFJKPF, All fluoride system

Gas evolution due to reaction of
components, non-reproducible
InP formation
Low Solubility of In,O,

Successful for InP electrodeposition
Solubility of In,O, only = 1 %
(Li,K)(PO, ,F) and (Li,Na)(PO,,F) InP deposits not obtained

with LiPO, and LiF,
Figure 25. Polycrystalline InP deposit on nickel foil (71 ).

Figure 26. Cross-section of fairly uniform layer of InP on CdS. (The

scratches are polishing marks on the CdS. ) (71 ).
The deposition of InP on doped [ 1001and [ 11 1 ] n-type InP wafers

was less successful. The preliminary work of Elwell, et al. (71), showed
that the deposition of InP films by the fused salt electrolysis method is
possible, but challenging.
Gallium Arsenide. GaAs is the next most commercially important
material after silicon. It is useful for light emitting diodes, diode lasers,
substrates for high speed integrated circuits and other optoelectronic
applications. It is surprising, therefore, that there has been only one
published paper to date on the electrolytic synthesis and deposition of
GaAs (DeMattei et al., Ref. 72). A molten salt solution was used which
contained NaAsO, and Ga,03 in a B,O,-NaF solvent at 720 - 760OC. The
specific source of As used was very important. NaAsO, is easily reduced
by conversion of As5+to As3+or Aso in the presence of either carbon,
metals,or GaAs. Melts usingAs,O, or NaAsO, are more stable but volatile
at high temperature. Following the work of Weiss (73), who employed
NaAsO, for the growth of MOASand reduced deposition temperatures
using a Na,B,O,-NaF flux, DeMattei, et al. (72), reduced the liquidus
temperature even lower to (720 - 760OC) by using 20.3 wt% NaF, together
with67.3wt%B,03in thesolvent,and4.3wt%Ga,O3and8.2wt%NaAsO3
as the solute. More than a stoichiometric amount of arsenic was needed
to prepare single phase GaAs deposits. The minimum deposition potentials
were 1.7 volts for GaAs cathodes and 2 volts in nickel at the deposition
temperatures used. The apparatus usedwas described by Zubeck, et al.
(74). Vitreous carbon crucible and gold anodes were used.
Epitaxial GaAs layers 10 p m thick were deposited on (100) n-type
(0.05 ncm) GaAs substrates.
4.3 IV-IV Compounds
Silicon Carbide. Silicon carbide is acomplexcompoundwhich has

a large number of structural polytypes. The Q or 6H (hexagonal)form is
the most commonly found while the cubic p-phase is of greatest technological
interest. The promise of silicon carbide as a high temperature, radiation
resistant semiconductor and blue-green light-emitting diode, has tantalized
researchers for at least three decades. The problem of growing large
crystals or epitaxial films with the appropriate structure and crystalline
perfection has been very challenging. Many approaches have been
pursued without much success until 1981 when Tairov and Tsvetkov (75)
developed a vapor-phase process to grow large bulk crystals.
While the electrochemical synthesisand deposition of various metal
carbides was first reported in 1946 (73), the one and only study on the
electrocrystallization of SIC (Elwell et al., Ref.76) was not reported until
1982. After experimenting with a variety of molten salt mixtures containing
Electrochemica1 Deposition of Semiconductors 45
alkali metal fluorides and sources of Si and C such as K,SiF,and Na,CO,,

they succeeded in electrodepositing single phase a-Sic. Best results
were obtained with Li,CO,/SiO, melts, which yielded thin layers of a-Sic
without excess C or Si. Deposition was on (0001) a-Sic single crystal
seeds (cathode) held at 1000 - 1050%. These seed crystals were obtained
from a commercial source and were the by-product of the carbothermic
reduction process for silicon dioxide.
The overall simultaneous cathodic deposition reactions involved in
SIC deposition were
Eq. (34) Li,CO, 4 Li,O + C (cathode) + 0, (anode)

and
Eq. (35) SiO, -* Si (cathode) + 0, (anode)

Electrodeposition was carried out in a furnace system having a
graphite heater and heat shieldsandawater-cooledstainless steel jacket,
as described by DeMattei, et al. (69). The potential difference used with
a polycrystalline Sic anode was -0.5 volts. Molten Li,CO, was found to be
very reactive with a variety of crucible materials. Vitreous carbon crucibles
were found to be the least reactive but were prone to cracking and,
therefore, zirconium metal crucibles which were almost as good, were
used routinely by Elwell, et al. (76). Care had to be taken so that the
temperatures were low enough ( < 12OOOC) to prevent the evolution of
CO, from the decomposition of Li,CO,. The upper temperature limit was
found to dependon the container and electrode materialsused. The lower
temperaturelimitwas relatedto the requirementthat the SiO,/Li,CO, ratio
be sufficiently high for the formation of Sic without excess C. Melt
compositions (expressed as [ 1 - x ] Li,CO, x SiO,), which contained x 2
70.3 were too viscous. Deposits from melts containing x < 0.1 had
irregular texture and poorly formed grains. A small quantity of LiF (1%)
added to the melt was found useful in lowering the viscosity. While the
lowest temperature used to deposit Sic from Si0,-Li,CO, melts was
95OoC,it was possible to electrodeposit this compound at 75OOC from a 5
m/o \SiF,-20 m/o Li,C03 mixture in a LiF/KF eutectic melt. Unfortunately,
the deposits from these melts had poor morphology.
Figure 27 shows aslow scan voltammogramfor Si0,-Li,CO, melts at
1OOO°C containing different compositions. Although no waves for Si and
C can be seen, an unexpectedly large increase in current with SiO,
concentration was observed. The best deposits (shown in Fig. 28a & b)
were made using potentials of -0.4 to -0.5 volts, deposition time of 3 hrs
and a melt composition of 0.8 Li,C03 and 0.2 SiO,. While conditions for
Figure 27. Slow-scan I-V plots for melts of different composition at

1000°C, using SiC electrodes: a) Li2CO3 only; b) 0.9 Li2CO3 + 0.1 SiO2;
c) 0.8 Li2CO3 + 0.2 SiO2 (75).
(a)
(b)
Figure 28. a) SEM photograph of SiC deposit on [0001] a-SiCfrom melt

of composition of 0.8Li2CO3; 0.2 SiO2, using-0.5voltsvs. SiC. Deposition
time, 3 hrs. b) SiC deposit under similar conditions to a) but at -0.4 volts
for 30 mins (76).
long-term deposition stability were found, attempts to produce thick

layers were hampered by delamination of the seed crystals.
5.0 CONCLUSION
It has been demonstrated that most of the technologically important

semiconductors can be synthesized and deposited as thin films by
electrolysis techniques. By far, the greatest amount of effort has gone into
the development of electrodeposition methods for low cost silicon solar
cells.
Although there are important advantages associated with
electrodeposition processes, the demanding requirementsfor purity and
structural perfection of materials for semiconductor devices and the
continued development of advanced alternative growth and deposition
methods, such as molecular beam epitaxy and metalorganic chemical
vapor deposition, make it unlikely that a strong interest to pursue electrolytic
processes will emerge during the next decade. If semiconductor films
with complex shapes become needed sometime in the future, then
electrodeposition techniques might become more important.
6.0 REFERENCES
1. Kunnmann, W., Preparation a n d Properties of Solid State

Materials, (R. A. Leferer, ed.), p. 1 , Dekker, New York (1971)
2. Wold, A. and Bellavance, D., Preparative Methods in Solid State

Chemistry, (P. Hagenmiller,ed.), p. 279, Academic Press, New York
(1972)
3. Feigelson, R. S., Crystal Growth by theElectrolysisof Molten Salts, in:

Solid State Chemistry: A Contemporary Overview, (S. Hold, J.
Milstein and M. Robbins, eds.), pp. 243, Amer. Chem. Soc., Washington,
D.C. (1980)
4. Wall, F. T., Chemical Thermodynamics, V. H. Freeman and Company

( 1 958)
5. Barrow, G. M., Physical Chemistry, McGraw-Hill(l961)
6. Kortum, G. and Bockris, J. O’M, Textbook of Electrochemistry,

Elsevier Publishing Company (1951)
7. Bokris, J. O’M. and Reddy, A. K. N., M o d e r n €lectrochemistry,

Plenum Publishing Corporation (1970)
8. Despic, A. P. and Popov, K. I., Transport-Controlled Deposition and

Dissolution of Metals, M o d e r n Aspects of €/ectrochemistry, (B.
E.Conway and J. O’M. Bockris, eds.), 7:99 (1972)
9. Despic, A. P., Diggle, J. and Bockris, J. O’M., J . Electrochem. SOC.
1 15:507 ( 1 968)
10. Kroll, W. J., Trans. €lectrochem. SOC.87:551 (1945)
1 1 . St Claire De Ville, H., Compt. Rend. Acad. Sci. Paris 39:323

(1854)
12. Monnier, R., Chimia 37:109 (1983)
13. Gore, C., Phil. Mag. 7:227 (1854)
14. Ullik, F., Ber. Akad. Wien 52:1115 (1865)

15. Minet,A., Compt. Rend. Acad. Sci. Paris 112:1215 (1891)
16. Warren, H. N., Chem. News 67:303 (1893)
17. Dodero, M., Compt. Rend. Acad. Sci. Paris 109:556 (1934)
18. Idem., Bull. SOC.Chim. France 6:209 (1939)
19. Monnier, R . and Barakat, D., Helv. Chim. Acta 40:204 (1957)
20. Monnier, R . and Giacometti, J. C., Helv. Chim. Acta 47:345 (1964)
21. Grjotheim, K. and Matiasovsky, K., Chem. zvesti 25:249 (1971)
22. Grjotheim, K., Matiasovsky, K., Fellner, P. and Silny, A., Can. Met.
Quart. 10:79 (1971)
23. Boe, G., Grjotheim, K., Matiasovsky, K. and Fellner, P., Can. Met.
Quart. 10:179,281,463 (1971)
24. Grjotheim, K., Matiasovsky, K. andFellner, P., Light Metals 1982, (J.E.
Anderson, ed.), p. 333, AIM€ (1982)
25. Monnier, R., and Barakat, D., U.S. Patents 3,219,561 (23 Nov 1965)
and 3,254,010 (31 May 1966)
26. Huggins, R. A. and Elwell, D., J . Crystal Growth 37:159 (1977)
27. DeMattei, R . C., Elwell, D. and Feigelson, R . S., J. ,Electrochem.

SOC.128:1712 (1981)
28. DeMattei, R . C., Elwell, D. and Feigelson, R . S., “Electrodeposition of

Molten Silicon,” U.S. Patent No.4,292,145 (September 29, 1981)
29. Johnson,K. E., High Temperature Technology, p. 493, Butterworths,

London (1967)
30. Davis, J. R., Rohatgi, A,, Hopkins, R . H., Blais, P. D., Rai-Choudhury,
P., McCormick, J. R . and Mollenkopf, H. C., /€€,E Trans. ED-27:677
(1980)
31. Olson, J. M. and Kibbler, A., Abstracts 5th Amer. Conf. Crystal
Growth, p. 248, San Diego (1981)
32. Hitachi Ltd., Japanese Patent 58,144,485, (27Aug 1983); see Chem.
Abs. 100:14515s (1984)
33. Ito, Y. and Yoshizawa, S., Advances in Molten Salt Chemistry,

V01.4, (G. Mamantov and J. Braunstein, eds.), p. 391, Plenum Press,
New York (1981)
34. Paulsen, E. R. and Hall, J. A., J . Metals 35:60 #6 (1983)
35. Lovering, D. G. and Williams, D. E., Molten Salt Technology, (D. G.

Lovering, ed.), p. 91, Plenum Press, New York (1982)
36. Cohen,U., J . Electron. Mater. 6:607 (1977)
37. Rao, G. M., Elwell, D. and Feigelson, R. S., Surface Technology

13:331 (1981)
38. Rao, G. M., Elwell, D. and Feigelson, R. S., J. Electrochem. SOC.

127:1940 (1980)
39. Idem., ibid. 128.1708 (1981)
40. Idem., ibid. 130:1021 (1983)
41. Olson, J. M. and Carleton, K. L., J. Electrochem. SOC.128:2698

(1981)
42. Sharma, I. G. and Mukherjee, T. K., Metall. Trans. 17B:395 (June

1986)
43. Boen, R. and Bouteillon, J. J. Appl. Electrochem. 13:277 (1983)
44. De Lepinay, J., Bouteillon, J., Traore, S., Renaud, D. and Barner, M. J.,
J. Appl. Electrochem 17:294 (1987)
45. Austin, A. E., U.S. Patent 3,990,953 (Nov. 9, 1976)
46. Bucker, E. R. and Amick, J. A., U.S. Patent 4,192,720 (Mar. 11, 1980)
47. Lee, C. H., and Kroger, F. A., J. Electrochem. S O C .129:936 (1982)
48. Rama Mohan,T. R. and Kroger, F. A., Electrochimica Acta 27:371

(1982)
49. Elwell, D. and Rao, G. M., J . Appl. Electrochem. 18:15 (1988)
50. Hodes, G., Manassen, J. and Cahen, C., Nature 261 :403 (1971)
51. Miller, B. and Heller, A., Nature 262:680 (1971)
52. Peter, L., Electrochim. Acta 23:165 (1978)
53. Panicker, M., Knaster, M. and Kroger, F., J . Electrochem. SOC.,

Electrochem. Sci. and Technol. 125:566 (1978)
54. Power, G. P., Peggs, D. R. and Parker, A. J., Electrochimica Acta

26:681 (1981)
55. Takahashi, M., Uosaki, K. and Kita, H., J. Electrochem. SOC.,

Electrochem. S c i . and Technol. 131:2304 (1984)
56. Takahashi, M., Uosaki, K. andKita, H., J . Appl. Phys. 55:3879 (1984)
57. Pandey, P. K. and Rooz, A. J. N., J . Phys. D : Appl. Phys. 19:917

(1986)
58. Russak, M. A., Reichmann, J., Witzke, H., Deb, S. K. and Chen, S . N.,
J . Electrochem. SOC. 127:725 (1980)
59. Bhattacharya, R. N., Rajeshwar, K.andNoufi,R. N.,J. Electrochem.

SOC.:Accelerated Brief Communication 131:939 (1984)
60. Baranski, A. S. and Fawcett, W. R., J . Electrochem. SOC.127:766

(1980)
61. Baranski, A. S., Fawcett, W. R., McDonald, A. C., de Nobriga, R. M.

and MacDonald, J. R., J . Electrochem. SOC.:Electrochem. Sci
and Technol. 128:963 (1981)
62. Darkowski, A. and Cocivera, M., J. Electrochem. SOC.:Electrochem.

Sci. and Technol. 132:2768 (1985)
63. Fatas, E., Herrasti, P., Arjona, F., Camarero, E.Garcia and Medina, J.
A., Electrochem. Acta 32:139 (1987)
64. Yamamoto, A, and Yamaguchi, M., Japan. J . Appl. Phys. 14561

(1975)
65.Baranski, A. S.,Fawcett, W. R., Gatner, K., MacDonald, A. C., MacDonald,

J. R. and Selen, M., J. Electrochem. SOC.: Electrochem. S c i .
and T e c h n o / . 130:579(1 983)
66.Bhattacharya, R. N. and Rajeshwar, K., Solar Cells 16:237(1986)
67.Hodes, G. and Cahen, D., Solar Cells 16:245(1 986)
68.Cuomo, J. J. and Gambino, R. J., J. ,Electrochem. SOC.115:755

(1 968)
69.DeMattei, R. C., Elwell, D. and Feigelson, R. S., J . Crystal Growth

44:545(1 978)
70.Andre, J. P., Hallais, J. and Schiller, C., J. Crystal Growth 31:147

(1 975)
71.Elwell, D., Feigelson, R. S. and Simkins, M. M., J . Crystal Growth

51:171(1981)
72.DeMattei, R. C., Elwell, D. and Feigelson, R. S., J . Crystal Growth

43:643(1 978)
73.Weiss, G., Ann. Chim. 1 :446(1946)
74.Zubeck, I. V., Feigelson, R. S., Huggins, R. A. and Pettit, P. A., J .

Crystal Growth 34:85(1 976)
75.Tairov, Y. M. and Tsvetkov, V. K., J . Crystal Growth 52:146(1981)
76.Elwell, D., Feigelson, R. S. and Simkins, M. M., Mat. Res. Bull.

17:697(1 982)
2
CHEMICAL ETCHING: PRINCIPLES AND
APPLICATIONS
Keshra Sangwal
1 .O INTRODUCTION
When a crystal is in contact with its undersaturated environment or

some other corrosive medium, it undergoes decrystallization (dissolution,
etching). Depending on the experimental conditions and the duration of
the process, the crystal may yield etch figures (depressions and elevations)
and dissolution layers on its surfaces without loss of its macroscopic
appearance,or may acquire a macroscopic form different from the initial
one. The former are produced when the process is carried out for short
durations, while a change in the macroscopic appearance of the crystal
takes place only after prolonged dissolution. Honess (1) gives a chronological
description of the literature prior to 1927 on etch figures produced on
crystal surfaces by corrosive solvents, while a recent monograph by
Sangwal(2) describes various etch figures from the standpoint of dissolution
theories. Heimann’s book (3), on the other hand, concentrates on the
theories of dissolution and dissolution forms of crystals.
A variety of etch figures are observed on etched surfaces of crystals.
In general, they are depressions and elevations. Among the former type
are pits, depression spirals and etch grooves, while among the latter are
crystallographically-orientedand noncrystallographic hillocks. Etch pits
are the most frequently observed figures on etched crystal surfaces,and
often they are formed at the emergence points of dislocations.
The present chapter treats the subject of etching of crystalsfrom the
viewpoint of electrochemical processes. The fundamental mechanisms
of dissolution and selective etching are first outlined in Secs. 2 through 7.
The morphology of etch pits and the reliability of etch figures in locating
the emergence points of dislocations are then briefly discussed in Secs.
53
8 and 9, respectively. Finally, the application of chemical etching in

producing etch profiles is presented in Sec. 10. Chemical etching techniques
are well known regarding the determination of the symmetry of a crystal
face from the shape of etch pits, the study of the density and distribution
of structural defects in crystals, the detection of nuclear tracks in solids
and in the fabrication of electronic devices. In the present chapter,
however, these aspects are either omitted completely or considered
marginally because they are adequately covered in the above monographs.
Similarly, all types of the theories of dissolution and etch-pit formation,
proposed so far, are not described.
2.0 MECHANISM OF DISSOLUTION
2.1 Driving Force for Dissolution: Some Basic Concepts and

Definitions
Under constant temperature and pressure conditions, any change in

a system proceeds from a state of higher to a state of lower Gibbs free
energy. Consequently, Gibbs function, G, is taken as a thermodynamic
potential, andthechangein thesystem asapassagefromastateof higher
to a state of lower potential. Thus the reaction,
Eq.(l) aA t bB+cC + dD tAG
is thermodynamically possible if the change in the free energy
Eq. (2) AG = (cG, t dG,) - (aG, + bG,) < 0

Since A G is related to changes in the Helmoltz free energy, AH, and
entropy, AS, by the equation
Eq. (3) A G = AH-TAS
a reaction possible at one temperature may not occur at another temperature.

If n,, n2,n3,..., ni represent the number of moles of components 1,2,
3, ..., i in a particular phase, we define chemical potential, p i , by,
It is the change in the Gibbs free energy of the phase with a change in the
number of moles of component i, while the temperature,the pressureand
the number of all other components are held constant. If ai is the activity
Chemical Etching Principles and Applications 55
of the i-th component, the chemical potential of the i-th component is

expressed by
Eq. (5) pi = p,* t RT In(ai)
where pi* is the chemical potential of i in some standard state of activity

ai*. Thus the difference in the chemical potentials of the i-th component
corresponding to two activities, a and a*, may be expressed as
Eq. (6) Ap = pi - p i * = RT In(ai/ai*)
The difference in the chemical potentials, A p i , is the driving force for

dissolution.
When a crystal is immersed in an etching solution, an electrical
double layer is produced. The double layer is developed as a result of the
adsorption of molecules of solvent, solute and reaction products at the
dissolving surface. The surface may be composed of the following:
1 . Charged ions of the same sign as on metal surfaces.
2. Oppositely charged ions of different sizes as on the surfaces
of halite-type structures.
3. Free dangling bonds as in the case of Il-VI and Ill-V
semiconductors.
It is believedthat the double layer is composed of a rigid layer of adsorbed
ions with a thickness equal to ionic diameters (the Stern or Helmoltz rigid
1ayer)followedby a diff use layer (Gouy-Chapmandiffuse layer) extending
up to several hundred microns into the solution. Below we describe the
surface potentials of this double layer.
A homogeneousmaterialcontaining an excess electrical charge and
situated in vacuum has an outer potential, or Volta potential, Q , which is
the amount of work necessary to bring a unit electric charge from infinity
to a point close to the surface where Q attains the maximum value.
The presence of charge on the surface of a material causes the
formation of a surfacedipole layer. Thus further work is needed to take the
test charge from just outside the surface to it. The potential change
resulting from the presence of the dipole layer is the surface potential x.
The inner, or Galvani, potential is defined as (4),
Eq. (7) 4 = Q tx
where I and x are mainly due to surface charge and surface dipoles,
respectively.
From the foregoing, it follows that the work done in taking the test
charge (or ion) from infinity to the material surface is chemical and
electrochemical in nature. Thus the electrochemical potential ,Ti for an ion

of charge zi may be given by (4)
where piis the chemical potential of the ions of component i (Eq. 5) and F
is the Faraday constant. The work required to take 1 mole of ions of the
component i across the double layer is given by
where Ap is expressed by Eq. 6.
2.2 Dissolution Process Controlled by Surface Reactions and

Volume Diffusion
We now consider the rates of dissolution of a crystal surface in

solution. The processes taking place during dissolution are essentially
similar to those in heterogeneouscatalysis. We may take acrystal surface
as a catalytic surface that facilitates the transfer of ions or molecules to
and from the etching medium. Thus, the dissolution process may be
viewed as a succession of steps:
(1) Diffusion of reactant ions/molecules to the surface.
(2) Reaction in the interfacial layer.
(3) Adsorption of reactant onto the crystal surface.
(4) Reaction between the crystal and the adsorbed reactant
species.
(5) Desorption of reaction products from the surface.
(6) Reaction of products in the interfacial layer.
(7) Diffusion of the ultimate reaction products away from the
surface.
The sequence of ultimate reactions does not necessarily include
steps 2 and 6, but such steps are indeed involved, for example, in the
decomposition or formation of complex ions in the solution phase before
or after a chemical reaction with the surface ion/molecules. However, for
the sake of clarity we did not include steps representing the formation of
activated complexes during the processes of adsorption and in the
chemical reactions in steps 2 through 6.
Any of the above consecutivesteps may control the etching process,
but it is always the slowest step that is rate limiting. Steps 1 and 7 of the
transport of the reacting species and reaction products are controlled by
diffusion kinetics, while surface reactions control the remaining steps.
Some of the above steps may be very important in certain systems, while
they may play no role at all in the etching process in others. For example,
steps 2 and 6 are of minor importance in the etching of water-soluble
crystals in undersaturated aqueous solutions, but the same steps, to a
large extent, control the etching of metals and semiconductors, in which
a supplementedreagent removesreaction products tending to adhere to
the crystal surface.
Whether etching is controlled by diffusion or by surfacereactions can
be determined by using a number of criteria (2). One of these criteria is the
activation energy, E, for dissolution, obtained from data on the temperature
dependence of dissolution rates. If the activation energy has relatively low
values (of the order of 20 kJ/mol or less) such that the dissolution rate
increases only slightly with temperature, dissolution is diffusion controlled.
In the case of dissolution controlled by surface reactions, E has relatively
high values of about 50 kJ/mol or more. Another criterion of diffusion-
controlled dissolution is that the activation energy for dissolution is of the
order of the activation energy for viscous flow. Dissolution rates controlled
by diffusion and the associated activation energies increasewith stirring,
while those of surface-reaction-controlled dissolution remain invariant.
The rates of dissolution of faces of different orientations are usually
anisotropic and isotropic in the case of surface-reaction- and diffusion-
controlled dissolution, respectively (see also Sec. 10).
Recalling that crystal-medium interfacial layer is composed of the
rigid layer 1 - 2 atomic diameterswide and the diffuse layer extends up to
some 10-2cmdeep into the solution, we may view the diffusion-controlled
dissolution process as one in which the reactant species and reaction
products have to traverse the wide diffuse layer in order to enable the
surface reactions to proceed. In this case, the rate of dissolution, expressed
as the amount, m, of material removed per unit time, t, is described by
Fick’s first law, namely:
Eq. (10) dm/dt = DA (co-c)/6
where D is the bulk diff usion coefficient of diffusing species in the medium,
A is the dissolving surface area, S is the thickness of the diffuse layer, and,
in solution etching, c and c, are the concentrations of solute in the bulk
solution and at the crystal surface, respectively. In “pure” chemical and
electrochemical etching, these concentrations may be replaced by chemical
(Eq. 5) and electrochemical potentials (Eq. 9),respectively.
In the case of surface-reaction-controlled dissolution, there is no
diffuse layer and the exchange of reacting species and reaction products
proceeds directly at the rigid layer. Depending on the type of surface

reactions (i.e. solvation, simple chemical or electrochemical), surface-
reaction-controlled dissolution process is different for different types of
materials. After a brief description of the classification of types of dissolution
in the following section, we shall discuss these mechanisms in Sec. 2.4.
2.3 Types of Dissolution
As mentioned above, the formation of activated complexes is a

prerequisitefor a chemical reaction. In order to attain the activated state,
the system requiressome source of energy. There are three main sources
by which the molecules can acquire the necessary activation energy:
1 . Thermal energy associatedwith translation of molecules and
their internal vibrations and rotations.
2. Energy supplied by an electric field.
3. Absorption of radiation.
Taking the above three sources as the basis, we can distinguish between
the traditional terminology of dissolution as follows:
Chemical etching (1)
(a) Solvation or solution etching.
(b) "True" chemical etching.
(c) Electrochemical etching.
€/ectro/ytic etching (2)
Photoetching
(a) Photoelectrochemical etching (1 + 3)
(b) Photoelectrolytic etching (2 t 3)
The subdivision of chemical etching stems from the types of reactions
taking place on a dissolving surface. In solvation or solution etching the
ions/molecules on a surface are severed by solvation, in true chemical
Eq (1 1) -
etching by double displacement reactions, e.g.,
MgO t 2HCI MgCI, t H,O
in which the oxidation state of the ruptured ionsremains the same as in the
crystal, and in electrochemical etching by charge transfer, e.g.,
Eq (12) Zno t 2HCI+ZnCl, + H,

in which the ruptured atoms/ions undergo a change of oxidation states.
Finally, it should be pointed out that since photoetching and electrolytic
etching are carried out at T > 0, there is always a thermal contribution to
the activation enegy.
2.4 Dissolution Kinetics in Terms of Interfacial Layer Potential
Dissolutionof Ionic Compounds in Aqueous Solutions (5). The

potential of a double layer developed on the surface of a crystal placed in
an aqueous solution is given by (cf Eqs 6 and 8),
Eq. (13) 4 = (RT/zF) In(a/a,)
where a and a,, respectively, are the actual and equilibrium solute activity
in solution. In theabsenceof anadditionalsource,other than diffusion,for
the activation energy for the addition of ions to the crystal, the exchange
rate, ve ,for ionsacross the double layer in a saturated solution is given by
Eq. (14) ve = Dc/S
Equation 14 gives a maximum exchange rate due to bulk diffusion of ions

in the solution in the absence of any additional source of activation
energy.
If the crystal is charged in the solution, the rates of removal, vr , and
deposition, v, , of each ion may be given by
Eq. (1 5a) vr = nkr exp(azF4 / RT}
Eq. (15b) v, = ckdexp{-(1 -a)zFd/RT}
where n is the number of ions per cm2, a is a transfer coefficient (see

below),and k's are rate constants. For pure diffusion, k, = D/S. Thus the
net rate of exchange (vr - v,) is,
Eq. (16) (vr-v), = nkrexp{azF4/RT} - ckdexp{-(1 - a)zF#/RT}

For a freely dissolving crystal MX in an undersaturated solution for
which, at the isoelectric point, the rates of exchange for Mt and X are
(v,, - val) and (vr2- v,), the net removal rates must be equal, Le.,
Eq. (17) n,kr, exp{alz,F4/RT} -

C l k d l exp{-(l - Q ~ ) Z , F ~ / R T }
= n2kr2exp{a23F4/W - c1kd2 expH1 a2)5F4/RT1 -
In this equation, the suffixes 1 and 2 denote cations and anions, respectively.
In the case of dissolution in pure solvents, Le., when the concentration
of both ions in the solution is zero, such that v, = 0, from Eq. 17 one gets
and from Eq. 18,
Eq. (19) 4 = [RT/F(a,z, -a,zl)l In[(n,kr,)/(npkr,)l
Using the values of from Eq. 19, one can estimate the dissolution rate
from Eq. 18.
For an undersaturatedstoichiometric solution, the potential may be
written from Eqs. 15a and 15b as:
Eq. (20) 4 = (RT/zF) In(nk,/ck,)
Equating Eq. 20 for both cations and anions, one gets,
Eq. (21) n2kr2= {n,kr,/n,k,,)z~~( z 1- z 2 )
* {(C2kd2)'i/(5- Zz)/(Clkdl)Z~/(Z1-
'2) }
where c, and c, are the concentrations of cations and anions, respectively,

in the stoichiometric solution.
SubstitutingintoEq. 15athevaluesof~from Eq. 19,oneobtainsthe
dissolution rate,
Eq. (22) v = n2kr2( n , k r , / n 2 k i 2 ) a ~ ~ ~ ~ ~ Z ~ ~ ~ ~ )
in pure solvents. Introducing Eq. 21 into Eq. 22, and takinga, = ap,one
gets,
in undersaturatedsolutions. Equation 23 gives the maximum dissolution

rate of a crystal, and reduces to Eq. 13 for c, = cp and k!, = ,k, .
Dissolution of Ionic Crystals in Acidic and Alkaline Media (5).
All ionic crystals are not soluble and many insoluble substances are
etched in acidic and alkaline media. The ideas of dissolution in terms of
double layer potentials can be extended to such situations in which it is
considered that protons or hydroxyl ions of an acid or an alkali are
complexingagentsfor anions or cations, respectively. For example, in the
dissolution of MgO in aqueous HCI, H' ions are thecomplexing agentsfor
02- such that OH' and H,O are produced.
Assuming that, for the dissolution of a hydroxide crystal, the rate of
removal of hydroxide ions as water is much greater than that of OH' so that
thelatter isnegligible, then,foIlowingVermilyea (5),wemaywritetherates

of removal of cations and hydroxyl ions as
Eq. (24a) vrl = nlkrl exp{a,z,Fd/RT}
Eq. (24b) v, = n2kr2ciexp{cr,z,Fd/RT}
where ci is the H' ion concentration. For afreely dissolving crystal with vrl
= vrp, from Eqs. 24a and 24b we obtain,
This equation shows that the potential 4 increaseswith H+ion concentration.

Substitution of Eq. 25 into Eq. 24a yields:
Thus if Q 1 = Q,, the dissolution rate increases as ci112,ci213,ci314,and ci4I5

for cations of charge 1 , 2, 3 and 4, respectively.
The dissolution mechanism of an oxide may be complicated. The
following situations are possible:
1. Surface oxide ions may first hydrate to form OH- ions. The
dissolution mechanism then is like that for a hydroxide.
2. Oxide ions may be removed by reaction with a proton to form
OH', which is subsequently removed as water as a result of
thereactionwithanother proton. Sincez, = 2, the dissolution
rate will increase as ci113,ci112,ci315,and ci213for cations of
charge 1, 2, 3 and 4, respectively.
3. Oxide ions may react directly with two protons to form water.
Then ci2will replace ci in Eq. 24b and the dissolution rate will
increase as ci213,ci, ci615,and ci413for cationsof charge 1, 2,3
and 4, respectively.
Dissolution of an oxide or hydroxide in an alkaline solution, in which
OH' ions are the possible complexing ions, may be treated similarly.
However, in this case the reversereaction for the crystal anion cannot be
neglected and the equations become complicated. The dissolution of
compounds other than oxides and hydroxides may also be treated in a
similar way.
Finally, two comments should be made in connection with the above
mechanism. First, theassumption Q 1 = mayyi yield higher rates than the
actual ones. Second, since hydrogen and hydroxyl ion concentrations
are related to the pH of a solution, the dissolution rate is expected to
depend on solution pH, as has been observed experimentally (6)(7).
Dissolution of Metals. A pair of eleclrodes immersed in an electolyte

solution and connected by an external metallic conductor represents a
typical electrochemical cell in which reaction at the surface of an electrode
takes place by the transfer of charge. If the cell is converting chemical
energy into electrical charge, it is a galvanic cell; if it is using externally
applied electrical energy to carry out a chemical reaction, it is called an
electrolytic cell. The distinction between a galvanic and an electrolytic
cell is given by the conditions (4):
Eq. (27a) A 4 < E galvanic cell,
Eq. (27b) A4 > E electrolytic cell,
whereA4 is the potential difference due to the electric double layer at the
electrode-electrolyte interface, and E is the cell emf.
A metal dipped in an electrolyte solution dissolves as a result of
electrode (electrochemical) reactions. Electrochemical reactions differ
from other heterogeneousreactions in that their rate constants dependon
the value of A4. According to the electrochemical nature of metal
dissolution, there are anodic or cathodic sites on the dissolving surface,
where oxidation of metal M and reduction of species N of the electrolyte
take place, respectively. Thus all metals give electrons and pass into the
ionized state.
Eq.(28) Mo +MXt + NX'

so that the metal itself is charged positively and acts as an anode (oxidation
process). Here x is a small integer. In order for the metal ions, Mx+,to be
removed from the surface, the reaction (reduction process),
Eq. (29) No + xe' + NX-
proceeds on the cathode, with the result that negatively charged ions are
formed. Equations 28 and 29 are called half-cell or partial reactions
occurring at the metal-electrolyte interface. Thus the following overall
reaction takes place:
Eq. (30) Mo + No+ Mxt + NX-

These Mxt and NX-ions then go into the solution. If the reaction product
is soluble, then an atom has been removed; if it is insoluble or poorly
soluble, then a complexing agent is required to remove it by an independent
subsequent reaction.
When positively charged metal ions, Mxt, leave the metal surface,
the metal surface remains negativelycharged. Theattraction of positively
charged metal ions in solution to the negatively charged metal surface
leads to the development of the double layer and the potential difference,
A4 , across it, as mentioned in Sec. 2.1.
Under nonequilibrium conditions, there is a net anodic or cathodic
current, i.e.,
Eq. (31) i = ic - ia
where iaand ,i are the currents in anodic and cathodic directions, respectively.
The electric potential difference between the terminals of the elecVochemical
cellthen departsfrom theequilibriumvalueA4 = E.TheactualvalueofA4
depends on the current density, i, at the electrodes. The difference,
Eq. (32) -
A4 (i) A4 = q
is the polarization (also called overpotential or overvoltage) of the cell,

and its value is determined by the potential necessary to overcome the
resistance in the electrolyte and the leads.
When a diffusion layer is present in a solution, the concentration of
the latter at the double-layer interface is c* rather than c, in the bulk.
Because of the concentration dependenceof chemical aff inity, a potential
difference is developed across the diffusion layer. This potential difference
is called concentration or d i f f u s i o n overpotential and is given by,
Eq. (33) qo = (RT/zF) In(c*/c,)
FromFick’sfirstlaw, therateofdiffusionofionstothedoublelayer isgiven

by!
Eq. (34) -dn/dt = DA (dc/dx) = DA(c, - c*)/S

Therefore the current density to the cathode is,
Eq. (35) ir = -(zF/A) (dn/dt) = -zFD(c,- c*)/6
Therefore for c = 0, a limiting current density,
Eq. (36) i,i, = -zFDc,/S
occurs. Using Eq. 36, Eqs. 33 and 35 may be written as:

Eq. (37) = (RT/zF) In{ 1 - (i/i,J}
Eq. (38) ir = ili, { 1 - exp(zFq,/RT)}
The values of il (see Eq. 36) can be increased by increasing the

values of D (by increasing temperature) and by decreasing the value of 6
(by stirring the solution). The dependenceof current density on electrode
potential showing concentration overpotential is schematically shown in
Fig. 1. The curve has three regions, which we will describe in Sec. 6 on
electrolytic etching.
At equilibrium, the currents of each ionic species to and from the
electrode should be equal, i.e.,
. .
Eq. (39) I, = I,= .I,
When A4 is different from its equilibrium value, A40, at low current
densities, the net anodic or cathodic current is associated with the extra
electric potential difference, called the activation overvoltage or
activation overpotential, given by
Eq.(40) Br = A4 - A 4 0
At equilibrium, the anode and cathode current densities are given
by:
Eq. (41a) ,i = ioexp{ -( 1- Q) zFq,/RT} , and
Eq. (41b) ,i = ioexp{~zFq,/RT}
where iois the exchange current. The net current density, therefore, is
given by
Eq. (42) i = ,i - ,i = io[exp{azFq,/RT} - exp{(l- Q) zFq,/RT}]

If qr > > RT/zF, we have the well known equations:
Eq. (43a) hi, = l n i o - [ ( l -a)zF/RT]q,and
Eq. (43b) hi, = hi,- [crzF/RT]q
Depending on the current density across an electrode, the surface reactions

or the transport of ions may become rate determining. At lower current
densities, with ionic reactants, the Galvani potential difference, A4 ,will
favor the transfer of an ion through the layer in one direction but it will
hinder its transport in the opposite direction. Thus in the process, only a
fraction, Q , of the electrical energy difference zFAtp is used. This fraction
is the transfer coefficient.
The electrochemical mechanism of dissolution is illustrated
schematically by the simplified polarization diagram shown in Fig. 2. The
open circuit potentials of the cathode process, E, and the anodic process,
E, are the equilibrium potentials of the corresponding partial reactions of
Eqs. 28 and 29. The dissolution current corresponds to the steady-state
rate of dissolution. The corresponding dissolution potential of the dissolving
solid lies between the equilibrium values of the cathodic and anodic
reactions. From the figure, it also follows that conditions which shift the
point of intersection of the anodic and cathodic polarization curves by
decreasing their slopes, lead to an increase in dissolution rate. Conversely,
an increase in the slopes of the curves lowers the dissolution rate.
It should be emphasizedthat activation overpotential, q , , is caused
by slow reactions in the solution close to the electrode (reaction overpotential)
and in the process of deposition of a solid product on the electrode
(crystallization or ohmic overpotential). Concentration or diffusion
overpotential, qD, on the other hand, develops when the activity of the
potential generating substance decreases substantially at the electrode
in comparison with that in the bulk solution.
Finally, we note that different faces of a crystal dissolve at different
rates. This aspect has been discussed by Despic'( 10). The difference has
been attributed to structural changes at the surface and the varying
participation of different crystal planes at a face of a given orientation.
Dissolution of Semiconductors. The mechanism of dissolution
of semiconductors is essentially similar to that of metals, described
above. The basic difference in the dissolution behavior of metals and
semiconductors lies in the concentration and type of charges responsible
for surface reactions. In semiconductors, the concentration of charge
carriers is much smaller than in metals because of the predominantly
covalent nature of bonding. The electron transfer process may involve
either valence band or conduction band electrons at the semiconductor
electrode while only conduction band electrons take part at metal eleclrodes.
Furthermore, the kinetics of dissolution of metals is determined by
electrochemical reactions occurring in the solution or at the solution-
metal interface, whereas the rate-determining process in the dissolution
of semiconductors may also involve phenomena taking place inside the
surface.
In contrast with metals, there are two types of carriers that can take
part in the anodic and cathodic partial reactions involved in the dissolution
of semiconductors: electrons in the conduction band and holes in the
+
C
0
L
L
3
u
7D
Potential E ( v o l t s )
Figure 1. Schematic illustration of the dependence of current density on

electrode potential showing concentration overpotential ( 8 ) .
I I
Logarithm current
Figure 2. Simplifiedpolarization diagram illustrating the electrochemical

mechanism of dissolution (9): (DP) dissolution potential, (DC) dissolution
current, (1 ) cathodic polarization, and (2) anodic polarization.
valence band. Thus in acidic solutions such as HNO,, the anodic reaction
is as follows (9):
Si t 3H,O H,SiO, t 4H t 4ae-, valence band,
Si t 3H,O -+ H,SiO, + 4H t 4( l-a)e-,conduction band,
or.
Eq. (44) Si t3H,O t4ae+(holes) +H,SiO, t 4 H t 4(1- a)e-
where a is the fraction of electrons entering the valence band. The

negative charges left in the silicon are used up in the cathodic reaction, but
they are not directly taken up by the acid. Insteadthey are used up by the
products of the acid that are stronger oxidizing agents. An initial reaction
is believed to take place between nitric and nitrous acids:
Eq. (45a) HNO, t HNO, -+2NO, t H,O
The charges are then taken up by NO,
Eq. (45b) 2N0, + 2e' -+ 2N0,'
The nitrite ions subsequently react again to yield more HNO, that feeds
back to reaction (45a):
Eq. (45c) 2N0,- t 2H++ 2HN0,
Therefore the overall cathodic reaction may be given by,
Eq. (46a) 2HN0, t 4H' t ce- (valence band)

-
t (1 c)e- (conduction band) -+ 2HN0, t 2H,O , or
Eq. (46b) 2HN0, t 4H+ t (1- c)e--+ 2HN0, t 2H,O t ce+
where c is the fraction of electrons that enter the valence band.

By combining the anodic and cathodic reactions given by Eqs. 44
and 46b, one gets
Eq. (47a) Si t H,O t 2HN0,

+H,SiO, t 2HN0, t -
4(c a)e- t 4(c - a)e+
In the steady state, sufficient HNO, is provided by reaction (45c) to feed
reaction (46c). Once a reaction initiates, it provides its own starting

material (autocatalytic reaction).
In place of the cathodic reaction of Eq. 46, an alternative reaction:
Eq. (48) HNO, t H' t (1- c)e'+NO t 20H- + ce-

has also been used (1 1)( 12). This equation gives the following overall
reaction :
Eq. (47b) 3Si t 4HN0, t 9H20

+3H2Si0, t NO t 4(c-a)e- + 4(c-a)et
The oxide H2Si03is insoluble. Therefore, HF or some other reagent
is used in the solution to convert it into a soluble compound, e.g.,
Eq. (49) H2Si0, t 6HF +H2SiF, t 3H20
Thus the overall dissolution may be written by combining this equation

with Eqs 47a and 47b, i.e.,
Eq. (50a) Si + 2HN0, t 6HF +H2SiF, t 2HN02 t 2H,O
Eq. (50b) 3Si t 4HN0, + 18HF +3H2SiF, t NO + 8H20

From Eqs. 47a and 47b, we note that the rate of production of
-
electrons and holes is proportional to (c a), and is equal to (c - a) times
the dissolution rate. For c > a, there is a net generation of electrons and
holes, implying, thereby, that the dissolution process is not controlled by
charge carriers (carrier control). For c < a, on the other hand, there is a
net consumption of electrons and holes by the dissolution reaction. This
means that, in this case, the supply of carriers controls the dissolution
process (diffusion control). It should be noted that electrons and holes
are produced in equal numbers during the electrochemical dissolution of
semiconductors (Eqs 47a, 47b).
If there is a net consumption of carriers during dissolution, then the
rate of consumption of holes by the anodic reaction exceeds their rate of
production by the cathodic reaction, so that the supply of holes by thermal
generationand diffusion could determinethe rate when an n-type material
is used for dissolution. Similarly, when the rate of consumption of
electrons by the cathodic reaction exceeds their rate of production by the
anodic reaction, the supply of electrons could be the rate-determining
step when a p-type material is used.
The stoichiometry of reactions 47 and 50 predicts the value of the
molar ratio, -y,,,*, of oxidant to complexing reagent (i.e. [HNO,]/[HF])

corresponding to the maximum rate of dissolution. The reaction of Eq.
50b, for example, gives the maximum dissolution rate for r,* = 0.22
(1 1)(13), while that of Eq. 50a predicts 7 ", = 0.33, but because of the
autocatalytic charater of the oxidation process, the ratio will probably be
lower (14). Equation 50 corresponds to the stoichiometry of the reaction
determined by surface reactions, but Eq. 47, involving reaction by
autocatalyticallyformed HNO,, indicates that the stoichiometry is determined
by reactions between the electric double layer and the bulk solution.
The reactions of Eqs. 50a and 50b correspond to complete oxidation
of the semiconductor as SiO,. However, if the depletion of the oxidizing
agent in the vicinity of the interfacial layer is fast, incomplete oxidation
takes place in which the number of charges involved in the oxidation
process is less than that predicted by these equations. For example,
corresponding to Eq. 50b for the dissolution of Si we have (15):
Eq. (51) Si t (z/3)HN03 + 6HF

+ H,SiF, +(z/3)NO +(2z/3)H20, z < 4
The value of z decreases in the order 4,3, 2 and 3/2 for SiO, , Si,O, , Si0
and Si, O,, respectively. Similarly, corresponding to the maximum rate,
the molar ratio also decreases for z = 4, 3, 2 and 3/2, in the order r,* =
0.22,0.16,0.11 and0.083.
The above basic ideas on the etching of elementalsemiconductors
are equally valid for compound semiconductors. However, in the latter
case, the partial reaction at the anode gives two oxidation products, e.g.,
(16)
Eq. (52) GaAs + 100H- + 6e' + GaO,' + AsO,~- + 5H,O

where OH- ions are formed by the well known electrolysis reaction:
Eq. (53) 2H,O + 2e- + 20H' + H,
The mechanism of dissolution of semiconductors usually involves

a number of discretesteps inwhich intermediatereactions corresponding
to surface states take place (17-25). For example, to describe the
dissolution kinetics of GaAs in Cr0,-HF solutions, the following oxidation
and reduction steps have been proposed (19)(25). Oxidation of GaAs
takes place in a series of consecutive steps:
\Ga:As t e+
I k 'Ga-As /
Eq. (54)
/ \ I \
>
\ / /
Eq. (55) / GaOAs + e + +
k* GaAs
\ \
In these equations, X- is a nucleophilic reagent. Thus, in all, six holes are

required to dissolve one GaAs molecule. The reduction reaction in which
holes are injected into the valence band, is given by:
Eq. (57) [CrV1- CrV'- CrV1]
44
+ -
+ 6e+
[Cr"' - Cr1I1 Cr1I']ads
where it is assumed that the trichromate ion [CrV1- CrV1- CrV1]is the
electroactive species. The model involves a depassivation of the adsorbed
complex with activated surface states from the GaAs oxidation process,
i.e.,
\ / k, \ /
Eq. (58) Ga-As + [Cr"'- CrV1-Crlll]ads3 GaAs + 2Cr"' + Cr" + 3ef
\ / \
Here k's are rate constants of the reactions. The overall partial reactions
are :
Eq. (59a) GaAs + 6e+ + Gall1+ As1"
Eq. (59b) -
[ CrVl- Crvl CrV1] Crll + 2Cr1I1+ 1Oe'
Thus the surface coverage due to the adsorbedfilm is important in etching

kinetics, and the injection of holes from CrV1and dissolution of GaAs via
nucleophilic attack are possible only at the free surface.
Information about the mechanism of dissolution of semiconductors
may be obtained from investigations of current-potential characteristics
of a semiconductor-electrolyte system (16)( 18-23), and from photo-
luminescence and electroluminescence spectra of semiconductor-
electrolyte interface (26-28). For details of surface reaction mechanisms,
the reader is referred to the above cited literature.
Figures 3a and b show the current-potential curves for p-type ( 100)
GaAs electrodes in 0.05 and 0.5 M K,Fe(CN), , respectively, at pH 13. It
is obviousfrom the figures that the total current-potential curves (1) show
three distinct plateaus. The partial cathode current due to Fe(CN),,+
reduction (curve 2) (obtained by using a ring-disk electrode) and the
partial anodic current due to GaAs reduction (curve 3) (obtained by
subtracting the cathodic current from the total current) reveal a number of
40 3
/----------
- /’ 1
41
I
0
40
Figure 3. Current-potential curvesfor p-type (100) GaAs electrode in (a)

0.05 M and (b)0.5 M K,Fe(CN), solutions at pH 13: (1) total curve; (2) and
(3) are the cathodic and anodic partial curves. Electrode rotation rate 200
rpm. From Ref. (16).
72 Electroc hemistry of Semiconductors and Electronics
features:
1. At relatively large positive potentials, anodic partial current is
constant.
2. The limiting anodic partial current is much larger than the
correponding cathodic current.
3. At higher electrolyte concentrations, the limiting current is
considerably larger than the corresponding anodic current..
There are two general aspects of the curves of Fig. 3. At voltages
greater than E,, the anodic partial current is constantat both low and high
concentrations. Dissolution, in this case, depends on the transport of OH'
ions to the electrode surface. At voltages smaller than E,, the cathodic
partial current is larger than the corresponding anodic current for higher
electrolyte concentration, while the cathodic current is smaller than the
corresponding anodic current at lower electrolyte concentrations. This
difference is due to the fact that, at low electrolyte concentration, the
dissolution kinetics are determined by the reduction of the oxidizing
agent, while at high concentration, they are controlled by the anodic
reaction (i.e. OH-diffusion), and the rest potential of the system is much
more positive than that at low concentrations.
2.5 Dissolution Kinetics in Terms of Surface Adsorption Layers
In the area of crystal growth from solution, the growth process is

considered to occur according to the following consecutive steps:
1. Transport of solvated growth units from the supersaturated
bulk solution to the crystal surface.
2. Adsorption of the growth units onto the surface to constitute
a loosely adsorbed surface layer.
3. Migration of adsorbed growth units over the surface.
4. Integration of growth units at energetically favorable sites
(Le. kinks) available at ledges, and release of solvent molecules
from the solvated growth units.
5. Transport of the released solvent molecules to the bulk solution.
These steps are very similar to those described Sec. 2.3. Steps 1 and 5 are
controlled by the usual diffusion kinetics, while steps 2, 3 and 4 are
controlled by the kinetics of surface adsorption, surface diffusion and of
integration of growth units at kinks.
The source of kinks is a crucial factor in the theories based on the
concept of surface adsorption layers. In the case of a perfect crystal, the
kinks may be supplied by two-dimensional nuclei forming on its smooth,
flat surface, but, if there are dislocations emerging on the surface, they
may serve as unending sources of kinks.
In dissolution, the sequence of steps is reverse to that in growth.
Therefore, as in growth, we have the following two types of models: two-
dimensional nucleation models, and surface diffusion models.
Two-dimensional Nucleation Models. Dissolution of a perfect,
flat surface can occur if there is a change in the free energy of the system
(Fig. 4), i.e.,
Eq. (60) AG, = 2rrh-y- r r 2hAp /a
where r is the radius of the developing circular disk-shaped nucleus of

height h, -y is the surface energy, Cl is the molar volume, and Ap is the
potential difference given by Eqs. 6 and 9. The critical radius r* of a
nucleus that can grow larger to give a dissolution ledge may be obtained
by taking dG, /dr = 0, i.e.,
Eq. (61) r* = @/Ap
and the corresponding free energy change (cf Eq. 60),
Eq. (62) AGp* = rh-y2n/Ap
The rate of two-dimensional nucleation (i.e. number of nuclei of radius, r*,

formed per unit area per unit time) is given by,
Eq. (63) J = J,exp{-AG,*/kT}
with pre-exponentialfactor J, = 10l2 /sec.

Depending on the rate of movement and the nature of formation of
two-dimensional nuclei, we have three types of two-dimensional nucleation
models (29). In the mononuclear model, the rate-limiting step is the
formation of a critically-sized nucleus of height, h, and once it is formed,
it spreads across the surface at an infinite rate, i.e., v_ = 0 0 . The surface
dissolution rate, vp, is given by
Eq. (64) V, = AJh
where A is the total area of the dissolving crystal. In the polynuclear

model, the lateral spreadingvelocity v_ = 0, but dissolution takes place by
the accumulation of critically-sized nuclei. The surface dissolution rate
then is
Eq. (65) v, = Jh
The mononuclear model predicts that vp depends on the area, A, of the

dissolving crystal, while the polynuclear model predicts that the dependence
of vp, on undersaturation, u = (c, - c) / c, = Ap / kT, should show a
maximum. Such dependences have not been observed experimentally.
I
I
I
r2 r*
r-
Figure 4. Schematic representation of the free energy change involved

during the formation of two-dimensional circular dissolution nuclei on: (I)
a perfect surface, and (11) at the emergence points of a dislocation.
Thus it is believed that these models probably represent extreme situatioris

of the rate of lateral spreading of the stable nuclei.
In the birth-and-spreadmodel, it is assumed that nuclei develop and
spread at finite velocities. Here the dissolution rate
Eq. (66) vP = hJ’i3 v-’/~
and the spreading velocity of a straight ledge is given by,
Eq. (67) v- = 2c,’D,n {exp(-&/kT)- l}pA/hA,
where c, ’is the equilibrium solute concentration on the surface, A, is the

interdiffusion distance, p and A are retardation factors connected with
interstep and interkink distances, respectively, and are always less than
unity,and D, is thesurface diffusion coefficient. Then from Eqs. 63,66and
67, one obtains
Eq. (68) vp = B ( A P / ~ T )exp{-

~ / ~ (AGp*)/(3kTAp)} ,with
Eq. (69) B = 2hi/6n5/s( V / X ) ’ /( ~~ , D , ~ A C ~ ’ / A , ~ ~ ~
where V and n, are the mean velocity and concentration of surface

adsorbed molecules/atoms, respectively.
According to Eq. 68, v = B(Ap /kT)5/6at relatively high values of
(Ap /kT), while v,/(Ap /kT)’/l is an exponential function of (Ap /kTyl at
low (Ap /kT) values.
Surface Diffusion Model. This model, developed by Burton,
Cabrera and Frank (30) postulates the presence of screw dislocations
which ensure a constant source of kinks by virtue of the geometrical
arrangement of atoms/molecules at their cores. Thus dissolution can
proceed by the removal of atoms from these unending sources of kink
sites. Since the unwinding ledge is fixed at the emergence point of a
dislocation and its inner parts move radially at afaster rate than the outer
parts, the ledge will yield a depression spiral.
The spacing, yo, between successive ledges emanating from the
dislocation is given by (31)
Eq. (70) yo = 19r*
and the velocity of the step is given by (30),

The dissolution rate normal to the surface is expressed as
Eq. (73) A p , = 197 n/%, and
Eq. (74) C =nSA co'D,/&'
Equation 20 predicts that vp = C (Ap / kT) for Ap > > Ap ,, and vp = (CkT/Ap
( b / k T ) * f ~ A p< < A p , .
Edge dislocations have their Burgers vector in the plane of the
dissolved crystal surface. Therefore they are not expected to act as
nucleation centers for dissolution. However, Keller (32), and Bauser and
Strunk (33)(34) showed that edge dislocations can also act as persistent
sources of steps for growth and dissolution. Bauser and Strunk (33)(34)
attributed the nucleation at an edge dislocation to its dissociation intotwo
partial dislocations with a possible reconstruction of the dislocation
emergence point to yield a step at the surface. Frank (35) suggested that
surface stresses of the same order of magnitude as their surface free
energies associated with an edge dislocation provide regions compressed
on one side and dilated on the other side. These surface stresses supply
the extra energy requiredfor the formation of nuclei even if no dislocation
component perpendicular to the surface is present.
Information about the actual dissolution mechanism based on
surfaceadsorption layers may beobtainedfromanalysisof kinetic dataon
the undersaturation dependence of dissolution rate and observation of
surface morphology of etched surfaces (36).
3.0 MECHANISM OF SELECTIVE ETCHING
In the preceding section, we were dealing with general dissolution

of a crystal without considering some specific points on its surface, which
may dissolve selectively. However, it is usually observed that, apart from
the general dissolution, etching also leads to the formation of etch pits or
hillocks at the sites of defects emerging on the crystal surface. All types
of defects ranging from one- to three-dimensionalones act as the sites of
preferred or selective etching. The resulting etch figures are characteristic
of the distribution and nature of defects revealed during etching. In this
section we describe briefly the mechanism of formation of etch pits at the
sites of dislocations.
The formation of etch pits at the sites of dislocations is possible
when the rate of nucleation of pits of unit depth along the dislocation line,
v, the rate of lateral motion of steps of these nuclei away from the source
of nucleation, v,, and the rate of dissolution of the defect-free surface, vp,
are properly related (Fig. 5). Etch pits are formed only when v, > vp but
their contrast is determined by the ratio v,/vt. Under an optical microscope,
etch pits are visible when v,/vt > 0.1 (37). In the other case, when v,/v,,
etch hillocks are produced. A decrease in v, may be caused by insoluble
or sparingly-soluble impurities segregated along the dislocation lines or
at random sites in the form of precipitates, or by insolubte or sparingly
soluble impurities supplied by the etching medium.
Revelation of different types of defects as etch pits is explained in
terms of either stresses associated with, or impurities segregated at,
them. In the former case, the energy localized at a defect site accelerates
the rate of nucleation of unit pits, while in the latter case the rate is
increasedasaresultofenhancedsolubilityofthe material. Impuritiesmay
also affect the rateof motion of steps, i.e., vt, by poisoning them. Thus the
whole process of etch-pit formation may be viewed to be comprised of two
aspects: thermodynamic and kinetic. The thermodynamic aspect, as we
shall see below, is associatedwith the free-energy change involved, and
deals with the rate of etch-pit nucleation. The kinetic aspect, connected
with the lateral motion of steps, involvesmutual interaction between steps
and adsorption of impurities at them.
In Sec. 2 the overall surface dissolution was considered. Except in
the case of surface diffusion, we assumed that a dissolving surface is free
from defects. Since real crystals usually contain dislocations and other
defects, it is necessary to know their effect on dissolution rates and the
mechanism of formation of etch pits at their emergence points.
3.1 Models of Etch-Pit Formation
From the geometrical arrangementof atoms around dislocations, it

is immediatelyobvious that screw dislocations should yield etch pits with
slope, m, given by:
Eq. (75) m = tan B = v,/v, = h/y, = h A p / I g r Q
where h is the height of the dissolution ledge, and yo is given by Eq. 70.
We can check the validity of this model, which is basically the BCF
growth model as applied to dissolution, from the experimentaldata on the
etching of the (101) and (100) faces of KH,PO, (36). In this case,
dislocation etch pits are produced at undersaturationCJ > 0.06 (i.e., A p >
1 J at T = 300OK). Taking 7 = 0.01 J/m2, h = 7 . 4 ~O-lo
2 . 3 3 ~0-22 1 m and Q
= 1 . 7 3 ~ 1 O - ~ oneobtains
~rn~, m = 0.00524,averysmallslope indeedwith
B = 0.3'. Such spiral etch pits cannot be visible. Moreover, according to
Figure 5. illustration of the mechanism of the formation of a dislocation

etch pit in terms of different etch rates.
this model, the height of dissolution ledges should be equal to the Burgers
vector of the screw dislocation being etched. Such etch spirals have not
been reported in the literature (2)(35-38). The observed spiral etch pits
have ledges of heights of several tens to several thousands of nm (2)(38-
40). These spiral depressions are formed by the etching of helical
dislocations (38)(39) or by the occurrence of a mistake in the early stages
of formation of atomic steps and subsequent bunching of these steps
during their motion (40). From the foregoing, the mechanismof formation
of visible etch pits at screw dislocations by the unwinding of their helicoidal
turns is ruled out. Moreover,etch pits are revealedalmost equally at edge
and screw dislocations. Thus there should be some other mechanismof
etch-pit formation.
Cabrera (31)(41) realized that the elastic strain energy associated
with a dislocation can lead to the successive generation of two-dimensional
nuclei at its point of emergence on the surface. According to Schaarwdchter
(42), the core energy of a dislocation ensures this preferred two-dimensional
nucleation at dislocations. Hoek, et al. (43), and Koziejowska and Sangwal
(44) treated the process of etch-pit formation using equations of total
dislocation energy. Here we describe the essential features of the formation
of dislocation etch pits according to Cabrera’s theory.
The change in the free energy, AG, , for the nucleation of a disk-
shaped cavity of depth, h, and radius, r, at the site of a dislocation is given
by:
Eq. (76) AG, = 2 r h y r - r r 2 hAp /Q - h(Gb2a/4r) ln(r/ro)

where the last term on the right representsthe dislocation energy outside
its core, the parameter Q = 1/( 1 - v )anda = 1, respectively, for clean edge
and screw dislocations (where Y is Poisson’s ratio), rois the radius of the
dislocation core outside which the elastic continuum theory is valid, G is
the shear modulus, and b is the Burgers vector of the dislocation.
The behavior of the free-energy curve is determined by taking the
firsttwo derivativesof AG,and equating them tozero. The first derivative,
d AG,/dr, gives a quadratic equation having two roots:
Eq. (78) f = (Gb2~/2~2~2Q)Ap
Figure 5 shows the dependence of free-energy change on the radius r of

a two-dimensional nucleus. For .$ < 1, the critical free-energy change for
nucleation of a unit pit at a dislocation is:
Eq. (79) AG,* = A G , (r2)- AG, (r,)
and using Eqs. 76 and 77, one gets:
Eq. (80) AGd* = (Irhy2Q/Ap)(l -<)’”

+ (hGb2a/4~)In{[l- ( 1 -<)’12]/[1 + (1 - < ) ’ I z ] }
<
which is valid for r > roand < 1 . Equation 80 gives the decrease in the
free energy of nucleation at a dislocation site with respect to that on a
perfect surface. When f < 1 , we get:
<
Thus for < 1 , there is an energy barrier for etch-pit nucleation at a
dislocation site but the value of this barrier is less than that at the perfect
<
surface. For = 1 , there is a double root:
Eq. (82) r** = r, = r, = -yQ /2Ap = r*/2 ,and
Eq. (83) AGd* = 0
The values of r* and AGp*of Eqs. 81 and 82 are given by Eqs. 61 and 62,
respectively.
<
When > 1 , there is no real root, which means that there is no
nucleation barrier for etch-pit formation. Thus, since <increases with Ap
(Eq. 78), there is a particular value of Ap above which dislocations open
up spontaneously. The critical value of this Ap is obtained from Eq. 78 for
f = 1 , i.e.,
Eq. (84) Ap* = 2n2-y2n/Gb2a
In the region of undersaturation corresponding to 0 < Ap < 4 *, stationary

spiral dissolution or a repeated two-dimensional nucleation process
favored by an edge dislocation may take place.
A test ofthevalidityof Cabrera’smodel is thecondition thatr,, r,and
r** should be greater than ro,which is believed to be 2 - 3 b. In the case
of KH,PO, (36),with the appropriate values of -y, Q and Ap given above,
one obtains r** = 3 7 . 2 ~ 1 0 -ml ~= 5 b. This means that here Cabrera’s
model holds true. For the electrochemical etching of (0001 } faces of Zn
(45), with Ap/kT = 3 0 x l d ’ J, -y = 0.1 J/mz andQ = 2 . 4 6 ~ 1 0 -m3,
~ ’ one
obtains r** = 4 . 1 ~ 1 0 -m‘ ~= 0.02 b. Thus Cabrera’s mechanism is not
applicable in this case. Schaarwachter (42) also found that etch-pit
formation on metal surfaces requires the consideration of dislocation
core energy. Here we shall not go into the details of the theory which is
given elsewhere (2) together with other models.
3.2 The Slope of Dislocation Etch Pits
The slope of etch pits is determined by the normal and tangential

etch rates, given by (46),
Eq. (85) v, = hvexp{-(AG,* +AH)/kT} ,and
Eq. (86) v, = 2pASv(Ap/kT) exp(-AH/kT)
where vis the frequency factor of the order of 1 Oi3/sec, and AH is the free
energy change for a molecule going from the crystal surface into the
etching medium. Then the slope,
Eq. (87) m = [h/2pAS(Ap/kT)] exp(-AG,*/kT)

= [h/2pAS(Ap/kT)] exp{-(nhy2Q /kTAp) (1 - f ) i / 2 }
It is easy to see from this relation that with an increase in Ap , the pit slope
first increases to a maximum value at A p *, and finally decreases again.
However, for Ap > A p *, depending on the behavior off?,the pit slope may
even increase with Ap . The factor p is largely determined by the dissolution
kineticswhichare governed by the potential differenceAp and adsorption
processes taking place on crystal surface.
The above considerations hold good for fresh dislocations. Several
experimentalresults, however, show that the segregation of impurities at
dislocationsfacilitates the process of etch-pit formation. This behavior is
associated with the fact that impurity segregation leads to an increased
potential difference and decreases the free-energy change required for
the formation of critically-sized two-dimensional nuclei. In such cases,
the size and slope of etch pits can vary from those at fresh dislocations, if
vp < < vn. This feature may be understood from the model given by lves
and McAusland (47).
If v, is independent of v, the rate of pit widening is
Eq. (88) dD/dt = V, - vP
where D is the diameter of a pit after time, t. Then, using the equation of
pit slope m = v,/v,, from Eq. 88 one obtains, after rearrangement,
Under steady-stateconditionswhen the etch pits are formed at a constant

rate, we have
The following two conclusions can be drawn from Eq. 90:

1. For pits at different defects (different v,) produced by the
sameetchant (constantv,andv,), aplotof magainst(D/t)will
be a straight line. The values of v, and vp may be estimated
from the slope and intercept of the straight line.
2. If vp < < v, the straight line will pass through the origin and all
pits on a given surface, regardless of the type of defect, will
havethesamewidth. If v,issignificant, steep pitswill bewider
than shallow ones.
Several experimentalobservationsin support of the above model are
available in the literature. For example, on the (1 00) face of NaCl etched
in ethanol containing CdCI, impurity (48), pits of different sizes are
producedat low impurity concentrationswhen vpis large, but the pits have
a tendency to acquire the same size when impurity concentration in the
etchant is high enough to decrease vp markedly. These features may be
noted from the etch-pit patterns and the kinetic data presented in Figs. 6
and 7, respectively.
According to the above model, the rate of pit widening, v,, is a
function of the composition of an etchant. Such dependences have been
noted in the case of Zn (47) and LiF (37)(49). In the case of LiF, the main
effect of the impurity FeF, was found to decrease v,. This is expected if it
is assumed that v, is unaffected by the impurity. In the case of NaCl(48),
both v, and v, decrease or increase simultaneously (Fig. 7). This means
that an impurity can not only change v, as a result of inhibition of ledges
and kinks, but also affects v, through surface energy as a result of
adsorption.
3.3 The Role of Impurities
Among the effects of impurities on selective etching, two deserve

particular attention. Impuritiespresent in an etching solution facilitate the
revelationof etch pitsat dislocation and often change the pit morphology.
The actual process of impurity adsorption may be quite complicated for
different crystal-etchant systems, but certain features are common for
different systems.
In the case of etching of the (100)face of NaCl in ethanol containing
CdCI, impurity, it may be seen from Fig. 7 that with an increase in impurity
concentration cimp,vp regularly decreases, but both v, and v, first slowly
increase and then, at high c,,~ values after attaining a maximum value,
Figure 6. Etch-pit patterns produced after 20 min at 25°C on the ( 100)

face of NaCI by C2H5OH containing: (a) 10.6M, (b) 10.5 M, (c) 10.4 M and
(d) 10.3 M CdCI2 impurity (48). Marker = 0.1 mm.
Figure 7 . Dependenceof etch rates (1) v, , (2) v, and (3) v, for the (1 00)
face of NaCl crystals on CdCI, concentration in C,H,OH (48). The arrows
indicate the relationship between impurity concentration and morphology
of dislocation etch pits.
show a slow decreasing tendency. The region of cim9in which v, and v,

increase with impurity content corresponds to the region in which pits of
different size are formed (Fig. 6); pits of the same size are produced at or
beyond impurity concentrations which give maximum values of v, and v,.
At very high impurity concentrations, the impurity starts depositing on the
dissolving surface (50).
The formation of etch hillocks on the (001) face of white tin with
increasing concentration of CuSO, impurity (51)(52) provides a close
similarity with the etching behavior of NaCI. The results on the morphology
of etched surfaces and the dependence of hillock size on the impurity
content in the etchant are presented in Figs. 8 and 9, respectively. The
dependence of hillock size on cimpshows three distinct regions, as indicated
in Fig. 9. In Region I the rate of hillock development increases, in Region
II it decreases, while in Region Ill the displaced metal precipitates epitaxially
and hillocks are not produced. As in the case of etch pits on the (1 0O)face
of NaCI, at low c,,~ the hillocks have different sizes, but at higher salt
concentrations, around the maximum rate, their sizes are uniform.
It is well known that dislocation etch pits on the surfaces of metals
are produced in solutions of salts of other metals as a result of contact
displacement reactions (53)(54). The size of pits formed by these solutions
depends on the concentration of a salt and the time of etching. However,
prolonged etching often leads to the precipitation of mono- or polycrystalline
displaced metal at relatively more active sites where dissolution is faster
than that at the rest of the surface. Subsequent etching can yield etch
hillocks, as observed in the case of etching white tin in acidic solutions of
CuSO,. Whether etch hillocks or etch pits will be formed at dislocation
sites is determined by the exchange kinetics at the electric double layer
and by the diffusion kinetics.
If A is the metal being etched and B is in the form of the salt, the
condition for the displacement reaction to occur is that the electrode
potential of B in the solution of Bn+ions is higher than that of the crystal A
in the solution of A"', i.e., E(B"/B) z E(Am+/A),or,
Eq. (91) Eo(Bn+/B)+ (RT/nF) In [a(Bn+)]

z Eo(Am+/A)t (RT/mF) In [a(A"+)]
where Eo(Am+/A) and Eo(Bn+/B)are, respectively, the standard electrode

potentials of A and B, and [a(Am+)]and [a(Bn+)]are the activities of the
cations in solution.
Taking displacement reactions as the basis and assuming that the
adsorption potential at the site of a dislocation (D), double kink, i.e., a pair
of neighboring kinks of opposite sign in a ledge (DK), ledge (L), and
surfaceterrace (T) changes in the sequence D z DK > K > L > T, one finds
Figure 8. Effect of addition of CuSO4 to an etching solution composed

from 10 mJ HCI + 10 g NH4NO3 + 50 mi H2O on the formation of etch
hillocks on the (001) face of white tin: (a) no impurity (b) 9.4x1 0.6 M, (c)
9.4x10-5 M, (d) 1.9x1 0-4 M, and (e) 9.4x10-3 M CuSO4 ; etching time: (a -d)
3 min. and (e) 30 sec. From Ref. (51 ).
10-5 10 - 4
CuSO4 concentration ( M I
Figure 9. Dependenceof the size of hillocks produced after 3 min etching

on the (001)face of white tin on the concentration of CuSO, in the etching
solution of Fig. 8 (51).
that the reaction is fastest at the dislocation site. Hence, in the absence
of deposition of products, dissolution is more along the dislocation line
than at the perfect surface. This mechanism also explains simultaneous
changes in v, and vt as well as etch pit morphology with changes in
impurity concentration. The effects of impurity on dislocation etch-pit
morphology are discussed in detail elsewhere (2).
4.0 COMPOSITION OF ETCHING AND POLISHING SOLUTIONS
Etching solutions reported in the literature for revealing dislocations

and surface polishing of crystals have been formulated by trial and error,
and are often widely different from each other. The concepts on the
chemistry of dissolution and the mechanism of etch-pit formation, described
in the preceding sections, are useful in explaining their etching behavior.
For the sake of clarity in discussion on the composition of etching and
polishing solutions, the crystals are classified below according to chemical
bonding and their dissolution behavior towards an etching solution.
4.1 Ionic crystals
Water-soluble Crystals. Typical etching solutions that produce

etch pits on the surface of ionic crystals, fairly-soluble in water, are
composed of a solvent with or without an added impurity. The solvents
commonly used are water, lower aliphatic alcohols such as CH,OH and
C,H,OH, andacids such as HCOOH and CH,COOH. However,the ability
of a solvent to reveal the emergence points of dislocations decreases in
the homologous series of organic alcohols and acids with the addition of
a CH, group (55-57). This behavior is due to the existence of a critical
undersaturation barrier (57) (Sec. 3.1, Eq. 84) whose value decreases
with a decrease in the value of interfacial tension 7 as a result of an
increase in the solubility of the crystal in a solvent. However, it is also
observedthat theetch pitsformedonacrystal surface byafast solventare
somewhat shallow. This is connected with the lowered value of
undersaturation in the immediate vicinity of the dissolving crystal because
of the impeded exchange of subsaturated solution with the fresh solvent
due to the diffusion-controlled nature of dissolution in fast solvents.
Acids, organic or inorganic, always reactwith the crystalline substance.
Therefore, addition of an acid can increase etch rates as a result of
chemical reactions (Sec. 2.4). However, by virtue of their low dielectric
constant, which determinesthe solubility of the substance, their addition
also decreases the rate. In the case of organic acids the latter effect is
predominent,while in strong inorganic acids the former effect is predominent.
Next Page
Previous Page
that the reaction is fastest at the dislocation site. Hence, in the absence
of deposition of products, dissolution is more along the dislocation line
than at the perfect surface. This mechanism also explains simultaneous
changes in v, and vt as well as etch pit morphology with changes in
impurity concentration. The effects of impurity on dislocation etch-pit
morphology are discussed in detail elsewhere (2).
4.0 COMPOSITION OF ETCHING AND POLISHING SOLUTIONS
Etching solutions reported in the literature for revealing dislocations

and surface polishing of crystals have been formulated by trial and error,
and are often widely different from each other. The concepts on the
chemistry of dissolution and the mechanism of etch-pit formation, described
in the preceding sections, are useful in explaining their etching behavior.
For the sake of clarity in discussion on the composition of etching and
polishing solutions, the crystals are classified below according to chemical
bonding and their dissolution behavior towards an etching solution.
4.1 Ionic crystals
Water-soluble Crystals. Typical etching solutions that produce

etch pits on the surface of ionic crystals, fairly-soluble in water, are
composed of a solvent with or without an added impurity. The solvents
commonly used are water, lower aliphatic alcohols such as CH,OH and
C,H,OH, andacids such as HCOOH and CH,COOH. However,the ability
of a solvent to reveal the emergence points of dislocations decreases in
the homologous series of organic alcohols and acids with the addition of
a CH, group (55-57). This behavior is due to the existence of a critical
undersaturation barrier (57) (Sec. 3.1, Eq. 84) whose value decreases
with a decrease in the value of interfacial tension 7 as a result of an
increase in the solubility of the crystal in a solvent. However, it is also
observedthat theetch pitsformedonacrystal surface byafast solventare
somewhat shallow. This is connected with the lowered value of
undersaturation in the immediate vicinity of the dissolving crystal because
of the impeded exchange of subsaturated solution with the fresh solvent
due to the diffusion-controlled nature of dissolution in fast solvents.
Acids, organic or inorganic, always reactwith the crystalline substance.
Therefore, addition of an acid can increase etch rates as a result of
chemical reactions (Sec. 2.4). However, by virtue of their low dielectric
constant, which determinesthe solubility of the substance, their addition
also decreases the rate. In the case of organic acids the latter effect is
predominent,while in strong inorganic acids the former effect is predominent.
Consequently,addition of small amounts of strong acids to slow solvents

is found to have a markedfavorable effect on the revelation of dislocation
etch pits, while in organic acid the etch-pit contrast is poor.
Addition of trace amounts of ionic salts to slow solvents is found to
be equally effective in producing contrasting etch pits as, for example, on
the surfaces of alkali halide crystals (see Fig. 6).
Water-insoluble Crystals. Aqueous solutions of mineral acids,
acidic salts and alkalies, and melts of alkalies and salts are the usual
dislocation etchants for water-insoluble crystals. However, in order to
reveal dislocations in some cases, particularly hard crystals, it is desirable
to manipulate etching conditions, such as etchant composition, temperature
and time of etching. Low concentrations and low temperatures are
generally suitable for revealing dislocations in these crystals by mineral
acids (58-60). At high concentrations and elevated temperatures, H,SO,
and H,PO, often exhibit polishing action.
Figure 10 shows a typical curve of the temperature dependence of
the polishing rate of the (1 11) face of gadolinium gallium garnet single
crystals in H,PO,. Etch pitsare observed in the temperature interval of the
ascending part of the curve but for higher temperatures,at which the rate
decreases, polishing takes place.
It should be pointed out that during the etching of water-insoluble
crystals, an insoluble or sparingly-soluble reaction product may deposit
on the crystal surface.
4.2 Molecular Crystals
The selectiveetching behavior of molecular crystals is similar to that

of ionic crystals. Etch pits can be revealed by solvents in which a crystal
is soluble or solutions which can react with the crystal (61).
4.3 Metallic Crystals
The etching of metals takes place according to one of the following

mechanisms:
(i) Simple reactions. Here the metal is directly converted into
reaction products. For example, etching of Zn in non-oxidizingacids and
alkalies takes place according to the reactions:
Eq. (92a) Zn + 2HCI + ZnCI, + H,

Eq. (92b) Zn + 2NaOH -* Zn(OH), + 2Na’
(ii) Redox reactions. In this case a metal atom is oxidized, while
90 Electroc hemist ry of Semiconductors and Electronics
.-
C
E
\
E
=L
a
>
Solution temperature ("c)
Figure 10. Dependence of the polishing rate of the (1 11) face of gadolinium
gallium garnet in H,PO, on etching temperature (60).
an atom or radical of the reagent is simultaneously reduced (Sec. 2.4).

Oxidation of the metal is caused by an oxidizing reagent such as HNO,,
Eq. (93a)
Eq. (93b)
2Cu
Cu
+ HNO,
+ 2FeCI, +
-
Br,, I,,or byacationofvariablevalency(e.g.,Cr3', Fe3', Cu2+,Ni2+, etc.):
Cu,O t HNO,
CuCI, + 2FeCI,
(iii) Contact displacement reaction. Here etching takes place
according to Eq. 91 :
Eq. (94) Cu t 2AgN0, + Cu(NO,), + 2Ag

Any metal can be etched by mechanisms (ii) and (iii), but mechanism
(i) holds true only for crystals of negative standard potentials. Thus
selective etchants for metals dissolving according to mechanism (i) are
confined to one-component systems. The composition of typical etchants
for metals according to the above mechanisms are summarized in Table
1. The role of complexing agents, given in the last column of the table, is
to dissolve insoluble reaction products forming on the crystal surface by
oxidants.
Finally, it should be mentioned that etch pits on metal surfaces are
produced when the concentration of the complexing agent in a three-
component etching solution, or the diluent in a two-component solution,
is moderately high. At low concentrations of the complexing agent or
diluent, polishing takes place. However, it should be remembered that
etchant pH is also an important parameter during the etching of metals
(62).
4.4 Semiconductors
The mechanism of etching of semiconductors is similar to that of

metals (Sec. 2.4). Consequently, the compositions of typical etchants for
metals and semiconductors are similar. Although melts of KOH and
mixtures of KOH and NaOH, and Br, or 1, in CH,OH are known as etching
solutions, the usually-employed etchants for elemental and compound
semiconductorsare based on an oxidant (HNO,, H,O, or CrO,), a complexing
reagent (HF, HCI, HBr, H,SO, or b S 0 4 )and a diluent ( Y O or CH,COOOH).
The salient features of the etching of semiconductors by redox
etchants are common for all three-component systems (25)(63)(64).
Thus, to describe the etching behavior of semiconductors, consider the
Cr0,-HF-H,O system, whose original components or chemically similar
components in different proportions constitute various well known etchants.
Si and GaAs, two semiconductors of great interest due to device application,
are used for discussion.
Table 1
Composition of typical etchants for metals
Mechanism Equation Reactant or Diluent or Complexing

oxidant medium agent
H,O, alcohols,
glycerine
H,O, alcohols,
glycerine
(ii) (93a) HNO, H20,alcohols, HCI, HF,

organic acids H,SO,
H,O, NH,OH, HF, H,SO,

organic acids
Alcohols
(93b) K,Fe(CN),
Cr3+,Fe3', H,O NH,OH

Cu2', Ni2+
(iii) (94) Ag'
Figure 1 1 shows the rateof dissolution, vp,of the (1 1 1)faceof silicon

asafunction of the molar concentration 7, = [Cr03]/[ HF] at 10 and 5OOC.
The rates first increaseand then, after passing through a maximum rate at
rm*,decrease. In theascendinganddescendingpartsof thesecurves, the
dissolution rate follows the relations:
Eq. (95) vP, = k,7,'/' for 7, < r , * and

,
where v and vp2,respectively, denote the rates in the ascending and

descend%g parts of the curves of the dependence of the overall dissolution
rate vp on y,, k, and k, are the corresponding rate constants, and p is a

constant. The temperature dependence of the rate constants is given by
the Arrhenius equation:
Eq. (96) ki = Ai exp(-E,/kT)
The overall dissolution rate for the whole range of the molar ratio is
expressed by the empirical equation:
Eq. (97) vP = (k,k,~m’/2)/(8.5k,~,P + k,)
which can be written in the form,
The form of Eq. 98 is typical of equations of processes involving

simultaneous surface reactions and volume diffusion. Thus it may be
believedthat at low concentrations of oxidant, the oxidation process is the
rate-limiting step, and that at high concentrations, the supply of the
oxidant to the surface is the rate-determiningstep. Information about the
rate-determining step can also be obtained from the value of activation
energy for dissolution calculated from Arrhenius-type plots of etch-rate
data at different temperatures, and from the effect of stirring on etch rates
(see also Sec.2.2).
Yang (65) investigated the etching behavior of the (100) and (1 11)
faces of Si in Cr0,-HF-H,O system using different ratios of 7m,given in
Table 2. It was observed that at low CrO, concentrations, rounded etch
pits are formed, but with an increase in CrO, concentration, the pits
become better defined. However, for the (100) and (1 11) faces, the
capability of Preferential etching decreases at 7, > 0.04 and 0.204,
respectively. At higher ratios, although the pits correspond to the emergence
points of dislocations, all dislocations are not revealed. These results
showed that preferential etching of dislocations takes place only at a
proper molar ratio of CrO, and HF. The effect of this ratio was more critical
on the (100) face than on the (1 1 1) face.
In order to describe the etching characteristics of GaAs, consider
the iso-etch-ratecontours for the (1 00) face as a function of composition
of Cr0,-HF-H,O system (Fig. 12). Dependingon the etching kinetics, the
ternary diagram can be divided into three main regions. In the region A
corresponding to the ratio 7, > 0.1, the contour lines of iso-etch-rate run
parallel to the H,O-CrO, axis. This means that the lines of iso-etch rate are
also the lines of constant HF concentration, and the rate depends on HF
concentration. In the concentration region B with 7, < 0.05, the etch rate
I
0.05 0.1 0.5 1
volar ratio r m
Figure 1 1. Dependenceof the overall etch rate of the (1 1 1) face of Si on

the molar ratio r,= [CrO,]/[ HF] at two different temperatures (1 5).
HF 0.05 10 croj
M o b r a t i o S,,
Figure 12. Iso-etch-ratecontoursfor the (10O)face of n-type GaAs in the

dark. The numbers on the contour lines denote etch rates in pm/min.
From Ref. (25).
increases with CrO, concentration, whereas in the region C in between

regions A and B with 0.10 > 7, > 0.05, the system follows mixed kinetics.
Table 2
Etching Behavior of (100) and (1 11) Faces of Si

in Cr0,-HF-H,O System [65]
Plane Etchant CrO,(M) y,- Etching Characteristics,
(100) Y1 0.5 0.010 Rounded etch pits.
Y2 1.O 0.020 Poorly defined etch pits.
Y3 1.5 0.031 Well-defined etch pits.
Y5 2.5 0.051 Capability of selective

etching decreases. Reveals
low etch-pit density.
Y6 3.0 0.062 Reveals very low etch-pit

density.
(111) Y1 0.5 0.01 0 Rounded etch pits.
Y17 10.0 0.408 Capability of selective

etching decreases. Reveals
low etch-pit density.
Y18 10.0 0.612 Reveals very low etch-pit

density.
In region A, with 7, > 0.1 and [HF] < 10 M, crystal defects are
revealed as defect-related hillocks. The surface between these hillocks
remainssmoothfor solutionswith-y, > 0.14. For solutionsof region Band
a part of A with -ym < 0.14, surface microroughness and defect-related
hillocks are observed. In the concentration region C with [ HF] < 10 M, in
addition to microroughness and defect-related hillocks, general
macroroughness accompanied by the formation of large hillocks unrelated
to crystal defects results. They are produced as a result of the inhibition
of etching by the adhesion of gas bubbles to the surface (66).
5.0 PHOTOETCHING
As discussed in Sec. 2.3, dissolution of semiconductors is

electrochemical in nature and takes place by the depletion of electrons
and holes at the surface. Therefore, the kinetics of etching may be
influenced by any mechanismthat creates or supplies extra carriers. This
can be caused by applying a potential with or without light (67)(68), by
heating the sample or by illuminatingthe surface during traditional chemical
etching (69)(70). In the case of illumination, the enhanced concentration
of photocarriers not only increases the rate of surface dissolution but also
influences the rate of reaction at the emergence points of dislocations.
Since dislocations are effective sites for the recombination of electrons
and holes, the concentration of carriers is reduced. This leads to the
formation of etch hillocks as a result of dissolution at dislocations rather
than at the perfect surface. Kuhn-Kuhnenfeld (69) showed that a photovoltage
is developed at the surfaces of GaAs crystals etched in 3H2S0, - 1H202-
1H,O under illumination, and attributed it to the separation of light-
induced electron-hole pairs in the built-inelectric field of the semiconductor
surface.
The general features of the effect of illumination on etching kinetics
may be described by referring to Fig. 12 of the etching of GaAs in CrO, -
HF solutions, studied by Ven, et. al. (25). For Cr0,:HF molar ratios greater
than 0.1 (region A in Fig. 12), the kinetics are independent of the CrO,
concentration but slowly depend on the HF concentration for both p- and
n-type GaAs (Fig. 13). For the etching of n-type GaAs in the dark, the etch
rate increases slowlywith the HF concentration at low concentrations, but
then it tends to increase linearly at high HF concentrations. Under
illumination, the rate increases linearly with HF concentration. The
dependence of etch rate of p-type samples on HF concentration is similar
to that of n-type crystals in the dark, but the etch rate remains unaffected
by illumination.
Theeffectof lightintensityon theetchrateofthe(1OO)faceof n-type
0 5 10
HF concentration ( M I
0 5 10
HF concentration [ M I
Figure 13. Etch rate vp of (a) p- and (b) n-type (1 00) GaAs as a function
of HF concentration for etchants from the HF-controlled region A of Fig.
12; solid and open circles are, respectively,for etching in dark and under
illumination (1 8).
GaAs in a solution containing 1.55 M HF and 1.2 M CrO, is shown in Fig.

14. The curve reveals that, after an initial steep linear rise, the rate tends
to attain a limiting value at high light intensities. The initial linear slope
suggests that one GaAs molecule is dissolved when three photons are
absorbed. Thus two holes are transferred for every photon absorbed
because GaAs dissolution is a six-hole process (see Sec. 2.4).
From analyses of current-potential plotsfor n- and p-type materials,
and the dependence of limiting cathodic current and cathodic peak
currents on HF concentration, it was shown (18) that a passivation layer
containing both Cr"' and Cr"' species is formed on the dissolving surface.
In the region of high CrO, concentration (region A), the surface coverage
by the film is high, but with the aid of HF the layer is removed by
intermediatesfrom the GaAs oxidation process. In the case of dissolution
of n-type GaAs under illumination, it was proposed (17)(19)(25) that
recombination of holes and electrons takes place via surface intermediates
of the GaAs oxidation process (see Eqs 54-56).
6.0 ELECTROLYTIC ETCHING AND POLISHING
By virtue of the chemical nature of the processes involved during

dissolution, electrolytic etching (also called electroetching or
electrochemicaletching) is used for metals and semiconductors. Electrolytic
etching and polishing take place only when dissolution kinetics are
controlled by volume diffusion and the general dissolution characteristics
usually observed on the surfaces of metals and semiconductors may be
described with reference to the current-voltage characteristics of Fig. 1 .
The current-voltagecurve of Fig. 1 can be divided into three regions.
In region AB the current density increases with the applied voltage, in
region BC the current density is independentof the voltage, and in region
CD the current density increases again with the voltage. It is found that in
region AB crystal defects are selectively revealed, in region BC polishing
takes place, while in region CD localized pitting, often unrelated to defects,
occurs. Thus electropolishing takes place in the plateau region of the
current-voltage curve.
The above features are illustrated in Fig. 15, which shows the
experimental current-voltage chatacteristics of an electrolytic cell containing
an electrolyte prepared by adding one part of a mixture consisting of equal
parts by volume of a saturated KOH solution and saturated Na,SO,
solution, to 5 parts of glycerol (71), used for the etching of the (100) face
of Pb,,Sn,,Se crystal. This solution gives a cell conductivity of 14 mA
ern-', but by adding small amounts of KOH-Na,SO, mixture, the cell
conductivity could be increased. High quality polished surfaces were
50 100 150 200

I [mqLrnZ 1
Figure 14. Etch rate vpof n-type (100) GaAs as a function of laser-light
intensity I for a solution containing 1.55 M HF and 1.2 M CrO, (25).
110 -
120 -
- 100 .
4
--E
c
Figure 15. Dependence of current on voltage for electrolytic cells of

different cell conductivities. The conductivities were measured at a cell
voltage of 6 volts. After Ref. (71).
obtained only for 4 and 6 mA cm-' cells in which there is little change in
-
current with the cell voltage (Le., in the 6 10 volt region). Use of 7.5 mA
cm-2cell gave non-uniform polishingwith slight pittingwhile 9 mA cm'* cell
yielded a badly pitted surface with no polishing.
Electropolishing is the elimination of surfaceroughness of height >
1 p m (macrosmoothing) and of height < 1 p m (microsmoothing) (72).
Macrosmoothingresults from the concentration of current lines on peaks
of a surface profile thus leading to a locally higher dissolution rate. The
rate of macrosmoothing, therefore, is equal to the differencein dissolution
rate between peaks and recesses of a rough surface. It is determined by
the current distribution present on the surface profile and, therefore,
depends on geometrical and electrochemical parameters. Microsmoothing,
on the other hand, resultsfrom the suppressionof the influence of defects
and crystallographic orientation on the dissolution process. Consequently,
it requires the study of surface kinetics and passivation behavior. In most
cases we deal with microsmoothing.
There is evidence (8)(21-24)(71-75)that during electrodissolution,
a passivationfilm is formed on the anode. The film is gradually eliminated
by diffusion when the current is switched off. The etching and polishing
behavior of an anode are perhaps associated with the chemical stability
of this passivefilm whose composition is usually speculative (21-23)(72)(74).
Avariety of electrolytic cells have been in usefor electrolyticetching
and polishing (73)(75-77). They all have a sample as the anode, a
platinum foil as the cathode, a vessel containing an electrolyte, temperature
controller, and a power supply. Additionally, stirring may be provided by
a magnetic stirrer or by bubbling an inert gas through the electrolyte.
Measurements of potential of an anode are made against a saturated
calomel electrode (SCE).
Figure 16 illustrates the current-voltagecurves for < 100 > oriented
n-type InP samples with different carrier concentrations. These curves
show that with an increase in carrier concentration the dissolution rate
increases, but the higher the carrier concentration the smaller is the
increase in dissolution rate. This behavior is due to a decreasing contribution
of holes to the anodic dissolution rate (77). It was observed that the
sample with the highest carrier concentration is simply polished both in
the dark and with illumination, but that structural defects are revealed as
etch hillocks in samples of lower carrier concentrations. The effect of
illumination in revealingstructural defects in n-type crystals is profound in
mild etching systems (75)(77-79). In mild etching electrolytes,
photoelectrolytic etching is also dependent on crystal orientation in n-
type materials (80). However, it is not known whether this anisotropy is
due to a difference inthe ratesof recombination at the surfaces due to the
surface structure or in the kinetics of dissolution of reaction products.
Obviously, in mild etching systems the dissolution kinetics are controlled

by surface reactions.
7.0 GAS-PHASE CHEMICAL ETCHING
Prior to epitaxy or chemicalvapour deposition, surfaces of crystals,

especially semiconductors, are frequently polished with chemically reactive
gaseous mixtures. In contrast with traditional wet chemical etching, the
procedure has the advantage of cleaning the surface by removing the
metallic impurities. The degree of polish depends on the ratio of components
in the reactive mixture, its total pressure, partial pressure of the reactant
and the temperature at which etching is carried out (81-86). Generally,
etching is carried out at atmospheric pressure (81-84), but low-pressure
etching has also been studied (86)(87).
Depending on the etching conditions, gas-phase etching of crystal
wafers reveals etch pits, bunch formation and surface polishing (81-85).
Figure 17 shows the surface morphologyof the (1 11)faceof silicon single
crystals etched in gaseous HCI with Ar or H, as a carrier gas as afunction
of temperature. The continuous and the dashed curves represent the
transition lines between smooth and bunched surface structures when H,
and Ar are used, respectively, as the carrier gas. When argon is used as
a carrier gas, the transition line is shifted to lower input concentrations of
HCI and higher temperatures. Also shown is the transition line from a
smooth to a pitted surface for HCI-H, system. In the region of high HCI
concentrations and low temperatures both etch pits and bunches are
produced. The shape of the observed bunches is determined by the
misorientation of the etched surface from the real (1 1 1) plane and by the
directions of the etching vectors in the (1 11) surface. It is also observed
(81)(83) that beyond a critical HCI concentration formation of block
patterns (facet formation) takes place on the (1 11) surfaces with very
small (<0.5O)misorientations. The block patterns are composed of
shallow triangular etch pits.
The etching behaviour of Si in HCI-H, mixtures at low-pressureCVD
conditions shows (85) that, at constant HCI concentration and total
pressure, the polishing effect improveswith an increase in temperature.
With an increase in the concentrations of HCI and H, or with an increase
in the total pressure, a roughening of the surface takes place. The effect
of temperature observed in this case is similar to that described above in
Fig. 17.
As in the case of wet chemical etching, the temperature dependence
of etch rates of crystals in gas-phase etching shows the reaction-rate and
diffusion controlled nature of the process. Figure 18 illustrates the
10 I I I
10-2 10’ 1
Current d e n s i t y h A / c m * 1
Figure 16. Curves of dissolution potential against current density for n-

type (100) InP of different carrier concentrations (77); carrier concentration
n: (1,4) 1x101ecm-3, ( 2 , 5 ) 4 ~ 1 0 ’ ~ c m(3,6)2x1
-~, 01Qcmm3;
(1,2,3)inthedark,
(4,5,6) under illumination.
1400 1450 1500 1550 1600

Temperature ( K ]
Figure 17. Surface morphology of the (1 1 1)face of Si etched in gaseous

HCI. Transition lines from smooth to bunched structures are for the H2/
HCI and Ar/HCI systems; the transition line from smooth to pitted structures
is for the HJHCI system. From Ref. (83).
Figure 18. Dependenceof etch rate vponthe reciprocal temperature 1/

Tat a pressure of 1.O torr and molar fraction of HCI of 0.1 (85).
Arrhenius-type plot of the dependence of etch rate of the (1 1 1)face of Si

on temperature,at apressure of 1.O torr and HCI molar fraction of 0.1. The
dependence clearly shows two regions. At temperatures below 1OOO°C,
where the temperature dependence is very strong, corresponding to an
activation energy of 105 kcal/mol, etching is kinetically controlled. At
temperaturesabove 1OOO°C corresponding to an activation energy of 16
kcal/mol, etching is controlled by diffusion. Etching at atmospheric
pressure also exhibits a similar behaviour (87)(88), but the values of the
activation energy corresponding to kinetically- and diff usion-controlled
regimes are 42 and 0 - 4 kcal/mol, respectively.
The change in the morphology of surfaces etched in the gaseous
phase can be correlatedwith the theoretical predictions (2). For example,
if we take the reactant concentration as a measure of undersaturation,the
transition from a region of smooth morphology to pitted structures is
predicted by the thermodynamic theories of dislocation etch-pit formation,
while the bunched morphology may be interpreted by the kinematic
theory of step motion. The kinetic data, on the other hand, have been
explained (82-84)(86) in terms of surface reactions, commonly used in
heterogeneous catalytic reactions, involving adsorption of molecules of
reactantand carrier gas onto the surface and subsequently,desorption of
different thermodynamically stable species generated at the surface. In
the case of etching Si in HCI-H, mixtures, the the following reactions have
been proposed (85)(86):
Eq. (99) (1/2)H2(g) 4. * + H*
Eq. (100) HCl(g) + * + HCI*
Eq. (101) HCI* t Si(,) + SiHxCl,-x*
where * denotes an adsorption site on the surface. Eqs. (99) and (100)
represent the adsorption reactions, Eq. (1 01) chemical reaction on the
surface, while Eq. (102) represents the desorption reaction. Below
1OOO°C, the rate-limiting step is the surface reaction given by Eq. (101),
while above this temperature the control of the surface reaction is diffusional
and may be determined by any of the steps of Eqs. (99),(100) and (102).
Dependingon experimentalconditions, which ultimately determine
the concentration of adsorbed hydrogen and chlorine on the surface, the
number of HCI molecules reacting with a silicon atom may be different,
forming different species containing S K I bonds (86). The stable species
forming at high and lowtemperatures are SiCI, and SiCI,, respectively, but
their stability also depends on pressure and CI/H ratio (Le,, HCI
concentration).
8.0 MORPHOLOGY OF CHEMICAL ETCH PITS
Etch pits always possess an external outline, which may be

crystallographic, rounded or circular. Etch pits with a crystallographic
outline are called preferential while those without a crystallographic
contour non-preferential. The process leading to the formation of
crystallographically-oriented(or preferential) etch pits is termed preferential
or anisotropic etching, while the one corresponding to the formation of
circular (or non-preferential)etch pits is called non-preferentialor isotropic
etching.
The external outline of etch pits depends on the conditions of
etching (i.e., composition of etchant and etching temperature). In the
case of etching water-soluble crystals of lower symmetry in simple solvents,
even a change of solvent leadstoa marked change in the geometry of pits
(56). On the surfaces of crystals of higher symmetry, on the other hand,
the outline of etch pits undergoes a change when the composition of the
etchant or temperature of etching is changed. This feature is illustrated in
Fig. 19 for etch pits produced on the (1 00) and (1 10) faces of MgO by
H,SO,. The figure shows that by changing the etchant concentration both
anisotropic and isotropic etching may be achieved.
It has been suggested (89) that isotropic etch pits are produced
when etching is controlled by diffusion, while anisotropic etch pits are
produced when the process is kinetically controlled. An analysis of the
experimental observations reveals that it is not always so. For example,
the process of etching of a crystal in a solvent in which it is highly soluble,
is expected to be diffusion-controlled, yet, as observed in the case of
etching of potassium dihydrogen phosphate (KDP) (57)and potassium
bichromate (KBC) (56) in water, the pits are anisotropic. Similarly, in the
case of etching of the (100) face of MgO, the kinetics at low acid
concentrations are controlled by chemical reactions but the pits produced
are isotropic (90).
The geometry of etch pits is determined by the relative rates of
removal of atoms along different directions. Factors, such as adsorption
of added impurities or reaction products, temperature of etching and
stirring, affect the absolute values of these rates and thus lead to a change
in the geometry of etch pits. In the absence of adsorption and diffusion
effects, a situation difficult but possible to be realized, the shape of etch
pits may be determined from a consideration of the binding energy
between nearest neighbors (2) or from Periodic Bond Chain vectors
Figure 19. Schematic representation of the geometry of etch pits on the

(1 00) and (1 10) faces of MgO at various H,SO, concentrations; dep:
dislocation etch pits. After Refs. (7)(58).
(PBCs) (2)(3)(91). The deviation in the observed morphology of etch pits

from the one predicted by the PBC theory is a consequence of adsorption
and diffusion processes.
Since the geometryof etch pits is determined bythe arrangement of
ions, atoms or molecules composing a crystal surface, the shape of etch
pits formed on crystallographically different or on opposite faces of a
crystal is different. This difference forms a basis for the distinction of
various faces of a crystal and, in many cases, serves as a quick methodfor
the identification of the faces or the directions on the face of a crystal.
Figure 20 presents dislocation etch-pit patterns on the (001) and (OOT)
faces of a KBC crystal. The difference in the shape of etch pits as well as
in the distribution of dissolution layers is evident. Similarly, the non-
equivalence of the [ 1101 and [ i l 01 directions in the (001) plane of zinc-
blende-type crystals leads to the formation of elongated or elliptical (93-
95) and rectangular etch pits (93)(96)(97). However, an appropriate
choice of the etching solution is a crucial factor in these observations, as
illustrated in Fig. 21 for the (001) face of InP crystals.
The side walls of dislocation pits with a crystallographic outline
usually correspond to high-index planes (47), but low-index planes of the
equilibium form are also observed (see Refs (47)(98-100) and literature
cited therein). The latter type of etch pits, customarily called crystallographic
etch pits, are observed on the surfaces of metals and semiconductors.
The sites for the origin of the crystallographic etch pits are either the
emergence points of dislocations or random points corresponding to
pinholes developed in a surface film formed during dissolution, permiting
localized etching action. These surface films are favorably formed by
redox etchants during the dissolution of metals and semiconductors.
Figure 22 presents the successive stages of the development of etch pits
onthe(11l)faceofSiinasolutionof3HNO3 + 1HF + 1.6CH3COOH. The
figure shows that most of the pits have turned flat-bottomed on successive
etching, indicating their non-dislocation origin.
Several factors, such as etchant composition, etching temperature
and surface treatment (e.g., prepolishing, surface damage, etc.) contribute
to the formation of etch pits by the pinhole dissolution mechanism and
different tests may be applied to demonstrate the validity of this mechanism.
If etch pits are produced by this mechanism, successive etching leads to
the development of shallower and more rounded etch pits which were
initiallysmall and relativelyflat (Fig. 22). The growth of etch pits with time
also shows a parabolic dependence (99). Furthermore, prepolishing may
greatly enhance the density of etch pits.
Finally, it should be mentioned that under certain experimental
conditions etch pits formed by this pinhole mechanism may be non-
preferential. Such shallow saucer-shaped pits, often called S-pits,of non-
Figure 20- Dislocation etch pits produced on the (001) and (001) faces
of potassium bichromate crystals by etching with water at room temperature
(92).
Figure 21. Etch pits of different shapes produced on the (001) plane of
InP; etchants: (a) 10HBr + 10CH3COOH + 1(1 N)K2Cr2O7 ' and (b) 2HBr
+ 2H3P04 + 0.07( 1 N)K2Cr 207. From Ref. (96).
Figure 22. Successive stages of the development of non-dislocation

etch pits on the ( 111) face of Si in 3HNO3 + 1 HF + 1.6CH3COOH etchant
(99); etching time: (a) 30 sec., (b) 90 sec., (c) 330 sec.
dislocation origin are frequently observed during the etching of

semiconductors.
9.0 CORRESPONDENCE BETWEEN ETCH FIGURES AND

DISLOCATIONS
All types of linear (dislocations), planar (stacking faults, twin

boundaries,grain boundaries,etc.) and volume defects (clusters of point
defects and impurities) emerging on the surface of a crystal lead to the
formation of etch figures (Sec. 3). Volume defectsare localized at random
positions in the crystal interior, planar defectsextend to large areas, while
linear defects do not terminate within a crystal. These geometrical
properties of defects may be used to distinguish between them by studying
etch patterns. Volume and linear defects produce etch figures at isolated
positions; stacking faults, twin boundaries and impurity striations produce
grooves or ridges, while grain boundariesyield rows of etchfigures. Etch
figures resulting from linear and volume defects may be recognized by
examiningthe etch patterns on a surface after successiveetching, or after
alternate etching and polishing, or by comparing etch patterns on matched
cleavage faces. Etch figures due to volume defects disappear continuously
with successive etching or alternate etching and polishing. If an etch
pattern persists after prolonged etching or alternate etching and polishing,
or if the etch patterns produced on matched halves are essentially mirror
images of each other, the etch figures correspond to the emergence
points of dislocations.
If a crystal contains tilt boundaries, the following tests can be
performedto identify etch figure sat the emergencepoints of dislocations.
In the case of an isolated low-angle tilt boundary, the distance between
two etch figures in the boundary is equal to that between neighboring
dislocations, D, as estimated by X-rays from the misorientation angle B in
accordance with the relation 8 = b/D of the dislocation model of a low-
angle tilt boundary( 101). Asimpler testwithoutthe measurement of angle
B is to look for three or four intersecting boundaries (Fig. 23). For
symmetrical low-angle tilt boundariesthere exists a relation between the
linear dislocation densities p i (i.e., number of etch figures per unit length)
between different branches (1 03), namely,
If a crystal can be deformed plastically, the appearance of slip

bands and indentation rosettes (Fig. 24) on a surface implies that the etch
figures locate the emergence points of dislocations. Comparison of etch
Figure 23. Low-angle tilt boundaries revealed by electrolytic etching on

the (0001) face of Sb -2.5% Bi alloy (102).
Figure 24. Indentation rosette revealed on the (100) face of MgO by

dilute H2SO~.
figures with dislocation patterns revealed by other techniques (e.g.,

decoration, X-ray topography, recombination techniques) also provides
evidence of the one-to-one correspondence between etch figures and
dislocations (Fig. 25).
10.0 ETCHING PROFILES
The fabrication of electron devices involves a sequence of steps,and

the production of etching profiles in the form of grooves and mesas in the
surface of a semiconductor by chemical etching is an important step. The
shape of an etching profile is determined by the composition of the
etching solution, the temperature of etching, the surface orientation,
degree of undercutting beneath the mask and the chemical aggressiveness
of the etchant towards the mask.
The surfaces for the formation of etching profiles for semiconductors
are (OOl), and single-crystal wafers of Si (105),GaAs (24)(64)(94)(106-
110) and InP (93)(94)(96)(111)( 112) and double-heterostructure (DH)
wafers of InAIAs/lnP (1 13), InGaAsP/lnP (96)( 114-117), and AIGaAs/
GaAs (1 18)(119) have been used. However, depending on the type of
wafer, and the chemical composition of the etchant used, the etching
profile produced on the surface is different.
Etching profiles on the surfaces of elemental semiconductors are
grooves, while those on the surfaces of the layers of compound
semiconductors are both mesa-shaped and reverse mesa-shaped,
depending on the direction in the (001) plane. This feature is shown in Fig.
26 for the profiles produced on (001) GaAs etched in different solutions.
The profiles were obtained by cleaving thewafers in orthogonal directions
along the (1 10) and (710) planes.
Figure 26 reveals that the profile walls are composed of individual
planes. In Fig. 26a, the reverse mesa-shaped and the mesa-shaped
planes form angles of 114O and 55O with respect to the (001) plane and
correspond to (221) (or { 1 v ) G a ) , and (71 1)Ga planes, respectively
(Fig. 27). In the case of Fig. 26b,c,d, the profiles on the (110) cleavage
planes show vertically etched planes with a small tilt from a direction
normaltothe surface(001)plane. Thesevertical planescorrespondtothe
(71 0) plane. Some etching solutions also produce profiles whose walls
on the (1 10) cleavage plane make different angles with respect to the
(001) plane. These profile walls correspond to {~2)111,( ~ 1 ) 1 1 1and
{ 071 } planes. Asummary of the etching profiles of semiconductor wafers
is given in Ref. (2).
The most commonly used etching solutions for producing etching
profiles on the (001) face of GaAs and InP are mixtures of various
Figure 25. One-to-one correspondence between (a) dislocation etch-pit

distribution, (b) cathodoluminescence images and (c) electron-beam-
induced conductivity images of the same region of a GaP layer (104).
Figure 26. Etch profiles obtained on (001) GaAs etched in different

etchants: (a) 1 H2SO4 + 1 H2O2 + 1H2O (25°C.30 sec); (b) 3(1 M)K2Cr2O7
+ 1H2SO4 + 1HCI (60OC. 1 min); (c) 3(1M)K2Cr2O7 + 1H2SO4 + 2HCI
(60°C.1 min); and (d) 3(1M)K2Cr2O7 + 1H2SO4 + 12HCI (25°C,2 min).
From Ref. (94).
I~ I
(110)
PlANE (TlO)
PLANE
Figure 27. Schematic representation of etch profiles of ( 110) and (110)
planes of a (001) InP crystal; the open and solid points represent III and V
atoms, respectively (93).
compositions of H,SO,-H,O,-H,O, HCI-H,O,-CH,COOH( H,PO,), HN0,-

CH,COOH, HF-HN0,-H,O(CH,COOH), HCI-HN0,-H,O, HBr-HN0,-H,O,
H,PO,-H,O,-H,O(CH,COOH), HBr (or HCI or H,SO,)-H,PO, (or CH,COOH)-
K,Cr,O, , Br,-CH,OH, and H,O,-NaOH (or KOH or NH,OH)-H,O. These
etching solutions are the same as those used for the polishing and
selective etching of these crystals. As in the case of selective etching,
these solutions form etching profiles with crystallographically oriented
walls or round profiles, and are called preferential (anisotropic) and non-
preferential (isotropic). In order to mask the surface of the semiconductor
crystals, either a photoresistor a layer of SiO, or Si,N, are used. Commonly
used photoresists are HNR-999, HPR-204 and AZ-1350.
A serious problem with most of the abovementioned etchants is a
tendency towards an enhanced etch rate of the crystalline wafer near the
edges of the mask (24)( 106)(1 1 1)( 120), probably due to the migration of
etchant molecules along the mass. In the case of a photoresist mask, this
effect results in the curling away of the masking layer. Both of these
phenomena lead to the lateral etching of the wafer beneath the edges of
the mask. This lateral etching is referred to as undercutting. The degree
of undercutting can be reduced by using a masking material that firmly
adheres to the substrate wafer. Metal coatings are often found to be
suitable as masking material (105). However, acid etchants corrode
metallic masks, while alkaline etchants tend to dissolve the oxide and
nitride masks (106). In some etching systems the long etching times
involved restrict their applicability (105).
For the reproducible production of etching profiles of good quality
it is of utmost importance that the etching solution does not erode the
mask material and the degree of undercutting is controlled. The first
problem can be overcome by choosing a proper etching solution and
masking material, while the degree of undercutting can be regulated by
improving the adhesion between the mask and the wafer. It has been
reported (1 11) that photoresist coated wafers stored for long durations
lead to a significant deterioration in the formation of etching profiles.
Like the rate of etching, v, of the wafer in a direction perpendicular
to the etching plane, the ratio of the rate of undercutting, v,, in the lateral
direction and the etch factor vt /v, depend on the composition of etching
solution as well as on the temperature of etching. The dependence of
these rates for the etching of (001) wafer of InP in a mixture of 3HCI t
1H,PO, on the etching temperature is presented in Fig. 28 in the form of
Arrhenius plots. The phenomenon of undercutting beneaththe edges of
the mask is shown in the insert of this figure. Obviously,the two rates are
activation controlled, but with an increase in temperature, the increase in
v, is relatively more than in v, because of a higher activation energy.
Consequently, the etch factor, which is equal to v,/v,, increases with an
increase in etching temperature, as shown in Fig. 29. The figure also

shows that the value of the etch factor decreases with an increase in HCI
concentration in the mixture. This fact means that with an increasein HCI
concentration etching becomes preferential.
When the dissolution rates for a crystalline solid are very high for all
crystallographic planes, diffusion in the solution is rate determining and
the etching is isotropic. Therefore, roundedetching profiles are predicted
at mask edges and the etch depth near the edge is increaseddue to a more
efficient supply of the etchant. However, it has been found that, while the
etch rates of various planes are diff usion-controlled, etching at mask
edges is anisotropic (24). Two examples of isotropic etching in diffusion-
controlled systems are illustrated by the etching profiles of (001) GaAs
etched in aqueous K,Fe(CN), solutions, shown in Fig. 30. Isotropic
etching profiles at mask edges may also be seen in Fig. 26.
Etching kinetics of semiconductors may be diffusion-controlled in
two ways (24). In the first case, the reduction reaction is diffusion limited
and controls the etching kinetics. Under these conditions a well defined
crystallographic facet is obtained. This behavior is observed at low pH. In
the second case of high pH, the rate of anodic dissolution of the
semiconductor wafer depends on the mass transport of OH' ions to the
electrode. Electroless etching based on this limitation shows rounded
profiles typical of diff usion-controlled dissolution.
Etching solutions employed to produce patterns in DH layers are
also similar to those used for etching profiles in elementaland compound
semiconductor wafers. However, appropriate choice of the composition
of an etching solution that will yield the desired, smooth profile in a DH
layer has to be made by trial and error. This is so because the etch rates
of two neighboring layers are different due toa difference in their chemical
composition. Etching profiles for InGasP/lnP and InAIAs/lnP DH layers
produced by various etching solutions are given by Adachi et al., (96)( 1 14-
117) and Stano (1 13), respectively, while a summary for different DH
layers may be found in Ref. (2).
11 .O ACKNOWLEDGEMENT
This work has been partly financed under Research Project CPBP
01.20.3.
2.8 2.9 3.0 3.1 3.2

10YT , li('l\
Figure 28. Arrhenius plots of v, and v, on temperature (1 1 1).
1.8
1.6
1.4
$1.2
.,
a' 1,o
+.
u
*
J?
. 0.8
?
Y
0.6
2O0C
I x
0.6
1 2 3 i 5
2'v
Figure 29. Dependence of etch factor VJV, on etchant composition at

2OoCand on etching temperaturefor 3HCI + 1H,PO, etchant. Based on
the data of Ref. (1 11).
Figure 30. Etching profiles produced on n-type (001) GaAs by (a) 0.05
M and (b,c) 0.5 M K3Fe(CN)6 solutions at pH (a,b) 13 and (c) 14 (16).
I
12.0 REFERENCES
1. Honess, A. P., The Nature, Origin a n d lnlerpretation of the

Etch Figures o n Crystals, Wiley, New York (1927)
2. Sangwal, K., Etching of Crystals: Theory, Experiment and

Application, North-Holland, Amsterdam (1987)
3. Heimann, R . B., Aufldsung von Krislallen - Theorie und

Technische Anwendung, Springer, New York (1975)
4. Moore, W. J., Physical Chemistry, 5th edition, Longman, London

(1972)
5. Vermilyea, D. A., J . Electrochem. SOC.113:1067 (1966)
6. Sangwal, K., J . Mater. Sci. 17:3598 (1982)
7. Borecka, B. and Sangwal, K., J.Mater. Sci. Lett. 1:364 (1982)
8. Lacombe, P., in: The Surface Chemistry of Metals a n d

Semiconductors, (H. C. Gatos, ed.), p. 244, Wiley, New York
(1960)
9. Gatos, H. C. and Lavine, M. C., in: Progress in Semiconductors,

Vol. 9 , (A. F. Gibson and R. E. Burgess, eds.), p. 1, Heywood,
London (1965)
10. Despid, A. R., Electrochemistry, in: Physical Chemistry Series

7, Vol. 6, (J. O’M. Bockris, ed.), p. 115, Butterworths, London
(1973)
11. Turner, D. R., in: The Surface Chemistry of Metals a n d

Semiconductors, (H. C. Gatos, ed.), p. 285, Wiley, New York
(1960)
12. Klein, D. L., and D’Stefan, D. J., J.Electrochem. SOC. 109:37

(1962)
13. Turner, D. R., J.Electrochem. SOC.107:810 (1960)

14. Schwartz, B. and Robbins, H., J.Electrochem. SOC. 108:365
(1961)
15. Heimann, R. B., J.Mater. Sci. 19:1314 (1984)
16. Notten, P. H. L. and Kelly, J. J., J.Electrochem. SOC. 134:444

(1987)
17. Kelly, J. J. and Notten, P. H. L., J.Electrochem. SOC. 130:2452

(1983)
18. van de Ven, J., van den Meerakker, J. E. A. M. and Kelly, J. J., J.
Electrochem. SOC.132:3020 (1985)
19. Kelly, J. J., van de Ven, J. and van den Meerakker, J. E. A. M., J .
20. Palik, E. D., Glembocki, 0. J. and Heard Jr., I., J.Electrochem.

SOC.34:404 (1987)
21 Glembocki, 0. J., Stahlbush, R. E. and Tomkiewicz, M., J.

22. Palik, E. D., Bermudez,V. M. and Glembocki, 0. J., J. €Electrochem.

S O C .132:135(1985)
23. Palik, E. D., Bermudez,V. M. and Glembocki, 0. J., J. Electrochem.

S O C . .132:871 (1985)
24. Raley, N. F., Sugiyama, Y. and van Duzer, T., J. Electrochem.

SOC.131:161 (1984)
25. van de Ven, J., Weyher, J. L., van den Meerakker, J. E. A. M. and
Kelly, J. J., J . ,Electrochem. SOC.133:799 (1986)
26. Beckman,K. H.,andMemming, R . J. Electrochem. SOC.116:368

(1969)
27. Decker, F., Pettinger, B. and Gerischer, H. J.Electrochem. SOC.

130:1335 (1983)
28. Decker, F., Abramovich, M. and Motisuke, P., J. Electrochem.

SOC.131:1173(1984)
29. Ohara, M. and Reid, R. C., M o d e l i n g Crystal Growth Rates
from Solutions, Prentice-Hall, NJ, (1973)
30. Burton, W.K.,Cabrera, N.andFrank,F.C., Phil. Trans. Roy. SOC.

L o n d o n A243:299 (1951)
31. Cabrera, N. and Levine, M. M., Phil. Mag. 1:450 (1956)
32. Keller, K. W., in: Crystal Growth and Characterization, (R.

Ueda and J. B. Mullin, eds.), p. 361, North-Holland, Amsterdam
(1975)
33. Bauser, E., and Strunk, H., J. Cryst. Growth 51:362 (1981)
34. Bauser, E., and Strunk, H., Thin Solid Films 93:185 (1982)
35. Frank, F. C., J. Cryst. Growth 51 :367 (1981)
36. Koziejowska, A. and Sangwal, K., J . Mater. S c i . 23:2989 (1988)
37. Gilman, J. J., Johnston, W. G. and Sears, G.W., J. Appl. Phys.

29:747 (1958)
38. Ives, M. B. and Ramachandaran, T. R., J . Appl. Phys. 38:2121

(1967)
39. Amelinckx, S., Bontick, W. and Dekeyser, W., Phil. Mag. 2:1264
(1957)
40. Lang, A. R., J. Appl. Phys. 28:497 (1957)
41. Cabrera, N., J . Chin?. Phys. 53:675 (1956)
42. Schaarwdchter, W., Phys. Stat. Sol. 12:375 (1965)
43. van der Hoek, B., van der Eerden, J. P. and Bennema, P., J. Cryst.
Growth 56:621 (1982)
44. Koziejowska, A. and Sangwal, K., Cryst. Res. Techno/. 20:455

(1985)
45 I Nanev, Chr. N. and Dicheva, K., f l e c t r o c h i m . Acta 28:933 (1983);

Acta Phys. Hung. 57:271 (1985)
46. Schaarwdchter, W., Phys. Stat. Sol. 12:865 (1965)

47. Ives, M. B. and McAusland, D. D., Surf. Sci. 12:189 (1968)
48. Sangwal, K. and Zaniewska, G.,J . Mater. S c i . 19:1131 (1984)
49. Ives, M. B., J. Phys. Chem. Solids 24:275 (1963)
50. Sangwal, K. and Urusovskaya, A. A., J . Cryst. Growth 41:216

(1977)
51. Honda, K. and Hirokawa, T., Japan J . Appl. P hys. 11:1763 (1972)
52. Hirokawa, T., Honda, K. and Shibuya, T., J . Cryst. Growth 24/
25:484 (1974)
53. Pandya, N. S. and Bhatt, V. P., J . S c i . Ind. Res. B19:363 (1960)
54. Popkova, E.G. and Predvoditelev, A. A,, Kristallografiya 15:91

(1970)
55. Baranova, G.K. and Nadgornyi, E. M., Krisfallografiya 18:1028

(1973)
56. Sangwal, K. and Szurgot, M., Cryst. Res. Techno/. 17:49 (1982)
57. Sangwal, K., Szurgot, M., Karniewicz, J. and Kolasinski, W., J .

Cryst. Growth 58:261 (1982)
58. Sangwal, K.and Sutaria, J. N., J . Mater. Sci. 11:2271 (1976)
59. Sangwal, K., Patel, T. C. and Kotak, M. D., J . Mater. Sci. 14:1869
(1979)
60. Takagi, K., Fukazawa,T. and Ishii, M., J . Cryst. G r o w t h 48:19

(1980)
61 McArdle, 6 . J. and Sherwood, J. N., in: Advanced Crystal Growth,

(P. M. Dryburgh, B. Cockayne and K. G.Barraclaugh, eds.), Chap.
7 Prentice-Hall, London (1987)
62. Predvoditelev,A. A. and Popkova, E. G., in: Rost Kristallov (Growth

of Crystals), (N. N. Sheftal, ed.), p. 259, Nauka, Moscow (1965)
63. Scwartz, 6 . and Robbins, H., J. €/ecfrochern. SOC.123:1903

(1976)
64. Mori, Y. and Watanabe, N., J. Electrochem. SOC. 125:1510

(1978)
65. Yang, K. H., J. Electrochem. SOC.131:1140 (1984)
66. Weyher, J. L. and van Enckevort, W. J. P., J . Cryst. Growth

63:292 (1983)
67. Greene, L. I., J . Appl. Phys. 48:3739 (1977)
68. Faktor, M. M., Ambridge, T., Elliot, C. R. and Regnault, J. C., in:
Current Topics i n Materials Science, V o l . 6, (E. Kaldis, ed.),
p. 1, North-Holland, Amsterdam (1980)
69. Kuhn-Kuhnenfeld,F., J . Electrochem. SOC.119:1063 (1972)
70. Weyher, J.L. andvandeven, J., J . Cryst. Growth78:191 (1986)
71. Qadeer, A., Reed, J. and Bryant, F. J., J. Electrochem. S o c .

129:2145 (1982)
72. Landolt, D., Electrochim. Acta 32:l (1987)
73. Barber, P. G. and Crouch, R. K., J . Electrochem. SOC.131:2803

(1984)
74. Laser, D., Silberman, J. A., Spicer, W. E. and Wilson, J. A., J .

75. Elliot, C. R. and Regnault, J. C., J . Electrochem. SOC.128:113

(1981)
76. Takahashi, K., Japan. J . Appl. Phys. 18:1741 (1979)
77. Yamamoto,A., Tohno, S. and Uemura, C., J . Electrochem. SOC.

128:1095 (1981)
78. Bhat, R., J . Electrochem. SOC.132:2284 (1985)
79. Carraba, M. M.,Nguyen, N. M. and Rauh, R. D., J . Electrochem.

SOC.134:1855 (1987)
80. Carraba, M.M., Nguyen, N. M. and Rauh, R. D.,J . Electrochem.

SOC.134:261 (1987)
81. Burmeister, J., J. Cryst. Growth 11:131 (1971)
82. van der Putte, P., Giling, L. J. and Bloem, J., J . Cryst. Growth
41:133 (1977)
83. van der Putte, P., van Enckevort, W. J. P., Giling, L. J. and Bloem, J.,
J . Cryst. Growth 43:659 (1978)
84. Bloem, J. and Giling, L. J., in: Current Topics in Materials

Science, Vol. 7, (E. Kaldis, ed.), Chap. 4, North-Holland,Amsterdam
(1978).
85. Dominguez, C., Pastor, G. and Dominguez, E., J . Electrochem.

SOC.134:199(1987)
86. Dominguez, C., Pastor, G. and Dominguez, E., J . Electrochem.

SOC.134:202 (1987)
87. Kuyer, Th. J. M., Giling, L. J. and Bloem, J., J . Cryst. Growth 22:29
(1974)
88. Wilkes, J. G., Kingsnorth, R . L. and Low, H. E., in: Proceedings of

the Fourth ,European CVD Conference, ,Eindhoven, The
Netherlands, p. 183, (1983)
89. Ives, M. 6. and Baskin, M. S., J . Appl. Phys. 36:2057 (1965)
90. Sangwal, K. and Arora, S. K., J . Mater. Sci. 13:1977 ( 978)
91. Hartman, P., in: Crystal Growth - an lntroduction, (P Hartman,

ed.), p. 367, North-Holland, Amsterdam (1973)
92. Sangwal, K., Szurgot, M. and Szczepaniak, M., J . Cryst. Growth

79: 185 (1986)
93. Adachi, S. and Kawaguchi, H., J . ,Electrochem. SOC.128:1342

(1981)
94. Adachi, S., Kawaguchi, H. and Iwane, G., J . Mater. Sci. 16:2449
(1981)
95. Caridi, E. A. and Chang, T. Y., J . Electrochem. SOC.131:1440

(1984)
96. Adachi, S., J . ,Electrochem. SOC.129:609 (1982)
97. Chu, S. N. G. and Jodlauk, C. M., J . ,Electrochem. SOC.129:352

(1982)
98. Spink, G. M. and Ives, M. B., J. Appl. Phys. 4251 1 (1971)
99. Heimann, R. B., Ives, M. B. and Zaya, P., J . Cryst. Growth 57:48
(1982)
100. Bojarski, Z. and Wokulska, K. B., Prace Naok. Uniw. Slaskiego,

No. 209, Fiz. Chem. Metali 3 5 6 (1978)
101. Vogel, F.L., Pfann, W. G., Corey, H. E. andThomas, E. E., Phys.

Rev. 90:489 (1950)
102. Doerschel, J., K r i s f . Tech. 7:197 (1972)
103. Amelinckx, S., Acta Metall. 2:848 (1954)
104. Werkhoven, C. J., Hengst, J. H. T. and Bartels, W. J., J. Cryst.

Growth 42:632 (1977)
105. Petit, B., Pelletier, J. and Molins, R., J. ,Electrochem. SOC.

132:982 (1985)
106. Gannon, J. J. and Nuese, C. J., J . ,Electrochem. SOC.121: 1215

(1974)
107. Adachi, S . and Oe, K., J . ,Electrochem. SOC.130:2427 (1983)
108. Adachi, S. and Oe, K., J . ,Electrochem. SOC.131:126 (1984)
109. Chen, J. A., Lee, S. C. and Ho, L. I., J . ,Electrochem. SOC.

132:3016 (1985)
110. Barycka, I. and Zubel, I., J . Mater. Sci. 22:1299 (1987)
111. Huo, D. T. C., Wynn, J. D., Napholtz, S. S . , Lenzo, F. R. and Wilt, D.

P., J . Electrochem. SOC.134:2850 (1987)
112. Uekusa, S., Oigawa, K. and Tacano, M., J . Electrochem. SOC.

132:671 (1985)
113. Stano, A., J . Electrochem. SOC.134:448 (1987)
114. Iga, K. and Miller, B. I., I € € € J . Quantum Electron. Q,E 78, p. 22,
Jan. 1982
115. Adachi, S., Noguchi,Y.andKawaguchi,H., J . Electrochem. SOC.

129:1524 (1982)
116. Adachi, S., Noguchi,Y.andKawaguchi, H., J . ,Electrochem. SOC.

129:1053 (1982)
117. Adachi, S., Kawaguchi, H. and Iwane, G., J. Electrochem. SOC

129:883 (1982)
118. Kobayashi, T. and Sugiyama, K., Japan. J. Appl. Phys. 12:619

(1973)
Gurevich, S. A., Mizerov, M. N. and Portnoi, E. L., Zh.

119. Alferov, Zh. I.%
Tekh. Fiz. 45:2602 (1975)
120. Shaw, D. W., J. ,Electrochem. SOC.113:958 (1966)

ELECTROCHEMlCAL PASSIVATION OF (Hg ,Cd)Te
Robert T. Talasek
1 .O INTRODUCTION
For device technology based on silicon, thermal oxidation has emerged

as the passivation method of choice. However, thermal oxides have
proved less effective for the passivation of Il-VI compound semiconductors
such as (Hg,Cd)Te, used in infrared imaging detectors. Heating causes
a loss of mercury from the semiconductor bulk (1)(2), changing both the
bandgap and type of the material near the interface. For this and other
reasons, anodic oxidation has become the most successful passivation
method for these materials: an instancewhere electrochemistry is having
significant impact on electronics technology.
1.1 Types of Infrared Detectors
Past generations of imaging detectors were based on photoconductor

technology, which relies on bulk material effects. For these detectors,
passivation, although necessary, has a relatively small effect on device
performance. Present and future generations of imaging detectors are
based on Metal-lnsulator-Semiconductor (MIS) capacitor or p-n diode
(photovoltaic) elements. These detectorsare significantly affected by the
properties of the passivation as well as the interface between the insulating
passivation and the semiconductor.
Photoconductor Devices. Photoconductive detectors (3) used
for infrared imaging are arrays similar to the basic structure in Figure 1.
The (Hg,Cd)Te is thinned to an appropriate thickness through a combination
of sawing and mechanical polishing, chemically treated to remove residual
127
mechanical damage and bonded to the thermally conductive substrate.

The exposed area is passivatedand coated with an antireflective coating.
Electrical contacts to the (Hg,Cd)Te are created by chemically removing
passivation and metal deposition, usually by evaporation. Many variations
of this basic configuration are possible, some of which place the detector
contacts on the side opposite illumination. No matter what the configuration,
the basic operating principle remains the same.
For the intrinsic photoconductor (4), the conductivity of the material
is given by:
where: (J = conductivity,
q = carrier charge,
p,, p , = carrier mobility, and
n, p = carrier concentration.
Since the narrow band-gap semiconductors used for this application are
easily saturatedwith carriers by thermal generationof electron-holepairs,
the detector temperature is lowered sufficiently to reduce this generation
to a manageable level. The detector is illuminated under bias, and the
optically generated electron-holepairs cause an increase in the conductivity
of the materialwhich can be sensed by the attached processing circuity.
By rastering an array of these structures, an image can be electronically
generated based on the infrared radiation striking each detector element
with respect to its scanning position.
MIS Devices. Imaging detectors based on arrays of MIS structures
( 5 )utilize variations of the basic structure shown in Figure 2. The surface
of the material is mechanically polished to allow all detector elements to
be in the same focal plane. The surface is then chemically polished or
treated in some other manner to remove the mechanical damage caused
by polishing grit. An insulator is deposited or grown on the prepared
surface through one or more possible techniques. Gate metallization is
deposited on the insulator and may be transparent to allow illumination
through the gate or may be opaque in the case of illumination through the
semiconductor.
The operation of the MIS structure is best explained in terms of the
energy band diagrams and capacitance-voltage (C-V)behavior of a
typical metal-oxide-semiconductor(MOS) capacitor (6).The energy bands
vs. distancefrom the surfacefor a structurewith zero surface potential (4s
= 0) is given in Figure 3. This condition is called the flat band. As the
surface potential varies with applied gate potential in the positive or
negative direction, the band positions are shifted at the surface, while
Electrochemical Passivation of (Hg,Cd)Te 129
ELECTRICAL
LEADS ---
RESERVOR
Figure 1. Principal elements of a typical photoconductive device. (With

permission of Academic Press.)
I ,,-METAL
1,
Tm
;"' I p/ I ,(
I,
SEMICONDUCTOR
/INSULATOR
Figure 2. Metal-insulator-semiconductor (MIS) structure. (With permission

of Academic Press.)
SEMICONDUCTOR
IDEAL M I S DIODE (n-TYPE SEMICONOUCTOR)
QX v:o
I
S E H I C O N DUCTOR
(b) IDEAL M I S DIODE ( p - T Y P E SEMICONDUCTOR 1
Figure 3. Energy band diagrams for ideal MIS structures at V = 0. (With

remaining unaffected in the bulk. In the space-charge region between the

surface and the bulk, the bands are bent to an extent dependent given by
a form of the one-dimensional Poission equation:
where: p(x) = space charge density, and

% = permittivity of semiconductor.
This equation, when integrated between the bulk and surface potentials,
yields a quantitative description of the band bending in the space charge
region. The results for three different potential conditions are shown in
Figure 4.
For an n-type non-degenerate semiconductor, when the surface
potential is positive, the conduction and valence bands are bent downward,
while the Fermi level of the semiconductor remains fixed. Since the
majority carrier density is dependent on the energy difference between
the conduction band and the Fermi level (E, - Et), this produces an
increase in the carrier density as compared to the bulk in the space-
charge region. This condition is known as accumulation. As the surface
potential is varied in the negative direction, E, - E, increases,producing a
decrease in carrier density from the bulk, called depletion. This lower
carrier density creates a “potential well” in which carriers have an extended
life due to this lowered density (recombination rates increasewith carrier
density). As the surface potential is increased further in the negative
direction, the hole concentration becomes larger than the electron
concentration at the surface, and the surface is described as “inverted”.
The band behavior for p-type semiconductors is similar, but the polarity of
the surface potential must be reversed.
The energy band behavior with respect to potential provides a
qualitative explanation of the C-V behavior of the capacitor (Figure 5).
More direct quantitative expressions have been derived, but are beyond
the scope of this text. One may view the structure as a series of two
capacitors, the fixed insulator capacitor (Ci) and a variable capacitor
whose value is dependenton the capacitance of the space-chargeregion,
often known as the depletion capacitor (CJ. In accumulation, the space-
charge region is heavilypopulatedwithcarriersand isessentially metallic.
The capacitance in this region is therefore that of Ci. As the surface
potential is varied in the negative direction (for n-type), the depletion of
carriers begins, and C,falls, causing a decrease in total Capacitance. The
minimum of the total capacitance is related to the carrier concentration by
the Poisson equation, since carrier density affects the extent of band
bending. As the onset of inversion begins, C, rises and total capacitance
- TYPE
n p
V Z O
A
n-TYPE
Figure 4. Energy band diagrams for ideal MIS structures when V z 0 for
both p-type and n-type semiconductors. (With permission of Academic
Press.)
increases until it reaches Ci again. This behavior is observed only when

the gate bias controlling the surface potential is d.c. or low frequency ax.,
and the resulting description is a low frequency C-Vbehavior. If the bias
frequency is increasedsuff iciently, the recombination-generationratesof
the minority carrier can no longer keep up with the small signal variations
and do not show an increase in capacitance. Other conditions, such as
deep depletion can also also be induced in this potential regime, but will
not be discussed here.
In the case of a photocapacitor (6) used for infrared detection, an
appropriate bias is placed on the gate to induce depletion. The potential
well is created in which carriers are generated optically. These carriers
can be storedfor a period of time, due to their extended life, allowing the
collection of signal for some “integration time.” The bias is removed,
which “collapses” the well, passing the carriers out to be processed by
appropriate electronics. This type detector can be operated in a scanning
mode similar to a photo conductor detector, or a “staring” mode. In a
staring detector, a two dimensional array of MIS structures “stare” at the
image without scanning for a specific integration time. The signal from
each individual pixel is then recreated electronically by the appropriate
circuitry.
Photovoltaic Devices. The basic structure of a photodiode array
(7) is shown in Figure 6. Surface preparation is typically the same as for
MIS structures, although structure thickness is probably a more critical
consideration for photodiodes. An implantation mask of photoresist or
photolithographically defined metal is then used to allow ion implantation
of boron or other sourcesto produce n-typeregions on a p-type su bstrate.
Other methods of creating n-type (Hg,Cd)Te, such as impurity diffusion
grown junctions or mercury in-diffusion or out-diffusion, are also used.
The surface is then passivated,and contacts are etched through to the n-
type islandswhich are connected to the silicon processing circuitry with
evaporated metal. This process can be modified slightly to yield a
backside illuminated device.
The basic operation of a p-n junction ( 8 )as an infrared detector can
be explained once again through the use of energy band diagrams(Figure
7). In the simplest case of an abrupt p-n junction, a region is formed where
the transition between energy levels occurs which is depleted of carriers
by recombination of holes from the p-side and electrons from the n-side.
This region, called the depletion (or space-charge) region, has a width
which is controlled by the difference in energy levels of the conduction
bands of the p and n type regions. The difference in the energy levels of
these two regions (and thus the width of the depletion region) is controlled
by (among other things) the potential applied to the diode. Under zero
bias conditions, a potential exists, known as the built-in potential (Vbi)
I
0
GATE VOLTAGE (V)
Figure 5. High frequency (solid line) and low frequency (dashed line)
capacitance-voltage behavior for a typical n-type MIS structure.
LCU
Epory Input Ncde
t dve r Silicon C C O C t i i p
5ilicon C C D C t i i p 'IIlpUI
Nodes
Figure 6. Schematic illustrations of the (a) planar processed and (b)

backside-illuminated hybrid mosaic focal plane schemes utilizing a silicon
CCD processing chip together with Hg,,CdxTe junction photodiodes. (With
Figure 7 . Schematic representations of depletion layer width and energy

band diagrams of a p-n junction under various biasing conditions. (a)
thermal-equilibrium condition, (b) forward bias condition, and (c) reverse-
bias condition. (With permission of Academic Press.)
I
I
I
I -
I -
I
I -
- -
*A
I
I HgCdTe
I
h--- 100 A-
OXIDE
Surface states of (HgCd)Te.
Figure 8. Surface states of (HgCd)Te. (With permission of Academic

Press.)
which is proportional to the difference between the two energy levels. In

the presence of a forward bias, the difference in energy levels is proportional
to the difference between the bias (V,) and Vbi, and the depletion width is
decreased. The inverse is true for the reversed bias (V,) case. Photodiodes
are operated either under zero bias or reversed bias. Photo generation of
minority carriers is possible in this case because of the reduced carrier
density of the depletion region. Because of the increased lifetime of
majority carriers, integration can be used in a similar manner to MIS based
devices, and both scanning and staring modes are feasible.
1.2 Electrochemical Passivation
Electrochemical oxidation of (Hg,Cd)Te has several shortcomings.

Various reports have indicated a lack of thermal stability of the oxide
(2)( 10-13). Also, the oxide-semiconductor interface and oxide near the
interface has relatively poor interface quality as compared to thermally
oxidized silicon (1 4-1 8). Anodic oxidation at an elevated temperature has
been used in one instance (9),and was reported to have increased
stability over room temperature anodic oxides. This idea has not been
further explored.
Oxide-relatedfeatures that can impair detector performance include
fast and slow interface states and fixed charges (5) (Figure 8). Fast
interface states (NFS)adversely affect the dark current by providing
indirect transition paths for tunneling (Figure 9). Slow states within
tunneling distance also contribute to dark current, acting as minority
carrier traps. These contributions are most significant in material with E,
5 0.1 2 eV. Positive fixed charge in the oxide near the interface can make
asignificant contribution to thesurface potential. Among other problems,
this charge drives the surface of p-type material to inversion, requiring an
isolating guard ring structure which allows the surface to be maintained in
a depletion mode for infrared imaging.
In photovoltaic devices (7), fast interface states at the insulator-
semiconductor interface can act as carrier generation-recombination
centers, and may make a significant contribution to dark current which
limits image contrast. Fixed oxide charge near the interface can modify
the surface potential which can cause an increase in dark current through
one of several mechanisms (Figure l o ) , including surface leakage.
Material Effects. Material also limits advanced infrared device
performance. Material parameters such as carrier concentration, minority
carrier lifetime, carrier mobility and crystalline perfection all have varying
effects on the performance of the finished device. Compositional
homogeneity also plays a significant role in device performance, since
relatively small variations in composition can produce significant shifts in
Figure 9. (a) Various tunneling mechanisms in (HgCd)Te, and (b)

tunneling via bulk Shockley-Read centers. (With permission of Academic
Press.)
.OXIDE/IN5ULAIOH ,hlllALLUl(LICAL
/ ' IUNLIIUN
Figure 10. Effect of fixed oxide/insulator charge on the effective junction

space-charge region: (a) flat-band condition; (b) positive fixed charge
causes inversion of the p side and formation of a n-type surface channel;
(c) negative fixed charge causes accumulation of the p side and a narrow
field induced junction at the surface; (d) larger amount of negative fixed
charge causes inversion of the n side and formation of a p-type surface
channel. (With permission of Academic Press.)
the band gap (E,)of the material. (Table 1 lists the variation of a number
of parameterswith composition (7)) All of these factors have resulted in
a significant body of work to produce improved materials.
Numerous bulk crystal growth techniques (1 9) have been investigated.
Although Czochralski growth methods have been investigated,technical
difficulties (e.g., Hg pressure, maintaining melt composition) have prevented
this technique from being as successfulas with other semiconductors. A
number of methods such as quench/recrystallization, often called solid
state recrystallization (SSR), both horizontal and vertical Bargeman
techniques, zone melting, travelling solvent zone melting and slush growth
all have been used tovarying extents. However, these methods generally
are limited to small crystals. Efforts to increase crystal size include bulk
techniques such as incremental quenching (20) and travelling heater
method (THM) (21) as well as epitaxial methods (22). These methods
include various liquid phase epitaxy (LPE) methods, vapor phase epitaxy
(VPE), controlled diffusion of Hg into CdTe, vacuum sputter deposition,
metal organic chemical vapor deposition (MOCVD) (22), molecular beam
epitaxy (MBE) (23)(24), and laser assisteddeposition (LADA) (25). (Note:
This list is only partial and is for illustration. The reader is referred to the
literature for a more complete review of material growth techniques).
Each of these methods produces material with different compositional
homogeneity and crystalline perfection.
Surface Preparation. Before passivation, various surface
preparations may be used to remove surface damage induced by mechanical
polishing which would otherwise significantly reduce carrier lifetime (26).
Chemical etching or chemical-mechanical polishing with various
concentrations of Br, in methanol (or other solvents) have been used, and
have been shown to leave the surface depleted of Cd, and to a lesser
extent, Hg (27-31). Electropolishing techniques (32) as well as oxidation
and etchback (27)(28) have been used toaddress this problem, and seem
to leave the surface closer to its original stoichiometry.
Such wide variances in material and surface preparation can be
expected to have a significant effect on a surface process such as anodic
oxidation. However, the literature is full of studies of anodic oxidation,
with each study using the researchers favorite material and surface
preparation, often with conflicting results. In the following pages, an
attempt will be made to summarize these studies and find a common
ground where one exists. Also, an attempt will be made to note material
and surface preparation differences when they have been reported, so
that conflicting results can be better understood.
With all of the problems associated with anodic oxidation, it is
inevitable that an attempt be made to develop alternative passivation
methods. Some, such as sulfidization, are electrochemical, others are
ElectrochemicalPassivation of (Hg,Cd)Te 141
not. Areview of these methodswill be given, with emphasis placed on the

electrochemical techniques.
Finally, there have been other electrochemical processes involving
Il-VI semiconductors, the most significant being electrodeposition. Among
the compounds deposited from solution are CdS (33), CdSe (34), (Cd,Se)Te
(35) (of various compositions), HgS (36), and (Hg,Cd)Te (37). (The
references listed are for example only). Various metal sulfides have also
been anodically produced from sulfide containing electrolytes (38-40).
These films have generally been used for solar cell application due to
limitations in crystalline perfection, compositional control, andfilm purity,
although a relationship between sulfide passivation and anodic deposition
of sulfide films may exist. Since this text will deal with infrared detector
technology, no further mention will be made of electrodepositionprocesses.
Electrochemical doping, primarily with copper, has also been described
(41). No published reports of device applications have been made,
however, and this process will not be discussed further.
2.0 ANODIC OXIDATION
Throughout the body of research on anodic oxidation, many different

electrolytes have been used. By far the most common is 0.1 N KOH in 90%
ethylene glycol/lO% water described by Catagnus and Baker (42). In a
few instances, the concentration of KOH was varied for one reason or
another. Hydrogen peroxide and 0.1 N KOH in methanol or water have
also been used studied in instances. Also, the use of acetate, borate, and
phosphate buffers have been used in investigative studies where it was
necessary to control pH, and at least one study exists where Hg, Cd, and
Te were added to the electrolyte. In almost no instancewas the total ionic
strength controlled. All of these factors as well as material and surface
preparation must be consideredwhen a comparison of various results are
made.
Much research has been conducted to determine the oxide composition
in the bulk and at the semiconductor oxide interface. Both calculated and
experimentally-determined phase diagrams of various types have been
reported. Also, numerous surface analysis and optical techniques have
been used to evaluate bulk and interface composition.
An equally active area of research has been determination of the
chemical processes involved in oxide formation. Many different
electrochemical measurement approaches have met varying degrees of
success in this endeavor. Various analytical techniques, such as those
used in surface analysis have, also been employed in attempts to follow
the progressive steps of oxide formation.
2.1 Oxide Composition - Phase Diagrams
Several attempts have been made to construct phase diagrams of

the Hg, Cd, Te, 0 system. The earliest diagram constructed, however,
contained only Cd, Te and 0 and were constructed from an effort to grow
and characterize bulk crystals from various mixtures of CdO and TeO,
(Figure 11) (43). Two compounds found were reported as cadmium
ditellurite (CdTe,O,) and cadmium mono-tellurite (CdTeO,). The later
was reported to have two phases with a transition occurring at 700OC.
In 1983, two different types of phase diagrams were reported
independently. First, a solid-state quaternary phase diagram was
constructedfor the Hg-Cd-Te-0system by Rhiger andKvaas (44) through
thermodynamic calculations of mutual stability as Gibb’s free energy of
various elements and compounds assumed to be possible components
of the anodic oxide. The four ternary diagramswhich represent the sides
of the quaternary diagram are shown in Figure 12. In order to construct
this diagram, a number of assumptions were made, as can be expected.
One significant assumption is the absence of CdTe,O, from consideration.
Another is that no compounds containing both Hg and Cd were considered.
Instead these compounds were considered as a two-phase mixture of
compounds in a ratio to arrive at the appropriate stoichiometry. (Note:
thermodynamic data for the quaternary compounds are not readily available).
The calculated phase diagramwas confirmed in some cases by attempted
reaction of mutually stable or unstable compounds as predicted by tie
lines within the phase diagram, and the composition of the reaction
products was confirmed by Raman spectroscopy. Where evaluated, the
phase diagram made correct predictions
A significant conclusion drawn from this diagram were that elemental
mercury, HgTe, (Hg,Cd)Te and CdTeO, are thermodynamically stable
with respect to mercury-telluriumoxides and mercury oxides. The prediction
was made that the instability of the mercury-telluriumoxides and mercury
oxides may decompose to yield Hg or HgTe at the oxide semiconductor
interfaces, with potentially disastrous results to device performance. This
instability has been demonstratedexperimentally (1)(2)(10-12)with mercury
loss and oxide compositional changes being demonstrated by various
methods after thermal annealing. In a least one case (12), a change in
semiconductor type as shown by capacitance-voltagebehavior has been
demonstrated. The presence of a multi-layeredoxide structure has also
been demonstrated by Auger and SAL1 (surface analysis by laser ionization)
profiling (13). The layer closest to the interface was concluded to be
CdTeO,, with the mercury depleted at the interface being present in the
outer layers as HgTe. Mercury enrichment as suggested by Rhiger and
Electrochemical Passivation of (Hg,Cd)Te
9 00
eo0
I LIO
-
-
900
BOO
t
U
w 700 - 700
3
.(
rx
w
n 600 -600
I
+.l
500 - 500
50 - 50
raol
IO 20 30 40 50
MOL O l a C d O -
60 70 8 0 9 0
CdO
Figure 11. CdO-TeO, (phase diagram, major compositions and

temperatures). 18 mol % CdO, 630 k 3°C eutectic; 33.3 mol % CdO,
708 k 3°C congruent point (A); 35.5 mol % CdO, eutectic 690 ? 3°C; 48
mol % CdO (approx.), 725 ?r 3"C, incongruent point (B); 50 mol % CdO,
795 ? 3"C, incongruent point (C), phase change in (C). (With permission
of Chapman Hall.)
0 0
TeOZ
A' uA Hqlc Te le Cd
cd w ca caic le
Figure 12. The four ternary phase equilibrium diagrams: (a) Hg-Te-0, (b)
Cd-Te-0, (c) Hg-Cd-0, (d) Hg-Cd-Te. Tie lines connect pairs of mutually
stable phases. (With permission of American Institute of Physics.)
Kvaas was only seen at the interface after thermal annealing.

Another type of phase diagram reported in 1983was constructed by
from depth profiling of anodic oxides by x-ray photoelectron
Davis, et al. (45),
spectroscopy (XPS) in conjunction with ion beam sputtering in Figure 13.
This type diagram, called a surface behavior diagram, represented the
oxide composition as a function of depth. This group concluded a bulk
composition of a combination of (Cd,Hg)Te, and (Cd,Hg)Te,O, with the
mercury to cadmium ratio being considerably lower than that of the
semiconductor. Mercury depletion of the interfacial region was also
reported agreeing with the experimental results of Stahle, et ai., cited
earlier. Similar studies were performed by Seelmann-Eggebertand Richter
by XPSwith dissolution used in place of sputteringfor depth profiling. The
studies report an accumulation of HgTe at the interface. The extent of this
accumulation is reported to be a function of electrolyte pH, anodization
voltage, and presence of an organic solvent in the electrolyte.
A year later another stability diagram based on thermodynamic
calculationswas reported. Based on the methods developed by Pourbaix
(48),this diagram presented regions of stability of various dissolved and
undissolved species in a potential vs. pH format (Figure 14) (49).
Thermodynamic calculations once again supported elemental mercury
being stable with respect to various oxide components, although calculations
were once again limited by the lack of data for the Hg-Cd-Te-0quaternary
compounds. Also, the mercury and cadmium ditellurites were calculated
to be electrochemically more stable than the monotellurite compounds.
Experimentally determined diagrams from electrochemical measurements
(potentiodynamic polarization measurements) made in various buffer
solutions were also constructed using n-type, x = 0.225 material which
had been mechanically polished (Figure 15). These measurements indicated
that the onset of passivation corresponding with a current maximum did
not occur until a much higher applied potential than what was predicted
thermodynamically, thus suggesting that the anodic oxidation process
might be significantly affected by the reaction kinetics.
2.2 Oxide Analyses
Perhaps the first investigation into the chemical composition of the

anodic oxide of HgCdTe was reported by Nemirovsky and Finkman in
1979 (50). Anodic oxide films were grown on n-type slush grown (Hg,Cd)Te
(Ee= 0.1 eV) in 0.1N KOH in methanol and various concentrations ( < 0.1
N) of KOH in 90% ethylene glycol /lo% H,O. The material surface was
mechanically polishedand etched in 20% Br, in methanol prior to anodization.
The refractive index and dielectric constant of the oxide were determined
to be closer to TeO, than CdO or HgO, so the authors concluded that the
Figure 13. (HgCd)-Te02-(HgCd)Te surface behavior diagram of

Hg,-&do,Te with a 1600 A thick anodic oxide. The superimposed depth
scale gives the equivalent thickness of oxide for an abrupt interface and a
mean free path of 16 A. The dashed line represents the depth profile path
for a stoichiometric oxide and semiconductor. (With permission of Elsevier
Science Publishers.)
Figure 14. Potential-pH diagram for Hg,,Cd,,Te constructed from

thermodynamic data.
1.2 \
\
\ PASS1VAT ION
1 .o
0.8
E
1
0.6
z
v)
>
vl \ \
3
0
0.4
0
2
W
0.2 0
0 IMMUNITY
\ 8
\
\
-0.2 \
\
\
-0.4 I I I\ I I I
2 4 6 8 10 12 14
PH
0 CORROSION POTENTIAL
x PRIMARY PASSIVATION POTENTIAL
Figure 15. Potential-pH diagram for mechanically polished HgCdTe

constructed from potentiodynamic polarization curves in various pH
buffers.
148 Electrochemistry of Setniconductors and Electronics
passivation layer must be primarily TeO, (no other oxides were considered).
The authors theorized that cadmium in the semiconductor was oxidized
to Cd2+ and dissolved, but mercury was trapped in the TeO, as an
elemental layer which is partially oxidized at the interface.
Shortly after this, however, the presence of cadmium in a Cd-Te-0
oxide compound was demonstrated (51)(52) using sputtering and XPS
depth profiling. Also, it was pointed out that the optical properties of
CdTeO, were suitable to explain the behavior observed by Nemirovsky
and Finkman. Finally, this group also reported the significant mercury
depletion in the bulk and interface region discussed with the surface
behavior diagrams.
Fromthis point on, various groups have argued the extent of mercury
depletion in the oxide. An extensive study (53) of mercury losses during
various combinations of sputtering or thermal desorption used in conjunction
with analytical techniques demonstrated that at least part of the mercury
loss was due to preferential sputtering of mercury during depth profiling.
However, the extent of this contribution to analytically determined
composition of the oxide is still disputed. In fact as late as 1982, Farrow,
et al. (54), still claimed that little mercury was present in the anodic oxide
asdetermined byXPS,while Morgenetal. (55), claimedonlythattheoxide
was depleted to some unknown extent through Auger electron spectroscopy
(AES). Also, Farrow claimed spectroscopic evidence that CdTe,O, could
not be a major component of the oxide. Christie, et al. (56), stated that the
oxide could be as much as 25% CdTe,O, while stating that this was not
necessarily in disagreement with Farrow since this would not be considered
a major constituent of the oxide. Seelmann-Eggebert,et al. (57), strongly
disputed Farrow’s measurement as erroneous, stating that they measured
the same oxygen Auger parameter (the parameter claimed by Farrow to
be different in CdTeO, and CdTe,O5) for both compounds. However, in
the same paper, Seelmann-Eggebert claimed the composition of the
oxide was primarily HgTeO, and CdTeO,, Also, they made the first claim
that the mercury to cadmium ratio in the oxide might be as high as in the
semiconductor, and was certainly no lower than 1 :1. Electron microprobe
analysis by Kaiser, et al. (58), also supported a significant mercury
concentration in the oxide, although somewhat less than Seelmann-
Eggebert. Surface analysis by laser ionization (SALI), Rutherford
backscattering (RBS) (59), and neutron activation analysis (NAA) (60)
have also established a significant mercury content in the oxide. Along
with the measurement artifacts which have been discussed in other
publications, perhaps Strong, et al. (61), have demonstrated another
significant cause of this wide disagreement concerning oxide composition.
By determining the composition of oxides grown in different electrolytes
(0.1 N KOH in 90% ethylene glycol / l o % H,O and H,O, unspecified
concentrations), significant differencesin oxide composition were found,

the later being depleted of Hg and Cd to a much larger extent than the
former. The oxide grown in alkaline solution showed significant quantities
of mercury and cadmium, although somewhat depleted with respect to
the bulk. In general, the oxide has been shown to be 30-40% depleted of
mercury with respect to the bulk, and only slight (if any) depletion of
cadmium. The multilayer model mentioned earlier (1 3) suggests that the
mercury depletion is primarily near the interface, and precludes the
conclusion of a single oxide composition.
2.3 Chemical Processes
Numerousattempts have been madeto study the interactionswhich

produce oxidation of the (Hg,Cd)Te surface. Several studies of the
interaction of molecular oxygen in ground and excited states have been
made with XPS and other surface analysis techniques (62-67). Pressure
conditions have varied from atmospheric to high vacuum in situ studies.
In one instance,AI was used insteadof oxygen to study the behavior of this
surface (68). These studies are of limited usefulness for studying anodic
oxidation processes since the oxygen-containing species is chemically
different and is contained in a significantly different matrix. (This is not a
criticism since most of these studies were intended to study gas phase
oxidation. In fact, in at least one case it was specifically concluded that
anodic oxidation and gas phase oxiddon must be significantly different
processes) (67). Nevertheless, an important conclusion can be drawn
from these works that should be applicable to anodic oxidation. Significant
evidence has been produced that surface preparation had a significant
affect on both the chemical processes involved in oxidation, and the oxide
composition for the oxides (63-70). It was also suggested that the
constituent elements may behave differently from one another and that
impurity absorption (specifically organic acids) may impede the oxidation
process (62). As will be shown, these conclusions will be supported by
more direct studies of the anodic oxidation process.
A number of different techniques have been usedto study the surface
preparationtechniques used prior to anodic oxidation. For device fabrication,
mechanical polishing is usually necessary to insure a relatively flat surface.
Some chemical treatment is necessary to remove the mechanical damage
induced by grit polishing. Often etching or chem-mechanicalpolishing is
used to remove this damaged layer. However, it has been shown that
chemical treatments do not leave a stoichiometric surface (27-30)(63). A
kinetic study of the reaction of 0.1N Br, in methanolwith (Hg,Cd)Te under
etch conditions found the relative rates of reaction of the individual
constituents to be:
Cd: 10
Hg: 3
Te: 1
and the disparity in reaction rates worsened with the addition of mechanical
action (29). Various attempts have been made to estimatethe depth of the
depletion layer caused by the inequity of the constituent reaction rates.
Estimatesaslowas0.7 nm (28)andas highas60 nm(27) have been made
either by optical techniques or a combination of oxidation-dissolution
combined with optical techniques.
This depletion has been shown to have an effect on the chemical
processes involved in anodic oxidation. Various authors have reported
the presence of an additional dissolution peak in potentiodynamic
polarization (I-V) curves of the etched or polished surface as opposed to
a mechanically polished surface, although the two preparation techniques
were only compared in case (49). Although no conclusive identification
has been made of this species, it has been speculated that this additional
peak could represent dissolution of excess tellurium on the surface or
dissolution of some bromine-containing species. Several efforts have
been made to develop surface treatment which removes this depleted
layer from the surface. Anodic oxidation followed by acid stripping of the
oxide with nitric acid, hydrochloric acid, or lactic acid in one or more steps
have been shown to yield a stoichiometric surface (27)(28)(69)(70), as
long as this process is stopped before consuming undepleted material
(32)(70)(72). Also, electropolishing has been used to treat the surface,
also resulting in a stoichiometric surface. Some limited description of the
effects of this surface treatment on the capacitance-voltage behavior of
an MOS capacitor fabricated from this surface were described as having
negative fixed charge (Figure 16) (32), although the results are not
improved with respect to fast sutface states, as compared to BrJmethanol
polished surfaces.
Various efforts have been made to study the anodization process
using common electrochemical techniques. One measurement used to
evaluate the corrosion/passivationproperties of alloys is potentiodynamic
polarization curves (74). Potentiodynamic polarization measurements
involve applying a voltage ramp (usually at a slow enough rate to approximate
static behavior) and measuring the resulting current from electrochemical
processes. Several characteristics of this curve are worth mentioning.
First is the open circuit potential (OCP), the potential at which the anodic
and cathodic currents are equivalent and total current is at a minimum.
Between the OCP and the current maximum peak is a region in which only
dissolution processes are occurring. The peak, the primary passivation
80.
70 -
60 - -
W
50- -
u
5 40-
c -
'
u
3 30-
-
20 - -
10 - -
0.
Figure 16. Capacitance voltage characteristic for electroetched MOS

structure at 109 K. (With permission of American Institute of Physics.)
potential, represents the onset of passivation. Beyond this peak is the

passivation region. Sometimes a drastic and sudden current increase is
observed, representativeof passivation breakdown. This is referredto as
the transpassive region. The first potentiodynamic polarization on (Hg,Cd)Te
appeared in 1982 (73). (See Figures 17-21). As is readily apparent, the
appearance of these curves vary significantly, and one must take into
consideration the wide variation in material, surface preparation, electrolyte,
and scan conditions to understandthis variation (See Table 2). However,
one consistent result is apparent. In the absence of any pre-treatment
designed to remove the depleted layer caused by bromine-methanol
treatment, two current peaks appear. One study (49) demonstrates that
one of these peaks appears at the same potential for mechanically
polished and chemically polished samples. This wasobserved in avariety
of electrolytes at a number of pH values. The second peak occurs at a
more cathodic potential. Both vary with pH at the same rate (See Figure
22).
A closely related experiment attempted to determine the effect of
hydroxide ion concentration on the rate of anodic processes in the
absenceof polarization (78). Thiswasdonewith aseriesofTafel plots (79)
(I-V relationship near the region where the OCP) in varying hydroxide ion
concentrations. The results indicated that the anodic current representative
of dissolution rate in the absence of overpotential is independent of
hydroxide concentration.
A second technique that has seen some use is ring-disk electrode
measurements (78). Ring disk electrodes generally consist of a disk
constructed of the material of interest (Hg,Cd)Te in this case) with a
concentric ring insulated from the disk. In this manner, the ring and disk
can be held or swept at separate potentials. Often the ring is split so that
each part can be held at separate potentials. Using an electrode of this
type allows various oxidation or reduction processes to be carried out
while various dissolved species of interest can be reduced at the ring(s),
producing an indication of increasedor decreasedconcentrations. Mass
transport processes can also be studied by rotating the electrode at
various angular velocities. Sakashista, et al. (72)(73), performed two
extensive studies of the first type with HgTe and (Hg,Cd)Te using a split
ring set-up. Among the significant points of this study was an observed
reduction of ring currents designed to demonstrate the decrease of Hg
and Te dissolution under appropriate conditions (Fig 23); cadmium cannot
be observed by the method due to its cathodic reduction potential. Also,
a distinct difference in the dissolution processes occurring during anodization
was demonstrated in an aqueous KOH electrolyte compared to the 90%
ethyleneglycol/l O%water mixture. (Figure24). It is readilyapparent that
there is an extreme difference in the extent of dissolution in the two
AI
I.&(1_1 I
- 1 6 I 2 3 1 5
Voltage (v) vs Hg/HgO
Figure 17
lor--
I I Y- UV
O W L - -..
0 10 20 30 LO
EdVH
Figure 19
Figures 17-21. Current potential behavior for HgCdTe under conditions

specified in Table 2. (With permission of The Electrochemical Society, Inc.)
“f n
n
-a> L o - - + - - - - y
*_I QO, -QJ------o;,om/ J
Current ( P A ) Current (PA)
Figure 20a Figure 20b
Voltage cv)vs S.CX
Figure 21
Figure 17-21 (continued). Current potential behavior for HgCdTe under

conditions specified in Table 2.
n
-
I
3
0
u
u 0
6OCn
QJ m
I
O N
m X ro
OY
> Y
.-C D0 t E z x
I I -
- N -u
1 - - 0
O \ * '4J
-
Y X I
x --
o o w
. uc m m n
' w u
OWLO
f
cnf
'- u
-+J h
o(I, Li
a
rc
o u X
-0
-C
(I,@
xa! mu
\
uz:
'- \
N. C N
L C
m .- m i
r m
rr
IT,
E
- U 6
E-
f f
4J e,
3 3
?
o m
u1
N 2
N O N
a J .
K %d nX o c no
a l .
a x u)X
3 + I1 rrcl II J + I1 Q)
- I
Lo c x
L n l
Lncx
- I
n ax
c
!-
Lc
-
I'.
0
N
m .
n
-. -
\I
1.2
\
\
\
1 .o \
0.8
0.G
W
I
z
v,
>
UJ 0.4
5
0
2
Lu
0.2
-0.2
-0.4
0 24 4 6 8 10 12 14
PH
Jf EXTRA PEAK FOR BRJMeOH POLISHED SAMPLES
0 CORROSION POTENTIAL
X PRlMAnY PASSIVATION POTENTIAL
Figure 22. Potential-pH diagram for Br,/methanol polished HgCdTe
constructed from potentiodynamic polarization curves in various pH
buffers.
0 03
0 02
-.
<
D
z
0 01
- 0.6 - 02 Q.1 0.6 1.o 1.C 1.3

ELI 1 V”
Figure 23. Quasipotentiostatic polarization curve of a CMT disk and the

related ring currents I, and ,I in pH 4.9 acetic buffer. Values after 10
min oxidation E ~ ,= -0.3OVl E, = 0.40V. (With permission of The
Electrochemical Society Inc.)
0.03
0.02
--.
4
I(
c,'
0.m
Figure 24. Quasipotentiostatic polarization curve of a CMT disk and

related ring currents I, and I, in 0.1 M KOH solid symbols (0,M, A)and
0.1M KOH t 90% EG open symbols (0, 0,A) afler 10 min oxidation, E ~ ,
= -0.68V, ,E = -0.08V. (With permission of The Electrochemical
Society, Inc,)
electrolytes. However, it is not clear from the data whether the difference
is due to the variation of total KOH concentration (0.1 N vs. 0.01 N total
concentration) or the difference in solubilities of the dissolved species
due to the presence of ethylene glycol. Both could be expected to make
major contributions. The ring-disk behavior of the oxidation of material
which had been potentiostatically reduced was also studied, but the
oxidation was carried out in pH 4.9 buffer, which has a significantly
different potentiodynamic behavior and the results of this portion of the
study are not directly comparable to others done in alkaline solution. A
second set of ring-disk electrode experiments performed as a part of a
kinetic study (79) has also been conducted to evaluate mass transport
effects. These experiments were performed with a solid gold ring and
incrementally quenched material under various potentiostatic conditions.
The electrolyte used was 90% ethylene glycol/lO% water with various
concentrations of KOH added. The current observed at the ring of a ring-
disk electrode is described by:
where: L,R = Limiting ring current,

n = Number of electrons transferred,
F = Faraday’s constant,
R3 = Outer ring radius,
R, = Inner ring radius,
D = Diffusion coefficient of active species,
w = Angular velocity,
v = Kinematic viscosity, and
CB = Bulk concentration of species being
reduced at ring.
Besides the factors that affect ring current that can be interpreted directly
from the equation above, it should also be noted that the kinetic half-life of
species being formed at the disk will affect the ring current, since it will
affect the population of the species surviving long enough to reach the
ring. Figure 25 represents the measured ring current at a function of
rotational velocity for several concentrations of KOH.
For a simple system in which this relationship is only affected by the
diffusion coefficient of the species being reduced, astraight line should be
observed. This is the case for 0.1 N KOH. However, as the KOH
concentration is dropped, non-linear behavior is observed which should
be indicative of a change of species being reduced at the ring, resulting
from an increased half-life of whatever species is reactingwith the hydroxide
ion.
1
0,L
Figure 25. Ring current vs rotational velocity for various concentrations of

KOH in 90% ethylene glycol/lO% H,O.
Another electrochemical study determined the potential behavior

vs. time for constant current anodization conditions (32). The material
used was n-type, slush-growth materialwith acomposition of x = 0.3. The
materialwas etched extensively in 5% bromine in methanol, to the extent
of removing approximately 20 microns of material. The electrolyte used
was 0.1 N KOH in 90% ethylene glycol/l 0% water. Figure 26 illustrates
the behavior observed for various current densities. One significant
aspect of this behavior is the period where potential is essentially independent
of time. This portion of the curve (called the induction period) is indicative
of no passivation being formed (no IR drop) with only dissolution supporting
current, a characteristic of dissolution-precipitation type passivation
processes. Furthermore, at sufficiently low current densities, the induction
period becomes infinite, suggesting that the precipitation portion of the
mechanism is prevented by mass transport of some species away from
the (Hg,Cd)Te surface.
A different technique was used to study the fundamental processes
involved in the initial oxide formation (76). A technique called a x .
impedance (80) studies the effects the system has on a x . signals of
varying frequency, thereby investigating the elementary mechanism steps
on the timescale that they actually occur. This investigation was conducted
on n-type material grown by solid state recrystallization wherex = 0.225.
The materialwas prepared by mechanically polishingwith 0.3 micron grit
only, so no effects of bromine etching were encountered. The electrolyte
used was 0.1 N KOH in 90% ethylene glycol/lO% water. Through
mathematical models which have been constructed for the impedance
behavior of various general mechanism types, it was concluded that the
impedance behavior of this system best matched that of a mechanism
consisting of three steps: an electron transfer, then chemical process,
and finally a second electron transfer (ECE mechanism) for the special
case where the first electron transfer is much faster than the second
electron transfer. Consideringthisas well as the dissolution-precipitation
effects presented earlier, the mechanismwas concluded to be:
k
Eq. (4a) M 2 Mn+ + ne' , k,,ha 1/R,
Eq. (4b) Mn+ t nOH- + M(OH),
(Authors Note: Corrections should be made in the hydroxide

stoichiometry to account for basic solutions. This does not change the
basic premise, however).
Another interesting component of the impedance that came to light
16 I I I I / I 1
15
d
0
I-.
P I[
z
9
r-
d
N
0
0
z
4
O!
0'
I I I
0 50 100 I50 2'30
TIME isecl
Figure 26. Anodization voltage vs time at constant current. (With

permission of American Institute of Physics.)
is a complex impedance component, the Warburg impedance, which is

normally attributed to mass transport effects. It had a significant potential
dependence, and essentially disappeared at 0.8 volt vs. a saturated
calomel electrode (SCE). A further conclusion from this fact was that
mass transport was the controlling effect at low overpotentials, being
taken over by charge transfer at some point. The rotating ring-disk
electrode data (79) demonstrating a relationship between hydroxide
concentration and mass transport suggests that the potential atwhich this
change occurs could be affected by hydroxide concentration or mass
transport rates, but this has never been concretely demonstrated.
This also tends to explain theobservation by Janosek, et al. (32),that
oxidation could be prevented at low current densities by stirring, but
would occur eventuallyat higher current densities in the constant current
studies. Further use of the a x . impedance technique used the assumption
that a component of the equivalent circuit, the charge transfer resistance
(RCT), is inversely proportional to the first step of the ECE mechanism, a
valid assumption as long as the first charge transfer is much slower than
the second. By saturating the electrolyte with one of the three constituent
elements, the dissolution of that element could be stopped, or at least
impeded significantly (79). Measuring R, could give relative dissolution
rates of the three elements. In fact, the relative dissolution rates were
shown to be:
Te : 1
Hg: 2.3
Cd: 10.3
which are very similar to those rates found for dissolution in bromine-
methanol under etch conditions.
An interesting approach to evaluating the processes involved in
(Hg,Cd)Te oxidation was taken by Lopes and Hart (81). They used
microelectroephoresis to determine the effect of pH on surface charge
density as expressed by zeta potential. As predicated by the experimental
method, powderedsamplesof (Hg,Cd)Te(p-typewithx = 0.2)wereused.
In order to cover large pH range, an aqueous solution of potassium
phosphate was used, and the pH adjusted accordingly with acetic acid,
hydrochloric acid, or potassium hydroxide. The measured relationship of
the zeta potential to pH is shown in Figure 27. Significantly, this behavior
agrees well with the dissociation of tellurous acid, H,TeO, up to a pH of
about 10. The authors associate this deviation with the decrease in
solubility of the cadmium species in this range (Figure 28). This diagram
demonstrates a similar behavior for mercury species, and this may be
significant since the surface concentration of mercury is four times higher
-sa - bO
- 20 - 3.0
-
N
5
>
E
-10
- 1.0 e-.
h,
-0
0
c 1.0
10
20 t
Figure 27. The zeta potential and surface charge density of Hg,Cd,Te
as a function of pH. The solid line is the best fit theoretical curve to the
acid dissociation model with pK = 12.7, pK, = 2.4, pK, = 7.6, and y =
0.30. The pK values correspond to the dissociation constants of the
different phases of tellurous acid and y is proportional to the total acid site
density. The surface oxide chemistry is HTeO,' (positive zeta potential,
H,Te03 (zero zeta potential), HTeO, (first step in the zeta potential), and
TeO,= (second step in the zeta potential). (With permission of American
institute of Physics.)
n -2
E
9
jci '4
a -6
-8
Figure 28. Solubility of various species containing Hg, Cd, and Te as a

function of pH. (With permission of American Institute of Physics.)
than cadmium. At any rate, significant evidence of pH affecting the

oxidation process is presented in this work.
A more limited amount of work has been done on the bulk growth of
anodicoxideon (Hg,Cd)Te. However, twoeffortsarepresented here. The
first involved the growth of oxide in the presence of water which has been
isotopically enriched with 18/8 0 (28). The profile presented in Figure 29
was concluded to suggest bulk growth by successivejumps of hydroxide
ions, although cation diffusion was not eliminated completely. The authors
also suggested that the hump in the profile near the interface suggested
that initial growth was by a separate mechanism, dissolution-precipitation,
and had a direct effect on the oxide-semiconductor interface. Another
series of studies (61)(83) were conducted using SIMS profiling of oxides
grown in both hydrogen peroxide and 0.1 N KOH in 90% ethylene glycol/
10%water using various markers for the original surface or intermediate
pointsin thegrowth. The materialusedwas SSR grown withx = 0.22. One
interesting sidelight of this study was the depletion of Cd at the surface of
the H,O, -grown oxide, and the lack of depletion in the basic electrolyte,
apparently the result of solubility factors mentioned earlier. (The depletion
of Cd was used as a marker in part of the study). Although the authors
presentedevidencefor Hg, Cd and Te dissolution in the electrolyte and for
metal out-diffusion in at least one instance, the bulk growth mechanism
was concluded to be by oxygen vacancy out-diff usion. The later paper
conceded the possibility of simultaneous oxide growth at the oxide-
electrolyte interface. This work also demonstrated that the bulk growth
mechanism could not be by dissolution-precipitation.
3.0 ALTERNATIVE PASSIVATION PROCESSES
A number of alternative passivation processes to anodic oxidation

have been investigated for (Hg,Cd)Te. The primary motivations for this
search have been the relatively high fixed charge associated with most
anodic oxides, and low thermal stability. The fixed charge is especially a
problem with p-type devices, since it produces surface inversion which
requires additional device structures to compensate. The thermal stability
has been pursued because of the likely association of thermal stability
with thermodynamic stability, with the ultimate result being increased
longevity and reliability of infrared imaging systems
3.1 Electrochemical
Perhaps the most significant development in alternative passivation

has been another anodic electrochemical process, sulfidization. The
feasibilityof this procedure was first published in 1984by Nemirovskyand

Burnstein (82), although the results were sketchy at this point. Later the
same group revealedmore details of the sulfidization procedure (83). The
electrolyte used was described as a “non-aqueous basic sulfide solution
in ethylene glycol.” Films were grown at constant current densities of
100 k 40 amp-cm‘2and the growth rate was reported on the order of 1 nm
min-’. The material used was p-type, x = 0.215,0.225, and 0.285, but the
crystal growth and surface preparation techniques were not mentioned.
The voltage-time characteristics were reported (Figure 30) but showed
little increase in potential as film growth occurred. The film composition
was originally reported as entirely CdS, the absence of substantial Hg and
Te being rationalized by the relatively high solubilities of the polysulfides,
HgSc and TeS,2. More recent works (84) have indicated that the depletion
of Hg and Cd, although severe, are not as complete as originally suggested.
The sulfide passivation has been shown to have lower fixed charge and
better thermal stability than the anodic oxide, and Nemirovsky,et al. (85),
have reported the device characteristics of a photodiode structure built
using this passivation process.
Little fundamentalelectrochemical information is presentlyavailable
about anodic sulfidization. One work (86) using oxide marker and
compositional profiling found the sulfide growth to be into the substrate,
as the same report concluded for anodic oxidation. The original reports
from Nemirovsky, et at. (83), suggested that the reaction was a six
electron process:
Eq. (5) Hg,,,,C&Te + [ 4 t ( l - x ) ] S2-

+ xCdS t TeSS2- + (1-x)HgS,2- t 6 e
where 0 IXI 1, basedon a CdS film composition and the solubility of Hg
and Te polysulfides mentioned earlier. This was later disputed and a two
electron reaction was suggested (87):
Eq. (6) Hg,,-,,Cd,Te + [ 1 t (1 - x)] HS- t -:S + [1 t (1 - x)] OH-

+ xCdS + (1 - x) HgS22-+ TeS,’- + [ 1 + ( 1 - x)]H20 t 2e-
where05 XI 1
Furthermore, this later work draws a possible connection between the

anodic sulfidization process and photocorrosion of Il-VI solar cells in
polysulfide electrolytes. If this connection could be established,a well of
knowledge could be applied to gain a better understandingof this passivation
process. (For example: Ref 88-90).
Anodic fluoride films have also been produced on (Hg,Cd)Te as an
SPUTTERING TIM[ ISBCI
Figure 29. Concentration of 0l8 (A)and 0 l 6 (V)as a function of

sputtering time in the sequentially grown film (growth current density =
300 pNcm2). The circles represent the 0'8:0'6 ratio for a film grown in
the 0l8anodization solution only. (With permission of American Institute
of Physics.)
io0 r
-
>
E
%IT------ ---
It 12
T I M E (min 1
-50
0
Figure 30. Electrochemical cell voltage as a function of the anodic

sulfidization duration for Hg,-,Cd,Te with x = 0.225. Current density is
130 pNcm2. (With permission of American Institute of Physics.)
alternative to anodic oxidation. Similar voltage-time behavior to that

observed in anodic oxidation was reported, and initial film formation was
concluded to proceed by dissolution-precipitation. The electrolyte used
for film formation was potassium fluoride of various concentrations in
ethylene glycol. The addition of water caused simultaneous incorporation
of oxygen and fluorine in the film, and small additions of hydroxide
produced oxides only. The film composition of the uncorrupted fluoride
was reported as cadmium fluoride within a matrix of mercury, cadmium,
and tellurium. The films were found to have a low positive fixed charge
density with low surface state densities. By varying the oxygen and
fluorine content of the film, the extent of band-bending at the interface
could reportedly be stable to temperatures of 105'.
3.2 Non-Electrochemical Passivation
Other non-electrochemical passivation processes have been explored

and have met with varying levels of success. Thermal oxidation of CdTe
has been demonstrated (91),although it has not successfully been extended
to the mercury-containing semiconductors. Plasma oxidation and
photochemical oxidation (94-97) have been used to produce oxides with
lower fixed charge than anodic oxides. However these passivation
techniques suffer other problems thought to be associated with radiation
damage caused by charged or excited oxygen species involved in the film
formation. Chemical vapor deposition (CVD) of silicon dioxide has also
been demonstrated both directly on (Hg,Cd)Te (98)and on thin native
oxide layers (99)(100). However, SiO, has a transmission cutoff of 7
microns, limiting its use in devices operating in the 8 - 12 microns
atmospheric window to those utilizing backside illumination. CVD deposition
of Cd-Te-0 compounds has been predicted to be thermally stable (1 Ol),
but successful experimental results have not been published to date.
ElectrocliemicalPassivation of (Hg,Cd)Te 169
4.0 REFERENCES
1. Dimiduk, K. C., Opyd, W. G., Gibbons, J. F., Sigmon, T. W., Magee,

T. J. and Ormond, R. D., J . Vac. Sci. and Technol., A 1 (3),pp.
1661-1 665 (1 983)
2. Opyd, W. G., Dimiduk, K. C., Sigmon, T. W. and Bibbons, J. F., J .

Vac. Sci. a n d Technol., A 3 (l), pp. 276-279(1985)
3. Broudy, R. M. and Mazurczyck,V.J., Semiconductors and Semi-

metals, (R.K. Willardson and A. C. Beer, eds.), Vol. 18,pp. 157-199,
Academic Press, New York (1 981 )
4. Sze, S. M., Semiconductor Devices - Physics and Technology,

pp. 278-281, John Wiley and Sons, Inc., New York (1985)
5. Kinch, M.A., Semiconductorsand Semimetals, Vol18,pp. 313-

378,Academic Press, New York (1 981 )
6. Sze, S. M., Physics of Semiconductor Devices, pp. 425-504,

John Wileyand Sons, Inc., New York (1969)
7. Reine, M. B., Sood,A. K. andTredwell, T. J., Semiconductors a n d

Semimetals, Vol. 18,pp. 201 -312,Academic Press, New York (1 981)
8. Sze, S. M., Physics of Semiconductor Devices, pp. 77-149,

John Wiley and Sons, Inc., New York (1 969)
9. Bertagnolli, E., Bettimi, M. and Gornik, E., Appl. Phys. Lett. 42 (9),
pp. 824-826(1 983)
10. Kaiser, U., Ganschow, O., Wiedmann, L. and Benninghoven,A., J.

Vac. Sci. and Technol., A 1 (2),pp. 657-661(1983)
1 1 . Takita, K., Ipposhi, T., Murakami, K., Masada, K., Kudo, H.and Sek,
S., Appl. Phys. Lett., 48 (13),pp. 852-854,March 31,1986
12. Stahle, C. M., Helmes, C. R., and Simmons, A., J. Vac. Sci. a n d
Technol., B 5 (4),pp. 1092-1096(1987)
13. Stahle, C.M., Helms, C. R., Strong, R. L., Simmons, A., Schaake, H.
F., Pallix, J. and Becker, C. H., to be published, J . Vac. Sci. a n d

Techol., A 7
14. Nemirovsky, Y. and Kidron, I., Solid State €/ec, 22 (1 0), pp. 831-
837 (1979)
15. Janousek, B. K., Daugherty, M. J. and Schouler, R. B., SP/€ Infrared

Detector Materials, 285,pp. 118-125(1981)
16. Schoolar, R. B., Janousek, 6. K., Alt, R. L., Carscallen, R. C., Daugherty,
M. J. and Fote, A.A., J. Vac. Sci. a n d Techno/., 21 (l),pp. 172-
177
17. Beck, J. D.,Kinch, M.A., Esposito, E.J. andchapman, R. A., J . Vac.

Sci. and Techno/., 21 (l),pp. 172-177(1982)
18. Anderson, W. W. and Hoffman, H. J., J. Vac. Sci. and Techno/.,

A 1 (3),pp. 1730-1734 (1983)
19. Micklethwate, W. F. H., Semiconductors a n d Semimetals, Vol

18,pp. 48-120,AcademicPress, NewYork (1981)
20. Columbo, L., Syllaios, A. J., Periaky, R. W. and Brau, M. J., J. Vac.
Sci. a n d Techno/., A 3 (1 ), pp. 100-104 (1985)
21. Triboulet, R., Nguyen Duy, T. and Durand, A., J . Vac. Sci. a n d
Techno/., A 3 (l),pp. 100-104(1985)
22. Schmit, J. L., J. Vac. Sci. a n d Techno/., A 3 (l), pp. 89-92(1985)
23. Faurie, J. P., Million, A., Boch, R. and Tissot, J. L., J. Vac. S c i a n d
Techno/., A 1 (3),pp. 1593-1597(1983)
24. Chow, P. 0.and Johnson, D., J . Vac. Sci. a n d Techno/., A 3 (l),

pp. 67-70(1 985)
25. Cheung, J. T. and Magee, T., J . Vac. Sci. a n d Techno/., A 1 (3):

pp. 1604-1607 (1 983)
26. Kannam, P. J., Murosaka, R. and Briere, T.,J . Electrochem. SOC.

129 (3),pp. 656-660(1 982)
27. Lastsras-Martinez, A., Lee, U., Zehnder, J. and Racrah, P. M., J . Vac.
Sci. a n d Techno/., 21 (l),pp. 157-160(1982)
28. Aspnes, D. E. andArwin, H., J . Vac. Sci. and Technol.,A2 (3),pp.

1309-131 5 (1 984)
29. Talasek, R. T. and Syllaios, A. J., J. Electrochem. SOC., 132 (3),

pp. 656-659 (1 985)
30. Talasek, R. T., Ohlson, J. J. and Syllaios, A. J., IRIS Detector

Specialty Group Proceedings (1 984)
31. Carey, G. P., Wahi,A. K., Silberman, J. A., Stahle, C. M., Spicer, W. E.
and Wilson, J. A., J . Vac. Sci. and Techno/., A 5 (5), pp. 3203-
3206 (1 987)
32. Janosek, 6. K. and Carscallen, R. C.,J. Appl. Phys., 53 (3),pp.

1720- 1 726 (1 982)
33. Basol, 6. M., Tseng, E. S. and Rod, R. L., U. S. Patent 4,388,483

(1983)
34. Kazacos, M. S. and Miller, B., J . €/ectrochem. SOC., 127, p. 869

(1 980)
35. Miller, 6. and Rosamilia, J. M., €/ectrochem. SOC.€xtended

Abstracts 86-7, p. 485 (1 986)
36. Baranski,A. S. and Fawcett, W. R., J . Electrochem. SOC., 127(3),

pp. 766-767 (1980)
37. Biter, W. J., Epps, G. F. and Harns, R. D., Electrochem. SOC.

Extended Abstract 86-7, pp. 487-488 (1986)
38. Miller, B., and Heller, A., Nature 262, p. 860 (1976)
39. Peter, L. M., €lectrochem Acta 23, 1965, (1978)
40. Peter, L. M., J . €/ectroana/. Chem. 98, p. 49 (1979)
41. Talasek, R. T., Ohlson, M. J. and Syllalos, A. J., J. Electrochem.

SOC., 133 ( l ) , pp, 230-232 (1986)
42. Catagnus, P. C. and Baker, C. T., U. S. Patent 3,977,018(1976)
43. Robertson, D. S., Shaw, N. and Young, I. M., J . M a t / . Sci., 13, p.

1986-1990 (1 978)
44. Rhiger, D. R. and Kvaas, R. E., J. Vac. Sci. and Technol., A 1 (3),
pp. 171 2-171 8 (1 983)
45. Davis, G.D., Buchner, S . P., Beck, W.A. and Byer, N. E., Appl. Surf.
Sci., 15,pp. 238-246(1983)
46. Seelmann-Eggebert, M. and Richter, H. J., Proceedings of the

78th I n t e r n a t i o n a l C o n f e r e n c e o n t h e P h y s i c s of
Semiconductors. Stockholm, Sweden, Aug 11-15,1986,World
Scientific, p. 283.(1 987)
47. Seelmann-Eggebert, M. and Richter, H. J., Proceedings of the

Optical Society of America, Seattle, Washington, Oct 19-24,
1986
48 Pourbaix, M., Atlas of Electrochemical Equilibria i n Aqueous

Solutions, Celebcor Publication F59,Pergamon Press (1 966)
49. Ohlson, M. J., Talasek, R. T. and Syllaios,A. J., Electrochem. SOC.

Extended Abstracts 84-7, p. 213 (1984)
50. Nemirovsky, Y. and Finkman, E., J. Electrochem. SOC.,126 (5),

pp. 768-770(1 979)
51. Sun, T. S.,Buchner, S. P. and Byer, N. E., J . Vac. Sci. and

Technol., 17 (5), pp. 1067-1073 (1 980)
52. Davis, G.D., Sun, T. S . , Buchner, S. P. andByer, N. E., J . Vac. Sci.

and Technol., 19 (3),pp. 472-476(1981)
53. Nitz, H. M., Ganschow, O., Kaiser, U., Wiedmann, K. and Benninghoven,
A., Surf. Sci. 104,pp. 365-383(1981)
54. Farrow, R. F. C., Dennis, P. N. J., Bishop, H. E., Smart, N. R. and

Wotherspoon, J. T. M., Thin Solid Films 88,pp. 87-92(1982)
55. Morgen, P., Silberman, J. A , , Lindau, I., Spicer, W. E. and Wilson, J.

A., J . Vac. Sci. and Technol., 21 (l), pp. 161-163(1982)
56. Christie, A. B., Sutherland, I. and Walls, J. M., Surf. Sci., 135,pp.
225-242(1 983)
57. Seelmann-Eggebert, M., Brandt, G. andRichter, H. J., J . Vac. Sci.

and Technol., A 2 (l), pp. 11-15(1984)
58. Kaiser, U., Ganschow, O., Wiedmann, L. and Benninghoven, A., J.

Vac. Sci. and Techno/., A 1 (4), p. 657 (1983)
59. Stahle, C. M., Thomson, D. J., Helms, C. R., Becker, C. H. and

Simmons, A., Appl. Phys. Lett., 47 (5),pp. 521-523 (1985)
60. Gnade, B., Simmons, A., Little, D. and Strong, R., Nuclr. l n s f r .and
Meth. Phys. Res., B24/25, pp. 1014-1016 (1987)
61. Strong, R. L., Anthony, J. M., Gnade, 6. E.,Keenan, J. A., Norbeck,

E.,Li, L. W. and Helms, C.R., J . Vac. Sci. and Techno/.,A 4 (4),
pp. 1992-1995 (1986)
62. Kowalczyk, S. P.andCheung, J. T., J. Vac. S c i . and Techno/., 18

(3), pp. 944-948 (1981)
63. Rhiger, D. R. andKvaas, R. E., J . Vac. Sci. and Technol., 21 ( l ) ,

pp. 168-171 (1982)
64. Morgen, P., Silberman, J. A., Lindau, I., Spicer, W. E. and Wilson, J.
A., J . Electr. M a t / . , I1(4), pp. 597-610 (1982)
65. Silberman, J. A., Laser, D., Lindau, I.and Spicer, W. E.,J . Vac. Sci.
and Techno/., A 1 (3), pp. 1706-171 1 (1983)
66. Silberman, J.A., Laser, D., Lindau, I., Spicer, W. E. andWilson,A., J.

Vac. Sci. and Technol., A 3 ( l ) , pp. 222-225 (1985)
67. Wager, J. F. andRhiger, D. R., J . Vac. Sci. and Techno/.,A 3(1),

pp. 212-217 (1983)
68. Daniels, R. R., Margaritordo, G., Davis, G. D. and Byer, N. E.,Appl.

Phys. Lett., 42 (1) pp. 50-52 (1983)
69. Arwin, H., Aspnes, D. E. and Rhiger, D. R., J . Appl. Phys., 54 (12),
7132-7138 (1983)
70. Aspres, D. E. and Arwin, H., J. Vac. Sci. and Techno/.,A2 (2), pp.
600-601 (1984)
71. Janousek, 6. K. and Carscallen, R . C., J . Vac. Sci. and Techno/.,

21 (2), pp. 443-445 (1982)
72. Sakashita, M., Strehblow, H. H. and Bettini, M., J . €lectrochem.

SOC., 129 (4),pp. 739-746(1982)
73. Sakashita, M., Strehblow, H. H.and Bettini, M., J . Electrochem.

SOC.,129 (8),pp. 1710-1716(1982)
74. Uhlig, H.H. and Revie, R. W., Corrosion and Corrosion Control,
3rd Ed., pp. 60-89,John Wiley and Sons, Inc., New York (1985)
75. Ahearn, J. S.,Davis, G. D. and Byer, N. E., J . Vac. Sci. and

Technol., 20 (3),pp. 756-759(1982)
76. Talasek, R. T., Ohlson, M. J. and Syllaios, A. J., J. Vac. Sci. and
Techno/., A 3 (l), pp. 218-221 (1985)
77. Dean, S.W., Electrochemical Techniques f o r Corrosion, (R.

Baboian, ed.) NACE pp. 52-60(1 977)
78. Bard, A. J. and Faulkner, L. R., Electrochemical Methods, pp.

280-311 , John Wiley and Sons, New York.
79. Talasek, R.T., Electrochem. SOC.ExtendedAbstracfs 85-I , p.

241 (1985)
80. MacDonald,D. D., Transient Techniques in E/ecfrochemistry,

Plenum Press, New York (1 981)
81. Lopes,V.C.,Hart,T.R.,J.Vac.Sci.andTechnol.,A5(2),pp.174-
178 (1987)
82. Nemirovsky,Y. and Burstein, L., Appl. P hys. Left., 44 (4),

pp. 443-
444 (1 984)
83. Nemirovsky, Y., Burstein, L. and Kidron, I., J. Appl. Phys., 58 (l),
pp. 366-373(1 985)
84. Strong, R. L., Luttmer J. D., Little, D. D., Teherini, T. H. and Helms, C.
R., J . Vac. Sci. and Techno/., A 5 (5),pp. 3207-3210(1987)
85. Nemirovsky, Y. Adar, R., Kornfeld, A. and Kidron, I., J . Vac. Sci.
and Techno/., A 4 (4),pp. 1986-1991(1986)
86. Strong, R. L., J . Vac. Sci. and Technol., A 5 (4),pp. 2003-2006

(1 987)
87. Ziegler, J. P. and Hemminger, J. C., Electrochem. SOC. Extended

Abstracts.
88. Hodes, G. and Miller, B., J. Electrochem. SOC.,122 (lo), pp.

2 177-2180 (1986)
89. Licht, S.,Tenne, R., Flaisher, H. and Manassen,J., J . Electrochem.

SOC.,133 ( l ) , pp. 52-59 (1986)
90. Licht, S. and Manassen, J., J . Electrochem. SOC., 132 (5),pp.
1076-1081 (1985)
91. Wang, F., Schwartzman,A.,Fahrenbruch,A.L., Sinclair, R., Bube, R.

H. and Stahle, C. M., J . Appl. Phys., 62 (4), pp 1469-1476 (1987)
92. Weiss, E. and Mainzer, N., J. Vac. Sci. a n d Techno/., A 6 (4), pp.
2765-2771 (1988)
93. Mainzer, N., et. ai., to be publishedJ. Vac. Sci. a n d Techno/., A 7
94. Nemirovsky, Y. and Goslen, R., Appl. Phys. Lett., 37 (9), pp. 813-
815 (1980)
95. Nemirovsky, Y., Goshen, R. and Kidron, I., J. Appl. Phys., 53 (7),
pp. 4888-4895 (1982)
96. Buchner, S. P., Davis, G. D. and Byer, N. E., J . Vac .Sci. a n d

Techno/., 21 (2), pp. 446-447 (1982)
97. Janousek, 6. K. and Carscallen, R. L., J . Vac. S c i . Techno/., A 3

( l ) , pp. 195-199 (1985)
98. Kao, T. M. and Sigmon, T. W., Electrochem. SOC. Extended

Abstracts.
99. Davis, G. A., Buchner, S. P. and Byer, N. E., J. Vac. Sci. a n d

Techno/., A 1 (2), pp. 670-671 (1983)
100. Wilson, J. A. andcotton, V. A., J. Vac. Sci. a n d Techno/., A 1, (3),

pp 1719-1722 (1983)
101. Rhiger, D. R. andKvaas, A. E., J. Vac. Sci. a n d Techno/., 21 (2),

pp. 448-452 (1982)
102. Janousek, B. K., Carscallen, R. C.and Bertrand, P. A., J. Vac. Sci.

and Techno/., A 1 (3), pp. 1723-1725 (1983)
103. Diehl, R., Nolang, B. I . , J . C r y s t . . Growth, 66, pp. 91-105 (1984)

4
PHOTOELECTROCHEMICAL PROCESSING OF
SEMICONDUCTORS
R. David Rauh
1 .O INTRODUCTION
Photoelectrochemistry is a general category encompassing light-

induced electrochemical reactions of semiconductors in contact with
liquid electrolytes arisingfrom the primary generationof minority carriers.
In the field of electronics and electro-optics, photoelectrochemical
processing has been a subject of considerable recent interest (1).
Photoelectrochemical etching has been used in the fabrication of diffraction
gratings, integral lenses for light emitting diodes, fiber optic couplers, and
via holes in a variety of semiconductor materials. Photoelectrochemically
induced electroplating has been employed for both ohmic and non-ohmic
contacts. Photoelectochemistry has also been used in the modification
of semiconductor surfaces to reduce surface state densities prior to
device fabrication.
When immersed in an electrolyte, a semiconductor undergoes an
exchange of electrons with the liquid at the interface to equalize the work
functionsof thetwo phases. The result isoften arectifying barrier between
the semiconductor and liquid that has properties similar to a Schottky
barrier. Like semiconductor-metal contacts, the barrier height can be
fixed by the semiconductor surface state distribution, or, in the case of a
low density of surface states, by the difference between the work functions
of the bulk semiconductor and the liquid. Similarly, a reaction between
the semiconductor surface and the electrolyte can produce a surface
layer that fixes the barrier height.
The contact barrier that is produced is characterized bya bendingof
the semiconductor’s conduction and valence bands. Such barriers are
177
well known to be photoactive. Thus, when light of energy greater than the
bandgap strikes the interface and is absorbed by the semiconductor,
electrons and holes are produced, with minority carriers being swept to
the semiconductor surface. In the case of n-type semiconductors, these
minority carriers are valence band holes which, upon arrival at the interface,
can be consumed by a chemical reaction with the electrolyte or by (non-
productive) surface recombination with majority carriers. For most
technologically significant semiconductors (Si, Ge, Ill-V and ll-VI
compounds), valence band holesare suff iciently energeticto decompose
the semiconductor in aqueous media. This phenomenonforms the basis
for photoelectrochemical etching/micromachining processes. In the
case of p-type semiconductors, photogenerated minority carriers at the
interface are conduction band electrons. Many semiconductors are
stable toward reduction by such electrons. However, metal ions added to
the electrolyte can, in some cases, be reduced to the parent metal, leading
to localized electroplating.
Light-localized, wet processing of semiconductors which is not
photoelectrochemical in nature is also known. Localized surface heating
can reduce the overpotential for electroplating (2) or accelerate etching,
for example. In addition, reactive intermediates can be produced by
photolysis of electrolyte species (e.g., BrJ, which then go on to react
chemicallywith the semiconductor surface (3). These nonphotoelectro-
chemical approaches will not be dealt with further in this chapter.
2.0 EXPERIMENTAL PROCEDURES
There are three major components in an apparatus for

photoelectrochemical processing. First is the cell for exposing the
semiconductor working surface to the electrolytewhile also accommodating
auxiliary electrodesfor controlling the semiconductor potential. Second
is the control apparatus, typically a potentiostat and a means for monitoring
current and charge. Finally is a light source, which can be broad band or
monochromatic laser radiation.
The choice of a cell will be largely dependent on the type of processing
to be accomplished. An example used in our laboratory for etching
diffraction gratings is illustrated in Figure 1. The cell body should be
machined from a polymer stable to the electrolyte of choice. Examples
are nylon for many aqueous electrolytes and Teflon for HF-basedetchants
or nonaqueous media. The semiconductor electrodes must be mounted
in such a way to prevent contamination of the ohmic contact (or other
corrosive elements) by the electrolyte. In the approach illustrated in
Photoelectrochemical Processing of Semiconductors 179
ILLUMINATION
r---------- Window
----------- R e l e r e m
:lor
act
Countor Ekctrodt e-------
Brass Dit&
PI MajhCylind~r
- Kel-FHdder
+=
Electrolyte
Ill
\
\
\
\
\
\
\
' Retaining Screw
Figure 1. Cell design and semiconductor mounting for laboratory

investigations of photoelectrochemical processing of semiconductors.
Figure 1, the semiconductor “working” electrode incorporates an ohmic

contact on the back side. The electrode is mounted onto a brass disk
using conductive silver-filled epoxy. The assembly is potted into a Kel-F
holderwith epoxy cement, save for a threaded hole in the rear of the brass
disk. A screw assembly is used both as a contact and to tighten the
electrode onto the cell via a liquid-tight O-ring seal. A cylindrical platinum
gauze counter electrode surrounds the working electrode, giving rise to
a uniform current distribution. A standard calomel (SCE) reference
electrode is placed as close as possible to the electrode surface without
occluding the light path.
When photoelectrochemical processing is to be used on a production
scale, the cell will be of utmost importance. In that case, special holders
would be necessary to incorporate multiple wafers. Nevertheless, the
basic design philosophy illustrated in Figure 1 should be retained.
The control apparatus would typically consist of a potentiostat with
a sufficiently large current output to support the total photoelectrochemical
process. Most examples of photoelectrochemical processing are carried
out at constant potential, but some examples exist of galvanostatic
control. The course of the photoelectrochemical process is monitored as
the integrated current, which may be related directly to the amount of
etched or deposited material. A coulometer or integrating computer may
be used for this purpose.
The light source for photoelectrochemical processing must put out
photons of energy greater than the bandgap of the semiconductor.
Lasers have the advantage of monochromaticity,thus permitting accurate
calculation of the photon flux from the intensity and the reflectance and
depth of penetration from the semiconductor’s optical constants. However,
since cw sources are usually desirablefor processing, lasers of sufficient
intensityfor uniform processing of large areas at a reasonablerate may be
prohibitively expensive compared to broadband sources. Lasers are the
sources of choice for etching interferograms (e.g., diffraction gratings)
into semiconductors, or for processes requiring a highly focused beam.
Broadband sources are frequently used for illumination of a photoresist
patterned electrode or for projection imaging. Broadband sources may
be obtained in standard illuminator configurations, as might be employed
for the exposure of photoresist. Table 1 summarizes the spectral
characteristics of several commonly employed light sources while the
wavelength equivalent of the bandgaps of some technologically significant
semiconductors are included in Table 2.
Table 1
Light Sources for Photoelectrochemical Processing
Source Spectral Range Major Lines

(nanometers) (nanometers)
NONCOHERENT SOURCES
Quartz Tungsten- 300 to > 2000 BB

HaIogen m a . at 900
Deuterium 180 to >400 BB, 480,660
Hg Arc, low P 180-350 185,253
Hg Arc, med. P 300- >800 313,365,430,540,580
Hg Arc, high P 300->800 same, BB (high intensity)
Hg/Xenon 250-2000 same, BB
Xenon 350-2000 BB
COHERENT SOURCES (CW)
Ar ion Blue-Green 351, 457.9, 476.5

488.0, 51 4.5
257 (doub)
Kr Red 530.9,647,676
He/Cd Blue-green 345,442
He/HeNe Red 543,632.8,1150
Semiconductor Red, NIR 750,780,820,1300,1550
Nd/Yag Green 532,1064,1318
Dye Tunable 350-950
BB: broad band

Table 2
Properties of Candidate Semiconductors

for Photoelectrochemical Etching
Semiconductor Bandgap (nm) n (e-/eq) Rate(pm/sec)

XI 0'4x1
BaTiO, 388 4,6 0.56

CdS 512 2 1.25
CdSe 729 2 1.38
CdTe 795 4 0.81
GaAs 873 6 0.38
GaP 549 6 0.34
Ge 1771 4 0.29
InP 919 6 0.42
Si 1107 4 0.25
Sic 41 3 8 0.13
TiO, 400 4 0.39
ZnO 370 2 0.61
ZnS 337 2 0.99
It is also important to note that many cases may be cited of

photoelectrochemical processing in which control is lacking, such as
using a two electrode arrangement (no reference electrode), or a non-
potentiostatic power supply. Similarly, many examples exist of photo-
electrochemical processing (primarily etching) of semiconductors simply
immersed in the electrolyte without external contact (1). Indeed, these
may be practical solutions to photoelectrochemical processing once the
systems have been characterized electrochemically.
3.0 P HOTOELECTR OCHE MICAL ETCHI NG

3.1 General Background
Photoelectrochemical etching occurs only when an optically produced

minority carrier at the semiconductor-electrolyte interface is sufficiently
energetic to inducea corrosion reaction. This condition is metfor valence
band holes for virtually all semiconductors used in electronics in contact
with aqueous electrolytes (4). In order for etching tooccur at a sufficiently
practical rate, kinetics must also be favorable. Processeswhich consume

minority carriers without corrosion, such as surface recombination and
reaction with dissolved electrolyte species, will detract from the overall
quantum efficiency of etching by kinetic competition. Mass transfer can
also kinetically inhibit the etching reaction. It is particularly important in
this regard to employ an electrolyte in which the corrosion products are
highly soluble. Examples of conditions which have been reported for
photoelectrochemical etching of a variety of semiconductors are given in
Table 3.
In a photoelectrochemical reaction with unit quantum yield, each
photon absorbed by the semiconductor will give rise to one electron in the
dissolution process. In a well-behaved photoelectrochemical cell, the
curve of photocurrent (or quantum yield) vs. applied potential is S-
shaped, as shown schematically in Figure 2. As indicated by the band
bending diagrams in Figure 2, when the potential is negative of the flat
band value (Vrb)for an n-type semiconductor, the bands are flat and there
is little electron-holeseparation, and hence negligible photocurrent. As a
positive potential is applied, the bands become bent due to a surface
depletion of electrons. A space-charge layer is thus formed, the width and
magnitudeofwhich growsas the potential becomes more positive. Above
a certain potential, the space-charge region reaches a limitingwidth. The
efficiency of separation of electron-hole pairs, and thus photocurrent
production, also increases with voltage due to the fact that more pairs are
created within the space-charge region and the driving force for their
separation increases. In some semiconductors, typically direct bandgap
materials like GaAs, only carriers produced within this space-charge
region are eff iciently separateddue to relatively low bulk diffusion lengths.
Indirect bandgap semiconductors with long diffusion lengths, like Si,
derive significant photocurrent from photons absorbed well beyond the
space-charge region. The plateau region of highest photocurrent quantum
yield is often referred to as the. -p
The beauty of photoelectrochemical etching is that the rate of etching
can be controlled by the photon flux and by the potential, due to the linear
relationship between light intensity and photocurrent. Since an ampere
is equal to a current of 6 . 2 4 ~O1l e electrons/sec, it can be shown that the
current equivalentof photons isequal to 8.065~1 05XI“photon amperes”.
The quantities X and I are the wavelength (angstroms) and intensity
(watts/cm2) of the light source. The rate of etching may be expressed
quantitatively as
Eq. (1) r(pm/sec) = 8.36x10-‘ (1 - R) 41 A I (mlnp)
where r is the normal incidence reflectance of the crystal, and 4 is the

cb
v ....
L.Y.-.ILyyu
1
+
vb
bV
Light
Figure 2. Idealized current-potential curves and band bending diagrams

for n- and p-type semiconductor electrodes in the dark and under illumination.
Conduction and valence bands are indicated (cb, vb). Voltage is applied
via an ohmic contact at the Fermi level, located in the band gap near the
cb (n-type) or vb (p-type).
quantum yield of conversion of photons to current in the dissolution

reaction. The quantity (mlnp) describes the etching reaction, where m is
the molecular weight of the dissolving layer, p is its density in g/cm3, and
n is the number of equivalents/mole in the dissolution process. The
relative rates of etching of some important semiconductors are compared
in Table 2, the differences arising from differences in density and the
stoichiometry of the corrosion process.
Table 3
Conditions for Photoelectrochemical Etching
Reported for Various Semiconductors.
Material Conditions Electrolyte Application Ref.
GaAs Ar+(488,513nm) 1 : 1.3:25 H,SO,: H,O,: holes, hologra- 21

SI HeNe(633nm) HZO phic gratings
lo.,- l o 3W / c m 2
n-GaAs Ar' (514,457nm) 1 : 1 : 100 H,SO,:H,O, holographic 45
(100),[01 11 300 m W / c m 2 HZO gratings
1OI8/cm3
n-GaAs A r t (514,257nm) 1 :20 HNO, holes. slits 22

var. faces 1 OmW-1k W / c m 2
focused
n,p,SI-GaAs A r t (257nm) 1 : 1 : 100 H,SO,: H,O,: holes, vias 10

1 OW/cm* H.0
focused 1 ;20(v) HNO,:H,O
1 :20(w) KOH:H,O
n-GaAs Ar' (488 nm) 1 : 1 : 50 H,SO,: H,O, holographic 11

(100).[011] 50mW/cmz H,O gratings
n-GaAs HeNe (633nm) 10% KOH character 24

(1 00) 16W/cm2 - writing
1 . 5 1~Ol8/cm3 16kW/cmz
n-GaAs 150W Xe 0.5M Tiron gratings, 14

(1 00) - 20mA/cm2 deep grooves
1017- 0.2V (SCE)
10i8/cm3 Metal mask
n-GaAs 150W Xe 0.5M Tiron symmetric 16

(100),[011] -20mA/cm2 Echelle gratings
-- Continued --
Table 3 (Continued)
Material Conditions Electrolyte Application Ref.
1017 - 0.2V (SCE)

1 O18/cm3 Photoresist mask
n-GaAs Hg-Xe,30mW/cmz 0.1 M KCI blazed Echelle 18

off ( 1 00) 0 . 4 v o l t (SCE) gratings
5x 101'/cm3 Photoresist mask
p-GaAs (11 1 ) HeNe (633nm) 0.1M H,SO,/O.lM holes 5

3-5x10I7/ 0.75W/cm2 NaSCN (and others)
cm3 -0.2, t 0.25V(SCE)
p-GaAs A r t (257nm) 1 : 1 :50 H,SO,:H,O,: holographic 13

(100),[011] 100mW/cm2 H,O gratings
9x10"/cm3
n-lnP Microscope lamp 40%(w) FeCI,/H,O mesas 25

(1 00) Au mask
i0'8/cm3
n-lnP 1OOWW-halogen 2M HF/0.5M KOH integral LED 28

(1 00) O.l-0.4vOlt (SCE) lenses
var. N, projection
n-lnP HeNe, Ar' 1 1 10 H,SO,:H,O,:

~ holographic 7,8
0 . 2 v o l t (SCE) H,O, etc. gratings
n-lnP W lamp 2M HF/0.5M H,S04; mesas, gratings 26

2x10"/cm3 ion beam mask 2M H,SO,
epi-InGaAs 0.25V (SCE)
epi-InGaAsP -O.l,m/min.
2~ io16/cm3
n-lnP Ar'(488nm) 1:1:20 HCI:HNO,:H,O holographic 12

(100),[011] 0 . 7 v o l t (SCE) gratings
6~ i018/cm3
n-Si Ar' (51 4 nm) 5% HF holographic 47

- 5~ i014/cm3 gratings
n-Si A r t (257 nm) 5 % HF fiber optic 23

focused coupler
n-Sic Hg-Xe arc 1M HF characters. 48

epi on Si; 1 volt (SCE) gratings
poly (CVD) Photoresist mask
Ostermayer et al. (5) have reportedthe photoelectrochemical etching

of p-type Ill-V compounds. Under cathodic polarization, the photogenerated
minority carriers (electrons) give rise to formation of the Ill element on the
surface along with the evolution of the group V hydride (e.g., Ga and
ASH,). Subsequent anodic polarization leads to dissolution of the surface
film by a majority carrier mediated reaction. As this process is repeated,
the light-localized etched profile is developed.
It has been noted that light can accelerate etching of n-type
semiconductors in oxidizing media, so that photoetching can be carried
out without electrochemical control (1). In this “corrosion mode”, the
light produces minority carrier holes at the interface, while the charge
compensating electrons are swept by the space charge field to adjacent
non-illuminated regions. Photoselective etching can be maintained if
reaction paths are availablefor the oxidation/dissolution of the lattice by
holes and also the (non-etching) consumption of the compensating
electrons. The oxidation and reduction currents must be balanced, so
that the rate of either reaction can limit the overall photoetching rate.
Electrolytes for photoselective etching may contain a weak or dilute
oxidizing agent to accelerate reaction of conduction band electrons, or
the reaction can be facilitated on a catalytic ohmic contact region of the
crystal. Typically photoselective etching of this sort is conducted in
known chemical etchants which have been diluted sufficiently to retard
the dark reaction.
3.2 Periodic Structures
One of the most useful aspects of photoelectrochemical etching is

the production of diffraction gratings and other periodic structures in
semiconductor crystals for optical and electro-optical applications. Because
of the different manner in which they are produced, these may be classified
as holographic and non-holographic structures.
Holographic Gratings. A diffraction pattern is produced when two
coherent, monochromatic beams are allowed to interfere. Today, the
light source is almost always a laser. Many optical arrangements are
possible to produce the desired interferogram, and the reader is referred
to the excellent review by Hutley (6) for a complete discussion of the
options. A typical optical arrangement is shown in Figure 3. The beam is
first spatiallyfiltered, expandedandcollimated. The beam is then split into
two separate beams of equal intensity, then recombinedat the interference
angle, 6 . What is produced is a sinusoidally modulated pattern of light and
dark fringes with a spacing, d, given by:
Eq. (2) d = A,/ 27 sin 6 cos S

Laser rn n II
SF
L c==,
Figure 3. Arrangement for photoelectrochemical etching of interfero-

metric gratings in semiconductor electrodes. Also shown is a weak or
non-absorbed probe beam for monitoring the etching process. BS =
beamsplitter; SF = spatial filter; L = collimating lens; M = mirror.
Here, Xo is the source wavelength, q is the refractive index of the medium

(electrolyte), and S is the angle of the substratewith respect to the bisector
or the interfering beams (usually OO).
There are several practical matterswhich must be confronted in the
photoanodic etching of holographic images. The laser generally must
have a large coherence length. Spectroscopic gratings are usually
fabricated using either an Ar ion or Kr ion source operating at 457.9 nm or
351 nm, respectively. These lasers must be equippedwith a Fabrey-Perot
etalon to produce the narrow spectral distribution required for a large
coherence length. If such lasers are unavailable, then special care must
be taken to equalize the pathlength of the two interfering beams to within
the coherence length. Another consideration is that the time of exposure
must be as short as possible to minimizethe effects of vibration (vibration
damped optical tables for holography should be used). Thus, higher
intensity lasers and small grating areas are most favorable; the exposure
time may be calculated from Eq. 1 , To obtain gratings with high frequencies,
it is necessary to increase B , although -6OOis considered to be a practical
limit. Using an ultraviolet line of an Ar or Kr ion laser, it appears possible
to etch holographic gratings with a pitch of 6000 - 7000 lines/mm.
The resolution obtainable for etching holographic gratings in a
crystal is a function not only of the optical arrangement, but also of
semiconductor transport parameters and the kinetics of the dissolution
reaction. The material parameters which influence the resolution have
been summarized by Ostermayer and co-workers (7). These are:
- The space-chargewidth, w, which is related tothe majority carrier
density, N, and the potential drop, cp ,across the space charge region by
the equation:
Eq. (3) w = (2 ecp / q ND)'"
where E is the dielectric permittivity of the semiconductor and q is the unit

charge.
-The diffusion constant of minority carriers, D, which is related to
their mobility, p ,by the Einstein relationship:
-The diffusion length of minority carriers, L, which is determined by

D and by their lifetime: L = [ D TI'/*
- The absorption coefficient of the semiconductor at the source
wavelength, a (cm-').
-The interfacial reaction velocity K (cm sec-').
The grating figure of merit is defined as the ratio of the grating

amplitude to the average depth etched, and may be evaluated from
micrographs of grating cross sections and from a knowledge of the
charge passed per unit area during the fabrication process. The maximum
value of this ratio is unity. A number of useful relationships between
grating quality and the material and electrolyte parameters were developed.
The following qualitative conclusions can be drawn from both the experiments
reported for n-lnP and the theoretical calculations:
1 . A high value of K favors high resolution, since if holes are
consumed more rapidly at the surface, they will have less opportunity to
diffuse laterally.
2. If K is high, then the resolution will be dependent ona, while if K is
low, as is most often the case for systems investigated to date, then the
resolution becomes a-insensitive. In the former case, lateral diffusion in
the bulk from the point of hole generation to the surface dominates the
resolution loss rather than surface diffusion.
3. All else being constant, increasingw, such as by decreasing N,
degrades the resolution. The decreased field in the space charge region
causes photogenerated holes to accumulate deeper into the surface,
making them less available for reaction and more susceptible to lateral
diffusion.
4. The quantum yields of etching many Ill-V compounds are often
near unity at the photon limiting potentials. Thus, under these conditions,
there is very little effect of bulk recombination, and hence little effect of
minority carrier diffusion length. However, higher diffusion coefficients,
D, decrease the overall resolution.
In a practical sense, the studies of Ostermayer et al. allow selection
of material properties which favor high grating resolution. The reaction
velocity is a dominant effect, and rapid dissolution of oxidation products
is necessary. Furthermore, a low lateral diffusion of surface holes is
fostered by high doping density and low D. If doping densities are too
high, however, then anodic dark currents may become problematic.
Numerous examples have been provided in the literature of gratings
and holograms fabricated in a variety of semiconducting materials. A
compilation by Gurevich and Pleskov (1 ) of examples primarily from the
Russian literature of the 1970’s cites gratings etched by both controlled
potential and light stimulated corrosion techniques. Diffraction gratings
have been produced in elemental semiconductors, Il-VI and Ill-V
semiconductors and alloys, and in oxides. Spatial frequencies of 1000 to
6150 /mm were reported in thisearlierwork. Later moreextensive studies
by the group at Bell Laboratories were all conducted under potential
control (8)(9). These studieswere aimed at fabrication of gratings for use
in distributed feedback lasers, and dealt primarily with InP and GalnAsP
alloys. Reaction velocities obtained from the resolution of gratings

etched in n-lnP are reported in Table 4. Podlesnik et al. (lo),employing
frequency doubled Ar' ultraviolet laser excitation, etched holographic
gratings in n-GaAs using light stimulated corrosion in 1 :1 :lo0
H,O,:H,SO,:H,O. The finest gratings had a spacing of 170to 200 nm and
a depth of 35 - 40 nm. The best depth-to-space ratio was 0.8 for a 2 p m
grating. The degradation of depth-to-space ratio, which is directly
proportional to the grating efficiency, is due to the factors limiting grating
resolution, discussed above. Matz (1 1)( 12) demonstrated a two step
process for producing high aspect ratio gratings in n-GaAs and n-lnP
usingasimilaretchant. In thefirststep, thecrystalof (1 00)orientationwas
illuminatedwith the diffraction pattern along the [Ol 71direction, resulting
in ashallow profile having the shape shown in Figure 4. Following this, the
etching was allowed to continue in the dark where the profile became
more triangular in shape due to exposure of slow etching (1 1 1)Ga planes.
Thus, the two step processwas not effective for other crystal orientations.
Gratings with 330 nm groove spacing and 150 - 170 nm depth were
prepared in this way. Matz (1 3) has also reported gratings etched in p-
GaAs using the aforementioned method of Ostermayer et al. ( 5 ) . Structures
with a spacing of 230 nm and a depth of 100 nm were reported using 257
nm light from a frequency-doubled Ar ion laser source. Crystalline
orientation was seen to affect the groove profile here as well, as has been
observedfor photoelectrochemical chemical etching using a photomask
(see below).
Table 4
Reaction Velocities for Photoelectrochemical Etching of InP

in Various Electrolytes (Ref. 7 ) .
Electrolyte Reaction Velocity, K (cm sec-')
2M HCI 890
2M HF:0.5M KOH 63
2M HBr 1500
1 : 1 : 10 H,SO,: H20,: H,O 2500
1 :1: 100 H,SO,:H,O,:H,O 15
1 : 10 H,SO,:H,O 1500
A detailed analysisof the effect on resolution of etching with external

control versus light enhanced corrosion has not been made. Qualitatively,
results seem similar for both techniques applied to Ill-V compounds.
Since the process can be followed optically, e.g., with a weak secondary
probe beam (9),coulometric monitoring may not be absolutely necessary.
Certainly the polarization of the crystal by external means often yields the
advantages of using a neutral or mild electrolyte and higher effective
quantum yields. At this time, more work is required to assess the effect of
externalfabrication variables on the actual performance of the gratings in
lasers and spectroscopic instrumentation.
Non-holographic Periodic Structures. The general approach for
making periodic structures of non-sinusoidal cross section has been to
employ a mask to define the photoetched pattern. The semiconductor is
then photoelectrochemically etched in the exposed regions, after which
the pattern mask is removed with a suitable solvent. In principal, the
etching will occur vertically with the edges sharply defined by the edges
of the mask. However, crystallographicorientation, the masking material,
the electrolyte and the bulk semiconductor properties can have profound
effects on the result.
Most of the results available at this time refer to Ill-V compounds,
particularly n-GaAs. This materialexhibits a strong orientational effect on
the photoanodic etching process. Orientational effects of photoanodic
etching of n-GaAs have been examined in this laboratory using (1 00)
crystals with a mask defining slots oriented along three primary directions,
as shown in Figure 5. In most electrolytes, the Ga-rich crystallographic
surfaces are preferentially exposed in grooves defined in the (100) surface
along the [ O l T ] direction (14)(15). The other directions exhibit more or
less vertical etching. The results are quite similar to oxidative chemical
etching processes in this regard, e.g., by Br,/methanol.
Effects of both orientation and doping on etching deep periodic
trenches in GaAs are shown in the photographs in Figure 6. These
structureswere etchedata biasof 0.5voltvs. SCE, anda broad band light
intensity sufficient to yield a dissolution current of 10 mA/cm2 in the Tiron
electrolyte (4,5-dihydroxy-l,3-benzenedisulfonic acid, a Gaf3 complexant).
The masking materialwas an adherent Cr/Au bilayer, patterned by liftoff.
The pointed bottoms of the grooves etched along the [OlT] direction
illustrate the exposure of the Ga-rich planes. Higher doping densities
decreasedthe degree of undercutting of the mask, as shown in the bottom
photograph. The reason for this undercutting is still uncertain, but may
relate to a greater propensity for diffusion of holes into the unilluminated
region beneath the growing lip of the metal mask. This explanation would
be consistent with the model of Ostermayer et al. (7), which predicts
greater lateral diffusion of holes as N, is reduced.
(al tbJ
Figure 4. SEM photographs of 330 nm gratings after (a) holographic

illumination, and (b) anisotropic dark etching (from Matz, ref. 12). Copyright
1986, IEEE.
Figure 5. Patterning of (100) n-GaAs crystal surface for groove etching

experiments.
Figure 6. Cross sections of grooves photoelectrochemically etched in

(100) n-GaAs. (0.5 M Tiron; 442 nm, 20 mW/cm2; 0.5 volt vs. SCE). Top:
[0011 direction, No = 3.2 x 1017/cm3; middle: [0111 direction, No = 3.2 x
1017/cm3; bottom: [0111 direction, No = 1.3 x 1018/cm3.Groove width is 10
pm (from Carrabba et al., ref. 15). Copyright 1987, The Materials
Research Society.
Photoelectrochernical Processing of Semiconductors 195
The crystallographic orientational effects on the photoanodicetching

profiles are particularly useful in fabricating Echelle-type diffraction gratings
(16 ) . The photoelectrochemical method permits the fabrication of these
gratings with a high degree of control of lateral uniformity and of processing
conditions leading to optimal smoothnessof the groovewalls. In general,
photoresist masking materials have been employed, e.g., Shipley 1350J
positive resist. The evolution of formation of the “teeth” is shown in Figure
7, in which the Ga faces become exposed and the mask becomes undercut
until it floats off as the apex is formed. The interior angle of the V-groove
can be influenced by the electrolyte composition,which in turn determines
the relative rates of dissolution of the Gat3 and Asf3 oxidation products
(17). For example, in KCI electrolytes, the angles are close to the
theoretical value of 70.54’ for low KCI concentrations, but become
significantly wider at higher concentrations. Apparently, high CI- enhances
Gaf3dissolution and the exposed planes have more of a mixed composition.
In order to produce blazed Echellegratings, it is necessary to cut the
GaAs crystal at an angle off the (100) plane toward the (01 1) plane (18).
As shown in Figure 8, orienting the photoresist lines in the [01T I direction
then still gives rise to structures with the interior angles governed by the
preferred Ga-rich surfaces. One advantage of photoelectrochemical
etching for producing these structures is that the process can be followed
coulometrically. The charge, Q ,required to etch the V-groove sawtooth
pattern is:
Eq. (5) tCOt(Q + p ) )

Q(C/cm2) = 3.54x103nN(0.5W2/cot(~-p)
Here, n is the electron stoichiometry (equivalents/mole) of the photoanodic

dissolution reaction, W is the width (cm) of each groove, Q is the angle of
the groove face with respect to the (100) surface, p is the angle of the
crystal slice with respect to the (100) surface, and N is the number of
grooves/cm.
An example from this laboratory demonstrates the
photoelectrochemical etching of blazed structures. Crystals of n-GaAs
(N, = 5x1Ol7/cm3)were cut and polished with (100), (100)-8Oand (100)-
18’ orientations. The other parameters in Eq. 5 were: n = 6, N = 500 and
W = 2x1 0*3cm. The electrolyte composition was 0.1 M KCI, adjusted to pH
3, and the light intensity was 30 mW/cm2. The potential was held at the
onset of the photon limited region, 0.4 volt vs. SCE. Initial structureswere
etched under the assumption that the interior angles were 70.54O, as
defined by the (1 11) Ga surfaces. However, we found under closer
examinationthat this angle was dependent on the etching conditions and
on the electrolyte, and was closer to 90’ under the present conditions.
This would correspond most closely to the (223)Ga-rich surface. With a
Figure 7. Scanning electron micrographs showing cross sections of

grooves photoelectrochemically etched in ( 100) n-GaAs, [011] direction,
at three stages of development ( 16) .Copyright 1986, The Optical Society
of America.
Figure 8. Schematic cross section of V-grooves arising from exposed

[ 111 ]Ga planes for GaAs cut at various angles off the ( 100 ) plane.
90’ interior angle, Eq. 5 predicts a charge of 10.6 and 8.6 C/cm2 required
to etch the gratings in the (100) and (1OO)-18°surfaces,respectively. With
coulometric monitoring, both unblazed and blazed gratings were produced
with pointed tops and bottoms and extremely smooth walls. A scanning
electron micrograph of the blazed structure from the (100)-18O surface is
shown in Figure 9. The blaze angle of 60° is slightly less than the expected
value of 63O, an error probably due to inaccuracies accumulated in the
cutting and polishing procedures.
Cummings et at. (19) have reported the use of a focused gallium ion
beam to produce damaged regions in n-GaAs which act as a mask for
photoelectrochemical etching due to a high rate of electron-hole
recombination. After exposure, the undamaged GaAs was
photoelectrochemically etched in 2 M H,SO, at -0.2 volt vs. SCE. Ion
beamscan befocusedtowell belowO.l pm. However,when0.2pm beam
was used to draw the mask, the resolution of the photoetched features
could be made no better than about 1 pm. The reasonwas determinedto
be the extremely low ion dose needed to deactivate the GaAs surface -
only about 1 ion per 1000 angstrom’. Hence, the gaussian tail of the ion
beam can deactivate the GaAs far from the beam center.
3.3 Focused Laser and Related Techniques
A focused laser may be used for “drawing” or for “drilling”

microstructures into semiconductor surfaces. In these cases, the laser
beam is focused into a spot onto the crystal surface. Specific patterns
may be generated by computer-controlled scanning techniques.
When using a focused laser beam to etch semiconductors
photoelectrochemically,it is important to consider the limiting light intensity
to preclude significant localized heating. In many situations, heatingwill
accelerate an etching process and hence degrade selectivity and resolution.
The approximate maximum temperature rise caused by an absorbed cw
laser beam has been derived by Lax (20) as:
Eq. (6) , ,T = P/27r1/’ n W
where P is the nominal laser power, nis the materialthermal conductivity,

and W is the beam width. For example,, T , caused by a 5 p m diameter
beam on GaAs (n = 0.46 W/cm deg) will be 1240 P degrees. In reality, the
temperature rise will be somewhat less due to the finite depth of penetration
of the light. Note that a 1 mW laser will produce a power density of 4000
W/cm2 under these conditions, and a, ,T of only 1.24OC.
Osgood and co-workers at Columbia University (10)(21)(22) have
investigated laser hole drilling in crystals of Ill-V compounds simply
Figure 9. SEM profilesofa60O blazed grating etched in (100)-180 cutn-

GaAs ( 18). Reprinted by permission of the publisher, The Electrochemical
Society, Inc.
immersed in etching electrolytes. As mentioned earlier, the light will

generate minority carriers, thus accelerating etching in the illuminated
zone. Most previous investigations of photoelectrochemical etching
under these conditions have shown it to be analogous to corrosion, and
a charge-balancing reduction reaction is expected to occur in neighboring
non-illuminated zones.
TheColumbiagroup hasobserved markedwavelengtheffectson the
hole drilling process. Many experimentswere carried out using an Ar ion
laser, comparing etching using fundamental (514 nm) and frequency
doubled ultraviolet excitation (257 nm), a 3 - 4 p m spot size and laser
powers ranging from 10 nW to 100 mW. Ultraviolet illumination of GaAs
immersed in pure water results in oxide growth, while no effect on the
oxide layer was notedwith visible light illumination. If the solution is made
acidic or basic so that the oxide is dissolved away, etching occurs with
both sources, but at a much higher rate with the ultraviolet source at the
same intensity. Etching with the ultraviolet source was observed for n-, p-
and semi-insulating material. Finally, holes drilled with the focused
ultraviolet source are narrower, deeper and formed more rapidly than with
the visible source. Usingthe 257 nm laser, holes 3 - 4pm in diameter have
been etched through 200 - 300pm thickwafers. An incident focused laser
intensity of 10 W/cm2 resulted in an etch rate of 0.15 pm/sec.
The highlyactinic natureof the ultraviolet light isascribed by Osgood
and co-workers to the fact that carriers are produced very near to the
surface ( < 5 nm) and thus diffuse to the interface and react before they
have a chance to relax thermally. Thus, holes are produced deeply
enough in thevalence band to reactwith redox levelsof purewater located
at -0.4 volt below the top of the valence band. The deep holes that can
be drilled with ultraviolet light are ascribed to a light guiding effect that is
promoted by the high angular dependence of reflectance at these
wavelengths.
One interesting application of laser drilled holes is in producing fiber
optic couplers for integrated electro-optic devices. Prucnal et al. (23)
reporteda coupler to Si, shown in Figure 10. Thewell was first etched into
n-Si using focused 257 nm laser light and a 5% HF electrolyte, then its
internal surface converted toa p-n junction photodetector. Thewell acted
as a receptacle for a 9 p m single mode optical fiber. Other applications
include vias for connecting the front and back sides of wafers, and light
guides for electro-optical circuits.
Rauh and LeLievre (24) examined the intensity and total delivered
energy dependence of hole diameters etched in n-GaAs with a focused
633 nm (HeNe laser) source. Unlike the results with ultraviolet excitation,
the hole diameters grew sharply with etching time and no “light guided”
structures were produced. However, shallow pits could be etched with
Single mode
fiber core
(9Pm)
I I Aluminum
Adhesive
Si02
Figure 10. Schematic cross section of prototype integrated fiber-optic

coupler prepared by photoelectrochemical etching (after Fig. 1, ref. 23).
high resolution. By raster scanning the stage holding the electrolyte

coated crystal (10% KOH), and by computer controlling the shutter,
complex patterns could be etched in relief in the surface. An example of
a character set produced in this way from 2 p m diameter “pixels” is
illustrated in Figure 11.
A common feature of the hole drilling experiments described above
is that they were performed without electrochemical control of the
semiconductor surface. As a result, quantum yields were always much
lessthan unityand thewidthof thespacecharge regionwas probablyvery
low. The latter factor doubtlessly magnified differences between quantum
yields observed for ultraviolet and visible light resulting from differences
in the penetration depth and hence efficiency of electron-hole separation.
To date, no studiesof the effects of electrode bias on the rates and profiles
of laser drilled structures have been presented.
3.4 Miscellaneous Applications of Photoelectrochemical

Etching
A number of examples have been reported of using

photoelectrochemical etching as a step in fabricating electronic and
electro- optical devices.
Lubzens (25) has demonstrated the photoelectrochemical etching
of mesa structures in n-lnP. The structures described were cylinders 50
p m in diameter by 18 p m high. They were prepared by first defining the
circular regions by standard lithographic/liftoff procedures. The patterned
crystal was immersedin a 40%FeCI,etching solution and illuminatedwith
a broadband collimated source. The etching proceeded smoothly in the
exposed areas between the mesas, although occasional removal of
etching productswith Br,/HBr/H,O was necessary. The aforementioned
ion beam masking technique has also been used to define mesas for
definition by photoelectrochemical etching, thus avoiding the need for
metallization (26). Mesa structures were demonstrated in InP, InGaAs
and InGaAsP alloys.
Forrestet at. (27) described theapplication of photoelectrochemical
etching to the fabrication of an annular photodiode, shown in Figure 12.
The central hole was photoelectrochemically etched through
photolithographically defined circular areas in the metallization. The
processwas carriedout under an external bias using an electrolyte of 0.75
N KF/0.75 N HF. The photo-process continued until the p+ region was
reached, after which it was completed by a dark chemical etch.
Ostermayer et al. (28) reported the photoelectrochemicalfabrication
of lenses onto light emitting diodes (LEDs). The spherical lenses were
formed by projection of the lens pattern onto an n-lnP/lnGaAsP double
Figure 11. Photoelectrochemically etched character set in unbiased

n +-GaAs using scanned 2jJm diameter HeNe laser spot(160W /cm2 , 0.2
sec/point, 1O%KOH) (24). Reprinted by permission of the publisher, The
Electrochemical Society, Inc.
Figure 12. Schematic cross section of annular p-i-n photodetector (after

Fig. 1, ref. 27).
heterostructureLED, which served as the electrode. The projection mask

for the 70 p m diameter lenses was produced from concentric Cr rings
lithographically fabricated from a thin Cr film evaporated onto a glass
substrate. The lenses are used for coupling the output of the LED directly
into optical fibers.
Cheng and Kohl (29) demonstratedthat fiducial patterns on one side
of an n-lnPwafer, defined in photoresist, could be photoelectrochemically
etched through to the other side of a wafer about 90 p m thick. The
metallized mask served the dual purpose of the electrical contact, which
was at the front surface in this case. The process was carried out by
illumination with a collimated beam from a HeNe laser and an electrode
polarizationof 0.2-0.4voIt (SCE)in2-4 N HCI. Bestresultswereobtained
at light intensities < 260 mW/cm2due to reduced effects of diffraction and
light scattering within the growing hole.
Lin and co-workers (30) have demonstrated photoelectrochemical
etching of very fine lines in n-GaAs with the tip of a scanning tunneling
microscope as the counter electrode. An electrochemically machined Pt
tip was used with an end < 0.1 p m in diameter. The tip was immersed in
the etching electrolyte, which also contained a reducible depolarizer to
avoid tip dissolution, and drawn across the GaAs electrode surface at a
controlled distance of 1 pm. The n-GaAs was anodically polarized, and
lightwas used to induceand controlcurrent flowwhile the tipwas moving.
A minimum line width of 0.3 p m was achieved.
4.0 PHOTOELECTROCHEMICAL DEPOSITION
Up to this point we have been concernedwith photoelectrochemical

processes for selective removal of the semiconductor substrate. As
stated earlier, light-localized deposition is also feasible. In general, this is
carried out by adding to the electrolyte the precursor of a material that
forms the deposit when reduced or oxidized. The precursor may be a
metal ion, an electropolymerizable monomer, or any compound or complex
that is soluble in one oxidation state and insoluble in another. Some of the
applications of photoelectrochemical deposition include formation of
ohmic or rectifying contacts, plating of via (throughwafer) holes, imaging
for information storage and display, and fabrication of optical elements.
We first consider the localized photoelectrodeposition of metals
onto semiconductor surfaces (31)(32). In almost every case the precursor
to a metal in a plating bath is a positive ion, so that the deposit is formed
by reduction. Thus, a p-type semiconductor is required for positive
photoelectrochemical imaging. If the deposition is done in an
electrochemical cell configuration, the semiconductor electrode should
Next Page
Previous Page
heterostructureLED, which served as the electrode. The projection mask

for the 70 p m diameter lenses was produced from concentric Cr rings
lithographically fabricated from a thin Cr film evaporated onto a glass
substrate. The lenses are used for coupling the output of the LED directly
into optical fibers.
Cheng and Kohl (29) demonstratedthat fiducial patterns on one side
of an n-lnPwafer, defined in photoresist, could be photoelectrochemically
etched through to the other side of a wafer about 90 p m thick. The
metallized mask served the dual purpose of the electrical contact, which
was at the front surface in this case. The process was carried out by
illumination with a collimated beam from a HeNe laser and an electrode
polarizationof 0.2-0.4voIt (SCE)in2-4 N HCI. Bestresultswereobtained
at light intensities < 260 mW/cm2due to reduced effects of diffraction and
light scattering within the growing hole.
Lin and co-workers (30) have demonstrated photoelectrochemical
etching of very fine lines in n-GaAs with the tip of a scanning tunneling
microscope as the counter electrode. An electrochemically machined Pt
tip was used with an end < 0.1 p m in diameter. The tip was immersed in
the etching electrolyte, which also contained a reducible depolarizer to
avoid tip dissolution, and drawn across the GaAs electrode surface at a
controlled distance of 1 pm. The n-GaAs was anodically polarized, and
lightwas used to induceand controlcurrent flowwhile the tipwas moving.
A minimum line width of 0.3 p m was achieved.
4.0 PHOTOELECTROCHEMICAL DEPOSITION
Up to this point we have been concernedwith photoelectrochemical

processes for selective removal of the semiconductor substrate. As
stated earlier, light-localized deposition is also feasible. In general, this is
carried out by adding to the electrolyte the precursor of a material that
forms the deposit when reduced or oxidized. The precursor may be a
metal ion, an electropolymerizable monomer, or any compound or complex
that is soluble in one oxidation state and insoluble in another. Some of the
applications of photoelectrochemical deposition include formation of
ohmic or rectifying contacts, plating of via (throughwafer) holes, imaging
for information storage and display, and fabrication of optical elements.
We first consider the localized photoelectrodeposition of metals
onto semiconductor surfaces (31)(32). In almost every case the precursor
to a metal in a plating bath is a positive ion, so that the deposit is formed
by reduction. Thus, a p-type semiconductor is required for positive
photoelectrochemical imaging. If the deposition is done in an
electrochemical cell configuration, the semiconductor electrode should
be operated under a negative bias to enhance the (downward) band

bending. A current-voltagecurve for a p-type semiconductor electrode in
a platingelectrolyte is shown schematicallyin Figure 2. At potentials more
positive than V, the p-type electrode passes positive current in the dark
by the injection of holes via the valence band. In an ideal junction for
semiconductors with bandgaps large compared to thermal energies,
cathodic current will be negligible in the dark and will be stimulated by
super-bandgap illumination. When the electrolyte contains metal ions,
their reduction by conduction band electrons produced by light excitation
will give rise to localized metal deposition.
An important aspect of photoelectrochemical deposition is to ensure
that the energy of the conduction band electrons reaching the interface is
sufficient toreduce the precursor ion in the electrolyte. In reality, there are
three possible situations relating the electrodeposition potential, Vdep,the
flat band potential, V,, and the semiconductor bandgap. In the case
where Vpepis more negative than the conduction band, photoelectrochemical
deposition will not occur. The optimal potential region for Vdepis within the
bandgap. When Vdeplies below V, precursor ions may be reduced in the
dark via electrons flowing in the valence band.
The progress of photoelectrochemical positive deposition of metals
depends on whether the deposit forms an ohmic or rectifying contact. If
the contact is ohmic, then the band bending will collapse during the
deposition of the first few monolayers and the light-driven reaction will
cease. The plating may continue in the dark, however, if the control
potential is sufficiently negative or if the process is being conducted
galvanostatically. If the deposited contact is rectifying, then the light
driven reaction will continue for up to a few thousand angstroms until the
growing contact blocks all of the incoming light.
The voltammetric behavior of p-Si in combination with different
plating electrolytes has been investigated in this laboratory (32)(33)(40).
It is seen that relatively little cathodic current passes in the dark in plating
electrolytes showing the more negative deposition potentials, exemplified
in Figure 13b by Zn. In CuSO,, where the metal deposits at a relatively
positive potential, considerable plating occurs in the dark. This is due to
the fact that the plating potential is relatively near to the valence band in
p-Si. Even though the plating current is accelerated by light for this
electrolyte (due to additional contribution from photogenerated conduction
band electrons), the selectivity is severely reduced.
The electrical characteristics of photoelectrochemically deposited
Au-Zn contacts onto p-lnP and p-GaAs has been reported by Kelly et al.
(34). Some electrolyte compositions gave very constant Zn/Au ratios
over awide range of deposit thicknesses. In general, once the depositwas
initiated photoelectrochemically, it could be continued in the dark. Rapid
-1
-0.1 0.1 0.4

v VS. SCE
Figure 13. Current-voltage for p-Si photocathode. (a) Hypothetical

curves showing overlap region (shaded) between anodic dark current
and cathodic photocurrent necessary for "electroless" imaging. (b)
Cyclic voltammograms for p-Si in Zn plating electrolyte and Cu plating
electrolytewith 0.5 M MS0,and pH = 3.1. Broken line, dark; solid line, 1.2
mW/cm2 tungsten-halogen light illumination (40). Copyright 1983, The
American Institute of Physics.
annealing to 425 - 45OOC in a mildly reducing atmosphere led to contact

resistances on the order of 1Oq50 cm2.
The resolution of photoelectrochemical deposits appears to be
determined by a number offactors, and its limitation is not yet understood.
Figure 14 shows results obtained in this laboratory for a projection image
of Au produced on p-GaAs under galvanostatic control. The minimum
feature size of the 0.4 p m deposit is 10 - 20 p m (32). Jacobs and Rikken
(35) have studied the nucleation and growth of Au photoelectrochemically
deposited on p-GaAs under potentiostatic control. They initially observed
nuclei of about 1nm in diameter at a surface density of 1O'O to 1012/cm2
which then grew and eventually coalesced. The dark current was observed
to increase during the initial stages of deposition, indicating that Schottky
barriers pinned at a low barrier height were formed at the nuclei. As a
result, some degree of growth of the pattern could be expected by
continuing the electrolysis in the dark following an initial photoelectro-
chemical "strike".
The best resolution was obtained by Jacobs and Rikken under the
-
following conditions. The p-GaAs was Zn doped (N, 2x10I7/cm3)and
etched in 100:2:1 H,O/H,O,/NH,OH, then dipped in 0.2 M KOH. The
plating bath consisted of 5x10'3 M KAu(CN),, 0.1 M KCN and 0.2 M KOH.
The deposition was conducted potentiostatically at -1.2 volts vs. SCE and
a light intensity sufficient to give a photocurrent of a few mA/cm2. The
patternwas obtained by projection imaging onto the electrode surface. A
pattern similar to that in Figure 14 was reported which was the result of a
10 second exposure, with resolution of about 2 pm. The average
photocurrent during deposition was approximately 1 mA/cm2,where the
growth rate of the deposit is approximately 1 nm/sec.
lnoue et al. (36) have demonstrated positive imaging of oxide layers
onto n-type semiconductors, where the oxide is formed by the oxidation
of a soluble precursor. The general reaction is:
M"+ t aH,O t bp+- YO, t 2aH+
The technique is suitable for depositing oxides of Pb, Co, Mn and Cr, and
has been reported primarily for wide bandgap n-type oxide semiconductors,
such as ZnO and TiO,. A number of the oxide deposits are electronically
conductive and can be used as electrical contacts. Although not
demonstrated, the process should also be suitable for semiconductors
more commonly used in electronics. Imaging via oxidative photodeposition
has also been reportedfor conductive organic polymers on ZnO and other
semiconductors (37). The polymer forms on electrochemical oxidation
of solutions of the monomer, e.g., pyrrole, thiophene and aniline. Since
the conductive polymers are also electrochromic, the process might be
Figure 14- Photomicrograph of a photoelectrochemical Au deposit on

p-GaAs, produced from the projected image of a USAF-1951 resolution
target (33)(34). Copyright 1983, The Materials Research Society.
useful in the fabrication of electrochromic display elements.

As with etching, it is also possible in some cases to observe
photoelectrochemical deposition of thin films without external control,
i.e., an electroless process (31)(38). In this case, the semiconductor is
simply immersed in the plating/precursor solution and the surface illuminated
with the desired pattern. The main requirement for the electroless
photodeposition of a metal to occur is to have a sacrificial reductant in
solution or on the semiconductor (e.g., the semiconductor itself). Thus,
when the semiconductor is illuminated with the light pattern, the metal
ionswill be reduced in the illuminatedzones and the reductant oxidized by
majority carriers in the dark zones.
One possibility is to employ the dissolution of the semiconductor
itself as the sacrificial oxidation reaction. Kohl et al. (39) have demonstrated
the filling of via holes through a 30 p m GaAs wafer by electroplated Au
using such a photoassisted process. As illustrated in Figure 15, nucleation
of the Au occurs as the dark reaction at the source pad of an FET bridging
the via hole. The minority carriers (holes) produced at the semiconductor/
electrolyte interface give rise to Ga+3andA~+~formation. The equivalents
of Au that are plated in the vias, which had a 6x1Om4mm2cross section, is
very low, so that < 1 p m of the 2” wafer surface is etched away.
A particular example of electroless plating with a controlled sacrificial
layer has been described by Roseet al. (40), illustrated in Figure 13. Here,
a crystal of p-Siwas contacted with metallic Zn on the back side, and then
immersedin 0.5 M ZnSO,. TheZn+2/Znpotential of approximately -1 volt
vs. SCE resulted in a significant band bending at the semiconductor/
electrolyte interface.
On illumination, conduction band electrons reduced Zn+2in solution,
giving rise to a light-localized deposit. To balance the charge, holes were
swept to the Zn contact where they participated in its anodic dissolution.
In other words, the sacrificial Zn layer provides an anodic dark reaction
which overlaps in potential with the cathodic photodeposition reaction,
thus allowing the process to proceed spontaneously under illumination.
The process also works for other metals with rather negative plating
potentials, such as Cd, but not for more noble metals, such as Cu or Ni. In
the latter cases, the metal potentials are not sufficiently negative to cause
significant band bending in p-Si. (Note that putting theZn-contacted p-Si
in the C U +bath,
~ for example, will not work since the Cu will simply plate
out onto the Zn contact by exchange). This approach may also be taken
to deposit an “activation” layer of Pd, for example, for later electroless
plating in the dark (see, for example, Stremsdoerfer et at. (41)).
Photoelectroclieniical Processing of Semiconductors 209
5.0 PHOTOELECTROCHEMICAL SURFACE PROCESSING
Illumination of semiconductors immersed in liquid electrolytes often

gives rise to formation of oxidation products on the surface. The nature of
these products can affect skongly the energy distribution and concentration
of surface states. Such surface states play an important role in the
operation of solid state Schottky and metal-oxide-semiconductor (MOS)
diodes. In general, processing techniques are constantly being sought
where the surface state densities are reduced to a point where the Fermi
level is essentially unpinned so that the band bending in the diode
becomes a sensitive function of the applied voltage over a wide range.
Silicon technology is successful because surface oxidation readily reduces
the surface state density to below a critical concentration, typically < loio/
cm2. The development of integrated circuits based on Ill-V compounds,
which generally have higher mobilities than Si, has been hampered
somewhat by the absence of surface processing techniques that can
reduce the surface state densities to below even 1O13/cm2(42).
The high surface state density in GaAs is generally considered to be
due to the formation of islands of elemental As. The As is formed
spontaneously by the reaction of GaAs with one of its native oxides,
Ga,O,. The oxidation of GaAs occurs spontaneously in air, as does the
subsequent evolution of surface As. Using the experimental apparatus
shown in Figure 16, Offsey et al. (43) demonstratedthat illumination of n-
or p-type GaAs while washing the surface with a jet of 18 MQ deionized
water increased greatly the intensity of photoluminescence, which is
proportional to the surface state density. The technique was used to
reduce surface state density to 4 x 1 Oii/cm2. Surface state densities
were measured using actual MOS devices prepared by coating the treated
surface with a thin nonreactivepolymeric insulator. It was speculatedthat
the light increases the hole concentration at the GaAs surface and hence
therateofformation of Asand GaandAsoxides,whicharesimultaneously
swept away by the stream of deionizedwater, leavinga clean surface. The
positive effect is transitory, however, as the surface gradually re-oxidizes
in air. Treating the surface with sulfide has been shown recently to
stabilize the surface against re-oxidation.
Woodall et al. (44) determinedthat photochemical oxidation of GaAs
immersed in 1:1 HCI/H,O resulted in only As surface films, about 5 nm
thick, if 0, was excluded and if the H,O was deionized (1 8 MQ ); without
these precautions, a mixture of As and Ga oxides was formed. The As
further prevented oxide formation on the surface, and could be removed
by a relatively mild heat treatment at 29OOC in vacuo. The surfaces so-
produced were superior to those formed by much higher temperature
heat treatments in vacuo that were necessary to clean typical oxide-coated
LIGHT
S
G SAPPHIRE
S D
GaAs - Gd3+ A ~ * ~ + 6 e -
Figure 1 5. Schematic diagram showing photoelectrochemical deposition

of Au in plasma-etched vias of a GaAs FET mounted on a sapphire
substrate (after Fig. 1, ref. 39).
PHOTOMU LTIPLIER
MIRROR
LASER
D.I. H 2 *
GaAs SAMPLE
Figure 16. Apparatus for photoelectrochemical unpinning of the Fermi

level in GaAs with simultaneous monitoring of the photoluminescence
(after Fig. 1, ref. 44).
Photoelectrocherriical Processing of Semiconductors 21 1
surfaces. The photoelectrochemical surface preparation was suggested

as a precursor to molecular beam epitaxial growth onto GaAs.
6.0 CONCLUSION
The application of photoelectrochemistryto processing semiconductor

materials is still at an early stage of development. Yet, numerous possibilities
can be foreseen. Optical elements may now be produced in hard,
thermally robust materials using simple holographic techniques that
previously applied only to soft photosensitive polymers. Diffraction
gratings may be produced in SIC, for example, for high efficiency in the
vacuum ultraviolet or for use with high power laser sources (48). Projection
photoetching of a variety of periodic structures should be feasible, such
as have been proposed for beam steering, coupling and focusing in
integrated electro-optics. Photoelectrochemistry has been shown to be
suitable for producing extremely high aspect ratios, superior in some
cases to alternative “dry” processes such as reactive ion etching. Thus,
it is expected to have its place in fabrication of waveguides and vias in
thick crystals, in machining of microminiature mechanical and sensor
devices, and in making high surface area solar cells for enhanced light
collection efficiency. The high degree of control obtainable in
photoelectrochemical etching should find applications in electronics as
well, e.g., in buried gate GaAs FETs which require a precisely etched
channel between source and drain. Future workers in this exciting area
will do well to maintain an interdisciplinaryapproach, since its full realization
will requirethe expertise of electrochemists,surface scientists, solid state
chemists, electrical and optical engineers.
7.0 ACKNOWLEDGEMENT
This chapter was prepared with the support of the Office of Naval
Research.
8.0 REFERENCES
1. Gurevich, Y.Y. and Pleskov, Y.V., Photoelectrochemistry of Semi-

conductors. Semiconductors and Semimetals (R.K. Willardson
and A.C. Beers, eds.), Vol. 19, pp. 256-328, Academic Press, New
York (1983)
2. Gelchinski, M.H., Romankiw, L.T., Vigliotti, D.R. and von Gutfeld,

R.J., Electrochemical and metallurgical aspects of laser-enhanced
jet electroplating. J. Electrochem. SOC. 132: 2575-2581 (1985)
3. Haynes, R.W., Metze, G.M., Kreismanis, V.G., Eastman, L.F., Laser-

photoinduced etching of semiconductors and metals. Appl. Phys.
Lett. 37:344-346 (1980)
4. Park, S.M. and Barber, M.E., Thermodynamic stabilities of

semiconductor electrodes. J . €lectroanal. Chem. 90: 67-75
(1 979)
5. Ostermayer, F.W. and Kohl, P.A., Photoelectrochemical etching of

p-GaAs. Appl. P h y s . Lett. 39:76-78 (1981)
6. Hutley, M.C., Diffraction Gratings Academic Press, New York

( 7 982)
7. Ostermayer, F.W., Kohl, P.A. andLum, R.M., Hole transportequation

analysis of photoelectrochemicaletching resolution. J . Appl. Phys.
5814390-4396 (1985).
8. Lum, R.M., Ostermayer, F.W., Kohl, P.A., Glass, A.M., and Ballman,
A.A., Improvements in the modulation amplitude of submicron gratings
produced in n-lnP by direct photoelectrochemical etching. Appl.
Phys. Lett. 47:269-271 (1985)
9. Bjorkholm, J.E. and Eichner, L., Monitoringthe growth of nonuniform

gratings written holographically by Gaussian laser beams. J . Appl.
Phys. 57:2402-2405 (1985)
10. Podlesnik, D.V., Gilgen, H.H., and Osgood, R . M . , Deep ultraviolet

inducedwetetchingof GaAs. Appl. Phys. Lett. 45563-565(1984)
11. Matz, R., Laser wet etching of diffraction gratings in GaAs for integrated
optics. J . Lightwave Technology LT-4:726-729 (1986)
12. Matz, R. and Zirrgiebel, J., Fast photoelectrochemical etching of

quarter-micrometer diffraction gratings in n-lnP. J . Appl. Phys.
6413402-3406 (1988)
13. Matz, R., Holographic photoetching of high-quality diffraction gratings

in p-GaAs for distributed feedback lasers. M a t . Res. S O C .Symp.
Proc. 75: 657-664 (1987)
14. Carrabba, M.M., Nguyen, N.M., and Rauh, R.D., Effects of doping
and orientation on photoelectrochemically etched features in n-
GaAs. J . €lectrochem. SOC. 134:1855-1859 (1987)
15. Carrabba, M.M., Nguyen, N.M. and Rauh, R.D., Crystallographic

effects of the photoelectrochemical etching of gratings in compound
semiconductors. M a t . Res. SOC.Symp. Proc. 75665-670 (1987)
16. Carrabba, N.M., Nguyen, N.M., and Rauh, R.D., Photoelectrochemical

fabrication of sawtooth gratings in n-GaAs. Applied Optics 25:4516-
4518(1986)
17. Li, J., Carrabba, M.M. and Rauh, R.D., Morphology control of
photoelectrochemically etched profiles in n-GaAs. Extended
Abstracts, Fall Meeting of the €lectrochemical Society,
October 15-20 (1989)
18. Li, J.J., Carrabba, M.M., Hachey, J.P., Mathew, S . , and Rauh, R.D.,
Photoelectrochemical etching of blazedechelle gratings in n-GaAs.
J . Electrochem. SOC. 135: 3171-3172 (1988)
19. Cummings, K.D., Harriott, L.R., Chi, G.C., and Ostermayer, F.W.,
Using focused ion beam damage patterns to photoelectrochemically
etch features in Ill-V materials. Appl. Phys. Lett. 48:659-661
(1986)
20. Lax, M., Temperature rise caused byafocused laser beam. J. Appl.
Phys. 4813919-3923 (1977)
21. Osgood, R.M., Sanchez-Rubio, A., Ehrlich, D.J., and Daneu, V.,
Localized laser etching of compound semiconductors in aqueous
solution. Appl. Phys. Letf. 40:391-393 (1982)
22. Podlesnik, D.V., Gilgen, H.H., Willner, A.E., and Osgood, R.M.,
Interaction of deep-ultraviolet laser light with GaAs surfaces in aqueous
21 4 Electrochemistry of Semiconductors and Electronics
solution. J. Op. SOC.Am. B 3:775-784 (1986)
23. Prucnal,P.R.,Fossum,E.R.and Osgood, R.M., Integratedfiberoptic

coupler for very large scale integration interconnects. Optics Letters
1 1 1109-11 1 (1986)
24. Rauh, R.D. and LeLievre, R.A., Microphotoelectrochemical etching

of n-GaAs using a scanned focused laser. J . Electrochem. SOC.
132:2811-2812 (1985)
25. Lubzens, D., Photoetching of InP mesas for production of mm wave

transferred electron oscillators. Electron. Lett. 13: 171-172 (1977)
26. Chi, G.C.,Ostermayer,K.D., Cummings, K.D. and Harriott, L.R., Ion

beam damage-induced masking for photoelectrochemical etching
of Ill-V semiconductors. J . Appl. Phys. 60:4012-4014 (1986)
27. Forrest, S.R., Kohl, P.A., Panock, R., Dewinter, J.C., Nahory, R.E.
and Yanowski, E., A long-wavelength annular InGaAs pin photo-
detector. / € € € Electron. Dev. Lett. EDL-3415:415-417 (1982)
28. Ostermayer, F.W., Kohl, P.A. and Burton, R.H., Photoelectrochemical

etching of integral lenses on InGaAsP/lnP light-emitting diodes.
Appl. Phys. Lett. 43:642-644 (1 983)
29. Cheng, J. and Kohl, P.A., The resolution of photoelectrochemically

etchedfeatures. M a l . Res. SOC.Symp. Proc. 29:127-132 (1984)
30. Lin, C.W., Fan, F.F. and Bard, A.J., High resolution photoelectro-
chemical etching of n-GaAs with the scanning electrochemical and
tunneling microscope. J. Electrochem. SOC. 134:1038-1039
(1 987)
31. Haisty, R.W., Photoetching and plating of GaAs. J. €lectrochem.

SOC. 1081790-794 (1961)
32. Rose, T.L., Micheels, R.H., Longendorfer, D.H. and Rauh, R.D.,
Photoelectrochemicaldeposition of metal patterns on semiconductors.
Mat. Res. SOC.Symp. Proc. 17:265-272 (1983)
33. Micheels, R.H., Darrow, A.D. and Rauh, R.D., Photoelectrochemical

deposition of microscopic metalfilm patterns on Si and GaAs. Appl.
Phys. Lett. 39:418-420 (1981)
Photoelectrochemical Processing of Semiconductors 21 5
34. Kelly, J.J., Rikken, J.M.G., Jacobs, J.W.M. and Vaister, A.,
Photocathodic deposition of gold alloys for ohmic contacts to Ill-V
materials. J . Vac. Sci. Techno/. B6:48-52 (1988)
35. Jacobs, J.W.M. and Rikken, J.M.G., Oxygen-diffusion size effect in

electroless metal deposition. J . €/ectrochem. SOC. 1352822-
2827 (1988)
36. Inoue, T., Fujishima, A. and Honda, K., Photoelectrochemicalimaging

processes using semiconductor electrodes. Chem. Lett. 1978:1197-
1200 (1978)
37. Okano, M., Kikuchi, E., Itoh, K. and Fujishima, A., Photoelectro-
chemical polymerization of pyrrole on ZnO and its application to
conducting pattern generation. J. Electrochem. SOC. 135:1641-
1645,(1988)
38. Silverman, S.J. and Benn, D.R., Junction delineation in silicon by

gold chemiplating. J . €lectrochem. SOC, 105:170-173 (1958)
39. Kohl, P.A., D’Asaro, L.A., Wolowodiuk, C. and Ostermayer, F.W.,

Photoelectrochemical plating of via GaAs FET’s. I € € € €/ectron.
Dev. Lett. EDL-57:7-9 (1984)
40. Rose, T.L., Longendorfer, D.H. and Rauh, R.D., Photoelectrochemical

deposition of metals onto p-silicon using an internal cell. Appl.
Phys. Lett. 42:193-195 (1983)
41. Stremsdoerfer, G., Martin, J.R., Garrigues, M. and Perossier, J.L.,

Localized photoactivation with Pd of p-lnP previoustoan electroless
gold plating. J . €/ectrochem. SOC. 133:851-852 (1986)
42. Robinson, A.L., Chemical coat helps semiconductor prospects.

Science 238: 27-29 (1 987)
43. Offsey, S. D., Woodall, J.M., Warren, A.C., Kirchner, P.D., Chappell,
T.I. and Pettit, G.D., Unpinned (100) GaAs surfaces in air using
photochemistry. Appl. P hys. Left. 48: 475-477 (1986)
44. Woodall, J.M., Oelhafen, P., Jackson, T.N., Freeouf, J.L. and Pettit,
G.D., Photoelectrochemical passivation of GaAs surfaces. J. Vac.
Sci. Techno/. B1:795-798 (1983)
45. Podlesnik,D.V., Gilgen, H.H. and Osgood, R.M., Maskless, chemical

etching of submicrometer gratings in single crystalline GaAs. Appl.
Phys. Letf. 43:1083-1085(1983)
46. Kohl, P.A., Wolowodiuk, C. and Ostermayer, F.W. Jr., The photo-
electrochemical oxidation of (loo),(1 lo), and (1 1 1 ) n-lnP and n-
GaAs. J. €lectrochern. S O C . 130:2288 (1983)
47. Dalisa, A.A., Zwicker, W.K., DeBitett0,D.J.and Harnack, P., Photo-

anodic engravingof holograms on silicon. Appl. Phys. Lefters 17:
208-21 0 (1 970)
48. Rauh, R.D. andcarrabba, M.M., Photoelectrochemicalfabrication of

gratings in silicon carbide. Extended Abstracts, F a l l Meefing of
the €lecfrochernica/ Sociefy, October 15-20(1989)
5
PHOTOELECTROCHEMICAL CHARACTERIZATION
Micha Tornkiewicz
1 .O INTRODUCTION
This chapter will cover possible applications of liquid junctions in

“noncontact” characterization of semiconductor electronic devices.
Wherever possible, we use examples from characterization of materials
that play a central role in the electronic industry, primarily Si and GaAs.
Review articles that focus on non-solar applications of semiconductor-
liquid junctions ( 1 ) and on experimentaltechniquesfor the study of such
junctions (2) were recently published. Conceptually, liquid junctions are
well suited for semiconductor characterization, since they have almost
“contactless” junctions and are most adaptable for on-line characterization.
Nevertheless, they have yet to significantly penetrate the electronics
industry. One reason is an apparent reluctance of the solid-statecommunity
to get involved with “wet chemistry”. Even the widespread use of “wet”
etching is now being replaced by “dry” Reactive Ion Etching (RIE) (3).
Another reason for slow acceptance is that liquid junctions are perceived
as more complex than their solid counterpart. Some aspects of this
complexity are real and involve problems of corrosion and chemisorption
that do not have a solid-state counterpart. However, much of the perceived
complexity reflects a traditional oversimplification of the solid state junctions
as abrupt junction devices. The simplest solid state contact, Schottky
barriers, are rich in uncontrollable chemistry during contact formation (4)
and only recently are attempts being made to control this chemistry either
through the use of liquid junctions (5) or through submonolayer metal
deposition under vacuum (4).
The science of semiconductor-liquid junctions has its roots in the
21 7
same work as that of solid state junctions (6)(7). However, the explosive
development of solid state devices, reflecting the growth in the volume
and diversity of their practical applications, has left their liquid junction
counterparts far behind.
Renewed interest in semiconductor-liquid junctions started to emerge
after it was demonstrated that a device based on partially reduced TiO, in
contact with an aqueous electrolyte can sensitize photoelectrolysis of
water ( 8 ) . lmmediatelyfollowing that work, the needtofind alternativefuel
sources rekindled interest in solar energy. As a result, the study of energy
conversion at semiconductor-liquid junctions grew to form almost a
separate discipline that was named “photoelectrochemistry.”
Before liquid-junctions can enter in force as an alternative junction
technique to MOS and Schottky barriers for the characterization of
semiconductors,a careful delineationof parametersthat are independent
of the nature of the junction will have to be compiled. To accomplish this,
more fundamental work has to be done in comparative studies of the
potential distribution under different junction configurations (9).
While liquid-junction devices cannot yet compete with solid-state
devices, stable photovoltaic liquid junctions devices can be made with
conversion efficiencies in the 12 - 15% range (10)( 1 1). Furthermore,
some niche applications are already emerging. Prominent among these
applications are the etching and photoetching of semiconductors (12)( 13).
At present, probably the broadest application of liquid junctions is for
characterizingsemiconductor surfaces: an eleckolyte provides a convenient,
reversible, junction medium in which the interfacial chemistry can be
modified and the transport and optical properties can be monitored in a
nondestructive way. Excellent examples of such applications can be
found in the literature. Lewis et al. (5) have used liquid junctions to probe
the origin of Fermi level pinning at metal junctions with Si and GaAs.
Similarly, Morrison et al. (1 4) have used liquid junctions to investigatethe
effects of dislocation loops on Si, on the transport properties. They
compared transport parameters, in averaged and scanned modes, in
both liquid junction and MOS devices, as a function of the surface
treatment that produces the dislocations.
Photoelectrochemistry is still an emerging discipline in that it is not
yet possible to write this chapter in the form of a tutorial in which one can
give a series of steps and techniques that, once applied to a given system,
results in a “complete” understandingof the system in terms of structure,
mechanism and kinetics. There is no single semiconductor/electrolyte
system of which our understanding is sufficient for its predictive power to
provide any degree of satisfaction. Afew systems such as Si and InP come
close.
Per definition, semiconductor/liquid junctions act as afocal point for
Photoelectrochemical Characterization 219
convergence of interdisciplinary methodologies of inquiry. Characterization

techniques are no exception. Almost every technique applied in
electrochemistry and solid state physics has found its way into
photoelectrochemistry. The scope of these techniques is too broad for
any attempt to cover them in a single chapter. We will exclude all the
techniques that primarily probe the semiconductor in an ex situ configuration;
this will include all the high vacuum surface spectroscopies and the ion
beam techniques. This is done in spite of recent progress in adapting
many of these techniques for “almost” in situ configuration by applying
sophisticated cell construction and transfer techniques (1 5). If one
includes the fundamental principles and recent relevant work, almost
every technique thatwill be mentioned in this chapter deserves achapter,
if not an entire book, for a comprehensiveexposure. This review will not
cover all available references, and we apologize for the omissions. It will
be somewhat tilted to our own work and priorities that recentlyfocusedon
the possible applications of some of these techniques for process
characterization in the electronics industry.
The chapter is organized as follows: We begin by discussing the
response of photoelectrodes to d.c. perturbation,focusing first on current-
voltage behavior of static and rotating electrodes and the applications of
various spectroscopic techniques. Next, we discuss responses to a x .
modulated electric fields, covering impedance, photocapacitance and
optical/spectroscopic techniques. The third area of focus is on time-
resolved techniques, in which short lived transient responses are considered.
Finally, brief mention is made of photothermal methods and of topographical
studies.
2.0 DIRECT RESPONSE
Figure 1 shows the combined energy diagrams of n-Si (100) in liquid

junction under short circuit conditions (1sa). The electrolyte is a methanolic
solution of ferrocene/ferrocinium with LiCIO, as the supportingelectrolyte.
The data for the energy diagrams were taken from impedance spectroscopy,
current-voltage measurements and electroreflectance, details of which
will follow. The electrolyte here is one of the simplest since it was
demonstratedthat Si is not corroding in the electrolyte and the ferrocene
is an outer sphere complex that, at least over a limited potential range, is
not absorbed on the surface of the electrode. Yet, in spite of this, the
complexity of the potential distribution at the solid-liquid interface, is
evident. One can compare this complexity with the relative simplicity of
the potential distribution of a Ti Schottky structure madewith materialwith
similar characteristics (16e). In both cases a smallfraction of a monolayer
is coveredwith surface states. These states tail from the conduction band
in the metal junction while they are more isolated in the liquid junction. In
addition, in the liquid junction, even after HF etching, a thin layer of a
porous oxide is observed. In the solid junction, the Fermi level is not
pinned until it reachesthe position of the surface states, while in the liquid
junction case, the Fermi level is pinned at reverse bias probably due to
surface accumulation of minority carriers, and then it gets pinned again
(or the band edges unpinned) due to charging of the surface states. The
filled surface states also catalyze precipitation of nearly a monolayer of
the electrolyte.
2.1 Current-Voltage
Figure 2 shows the current-voltage characteristics of the cell for

which the energy level diagram was shown in Fig. 1 (16a). The figure
shows current-voltage characteristics of the semiconductor in a three
electrode configuration: one electrode (Pt) serves as a counter electrode
and the third electrode serves as a referenceelectrode. The potentials are
givenagainst the solution potential. The photocurrent was measuredwith
a lowfrequencychopped light source (usinga hand asashutter) such that
one can observe the light induced current and the dark current on the
same chart. One also can see that the most negativepotential (for n-type
materials), in which light induced charge separation starts, is around -0.6
volt. This potential is usually a rough estimate of the position of the flat
band potential. In addition, one can observe the irreversible reduction in
the dark of the electrolyte, shown also in the energy diagram. This
information, combined with knowledge of the band gap, flat band potential,
doping level, interface states and presence or absence of an interfacial
layer, is essential for construction of a band diagram such as the one
presented in Fig. 1. Such a diagram serves as a starting point in understanding
the semiconductor-electrolyte interface and in the interpretation of the
results from the dielectric techniques that will be described below.
2.2 Rotating Ring Disc Electrodes
In a typical experiment with semiconductor-liquid junctions, one of

the most important experimental problems is the differentiation between
reactions that involve chemical changes at the semiconductor electrode
(corrosion with insoluble products)and chemical changes in the electrolyte
that might be subject to mass transfer limitations. The technique of
Rotating Ring Disc Electrode (RRDE) (17-19) provides an opportunity to
differentiate between these two types of reactions under controlled
hydrodynamic conditions. In its simplestform, the metallic ring is isolated
Photoelectrochemical Characterization 22 1
BAND DIAGRAM
0.5
I .o
Figure 1. The proposed band structure diagram of the Si/methanol that

is described in the text (1 sa).
I I I I
-0.6 -0.4 -0.2 0.'
Potentia! ( V vs. P t )
Figure 2. Current-potentialresponse with chopped white light for unetched
< 100 > n-Si in methanolic solution of oxidize and reduced dimethylferrocene
in thefollowingconcentrations: 0.2 M FeCp,, 1 mM FeCp,' and 1 M LiCIO,
supporting electrolyte under nitrogen atmosphere. Tungsten Halogen
ELH lamp with light intensity at the electrode surface of 100 mW/cm2.
Sweep rate: 10 mV/sec (1sa).
from the semiconductor disc. Both rotate at a certain angular frequency.

The voltammogram of the disc is similar to the one we have previously
described. The metallic ring is kept at a potential characteristic of one of
the anticipated reaction products. The ring current, at this potential, is
being monitored simultaneouslywiththe disc voltammogram. The procedure
can be repeatedwith other products to provide sufficient information for
construction of a unique reaction mechanism.
2.3 Absorption, Reflection and Photoluminescence

Spectroscopies
These techniques are covered in standard text books (20), review

articles (21) and books (22) dedicated to the individual techniques. The
subject of applications of any of these techniques in the study of properties
of semiconductors is too broad to be covered here. We will just provide
ashort outline focused on specific applicationsfor study of liquid-junction
devices.
The use of direct absorption spectroscopy, with photon energies
above the band gap, for study of liquid junction devices is limited due to
the special requirements of thin, optically transparent, crystals. Electric
field modulated absorption was reported (23). Reflection spectroscopy
is of wider use but its application is also limited, primarily due to advances
in techniques such as Electrolyte Electroreflectance, Photoreflectance
and Ellipsometry which will be described later, techniques that can provide
the same information and offer higher spectral sensitivity and higher
surface sensitivity. Nevertheless rotating light-pipe reflectivity (RLPR)
studies were used for in situ study of the growth of anodic oxide films on
GaAs (24). One can greatly increase the sensitivity of reflection studiesto
interfacial phenomena by employing subband gap irradiation and by
giving the crystal a shape that allows multiple internal reflection (25). An
example of the applications of this technique in detecting surface states
on cleaved Ge and Si can be found in Ref. 26. The technique was also
applied in the presence of an electrolyte (27). The sensitivity of this
technique to the interface can also be greatly improved by modulation of
the electric field and will be described in the section that describes
modulation techniques.
Among the direct optical spectroscopic techniques, photo-
luminescenceis probably the most widely used in both in situ and ex situ
configurations. Since, as with many other techniques, the temperature
window in which one can measure in the in situ configuration is limited,
one can usually find in the literature comparative studies between in situ
and ex situ conditions. A short review article relevant to applications in
liquid junction devices was written by Ellis (28). The technique also can
be applied in an electric field modulated form and in the time-resolve form

that will be addressed later. In addition to its utility as a characterization
tool that provides informationon the electronic states involved in emissive
transitions and the kinetics of excited state processes, photoluminescence
can be used as an in situ diagnostic tool to monitor effects of etching (29),
aging of photoelectrochemical devices (30) and much more. An interesting
application being pursued recently is the ability of photoluminescenceto
provide information on the distribution of electric field at the interface.
This is based on the assumption that the electric field in the space-charge
layer provides a strong enough driving force to separate the electron-hole
pair, thus preventing them from radiative recombination. This gives rise
toa “dead layer” inwhich the luminescenceis totally quenched. Often the
“dead 1ayer”coincideswith the spacecharge layer (31)(32).Under these
conditions the intensity of the PL signal will be given by:
where @ is the quantum efficiency, Q the absorption coefficient, d the

thickness of the “dead layer” and la the intensity of the incident irradiation.
If d is associated with the thickness of the space-charge layer, it will
change with the electrode potential according to:
d - [(2Ke/eN,)(U - Ufb-kT/e)]’/’
where E is the permittivity of free space, K is the relative dielectric

constant, e the electronic charge, N, the effective dopant concentration,
k the Boltzman constant, T the absolute temperature, U the electrode
potential and U, the flat band potential. Application of these equations to
the potential dependence of the PL intensity should provide information
on the flat band potential and the doping level. This was demonstratedon
a number of systems.
A variant of photoluminescence that in some cases requires the
presence of a liquid junction is electroluminescence (EL) (28). In this
case, the luminescenceis not induced by optical excitation but bycharge-
transfer processes. The most widely used electrolyte for this purpose is
alkaline peroxydisulfate (33) in which the S,O,-’ ions are reduced by the
conduction band electrons of an n-type semiconductor to yield the highly
oxidizing sulfate radical anions that can inject a hole into the valence
band. Recombination of an electron with the injected hole yields the
luminescence. This process is much more localized to the surface as
compared with photoluminescence. Comparison between the two can
yield valuable information about location and energeticsof traps localized

near the interface (31). Under strong reverse-bias conditions, electro-
luminescence due to avalanche breakdown was also observed (34).
2.4 Ellipsometry
Ellipsometryis areflectance method in which the change of the state

of polarization upon reflection is determined. This change is described in
terms of two ellipsometric angles A and @ given by:
@ = tan (A, , r Ali / AI A);
where 6 and A are the phase and amplitude of the parallel and perpendicular
components of the electric field of the incident (i) and reflected (r) beams.
One can show that these angles can be correlated with the real and
imaginary parts of the dielectric constant of the reflecting medium. In this
form the results are analogous to other dielectric techniques such as
impedance with its model sensitive interpretation. The advantage of this
technique is the high sensitivity to the surface. Spectroscopic ellipsometry
(35) takes out much of the guesswork, at least concerning the chemical
composition of the surface. One can analyze the data in terms of energy
dependence of the dielectric constant and compare the results with
electroreflectance. The results of such a study were interpreted in terms
of extension of the low field theory of electroreflectance to include interactions
of the modulating electric field with defects (36). The technique is widely
used for in situ characterization of surface modifications either through
adsorption or under electrochemical conditions in a liquid junction
environment (37-40).
3.0 ELECTRIC FIELD MODULATION OF SYSTEM’S

RESPONSE
These techniques are based on the sensitivity of the system’s response

to the electric field. Since the electric field is localized at the interface,
phase sensitive detection of the component that is modulated by the
electric field can differentiate between interfacial response and bulk
response.
3.1 impedance
The system’s response here is the dark current. Excellent recent

reviews on impedance spectroscopy, that cover also many aspects of
semiconductor-liquid interfaces, are available (41 )(42).Detailsabout the
instrumental setups can be found there. In principle, if the frequency
range is broad enough, the technique is sensitive to all the charge
accumulation modes and to the charge transport kinetics and mechanism.
The difficulty, as with most of the dielectric techniques, is to establish the
uniqueness of the proposed model for the interface. The most heavily
used practice is to measure the “capacitance” or the “conductance” at a
fixed frequency and to deduce the physics from the voltage or temperature
dependence. The next degree of complexity is found when the real and/
or the imaginary part of the impedance are measured over a wide frequency
regime, an equivalent circuit is proposed and the physical nature of the
elements in the equivalent circuit are determined from their dependence
on external parameters such as voltage, temperature, electrolyte, etc.
Alternatively one can model the interface as a capacitor with a composite
dielectric and apply some form of an effective medium model to characterize
the interface (43). Another widely used practice, particularly among
electrochemists, is to measure the impedance in a voltage regime in
which charge transfer takes place, propose a mechanism, solve the
kinetic equations in terms of frequency dispersion of the impedance and
getthe kinetic parametersfrom the fittotheexperimentalresults(41). The
main difficulty in all these schemes is the necessity to show the uniqueness
of the original assumption, be it the equivalent circuit, the form of the
effective medium theory or the proposed charge transfer mechanism.
The technique described here attempts to satisfy this requirement by
analyzingthe data in terms of a generalizedequivalent circuits, the details
of which emerge from the analysis. Figure 3 shows a generalized equivalent
circuit for a simple case where one measures the impedance of a single
dielectric (abrupt junction between a semiconductor and a metal or a
concentrated electrolyte). C, is usuallythe capacitance due to the space
charge layer, assumed to be the fastest relaxing element. Ci and ri = Ci Ri
are the capacitances,and their corresponding relaxation times, of charge
accumulation modes such asvarious surface states, minority carriersand
bulk states. 2, is a generalized CPA (Constant Phase Angle) element
given by:
Eq. (5) 2, = N(l + JOT)-“

whereo is the angular frequency and N, rand n are parameters. R, is the
series resistance and R, is the shunt resistance.
iRs
Figure 3.The generalized equivalent circuit of a single interface. R, is the
resistance associated with the Faradaic current flow, 2, is a generalized
impedance associated with disorder either in the structure or in the
dynamics (diffusion), Ci and Ri are associated with parallel charge
accumulation modes with different relaxation times than the majority
carriers such as surface states or minority carriers, C, is the space charge
capacitance and R, is the series resistance (1sa).
The inclusion of Z, extends the previously reported procedure of

RelaxationSpectrumAnalysis (44). 2, in this form can include contributions
from static disorder such as porosity (45), random mixture of conductor
and insulator that can be described by the effective medium approximation
at percolation (46), or an interface that can be described by a fractal
geometry (47). It can also include contributions from dynamic disorder
such as diffusion. To provide one specific example: if 2, originates from
diffusion capacitance in the semiconductor, then r is the minority
carriers’ diffusion time, n = 0.5 and
Eq. (6) N = (kT/q) (Lp/qDpP,,) exp(-qV,/ kT) (see Ref. 48)
where T is the temperature, k the Boltzmann constant, q the electronic

charge, L, the minority carriers’ diffusion length, Dptheir diffusion coefficient,
P, their concentration in the bulk and V, is the forward bias. As with all the
other elements, the origin for this contribution can be determined by the
voltage and the temperature dependence.
If one subtracts the frequency-independent,high frequency resistance
and calculates the resulting admittance, it is easy to show that:
Eq. (8) -
1 - 1 + JwC,
- t w2
Ci ri + Ju) Ci
z’ R, 1tu y * 1 t w%;2
+ Nql(l t w 2 r i 2 ) n / 2(cosnd t Jsinnd)

where tan 0 = r . For r > > 1, Eq. 8 can be viewed as superposition of a
constant term, linear term, power law term and superposition of Lorentzian
terms with respect to the frequency. If the separation between the time
constants is large enough one can isolate the respective terms and obtain
all the parameters directly. The contributions of the parallel RC elements
can be most conveniently obtained through the following two functions:
Gss Ci ri
Eq. (9a)
w i 1 + w2ri2
Bss is sensitive to the fast relaxing elements and Gss to the slow elements.
If the separation between the time constants is large enough one can
isolate the respective terms and obtain all the parameters directly. If it is
not large enough and/or the system cannot be represented in terms of a
simple abrupt junction due to film formation, multiple junction etc., some
data fitting will be necessary. In the latter case, a degree of confidence is
obtained by fitting the real or imaginary component and checking it
against the other component. The uniqueness of the interpretation is then
checked against other techniques that provide complementary information.
Figure 4 illustrates the impedance spectra of n-Si in contact with the
methanolic solution (a) for which the band diagram was presented in Fig.
1 (1sa), (b) n-Si/Ti Schottky structure (16e), and (c) n-Si MOS device
(16e), together with the equivalent circuits and the corresponding calculated
spectra. The equivalentcircuits of the solid-state devices are as expected
from ideal structures. The equivalent circuit of the liquid junction is much
more complicated (16a). 2, is due to a porous oxide layer that can be
almost completely removed by HF etching. R, and C, are due to absorption
of the electrolyte at potentials negativeto the surface states. Figures 5(a)
and (b) depict the Mott-Schottky plots (49)(50) of the fastest relaxing
element, the space charge layer capacitanceC, for the HF-etchedSi and
for the unetched one, in the liquid junction configuration. The flat band
potentials, obtained from the intercept, are -0.76 volt vs. Pt for the HF-
etched sample, and -0.86 volt vs. Pt for the unetched one. The doping
levels, obtained from the slopes, yielded 4 . 7 ~O15/cm3
1 for the etched
sample and 22x1O15/cm3for the unetched sample. The apparent difference
in the doping level is due to the increasein roughness due to etching. The
pinning of the Fermi level around -0.4 volt vs. Pt can be easily observed.
Figure 6 shows a typical Gss/o and Bss spectra of n-lnSeelectrode
in iodide solution (51), from which two surface state elements were
assigned, with relaxation times of 1 ms. and 10 ms., respectively. An
example of the dependence of the surface state capacitances on the
electrode potential, is shown in Fig 7. The data were fitted to a Gaussian
distribution of surface states (52). For the fast surface state, the distribution
is centered around -0.23 volt (vs. Pt) with area density of 1 . 0 5 ~013/cm2
1
and width of 0.11 eV. For the slow surface state, the distribution is
centeredaround-0.21volt(vs. Pt)with areadensityof 3.1 2x10i3/cm2and
width of 0.13 eV. If we take 1O”/cm2 as typical density for a monolayer,
both states occupy a small fraction of a monolayer. Similar analysis of the
impedance data was done on other semiconductor/electrolyte systems
that include TiO,, CdSe, Culn,Se, etc.
An interesting variance of impedance spectroscopy is the time
domain spectroscopy (53) in which a voltage function (a Heavyside
function, for example) is sent along a line of fixed constant impedance. At
2
Q 1
-I
- -2 -
- 2 0 2 4 6 2 3 4 5 6 7 - 2 0 2 4 6
Figure 4. impedance response curves for n-Si in three different junctions:

a. HF etched n-Si in a methanolic solution of oxidized and reduced
dimethylferrocene in the following concentrations: 0.2 M FeCp,, 1 mM
FeCp,' and 1 M LiCIO, supportingelectrolyte under nitrogenatmosphere.
Potential: 0.2voltvs. Pt; b. Ti/Schottky-barrier at 0 volt; c. Ti/ MOS device
at -0.5 volt. Symbols: experimental data for the real and imaginary parts
of the impedance. Solid lines: numerical fits to the equivalent circuit
shown in the inserts (16a and 16e).
Potential (V vs. Pt)
Figure 5. Mott-Schottky plots of n-Si in the methanolic Solution described

inFig.2(a). (a)HFetchedsample. V=, -0.76voltvs.PtandN,= 4.7x1Ol5/
cm3. (b) original sample, V, = -0.86 volt vs. Pt, N, = 2 . 2 ~Oq5/cm3
1 (16a).
PhotoelectrochemicalCharacterization 231
3
CL)
0
2 *
v)
2
I
0
0 I 2 3 4 5 6 7
Log f (Hz)
Figure 6. Gss/o and Bss spectra of n-lnSe in the electrolyte: 1 M KI/

0.05 M 1,/2 M H,SO,, electrode potential: 0 volt vs Pt. The capacitive and
resistive elements of the surface states are evaluated from the amplitude
and the position of the peaks (51).
2 .o
1.5
N
E
<
LL
vx 1.0
-
ln
ln
0
0.5
0.0
-0.4 -0.2 0.0 0.2 0.4
Potential (V v s Pt)
Figure 7. Variation' of Css, of the slow surface states, with electrode
potential for (a) n-lnSe and (b) modified InSe. The solid lines are theoretical
fits to Gaussian line shapes (51).
the point of intersection between this impedance line and a different

impedance (the electrochemical cell) part of the original voltage function
is reflected back along the line and provides information about the
encountered impedance. The promised advantageof this technique is its
high speed. In principle the technique should provide the same information
as the frequency domain analog.
3.2 Photocapacitance
Under this heading one can include all the techniques that use
impedance spectroscopy under illumination. With band gap illumination
one has the usual difficulty of a unique interpretation, particularly the need
to separate the light-induced kinetic effects from changes in the light
induced “equilibrium” potential distribution. Nevertheless routine analysis
of the high frequency capacitance in terms of the Mott-Schottky relation
was interpreted in terms of light induced shift in the flat band potential
and/or change in the dopant distribution (54). A technique that does not
suffer from these difficulties is the photocapacitance or its liquid junction
version Electrochemical Photocapacitance (EPS) (55). In EPS, the
capacitance of a reversed biased semiconductor electrode is measured
as a function of the wavelength of incident sub-band gap illumination. It
is assumed that the origin of the capacitance signal (imaginary part of the
impedance) is the space-charge layer and that the subband gap illumination
is absorbed by the deep traps and interface states and induce exchange
of charge with the space-charge layer that results in capacitance changes.
Typical plots of capacitancevs. wavelengthyield aseriesof plateausand/
or peaksthat correspond to population/depopulation of band gap states.
The results usually complement those obtained using Deep Level Transient
Spectroscopy (DLTS) (56). Since, as a rule, the available temperature
window in liquid junction devices is very narrow, EPS is much more
suitable than DLTS for evaluation of deep traps in these devices.
3.3 Optical Techniques
A cost effective experimental setup for optical modulation experiments,

recently built in our laboratory, is shown in Fig. 8 (57). Similar setup was
recently reported by Tian et at. (58). Experiments performed with this
system include: photoreflectance (PR), electrolyte electroreflectance
(EER), surface photovoltage spectroscopy (SPV), 1st. and 2nd. harmonics
photoinduced current-voltage characteristics, spectral response and
d.c. current-voltagecharacteristics. One can switch electronically between
experiments and perform any number of techniques without moving the
cell or removing the electrode from the electrolyte. A variable neutral
POTENTIOSTAT
OSCILLOSCOPE
'I
FUNCT'ON
GENERA'Cr7
- COMPUTER
Figure 8. The experimental arrangement of the modulation spectroscopy

workstation (16b).
density filter, placed between the monochromator and the sample, is

adjusted by a computer controlled motor to fit the requirements of each
experiment. For EER and PR experimentsthefilter isadjusted to keep the
d.c. reflectivity constant. For SPV experiments the filter is adjusted to
keep the open circuit voltage constant, and for spectral responsethe filter
is adjusted to keep the intensity of the probe beam constant enabling
direct measurements of unnormalized quantum efficiencies.
Electroreflectance. The system’s response here is the surface
reflectivity. The technique is based on modulating the reflectivity of the
semiconductor with low frequency bias voltage across the device. If the
junction is being formed with an electrolyte, the technique becomes
known as Electrolyte Electroreflectance (EER). Excellent review articles
cover the applications of this technique for analysis of optical properties
of semiconductors and in the determination of the potential distribution
across the interface of a solid or a liquid junction (59).
Often one can work within a theoretical framework known as the
“low field regime” (60). If one includes the possibility that the Fermi level
is pinned at the surface (61), then the ratio between the modulated
reflectance (AR) and the reflectivity (R) is given by:
Eq. (10) AR/R = -(2eN,,/~) 11 - (e/C,)(dN,,/dV)] -aV-L(E)
where e is the electronic charge, E the energy, N, the doping level, E the
dielectric constant, C, the capacitance of the Helmholtz layer (that
corresponds to the entire surface coveredwith states that can be charged
and discharged upon modulation of the surface potential), N, is the
densityof thesurfacestates,AVistheamplitudeof themodulatingvoltage
and L(E) is a spectral line-shapefunction given by:
L(E) = Re[C(E - E, t i I?)-” exp(iB)]
where C and B are amplitude and phase factors, n is a number characteristic

of the interband critical point, E, is the energy gap and I’ is a broadening
parameter related to the lifetime of the majority carriers.
From analysis of the spectral line-shape one can get information
about the optical properties through evaluation of n and E, and about
properties of the majority carriers, hence presence of defects that act as
scattering centers, through evaluation of B and I’.
From analysis of the variations of the spectral amplitudewith the d.c.
bias voltage and the modulation amplitude (Eq. lo), one can obtain a
measure of the flat band voltage that directly relates to the properties of
the surface and serves as a reference point in any attempt to understand
the distri bution of the electric field at the interface. One also can obtain the
doping level through evaluation of N, and actively trace the shifts of the
Fermi level with the applied bias. When (e/C,)dN,,/dV = 1, the Fermi
level is completely pinned, all the potential drop is at the surface and the
electroreflectancesignal reduces to zero. If, on the other hand, dN,,/dV
= 0 either because there are no surface states or because their energy is
such that they do not charge and discharge as the surface potential is
changed, all the potential drop falls on the space-charge layer, the
junction behaves as an abrupt junction and the electroreflectance signal
is independent of the d.c. bias. From the quantitative evaluation of the
functional dependence of the spectral amplitude with the d.c. bias, the
density and energy of the surfacestates responsiblefor the pinning of the
Fermi level can be obtained (61). An example, seen in Fig. 9, was taken
from the dependence of the E, transition of Si in the methanolic solution,
previously discussed (16a). The flat band can be seen at the potential in
which the signal inverts sign and two potentials ranges can be seen in
which the Fermi level is pinned: one narrow range around -0.4 volt vs. Pt.
due the surface states mentioned before, and another one at potentials
positive to -0.2 volt vs. Pt, probably due to accumulation of minority
carriers at the surface. Recently it was demonstrated that one can
determine the potential distribution under illumination by performing the
experiments with a second light source (62).
The power of the technique can be appreciated if one considers that
in a multicomponent system one can obtain the electric field distribution,
with all its implications, for each component, if the spectral transitions can
be resolved.
Photoreflectance. This technique is almost identical to
electroreflectance except that the electric field is not modulated by
changing the bias voltage across the device, but instead, the system is
irradiatedwith a second light source that induces a photovoltage between
the surface and the bulk. Since the photovoltage can be correlated with
the light intensity, it is possible to analyze the data using the theoretical
machinery that was developed for electroreflectance. This is demonstrated
in Figs. 10 and 1 1. In Fig. 10 we compare photoreflectance in air with the
electroreflectanceand the photoreflectance in the methanolic solution of
the E, transition of Si (16b). One can observe some differences in line
shape due to the difference in the penetration depth of the electric field
and the light, but the similarity is obvious. Figure 11 shows the variation
ofthePR amplitudeof Siwith the intensityofthepumping beam (16b). The
logarithmic relationship, the same as the one observed for the photovoltage,
is evident. Although this method is somewhat indirect compared with
electroreflectance, it offers a unique opportunity to obtain electrical
transport properties without the need to form a junction. The power of
-1.0 -0.5 0.0 0 . 5 -I

Potential ( V vs. Pt)
Figure 9. Variation of the amplitudes of the 3.4 eV EER peaks with the
electrode potential: (a) after HF etching; (b) before etching. The potential
sweep rate was 5 min. per experimental point. The lines are drawn for
convenience of inspection. The arrows indicate the direction of the
potential sweep (1sa).
1.01 1
m
0
x
lx
1
0.0
lx
a
-1.0
3.0 3.2 3.4 3.6 3.8
Energy (eV)
Figure 10. Comparison between Photoreflectance (PR), Electrolyte-

Electroreflectance(EER) and Photoreflectancein an electrolyte (EPR) of
n-Si. The EPR and the EER were measured in the methanolic solution of
oxidized and reduced dimethylferrocene in the following concentrations:
0.01 M FeCp,, 5 0 p M FeCp,' and 1 M LiCI0,supporting electrolyte under
nitrogen atmosphere. The EPR was measured under open circuit conditions
and the EER at a potential of 0.0 volt vs. Pt. The chopping frequency in all
I I
0.oL
I to-' lo-* IO-^
Intensity of Furnping Eeam (Arb. Units)
Figure 11. Dependenceof the PR signal of n-Si, on the light intensity of
the modulated beam (16b).
photoreflectance as a nondestructive spectroscopic tool can be seen in

Fig. 12, which shows the spectrum of the various confined and unconfined
PR transitions of a multiple quantum well, made of GaAs/Gao,82Alo,,8As,
with well and barrier widths of 71 A. Also shown is the theoretical fit of the
line shape to superposition of Gaussian derivatives that were found to be
more suitable for quantumwell structures than the Lorenztianline shapes
of Eq. 1 1 (63). Figure 13 shows the PR spectra of Si samples that were
subject to various RIE etchants (16c). The shift in peak position was
interpreted in terms of strain induced by the damage to the Si surface by
the treatment.
Since PR is a contactless technique, one cannot induce electric field
with a power supply, instead the space-charge layer is being formed by
equilibration of charge between the bulk and surface or interface states
(64). The surface states can either be traced to modified intrinsic states
or be induced by absorption of impurities. Since this is a contactless
technique, at equilibrium charge neutrality requires that the amount of
charge needed to form the space-charge layer will originate from the
surface (65). From Eq. 2, within the Schottky approximation for an abrupt
junction, the barrier height at the surface is given by:
where e is the electronic charge, N, is the concentration of the donors, d

is the thickness of the space charge layer and 6 is the dielectric constant.
The number of charges per unit area on the surface, N, arising from
electrons or holes exhausted from a distance d from the surface, is given
by :
N, = -N,d
Combining Eqs. 12 and 13 results in:
V, = eNsS2/2cN,
Weisz(66)wasfirst topointoutthatforaconstantV,, since N,,varies

as the square root of the impurity concentration, no matter how impurethe
sample, one is limited to a fraction of a percent of a monolayer of
equilibrium ionosorption, when depletion layer is present.
The last argument can also be taken in reverse:for charge neutrality
Energy (eV)
Figure 12. Photoreflectance spectrum at 300K (dotted line) of GaAs/

Gao,,,AIo,,,As multiple quantum well, with well and barrier widths of 71 8.
The solid line is a least-square fit of the experimental data to a first-
derivative of a Gaussian line shape form. The arrows indicate the energies
of the various transitions as obtained from the fit (63).
Energy (eV)
Figure 13. Photoreflectance spectra of n-Si subjected to various RIE
treatments. Temperature 77OK. The modulation source is 7.5 mW He-Ne
laser; modulation frequency: 750 Hz (16c).
to be maintained, only small modulation of the surface charge can produce

a very large modulation of the surface potential and produce a large PR
signal. To our knowledge, this aspect of photoreflectance has not yet
been credibly demonstrated. Both, EER and PR can also be used for
measurements of the impedance of the system. It was shown (67) that if
one measures the amplitude of an EER signal as a function of the modulating
frequency over a broad frequency regime, it is possible to correlate the
results with the impedance response over the same frequency regime.
Surface Photovoltage. This technique was first proposed by
Goodman (68), and has since been used both with solid and liquid
junctions. It is based on the observation that for most “good” semiconductors
the photovoltage is related to the minority carriers’ diffusion length through
the following expression:
Eq. (1 5) lau/(l t uLp) = f(V,)
where la is the light intensity, Q the absorption coefficient, Lp the minority

carrier diffusion length and V ,, the photovoltage. la is being adjusted at
various wavelengths to give the same photovoltage. From Eq. (1 5), the
plot of the light intensity against the absorption coefficient directly gives
the value of the minority carriers’ diffusion length that in turn correlates
with presence of defects in the bulk of the semiconductor. Direct comparison
between results taken with AI Schottky barrier and liquid junction on
silicon ribbon gave identical results for the minority carriers’ diffusion
length (69).
A variance that usually appears in the literature as Surface Photovoltage
Spectroscopy (21) uses either a vibrating metal electrode in close proximity
to the surface of the semiconductor (Kelvin probe), or a semitransparent
metal electrode with a modulated light source to make contact with the
semiconductor, and monitors the spectral response of the photovoltage
with subband gap illumination. This technique can also be viewed as a
non-contacttechnique that can be used for in situ characterization during
device fabrication.
Electromodulated Infrared Spectroscopy. The form mostwidely
in use is the electromodulated attenuated-total-reflection-spectroscopy .
This is the electromodulated form of the multiple internal reflection
spectroscopy that was previously discussed. The practice here is to
modulate, by switching the potential between twovalues, and monitor the
absorption changes by lock-in techniques. Description of this technique
with an emphasison metal-electrolyteinterfaces can be found in (70). The
modulation here is different from all the other techniques that were
mentioned in the sense that the modulation is not a small perturbation of
an equilibrium state but a shift between two equilibrium states. Whether
or not one achieves equilibrium in the two states depends on the modulation
frequency. But since this is a true spectroscopic technique in which one
can resolve vibrational modes at the interface and the modulation is used
primarily to increase sensitivity to voltage-induced chemical change at
the interface, the issue of switching between two real equilibrium states
might be irrelevant as long as information is not being derived from line-
shapes.
The semiconductor/electrolyte system most thoroughly investigated
using this technique is the Si/acetonitrile (71-74). The researcherswere
able to observe vibrational transition due to Si-H, CH, and distinguish
between isolated and hydrogen bonded Si-O-H. They were able directly
to observe absorption due to surface states, correlate these states with
the surface chemistry and with treatments such as etching, aging and
oxidation (72). They were able to estimatethe reorganization energy (71)
of the surface states and estimate the potential distribution from the free
carriers’ absorption (74). By itself this series of papers constitutes a tour-
de-force. When and if theseor similar datawill be correlatedwithtransport
properties and UHV spectroscopies, they will provide a solid framework
for our understanding of semiconductor-liquid interfaces.
Other Modulation Techniques. Electric field modulated
photoluminescence in a liquid junction device was reported in Ref. 75.
The results agreed with the “dead layer” model previously discussed.
Quantitative comparison of differentiation of Eq. 1 with respect to the
electrode potential and impedance measurements yielded an agreement
between the “dead Layer” and the space-charge layer.
Electroabsorption was demonstrated in Ref. 76 and the line shape
was found to fit Eq. 1 1. Theoretical justification for this fit was not given.
A summary of other modulation techniques such as wavelength
modulation,temperature modulation and stress modulation can be found
in Ref. 77. These modulation techniques have not yet found any significant
applications in the study of semiconductor-liquid junction devices.
4.0 TIME RESOLVED TECHNIQUES
The overall charge separation process across a semiconductor

liquid junction involvesvarious contributions to the mechanism, details of
which are still lacking. These include the light induced electron-hole
formation, their various recombination mechanisms and their transport
across the interface to react with the electrolyte. Time resolved techniques,
that can now reach the femtoseconds time scale, should be a powerful
tool to elucidate many mechanistic and kinetic aspects of these processes
and provide the best interface with theory. Care should be exercised in
correlating many of the pulsed experiments with their steady-state analogs:

typical laser power of 1 mJ/cm2 per pulse (30 ns) corresponds to steady-
state illumination power of 33 kW/cm2, close to five orders of magnitude
higher than a typical steady-state measurement (78).
4.1 Current
Perhaps the simplest of these techniques are the potentiostatic

photocurrent transients (79) that were shown to be sensitive to the
semiconductor electrodes down to 1 ns. (80) and below (81). Often the
time resolution is limited by the RCof the system and the technique is most
valuable in the longer time scales for identification of intermediates and
products of photo redox reactions (79). The interpretation of the data
follows the routine in some of the methods that we have explored to
interpret impedance data, i.e., assume an equivalent circuit and analyze
the decay as a superposition of exponential decays where the time
constants are correlated with the elements of the equivalent circuit
(79)(80)(82). The time constant that was associated with the space
charge layer was in reasonable agreement with the Mott-Schottky data
(79)(80). The time-scale of the predicted response (83) is much faster
than the one observed by the authors of Ref. 79, but the much faster
resolution reported in Ref. 81 was in agreement with the time-dependent
version of Gartner’smodel. Etchingwasfoundto havea largeeffect on the
amplitude and decay time of the transients (82). This method was also
applied to the study of dye sensitization and the role of a “super sensitizers”
in these systems (84).
4.2 Potential
An alternative method is the coulostatic-flash technique (85) in

which the system is perturbed by a pulse of charge induced by a pulse of
supra-band-gap photons. The subsequent time-resolved return of the
electrode potential toequilibrium is observed at open circuit. In principle,
in this mode, the system is not limited by the RC of the interface but only
by the parasitic capacitance of the setup. The correlations with the
kinetics of charge transfer are less direct compared to the potentiostatic
mode. The results with TiO, in aqueous electrolyte were in general
agreement with the theory (86) of double layer relaxation. These experiments
were extended to non-aqueous electrolytes (87) where it was found that
surface modification through silanation has a significant effect on the
slower components of the time evolution of the photopotential that was
interpreted in terms of induced surface states that can enhance the
charge transfer to the electrolyte. An algorithm that relates the number of
photogenerated carriers with the observed transients was developed

(88). This theory was applied to the investigation of the time-resolved
photopotential of one of the best characterized semiconductor-liquid
junctions, InP both n and p types, in aqueous and non aqueous electrolytes
(89). In this system one can control to a certain degree the distribution of
surface states by controlled deposition of various metals and the system
appears to be stable to corrosion. This method was also used to determine
barrier height of a leaky Schottky barrier junction between SnO, and p-Sic
(90).
4.3 Photoluminescence
Perhaps the most informative variant of the time-resolved techniques

is the one based on photoluminescence (91). The time resolution can
extend below the picosecond regime. In addition to band-gap emission
one can monitor the emission due to specific recombination centers. The
disadvantages are that one is confined to materialsthat emit light and the
sensitivity is such that most of the work is reported under high injection
conditionswhere the system was driven to flat band. The one dimensional
continuity equation under these conditions was solved and the experimental
results with CdS were analyzed to yield the surface recombination velocity
that was found to be affected by the choice of electrolyte (91). The time-
resolved study of cathodic and anodic electroluminescenceof ZnO in the
psec time scale was reported (92).
4.4 Microwave Conductivity
The method uses the proportionality between the relative change of

reflected microwave power and the conductivity induced by a light pulse
in a solid embedded in a waveguide system (93). Time resolution of
nanoseconds was reported.
4.5 Surface Restricted Transient Grating
This powerful method was only recently applied to semiconductor-

liquid interfaces by R.J. Dwayne Miller et at. (94). Two time coincident
supra-band-gap beams are used to image an optical interferencepattern
on the semiconductor surface. The excitation promotes electron-hole
pairs that decay into free carriers. The spatial modulation of the carrier
concentration creates a periodic variation in the material index of refraction
that mimics the optical interference pattern. The holographic diffraction
grating thus formed is being probedwith avariablydelayedsub-band-gap
probe. The grating image decaysaccording to the carrier population and
thus measures directly the dynamics of the carriers. The strong electric
field at the solid-electrolyte interface can focus the pattern at the interface
by driving minority carriers to the surface and majority carriers away from
the surface in a time scale that is short compared to the measurement. In
addition the grating image excites a surface acoustic mode (94) with its
own temporal and spatial resolution that increases considerably the
information contents of these experiments. The time scales that were
probed are from subpicoseconds to nanoseconds. The experimental
work is supported by Monte Carlo computer simulations (95) of the
dynamics of some of these systems to offer the hope that our understanding
of the dynamics of charge carriers at the solid-liquid interfaceswill approach
our understandingof charge transport processes in homogeneous systems.
5.0 PHOTOTHERMAL METHODS
During the course of photoelectrochemical reactions only a portion

of the light energy absorbed by the semiconductor can be utilized in the
electrode reaction. The remainingenergy is being dissipated as heat. The
heat should cause a local temperature change that can be monitored by
placing a suitable sensor near the surface of the electrode (96).
The simplest sensing device isa thermistor that can directly measure
the temperaturechanges (97). The temperaturechangeswere correlated
with the quantum efficiency. A serious deficiency of this method is the
slow equilibration time of the thermistor ( - 1 sec.). The equilibration time
can be considerably reduced by shifting to photoacoustic detection
either in the form of a microphone (98) or a piezoelectric transducer (99).
These developments follow the success that photoacoustic spectroscopy
had in determining absorption spectra of solids (100) and efficiency of
photovoltaic devices (1 01).
Variants of these techniques are the Photothermal Deflection
Spectroscopy (PDS or “Mirage” effect) and Photothermal Displacement
Spectroscopy (1 02). These techniques are based on deflection of a light
beam due to refractive index gradients either in a fluid (or air) in contact
with a light absorbing solid or in the solid itself. If the fluid is inert the
technique can be used to measure absorption spectra of solid materials
and transport properties. A version of these techniques was applied to
electrochemical and photoelectrochemical systems (1 03). The authors
describe the experimental conditions needed to separate the contributions
from the temperature and concentration gradients. Once this is done the
results can be correlated with the kinetics and mechanism of the
electrochemical reactions.
6.0 TOPOGRAPHICAL STUDIES
One of the most convenientfeatures of liquid junctions is that almost

every one of the various techniques mentioned so far can be used in
conjunction with a computer-controlledX-Y stage to provide a topographical
image of a particular parameter to which the technique is sensitive. The
simplest one is the photocurrent (104). Since many of the corrosion
products are semiconducting oxides, this technique was used to detect
corrosion on metal structures (105). Figure 14 shows an example of an
electroreflectance scan ( 106) across the surface of GaAs. From Eq. 10,
one can see that the amplitude of the EER signal is proportional to the
doping level. The scan was interpreted to represent contours of doping
levels. The resolution reported in most of these studies is of the order of
fractions of a millimeter. Yet, the optics and the stage can bring the
resolution down to that of an optical microscope, that is, in the p m range.
The resolution limitingfactor in many of the modulated optical techniques
is the low signal to noise ratio. For a typical EER signal of 10-4,the signal
accumulation time is about 1 sec. that will amount to a resolution of 0.2
mm for one hour experiment on a 1 cm2crystal. If, in addition, one has to
monitor the peak position in every point the spatial resolution will be
further reduced.
7.0 ACKNOWLEDGEMENT
This work was supported by ONR.

GaAs 2756 ( 8 3 7 0 ; )
Figure 14. Contours of thevariations in carrier concentrations across the

surface of GaAs utilizing the electrolyte electroreflectance signal at 8370A.
The numbers in the contours indicate percentagevariationsrelative to the
lowest concentration (1 06).
8.0 REFERENCES
1. See for example:

Tomkiewicz, M., P r o c . SPl€, 452: 120 (1984)
2. See for example:

Chazalviel, J-N., € / e c t r o c h i m i a Acta, 33:461 (1988)
3. See for example:

Oehrlein, G.S., Phys. Today, 39:26 (1986)
4. See for example:

a. Brillson, L.J., J . Vac. S c i . Techno/., 16:1137 (1979)
b. Spicer, W.E., Chye, P.W., Skeath, P.R., Su, C.Y. and Lindau,
I., J . Vac. S c i . T e c h n o / . , 16:1422 (1979)
c. Freeouf, J.L. and Woodall, J.M., Appl. Phys. Lett., 39:727
(1981)
5. See for example:

Lewis, N.S., Rosenbluth, M.L., Casagrande, L.G. andTufts, B.J.,
in: Homogeneous and Heterogenous Photo-catalysis, (E.
Pelizzetti and N. Serpone eds.), 343, D. Reidel Pub. Co. (1986)
6. For references that cover the early work at Bell Telephone Laboratories
see: Brattain, W.H. and Boddy, P.J., J . Electroch. SOC.,109574
(1962)
7. See for example:

a. Gerischer, H. in: Physical Chemistry: An Advanced
Treatise, (H. Eyring, D. Henderson and W. Yost, eds.), p.
463, Academic Press, New York (1970)
b. Memming, R., J . f / e c t r o c h . S o c . , 116:785 (1969).
c. Myamlin, V.A. and Pleskov, Y.V., € / e c t r o c h e m i s t r y of
Semiconductors, Plenum Press, New York (1967)
8. Fujishima, A. and Honda, K., Nature, 238:37 (1972)
9. See for example: Horowitz, G., Allongue, P. and Cachet, H., J .

Electroch. S o c . , 131:2563 (1984)
10. Rosenbluth, M.L. and Lewis, N.S., J . Am. Chem. SOC.,108:4689

(1986)
11. Licht, S., Tenne, R., Dagan, G., Hodes, G., Manassen, J., Triboulet,
R., Rioux, J. and Levy-Clement, C., Appl. Phys. Lett., 46:608
(1985)
12. See for example:

Kern, W. and Deckert, C.A., in: Thin Film Processes, (J.L.
Vossen and W. Kern eds.), p. 401, Academic Press, New York (1978)
13. a. Tenne, R., Appl. Phys., 25:13 (1981)

b. Ostermayer, F.W. Jr., Kohl, P.A. and Burton, R.H., Appl. Phys.
Lett., 43:642 (1983)

Divigalpitya, W.M.R. and Morrison, S.R., J. Appl. Phys., 60:406
(1986)
15. Yeager, E., Homa, H., Cahan, B.D. and Scherson, D., J. Vac. Sci.
Tech., 628 (1982)
16. Series of five papers by the common title “Liquid Junctions for
Characterization of Electronic Materials” in J. Appl. Phys. (1989)
a. Fantini, M.C.A., Shen, W.M., Tomkiewicz, M. and Gambino, J.P.,
J . Appl. Phys., 65:4884 (1989)
b. Shen, W.M., Fantini, M.C.A., Tomkiewicz, M. and Gambino, J.P.,
J . Appl. Phys., 66:1759 (1989)
c. Shen,W.M., Fantini, M.C.A.,Pollack, F.H.,Tomkiewicz, M.,Leary,
H . and Gambino, J.P., J . Appl. Phys., 66:1765 (1989)
d. Fantini, M.C.A., Shen, W.M., Tomkiewicz, M. and Gambino, J.P.,
J . Appl. Phys., 66:2148 (1989)
e. Fantini, M.C.A., Shen, W.M., Tomkiewicz, M. and Gambino, J.P.,
J . Appl. Phys., 66:4846 (1989)
17. Bard, A.J. and Faulkner, L.R., in: €lectrochemical Methods, p.

298., John Wiley & Sons, 1980
18. Memming, R., B e r . Bunsenges. Physik. Chem., 81:732 (1977)
19. Miller, B., Menezes, S. and Heller, A., J. Electroch. SOC.,127:1483

(1979)

Burns, G., in: Solid State Physics, Ch. 13, Academic Press
(1985)
21. Luth, H., Appl. Phys., 8:l (1975)

Luminescence of Inorganic Solids, (P. Goldberg, ed.),
Academic Press (1966)
23. Pujadas, M., Gandia, J., Salvador, P. and Decker, F.,J . €lectroanal.
Chem., 218:347 (1987)
24. Silberstein, R.P. and Pollak, F.H., J . Appl. Phys., 51:1743 (1980)
25. Harrik, N.J., in: lnternal Reflection Spectroscopy, John Wiley

and Sons (1 967)
26. Chiarotti, G., Nannarone, S., Pastore, R. and Chiaradia, P., Phys.
Rev., B4:3398 (1971)

a. Palik, E.D., Holm, R.T. and Stella, A., J . Appl. Phys., 57:843
(1 984)
b. Spitler, M.T., in: ,Electrochemistry: Fundamental Processes
andMeasurement Techniques, (W.Wallace, A.Nozikand S . Deb
eds.), p. 282. Electrochemical Society (1 982)
28. Ellis, A.B., J. Chem. Educ., 60:332 (1983)
29. Garuthara, R., Tomkiewicz, M. and Tenne, R., Phys. Rev., 831 :7844
(1 985)
30. Garuthara, R . and Tomkiewicz, M., J . Appl. Phys., 58:1662 (1985)
31. Streckert, H.H., Tong, J., Carpenter, M.K. and Ellis, A.B., J . Electroch.
SOC.,129:772 (1982)
32. Garuthara, R., Tomkiewicz, M. and Silberstein, R.P., J. Appl. P hys.,

54:6787 (1983)
33. Memming, R., J . € l e c t r o c h . SOC.,116:785 (1969)
34. Gautron, J., Delbera, J-P. and Lemasson, P., Surf. Sci., 99:300
(1980).

Aspnes, D.E., J . Phys., 44:C10-3(1983)andSurf.Sci., 101:84
(1980)
36. Raccah, P.M., Garland, J.W., Zhang, Z.,Lee, U., Xue, D.Z, Abels,
L.L., Ugur, S . and Wilinsky, W., Phys. Rev. Lett., 53:1958 (1984)
37. Meyer, F., Phys. Rev., B9:3622 (1974)
38. Dorn, R. and Luth, H., Phys. Rev. Lett., 33:1024 (1974)
39. Mclntyre, J.D.E.andAspnes, D.E., Surf. Sci., 24:417(1971)
40. Mu, X.C., Fonash, S.J., Yang, B.Y., Vedam, K., Rohatgi, A. and
Rieger, J., J . Appl. Phys., 58:4282 (1985)
41. MacDonald, J.R., in: Impedance Spectroscopy, John Wiley &

Sons (1987)
42. Raistrick, I.D., Ann. Rev. Mater. Sci., 16:343 (1986)
43. Shen, W.M., Aurian-Blajeni, B., Tomkiewicz, M. and Cahen, D., J .

€lectroch. SOC., 133:930 (1986)
44. Tomkiewicz, M., J . Electroch. SOC.,126:2220 (1979)
45. Kramer, M. and Tomkiewicz, M., J . Electroch. SOC., 131:1283

(1984)
46. Tomkiewicz, M. and Aurian-Blajeni, B., J . Electroch. SOC.,(1988)
47. Liu, H.S., Phys. Rev. Lett., 55:529 (1985)
48. Sze, S.M., Physics of Semiconductor Devices, John Wiley &

Sons, New York (1981)
49. Schottky, W., Z.Phys., 113:367 (1939); 118:539 (1942)
50. Mott, N.F., Proc. R. SOC.London, Ser A, 171:27 (1939)
51. Shen, W.M., Tomkiewicz, M., Sedaries, D. and Levy-Clement, C., J .

€/ectroch. SOC., 137:2656 (1990)
Photoelectrochernical Characterization 251
52. Tomkiewicz, M., Surf. S c i . , 101:286 (1980)
53. Braun, C.M., Fujishima, A. and Honda, K., S u r f . S c i . , 163:369

(1985)

a. McEvov, A.J., Etman, M. and Memming, R., J . Electroanal
Chem., 190:225 (1985)
b. Chandrasekaran, K., Kainthla, R.C. and Bockris, J. O’M.,
Electrochimica Acta, 33:327 (1988)
55. Haak, R. and Tench, D., J . € l e c t r o c h . SOC.,129:275 (1984)
56. Lang, D., J . Appl. Phys., 45:3023 (1974)
57. Tomkiewicz, M. and Shen, W.M., in: Photoelectrochemistry and

€ l e c tr osynrh esis On Semiconducting Ma terials, (D.S. Ginley,
A.J. Nozik, N. Armstrong, K. Honda, A. Fujishima, T. Sakata and T.
Kawai eds.), The Electrochemical SOC.,Inc. (1987)
58. Luo, J., Lin, Z.H.andTian, Z.W., Photochemical Conversionand

Storage of Solar Energy, Abst. E41, Proc. 7th. Internat. Conf.,
Evanston, Ill (1988)

Pollak, F.H., in: Photoelectrochemistry: Fundamental
Processes and Measurement Techniques, (W.L. Wallace,A.J.
Nozik, S.K. Deb and R.H. Wilson, eds.), The Electrochemical SOC.
(1982)
60. Aspnes, D.E., Surf. S c i . , 37:418 (1973)
61. Tomkiewicz, M., Siripala, W. and Tenne, R.. J . Electroch. SOC.,

131:736 (1984)
62. Tafalla, D., Pujadas, M. and Salvador, P., Photochemical

Conversion and Storage of Solar Energy, Abst. E l 3 , Proc. 7th.
Internat. Conf., Evanston, Ill (1988)
63. Shen, H., Pan, S.H., Hang, Z., Pollak, F.H. and Sacks, R.N., Sol. S t .
Comm., 65:929 (1988)
64. Risch, L., Phys. Slat. Sol., 88:111 (1978)

65. Morrison, S.R., in: The Chemical Physics of Surfaces, Plenum

Press, NY (1977)
66. Weisz, P.B., J. Chem. Phys., 21:1531 (1953)
67. Silberstein, R.P., Lyden, J.K., Tomkiewicz, M. and Poilak, F.H., J .

Vac. Sci. Techno/., 19:406 (1981)
68. Goodman, A.M., J . Appl. Phys., 32:2550 (1961)
69. Micheels,R.H.andRauh,R.D.,J . Electroch. SOC.,131:217(1984)
70. Bewick, A. and Pons, S., in: Advances in I n f r a r e d a n d Raman

Spectroscopy, Vol. 12, (R.J.H. Clark and R.E. Hesler eds.), p. 1.,
Wiley-Heyden, Chichester (1985)
71. Chazalviel, J-N. and Rao, A.V., J . Electroch. SOC., 134:1138

(1987).
72. Rao, A.V. and Chazalviel, J-N., J . Electroch. SOC., 134:2777

(1987)
73. Rao, A.V., Chazalviel, J-N. and Ozanam, F., J . Appl. Phys., 60:696
(1986)
74. Tardella, A. and Chazalviel, J-N., Phys. Rev., 832:2439 (1985)
75. Garuthara, R., Tomkiewicz, M. and Silberstein, R.P., J. Appl. Phys.,

54:6787 (1983)
76. Pujadas, M., Gandia, J., Salvador, P. and Decker, F., J. Electroanal.
Chem., 218:347 (1987)
77. Cardona, M., in: Modulation Spectroscopy, S o l i d State Physics,

Supp. 11, (F. Seitz, D. Turnbull and H. Ehrenreich, eds.), Academic
Press (1969)
78. Jaegermann, W., Sakata, T., Janata, E. and Tributsch, H., J.

€lectranal. Chem., 189:65 (1985)

Deutscher, S.B., Richardson, J.H., Perone, S.P., Rosental, J and
Ziemer, J., Far. D i s c . 70:33 (1980)
80. Willig, F., Bitterling, K., Charle, K.P. and Decker, F., Ber. Bunsenges.
Phys. Chem., 88:374(1984)
81. Bitterling,K.andWillig,F., J . €lectroanal. Chem., 204:21l(1986)
82. Harzion, Z.,

Croitoru, N. and Gottesfeld, S., J . Electroch. SOC.,
128:551(1981)
83. Laser, D. and Bard, A.J., J . € l e c t r o c h . SOC.,123:1837(1976)
84. Kojima, T.,

Ban, T.,
Kasatani, K.,Kawasaki, M. and Sato, H., Chem.
Phys. Lett., 91:319(1982)
85. Perone, S.P., Richardson, J.H., Deutscher, S.B., Rosenthal, J. and

Ziemer, J.W., J . € l e c t r o c h . SOC.,127:2580(1980)
86. Feldberg, S.W., J . Phys. Chem., 74:87(1970)
87. Kamat, P.V. andFox, M.A., J . Phys. Chem., 87:59(1983)
88. Gottesfeld, S.and Feldberg, S.W., J . Electroanal. Chem., 146:47

(1 983)
89. Cook, R.L., Dempsey, P.F. and Sammells, A.F., J . € l e c t r o c h .

SOC.,133:2287(1986)
90. Itoh, K., Matsumoto, H., Kobata, T. and Fujishima, A., Appl. Phys.
Lett., 51:1685(1987)
91. Evenor, M., Gottesfeld, S., Harzion, 2. and Huppert; D., J . Phys.
Chem., 88:6213(1984)
92. Fichou, D. andKossaney, J., J . Electroch. SOC.,133:1607(1986)
93. Kunst, M., Beck, G.andTributsch, H., J . Electroch. SOC.,131:954

(1 984)
94. Kasinski, J.J., Gomez-Jahn, L.A., Faran, K.J., Gracewski, S.M. and
Miller, R.J. Dwayne, J . Chem. Phys., 90:1253(1989)
95. Zhou, X., Hsiang, T.Y. and Miller, R.J. Dwayne, J . Appl. Phys. -- in
press
96. Fujishima, A., Maeda, Y. and Honda, K., Bull. Chem. SOC.Jap.,
2735 (1 980)
97. Fujishima, A,, Maeda, Y., Honda, K., Brilmyer, G.H. and Bard, A.J., J .
Electroch. S O C . ,127:840(1980)
98. Masuda, H.,Fujishima,A.andHonda,K., Chemistry Letters, 1153

(1 980)
99. Yoshihara, S.,

Aruchamy, A. and Fujishima, A., Bull. Chem. SOC.
Jap., 61:1017(1988)
100. Rosencwaig, A., Opt. Comm., 7:305(1973)
101.Cahen, D., Appl. Phys. Lett., 33:810(1978)
102. Amer, N.M., Proc. SPIE, 294:ll (1987)
103.Decker, F. and Fracastoro-Decker, M., J . Electroanal. Chem.,

243:187 (1 988)
104.Furtak,T.E., Canfield, D.C. and Parkinson, B.A., J . Appl. Phys.,

5:6018(1 980)
105.See for example:

a. Butler, M.A., J . €/ectroch. SOC.,130:2358(1983)
b. Shukla, D. and Stimming, U., Werkstoffe und Korrosion,
40:43(1989)
106.Pollak, F.H., Okeke, C.E., Vanier, P.E. and Raccah, P.M., J. Appl.
Phys., 50:5375(1979)
6
ELECTROCHEMICAL MIGRATION
Giulio Di Giacomo
1 .O INTRODUCTION
Electrolytic metal migration is a mechanism that affects microelectronic

devices causing failure (1). The failure consists of dendritic growth
bridging across adjacent conductors due to metallic ions. The migration
occurs between biased lands and under conditions conducive to
electrocrystallization (2)(3).
A number of conditions must be satisfied to form a dendrite: 1) the
current density must reach high levels at the tip of the dendrite (e.g., 1O3
A/cm2), e.g., achieved through spherical diffusion: 2) a sufficient liquid
medium (polar), such as condensed water, must be present in the migration
path so that adequate ionic flux is developed to sustain the required
dendritic growth rate (3)(4); and 3) the applied voltage must exceed the
sum of the anodic and cathodic potential in equilibrium with the electrolyte.
The conditions dictate that the materialssurfaces or interfaces must have
the physical and chemical properties for adequate water condensation to
the extent that the current density requirement at the tip of the growing
dendrite is satisfied (2).
Based on literature findings, there is a minimum rate below which a
dendrite cannot continue to grow because of lattice poisoning at the
surface by contaminantswhich deposit on the crystallographic facetsand
obstruct the process.
To produce a metal dendrite, the current density at the whisker’s tip
must be orders of magnitude greater than the average current. The
growth is possible only through spherical diffusion of the ionic current on
the dendrite tip, which can be expressedin terms of its radiusof curvature,
r. (The current density, however, must be above a critical value, Jc.)
255
To achieve dendritic growth successfully, the mass flux density Qtfp

must be equal to or greater than the minimum rate below which dendrite
poisoning by contaminants occurs:
where: M = Atomicweight,
d = Distance between electrodes,
U. = Density of dendrite,
,t = Maximum time to bridge across electrodes
beyond which poisoning takes place.
Since the impurities and metal ion concentrations are probably not the
same from site to site, i.e., within a module, the critical current value will
vary accordingly.
It must be emphasized, however, that the magnitude of the critical
current depends not only on the level but also the type of impurities
(among other things) and therefore its value must be considered a
characteristic of the particular process.
According to Price et al. (3),the minimum time requiredto completely
block the dendrite’s growth can beapproximatedfrom the rate of impurity
diffusion to the dendrite’s tip and the number of molecules needed to
block a unit area. The time, therefore, will vary from impurity to impurity
and from metal to metal.
The strong effect of impurity atoms and molecules on the dendrite
crystal growth can be explained in terms of the edge energy of the lattice
step per unit area ( s ) and the free energy decrease per unit volume (AG)
accompanying the deposition process (3). The atomic step can advance
(during deposition) between two strongly adsorbed impurities only if the
distance between them is greater than twice the critical radius of curvature,
Rc, of the advancing step. In terms of energy, the steps that will advance
and add to dendritic growth are those having radii of curvature greater
than the critical value (2) given by Rc = s / A G . Schematically,
Advancing
SteD
On a statistical basis, as the impurity concentration increases, the

critical current density of the metal ions, Jc, required to sustain dendritic
Electrochemical Migration 257
growth also must increase. For a given metal ion concentration, type of
impurity, temperature and dendrite radius, the critical current density for
dendrite growth is proportionalto the cube root of the impurity concentration.
As lower voltages across the gap reduce the current density below the
critical value in more impure sites, the average dendrite growth will
decrease approaching zero with the applied voltage and will show a
superlinear growth dependence on voltage.
Another factor is the capability of the adsorbed layer ofwater to carry
the mass of ions required by the dendritic growth rate which defeats
poisoning. If t, is of the order of one minute and the electrodes are Ag
lands on a substrate 0.1 mm apart, then the required ionic flux to bridge
1 g/cm2 s. For a Ag' mobility of 7x10-4cm2/s V
across the gap is 1 . 7 5 ~0-3
and 5 volt bias, an ionic Ag' concentration of 5 ~ 1 Og/cm3 -~ would be
required to meet the flux density rate and therefore the ,t requirement.
This assumes the electric field to be parallel to the dendrite and no
distortion around the dendrite tip. In reality, the field focuses on the tip due
to the spherical shape of the equipotential lines around it, and therefore
the flux density is larger and the Ag' concentration required lower. On the
other hand, if the thickness of the adsorbed water film is thinner than the
minimum dendrite thickness which is possible,then theoretically not only
the ionic concentration must be greater but the dendritic growth will be
impaired. For a dendrite to succeed, therefore, all these factors must be
in tune and satisfy the requirement of dendrite growth without poisoning.
But even when the ionic concentration is sufficient to start a dendrite, it
may be that the ionic supply is source limited. This would be the case
when just enough ions were produced during a period of incubation but
not enough to sustain a dendritic growth to its completion or to prevent it
from being poisoned. A periodof incubation is usuallyrequired beforeany
dendrite is discovered between lands in test vehicles. The time varies
depending on the chemical environment created by process and test
conditions.
Dendrite failures have been observed for most metals used in the
electronics industry during accelerated tests and under field conditions.
The empirical models are usually a product of three factors (which are
assumed to be separable), each being afunction of one variable. These
factors are voltage, temperature, and relative humidity. This last one has
assumed differentformsvaryingfrom a simple exponential, eBRH, to B.E.T.
form, CRH/(1 -RH)[ 1 t (C- l)]RH(5)(6),and to the statistical pore distribution
erfc (1/2a)In[ K T F ~ ~ R H / ( ~ V , where
~ ) ] ~ , RH is the relative humidity, Band
Care process parameters,r is the poreradius, -y the surfacetension,v, the
molar volume, and a the pore distribution sigma. The parameter C is a
function of temperature, heat of adsorption and the condensation (5-7).
The other two factors usually take the form of and V" where n can
vary from 1 to2. Because of the migration,strong dependenceon RH and

the long testing time required at the low RH levels, most of the test data
available to date were obtained above 50% RH; therefore, the empirical
models do not cover the dry field condition within the module, towhich the
data is projected by extrapolation. When contaminants and process
residuesare present,the migration dependenceon relative humidity may
be shifted to lower RH values because adsorption of moisture by reaction
products and hygroscopic particles will allow condensation to occur at
RH values lower than saturation.
Another aspect of migration that is least explored, therefore least
known, is its behavior in the presence of polymer coatings (4)(8)(9).
Polymers absorb water and other pollutants from the environment and,
under the driving force of an electric field, current leakages are generated
between electrodes. Consequently, the insulation resistance of the polymer
is curtailed. Soluble ionic species within the polymer (e.g., hydrolyzable
chloride...) contribute to the mechanism. Water not only may be the
solvent and vehicle for ionic transport but it can also participate pervasively
in the conduction through electrolysis, especially at the higher voltages.
First, however, the species must be transported to the vicinity of the
electrodes (within the sphere of the electric field influence) via diffusion
under the driving force of the concentration gradient before participating
in the leakage process, thus closing the electrochemical cycle. A steady
state is eventually achieved which will determine the local concentration
of the species and the limiting ionic current across the electrodes. Generally,
at steady state, such transport mechanisms are controlled by diffusion,
which tends to simplify the analysis; however, the process can be very
slow, which makes it necessary to deal with the transient aspect and time-
dependent factors.
In practice, the interfacial insulation resistance of polymers degrades
as the films are exposed to humidity under electric bias (8).Indeed, the
resistance tends to level off with time unless defects in the film or at the
polymer-ceramic interface act as shorts. The process of interfacial
adhesion degradation can continue as a function of time as the bond
weakens under the degrading action of some of the electrolytes. Under
these conditions the leakage steadily increases with time. Much higher
leakages between biased electrodes can occur because of polymer bond
degradation in humid environments rich in sulfurous and nitrous pollutants.
Water drop tests are not very revealing and do not necessarily
predict the migration susceptibility of metals in microelectronics under
field conditions (1). The reason is that metal migration and dendritic
formation depend on the state of the metal surface,whether it is oxidized
or not, and on whether the oxide passivates the surface or it dissolves in
water. For instance, copper will rarely produce dendrites when it is
oxidized due to the oxide tenacity and insolubility in water, but it will
migrate if the oxide is removed. On the other hand, silver will migrate
whether it is oxidized or not since the oxide is water soluble. Insoluble
oxides (and other insoluble compounds) tend to prevent metal migration
because they limit or stop the ionic and electronic exchange required at
electrodes to fulfill the electrochemical cycle (1). Another element that
migrates massively is molybdenumwhich forms water soluble oxides and
hydroxides. The migration, however, does not result in distinct dendrites
but rather mass growth along the entire electrode. Under field conditions
of relative humidity, temperature, and contamination, the propensity for
dendritic formation depends on a host of other variables: process variables,
materials variables, and test condition variables. Therefore, the water
drop test is only an indicator for migration susceptibility between metals
for different surface conditions, to screen out those which would perform
badly in the field under avery high relative humidity condition. Even under
theserestrictions, metalswhich would perform well in the field can also be
rejected by the test.
Another key variable in metal migration is the critical current density,
Jc, which determines whether an electrode will passivate or be inhibited
by adsorption (10-1 3).
It is known that dissolving anodes begin to passivatewhen the value
of J exceeds a critical value ( Jcr)which depends on the natureof the metal,
the composition of the electrolyte, and temperature. According to
Savchenkov and Uvarov (1 l ) , at current densities below J, of the metal,
all the anodic current is spent on the dissolution of the metal, while at
current densities > J,, the metal may be in the active state only a limited
time, t, before the metal changes to passive state. Usually the higher the
current densityof the anode, the smaller the value oft, which is referred to
as the transition time. The length of the transition time depends on the
ratio between the rates of formation and removal of the passive layer
products. Most investigators present this relationship as:
( J - J,,)tn = K
where K is a constant for values of J - J, I10 J,, and n may vary between
0.5 and 3. The value of J, is obtained by measuring t for different values
of Jandextrapolating J to 1/t" = 0. Thevalueof J,,represents thecapacity
of the metal for active dissolution. For metals of the iron group in acid
solutions at 25OC, the value of J, increased from Ni to Co to Fe (1 2).
According to Okamoto et al. (13), the pH of the solution has a significant
effect on J,, expressible (at 25OC) as:
Equation3 indicatesthat the more acidic the electrolyte is, the higher
the value of J, and the harder it is to achieve passivationwhich requires
current values > Jcr. For a water drop test at room temperature (pH 7), J,
is 1 . 3 ~O-1' A/cm2 which allows passivation to occur.
Savchenkov and Uvarovfound that the dependence of the transition
time of Ni passivationon the current density at 25OC can be expressed as
(J-J,,)t = 0.52,andat160°Cas(J- Jcr)t0.5 = 1.64,whereJisinA/cm2and
t in seconds. At room temperature for current densities of the order of
3 . 8 ~0-4
1 A/cm2, the transition time is 23 minutes. At the higher temperature,
a much longer transition time is required to begin passivating as J,
increases with temperature. According to Savchenkov, this behavior
indicates that the electrode is in the active state at a higher temperature
(for which T<T,,), which may be due to the decrease of the degree of
irreversibility of adsorption of foreign particles (hydroxide, oxide) on the
Ni surface.
According to Vagramyan and Uranov (14) at lower temperature the
rate of inhibition is high because the adsorbed particles bind themselves
firmly to the electrode surface so that, in effect, reduction of the metal ions
proceeds through a continuous layer of foreign particles that form a
barrier (cathode). Because this diffusion process requires a relatively
high activation energy, it will control the deposition of Ni which occurs
uniformly. At higher temperature, however,the inhibitioneffectsdiminish,
the discharge of Ni ions occurs at very low overvoltages, and the deposition
of the metal takes place only on the active areas of the cathode surface.
As a result, isolated aggregates of Ni form whose rate of deposition
depends primarily on the supply of the Ni ions to the electrode. The high
polarization during Ni deposition is due mainly to oxides, hydroxides,
hydrogen, and other foreign particles.
Vagramyan and Zhamagortsyants(1 0) have indicated that the sharp
decrease in the overvoltage of cathodic and anodic processes with
increase in temperature may be due to two cases: (1) the direct effect of
temperature on the electrochemical reaction, and (2) the elimination of
the inhibiting effect of adsorbed foreign particles by means of temperature.
The overvoltage for Ni at room temperature for 1 N MeSO, solution is
about 3/4 of one volt ( l o ) , while for Ag, Cu, and Pb, the overvoltage is
about an order of magnitude lower.
2.0 MODEL
2.1 Current Density through an Electrolyte
The current density, J, can be expressed as a function of the drift

velocity, vd,and the ionic concentration C (4):
where F = Faraday's electrochemical equivalent, and Z = valence. The

drift velocity is the difference between the ions' velocity in the direction of
the field (if) and their velocity against it (5):
vd = 2- y' = I [KT/h] e(-MD/U) [e(ME/KT) - ~(-AHE!K'V]
Eq. (5) vd = 2 I [KT/h] e(-%'U) sinh(AH,/KT)
where: I = jump distance,

A HD = diffusion component of free energy of
activation,
AH, = 1/2 ZFIE, electric component of free
energy of activation representing the
electric work from the equilibrium position
to the top of the barrier,
h = Planck's constant,
K = Boltzmann constant.
Substituting vd from Eq. 5 into Eq. 4:
J = 2ZFCI [KT/h] e(-&HdU) sinh ZFIE/2RT
For low electric fields such that 1/2 ZFIE/RT < < 1:
J = (2F)'C [ I'KT/h] E/RT
J = {[(ZF)'CE]/RT} Do
Eq. (6) J = [(ZF)2CDE]/ RT
For higher fields,
Eq. (7) J = (2ZFCD/I)epE
where p = ZF1/2RT.
Water Availability as a Function of RH. The ionic current is rate-

controlled by the availability of water in the form of an adsorbed layer or
condensate within interconnected pores (15). There exists a probability
that the current density J, and therefore the dendritic growth, will take
place. The probability is assumed to be proportional to the adsorbed or
condensed water film cross section characterized by its thickness (such
as given by the B.E.T. adsorption isotherm) or based on pores cross-
sectional area distribution (15). A time-to-failure, t,, will, therefore, depend
on the product of the current density, J, and the relative humidity or
probability function, P(RH) which can be termed statistical J,:
Eq. (8) t, = A/JxP(RH) = A/J,
where A is a proportionality constant. It is assumed that t is inversely

proportional to J, and, for simplicity, that the current density is entirely due
to the metallic ions. The migration process in humid environment, therefore,
must rely on the presence of a water film as the medium in which the metal
ions will dissolve and through which they will migrate under the influence
of an electric field. The water film thickness at any point dictates the ionic
flux density and concentration requirementsfor dendrite growth without
being poisoned by impurities. The average thickness of the water film
adsorbed on a substrate is a function of relative humidity and temperature,
which can be expressed by adsorption isotherms. It is assumed that the
probability of successfully growing a dendrite across an electrically
biased gap between the two conductors is proportional to the average
thickness of the water film or its cross section participating in the ionic
transport for the dendritic growth. As the water film grows with relative
humidity, more and more sites within a gap will achieve adequate film
thickness to meet the physical and chemical requirements to satisfy the
dendritic growth demands. The concept of film thickness applies equally
towater films at interfaceswhen porous polymeric coatings are applied at
the surface.
B.E.T. RH -Function. The most used adsorption isothermfunction
is the B.E.T., named after Brunauer, Emmett, and Teller (5). It expresses
the number of water monolayers or cross-sectional area, S , as follows:
Eq. (9) S a [m RH]/{(l- RH)[l + (m - 1)RHI) = P(RH)
where: m = e[-(AHv-AnA)’RT]determined experimentally,

AH,, = Heat of water vaporization, 10.5 kcal/mol,
AHA = Heat of adsorption which depends on the
material and surface conditions. It is
determined from the knowledge of m and
AH,.
Equation9 hasa singularity at 100% RH; however,it is suitable for relative

humidities below 90% where the adsorption theory reasonably applies.

B.E.T Tirne-to-Failure Model for Dendrites. Substituting Eqs. 6
and 9 into Eq. 8, we have,
Eq. (10) ,t = A/{[(ZF)'CDE]/RT) mRH/{(l- RH)[1 t ( m -1)RHI)
where: D = ionic diffusivity in water,

A = a function of the material and process.
Equation 10 allows the determination of AH, by experimentally determining

the time-to-failure, t, at two or more temperatures with fixed RH and E,
assuminga constant C in the process. On this basis, Eq. 10 is rewritten as:
Eq. (1 1) t, = B / mRH/{(1- RH)[1 t ( m -1)RHI) /TI-E
where 6 = AK/(ZF)'CD,, experimentally determined. For known C, the

constant A can be determined and C could beexercised as a variable if so
desired; otherwise C is an unknown in practice and plays no role except
if it represents ionic contaminants, in which case it would have to take the
appropriate empirical form from a suitable experiment. The time-to-
failure is not the time that a dendrite takes to bridge across a gap after it
has succeeded to start, but rather the time determinedfrom the statistical
current densitywhich is basedon theprobabilityof havingsufficientwater
and propitious conditions for nucleation and growth. The probability of
occurrence among the many sites increases as the test conditions become
more stressful.
The model expressed by Eq. 11 is used to project the time-to-failure
from test to field conditions, making use of the failure acceleration due to
relative humidity, temperature, and electric field. The model has been
verified experimentallyand has been consistent. The values of B, m, and
A H,will varywith contaminants,materialand surface condition. The ionic
impurity environmentisassumed not to vary among all samples under test
within the same experiment. The metallic ion concentration will vary with
the samples' contamination due to process residues and other ionic
species that may be present in the system. Adding chemicals to enhance
the oxidation/reduction process, therefore the electrodeposition, is another
way to acceleratethe formation of dendrites, however, it would be difficult
to relate the results to field conditions if a new mechanism is introduced.
RH Function Based on Pore Distribution. Another way of expressing
the relative humidity effect on metal migration is derived from the assumption
that a surface or an interface between a film and a substrate can be
imagined as consisting of a log-normally distributed pore-network in
which water condensation will first occur in the smaller pores as the
relative humidity is raised. The condensation in pores occurs according

to the Kelvin equation which expresses the vapor pressure, p, at which
condensation occurs within a cylindrical pore as a function of the pore
radius, r, the surface tension, 7,the molar volume, urn,and temperature, T:
where p, is the saturated vapor pressure above a flat surface (r = ), and

9)
p/po is the relative humidity at which condensation occurs in a pore of

radius r (5)(6). For a given material and pore frequency distribution, the
ionic current density and mass flux density are assumed proportional to
the cumulative cross section of the pores where water has condensed.
For a log-normal distribution, the fractional area P in which condensation
has occurred (small pores first) can be written as (1):
Eq. (13) P = (I /2) erfc (1 /2a) In (F/r)2
where: u = In (rs0/ rI6), which defines sigma,

rso = marks 50% of cumulative cross-sectional
area made up by pores having r < rs0,
= marks 16% of the area.
r16
Equation 13 is assumed to be the probability to have a water path.

SubstitutingrfromEq. 12intoEq. 13,weexpresssuchprobabilityinterms
of relative humidity, temperature, water surface tension:
Eq. (14) P(RH) = erfc (1/2a) In [KTiIn RH / (2 7u,,)I2
Note that since RH replaces the variable r and full condensation is

assumed in all pores (not just up to F)when RH = 1, Eq. 14 had to drop the
1/2 coefficient, thus conforming to theconditions: P(RH) = 1 when RH =
1, P(RH) = 0 when RH = 0.
t,for Dendrites Based on Pore Distribution. Substituting Eqs. 6
and 14 into Eq. 8 , we have,
Eq. (15) t, =B'/E.(e(-"Hdm)/T)*erfc(l/2~)In [KTrlnRH/(27vm)I2
where: 6' = A' K/(ZF)2CDo

E = V/d
d = gap between conductors.
The parameters B, AH, C and u are experimentally determined from

the time-to-failure (i.e., go)as afunction of one variable at a time: RH, T,
and E. The parametersr andaare determinedfrom the t,,values obtained

from two or more test cells run at different relative humidities and fixed T
and E. Similarly,AH, is determinedfrom the ,t results obtained at two or
more different temperatures,while RH and E remain constant. The same
is done with the electric field as a variable.
Both models, Eqs. 11 and 15, are functions of the same three test
variables and both have been used to interpret the results of Ag migration
in vehicles having epoxy and silicone backseats. The models have been
used to project electric failures (due to Ag dendrites grown under accelerated
test conditions) to field conditions, thus allowing the calculations of failure
rate based on a distribution of fails in time.
2.2 Current Density through a Polymer Coating
The model must deal with the situation encountered when the electrodes
are coated with a polymer and ionic leakage eventually causes failure.
Ionic species and molecules diffuse to the electrodes as migration takes
place under the effect of the electric field, the latter governed by the
electrolytic process. Before steady state is attained, there is a transient
period during which the incoming diffused species and the migration
component establish concentration profiles and an equilibrium ionic
concentration at the electrodes.
At equilibrium, the rate of ionic transport due to the diffusion component
,i from the sample surfaceequals the transport component iEbetween the
electrodes.
The mass flux density, Q, ofwater and other species that diffuse from
the outside through the polymeric film, destined to be electrolized on
arrival at the electrodes, can be expressed as:
Q = D(C,-Ci)/L
where C, is the molar concentration of the species at the polymer surface,

Ci is the prevailing concentration near the electrodes, and L the average
distance traveled to reach the electrodes driven by the concentration
gradient. Upon electrolysis, the ionic current density, J, produced from
such mass flux is equivalent to:
J, = ZFD (C, - Ci) / L
at steady state, the rate of ionic transport due to the diffusion component,
J, equals the transport rate between the electrodes, J, given by Eq. 6.
However, steady state is achieved after a transient period during which
the ionicconcentration,Ci, within the polymerfilm between theelectrodes
tends to approach a constant value at a rate of dC,/dt.

The following differential equation describes the process to a first
approximation :
Eq. (18) J,- J, = -

[ZFD (Co-Ci)/L] [(ZF)*DEC,/(RT)] = ZFd (dCi/dt)
from which the solution, given the fact that ZFDE/(RT) > D/L, is:
Eq. (19) J = (ZFDCJL) { 1- exp[-ZFDE/(RTd) t]}
where D = Doe(-AHdRT)
.
The statistical function, J, is:
Eq. (20)
J,a i = A RH e(-ddRT)/{ (1- RH) [ 1 t (m -1 ) RH]} { 1- exp[ -ZFDE/(RTd)t]}
The time-to-failure, t, (which is implicit in Eq. 20) occurs when the current
reachesafail criterion i,. For ZFDE/(RTd) t < < 1,which requires t < 1000
hrs at test temperature, Eq. 20 can be rewritten as:
t,- 6 / RH / { ( l - RH)[l t ( m -1)RH]}-[e(-2AH~’KT)/T]

E
3.0 EXPERIMENTAL
3.1 Parameters for Dendrite Model
The work deals with Ag-migration between tinned Ag-Pd lands (free
Ag along edges) in an encapsulated vehicle consisting of a polyimide-
amide surface coating, aluminum cap and silicone rubber or epoxy
backseal. Figure 1 shows the test site.
The tests were conducted in temperature-humidity chambers. The
lands were biased with 5, 10, and 20 volts, and in-situ voltage-drop
measurements (across 10 M n limiting resistor in series with each gap)
were made at intervals to generate a cumulative fail distribution as a
function of time. A fail is defined as a resistance 5 5x106n.
Failure analysis was performed on each failed unit. Leakages were
first analyzed electrically to ascertain that they were caused by dendrites
(and not other conductive paths), and then by an electron microprobe to
determinethe elements responsiblefor the high conductivity. Typical Ag
dendrites are shown in Fig. 2. Those obtained at the lower voltage are
‘Backseal
Figure 1. Vehicle (I), 0 1982, IEEE

95°C/80%RH 85°C/80%RH
Figure 2. Typical Silver Dendrites

somewhat heavier than those grown at the higher bias.

The cumulative percentage of failures obtained from each stress
condition were plotted versus time on lognormal paper. Figures 3 and 4
show the fail distribution for silicone rubber and epoxy backseals respectively
at each voltage.
The reciprocal t, values for both backseals are plotted against the
applied voltage in Fig. 5 showing a linear dependence and an intercept of
four volts. The results suggest that the current density distribution may be
falling to subcritical values.
There is a notable difference between the results obtained with the
two backseals. The modules encapsulated with silicone rubber show a
-
constant sigma ( 0.9) independent of the bias,while those encapsulated
with epoxy have a sigma ranging from 0.7 to 2.5, increasing as the bias
falls from 20 to 5 volts so that, at low fallout, the cumulative percentageof
failures is practically independent on the applied voltage.
Based on the results at 85°C/80%RH/20V and 95°C/80%RH/20V,
which are presented in Fig. 6, the activation energy for the migration
process is -0.1 5 eV, in agreementwith literature valuesfor ionic diffusion
(17). The time-to-fail dependence on relative humidity is seen in Fig. 7,
which best satisfies the model for an average pore size r-= 17.5 nm and a
lognormal-distribution u = 1.45. Figure 8 shows the humidity factor
versus relative humidity (RH) calculated from the model and how well it
agrees with the experimental data and the B.E.T. adsorption isotherm.
The actual pore-size distribution should, however, be measured
experimentally to provide materials characterization and direct model
verification. Meanwhile7and u are just parameters.
The experimental results show that below 4 volts no failures were
observed,which is consistent with a critical current. Based on the&, data
for the silicone rubber backsealed modules, the parameters in the pore-
based model, Eq. 15, are:
B' = 0.88 hV/cm°C

u = 1.45
= 175x10'~cm
AH, = 0.15eV
E = (V-4)/d
The parameters for the B.E.T.-based model, Eq. 11 based on the

same data, are:
B' = 0.88 hV/cm°C
m = 0.0065
AHD = 0.15eV
E = (V-4)/d
85"C/8O0/oRHi20, 10, 5V
20v
10 100 1000 10,000
Hours
Figure 3. Migration-Fail Distribution, Silicon Rubber Backseal (l), @ 1982,

IEEE.
85"Cl80%RH:20, 10, 5V q+-+3a

3a
- 30
--Il b
Hours
Figure 4. Silver Migration, Epoxy Backseal, Voltage as a Parameter (l),

o 1982,IEEE.
6ol
9
x
c 50
/
Applied Potential, Volts
Figure 5. Reciprocal Time-to-Fail (tS0) vs. Applied Voltage (l), @ 1982,

IEEE.
85"Ci80%RH/20V and
95"C/80%RHi20V 1*"
+2u ,"
5
c
E
-+a Q,
1 H = O . 1 5 ev i
n
-50 Q,
.-c>.
-m
--u
-20 0
z
J
------1i 7-3a
Figure 6. Silver Migration, Epoxy Backseat, Temperature as a Parameter

(l), @ 1982, IEEE.
8S°Ci80%R H,
85"C/6Oo/'RH,
{ +3a
25'14OYoRH
- +2a
20V Bias
B -+a
.--ro
u.
8
-50 E
- -a 0'
--2u
I* I I I 4-3a
10 1o2 1o3 1o4
Hours
Figure 7. Silver Migration, Silicon Rubber Backseal, Relative Humidity as

a Parameter ( l ) , @ 1982, IEEE.
14 -
0
12 -
N
I
0 Silicon Rubber Data
0
7
10- o Epoxy
X
b
4-
8-
0
2 6-
k
4-
2-
0 ' F
0 ,1 .2 3. 4. .5 .6 .7 .8 .9 1
Relative Humidity
Figure 8. Relative Humidity Factor vs. RH ( l ) , 0 1982, IEEE.
Theepoxy back-sealedmodulesshowsimilart,,resultsforwhich the

parameters in Eq. 15 are:
B1 = 2.0 hV/cm°C
u
- = 1.70
r = 175x1O'*cm
E = (V-4)/d
However, the statistical sigma increases as the voltage decreases

such thatfor < 1%cumulativefails, there is no distinctionamong the three
voltages, that is, the time-to-failure becomes independent of voltage (See
Fig. 3A). This behavior is attributed to a second mechanismwhich must
prevail at the lower voltages and must not depend on the voltage itself.
Species which enhance metal migration may be diffusing from the epoxy
backseal. The result is a larger sigma at the lower voltages.
3.2. Parameters for Leakage Model
The experiment deals with the measurement and interpretation of

electric leakage between parallel tinned and non-tinnedCr/Cu lands on a
ceramic substrate, covered with an epoxy overlay about 1mm thick. The
lands are 0.254 cm wide and spaced 0.0127 cm, 0.0254 cm, and 0.038
cm, while the thickness of Cr, Cu, and Pb-1O%Sn in the land structure are
-
100 nm, 800 nm, and l o 4 nm, respectively. Chromium is used for
adhesion.
The experiment is conducted under various stress conditions to
develop leakage kinetics to verify the model and to make projections to
field conditions as a function of temperature, relative humidity, voltage,
electrode gap, geometry, diff usivity parameters, and time.
From a mechanistic point of view, the process is apparently controlled
by the permeation of moisture and other soluble species that participate
in the leakage. The tests were conducted under three electrical biases (50
V, 30 V, and 10 V) and three electrode gaps (0.0127 cm, 0.0254 cm, and
0.038 cm) for the following temperatureand relative humidity conditions:
7OoC at 70%RH, 7OoC at 81%RH, and 85OC at 81%RH. The leakage is
measuredin situ and plotted versus time. Figures 9 and 10 show the data
from the three test cells (solid lines) which are used to determine the
model's parameters and, therefore, the leakage dependence on temperature,
RH, and voltage from which the time-to-failure dependence is determined
based on a fail criterion. Initially, the leakage increases linearly with
voltage but become sublinear as the test progresses,which implies ionic
depletion. (See Fig. 1 1 showing leakage versus voltage after 1500 hours
at 85OC.) The trend indicates that the voltage dependencewill continue to
- 50 Volts Experimental
- A 30 Volts Experimental
- 0 10 Volts Experimental
-
-
-
1 I I I I I I 1
Figure 9. Electric Leakage Current Versus Time on Test at 70°C/81% RH

and 85OC/81% RH for Three Different Voltages (8),@1985, IEEE.
Model
w 50 Volts Experimental
A 3 0 Volts Experimental
0 10 Volts Experimental
103 t
0
L
3
t
11 I 1 I I I I I
100 300 500 700 900 1100 1300 1500

Time, Hours
Figure 10. Electric Leakage Current Versus Time on Test at 7O0C/7O% RH

for Three Different Voltages (8),0 1985, IEEE.
erode as the system approaches a steady state, consistent with the

model's transient form with respect to voltage, whose effect is, therefore,
time dependent.
The parameters in Eq. 20 were determined experimentally:
m = 0.0065
AH, = 1.0eV
A = 1 .83x1Ol5nA
Using these parameters,the leakage criterion i,can be expressed as

a function oft,:
RH -1.16~10' ZFV -1.16~10'

___ e T
i = 1 . 8 3 ~O1I 5(1- {1-exp[-8.87x107-
RH)' RT e tl)
To reach afail criterion of 1O'8nfrom an insulation resistanceof 10"

n , it will take 1O5 hrs. at 4OCC/40%RH/1OV. The leakage dependence on
d is practically nil. This may be the result of the polymer thickness being
considerably larger than the gaps.
3.3 Water-Drop Migration.
Ag, Pb, and Cu Films. EvaporatedAg, Pb, and Cu land patterns on

ceramic substrateswere tested in water-drop experiments under 10 volt
bias: as-evaporated, after being oxidized, "sulphized" and exposed to
HCI dilute solution to form metal chlorides. Five samples for each metal
and surface condition were stressed. The reaction surface layers were
I50 nm thick and rather spotty (discontinuous) as determined by electron
microprobe. The oxide and sulphide layers either halted migration or
reduced its occurrence dramatically. It was obvious that dendrites,
whenever they occurred, were associated with incomplete passivation
and, in general, the surface treatment was very effective.
Metal chlorides did not improve the migration resistance. Actually
the chlorides made it worse, which can be explained by their high solubility
in water. Table I summarizes the results in terms of the time to grow the
dendrite and the current recorded during the process. The results show
how important the surface condition is and explainswhy metals that grow
stubborn insoluble oxide (e.g., Cu) are much more resistant to metal
migration than those whose oxides are water-soluble (e.g., Ag). These
differences reflect in the pre-exponentialfactor Awhich is afunction of the
material and process. These affect the critical current requirement and
7r
0
0 10 20 30 40 50 60 70 80
Applied Voltage, V
Figure 11. Voltage Dependence Drift (81, 0 1985, IEEE.
Table 1. Migration Across Evaporated Lands (Water-Drop) (l),

1982, IEEE.
@
Current
Metal Untreated Chloride Oxide Sulfide for Migration
Ag 50Sec 53Sec 45Sec 23Min 10-19p.A
cu 7 Min 5 Min None None 10-16pA
Pb 1 Min 30Sec 20Min 30 Min 18-47 FA
Cu (Wire) None
Cu (.Ole% H2S04)30 Sec
determine whether a dendrite will or will not grow in given materials

following certain processes through which contamination occurs.
In addition, when the experiment is conducted with acidic solutions,
e.g., l0-3M H,SO, solution, the Cu will readily migrate all over, across the
electrodes. Figure 12 shows a number of dendrites across parallel Cu
wires (0.005 cm diameter) in water drops having a pH of 3.2. No dendrites
were obtained with water having pH 7. Therefore, acidic environments
that can dissolve the oxide will promote metal migration.
It appears that as long as the metal is protected by an insoluble layer
of oxide, sulphide, etc., metals will not be able to migrate as the exchanging
of ions and electrons across the anode and cathode interfaces (essential
for the oxidation-reduction process) will cease.
In the present experiment we have observed threshold voltages of
0.5 to 1 volt for Cu, Ag, and Pb utilizing evaporated film electrodes
separated by a gap 20 mils. Similar limiting voltages were observed with
test vehicles having Ag paste and evaporated Cu patterns, all lands
spaced 3 mils apart. Cr had been etched off the Cu lands. When a 0.01 M
electrolyte solution of Cu(NO,), was used, however, the dendrites between
the Cu landswere producedwith 0.1 V bias. This reaffirms the fact that the
threshold voltage varies inversely with concentration in maintaining a
critical current or a certain dendrite growth rate for a given set of conditions.
In fact, literature data shows that, for low current densities, the dendrite
rate of formation is linear with the ion concentration and the applied
voltage.
Therefore, for very low ion concentration, one requires a
correspondingly higher voltage to produce the current density necessary
for dendritic growth.
Ni Films. This section deals with nickel migration and how it
compares with the migration of Ag, Cu, and Pb. Ni films were evaporated
on ceramic blank substrates leaving a 15 mil gap between two halves of
-
the film. The gap was covered with a water drop ( 1 mm diameter) and
biased at various voltages ranging from 0.5 to 15 volts. Ni dendrites
formed under these conditions. Dendrite growth rates were calculated
from the time that each dendrite took to bridge the gap. The ionic current
density was 0.1 A/cm2.
Results at room ambience show that a voltage threshold exists at
1.60 volts, below which no dendrite grew (see Fig. 13). This voltage
threshold is much higher than those of Ag and Ag 20%Pd pastewhich are
0.80 and 0.75 volt respectively (1 7). For Cu and Pb, it was 0.5 volt ( 6 ) .
The dendrites which were produced under higher voltages are heavier
and have a ramiform structure, while those at lower voltages are thinner
and occur in fascicles aligned with the field. The growth rate is nearly
linear with the voltage in excess of the threshold and has avalue of -0.17
Figure 12. Cu Dendrites Across Parallel Copper Wires

0.28
.-
I
E 0.2
0I Y I I I I I I 1 1
1 Vr 2 3 4 5 6 7 8 10
Applied Voltage, Volts
Figure 13. Ni Dendrite Growth Rate vs. Applied Voltage

mils/sec x volt. Ni dendrite growth rates of 1.5 mils/sec x volt have been
reported in the literature (17) at 105OC using aqueous solution of NiCI,
(18). The two results are consistent if it is assumed that the activation
energy is 0.3 eVand the electrolyte has no effect or has a lower activation
and only some electrolyte effect. Oxidized Ni, however, did not migrate
under the same conditions. Alldendrites occurred in the first few minutes,
and no additional ones formed after 23 minutes. This indicates that
passivation takes place under the low temperature test conditions and
neutral pH. Based on literature findings and the results of this experiment,
it is evident that the high-threshold voltage must be due to the high
overvoltage resulting from the effect of surface contaminants and oxide/
hydroxide formation. Indeed the literature shows that above a critical
current density, J, the Ni cathode passivates such that Ni reduction or
deposition is governed by the activation energy of Ni++diffusion through
the passivatingimpurity layer ( - 10.5kcal/mol). Since the critical current
increaseswithtemperature, the critical value, J, is more easily exceeded
at room temperature, thus satisfying the conditions for passivation.
Therefore, the voltage threshold is expected to increase at higher
temperatures as it will depend on the critical current density which limits
the process. In addition, drastic reduction in the migration of Ni is,
therefore, possible by introducing substanceswhich decrease the acidity
and increasethe rate of impurity deposition at the cathode. It appears that
oxidation or the production of a layer of hydroxide at the cathode are very
effective in reducing Ni migration. Because dendritic growth is occurring
at both temperatures,one may assume that the mechanismof passivation
is not significant in both cases.
It should be noted that Fe and Co behaved quite similarly. These
three elements (Fe, Co, Ni) are called metals of the iron group. The outer
electrons' free atoms of iron, cobalt, and nickel have configurations d6S2,
d7S2,and d8S2respectively. Their high chemical activity,is due to the
presence of unfilled d-levels. Among the characteristics of these metals
is high overvoltage during deposition and dissolution of the metals in
simple salt solutions. Because of their high reactivity, when one of the
metals comes into contact with a solution, foreign particles are adsorbed
on its surface leading to the passivation of the electrode.
Cu-15%Ag-2.5%P Wires. The wire is 10 mils in diameter and has
a Cu core. Other wires used in the experiment were pure Cu and Ag-
coated Cu.
The preliminary experiment consists of placing sets of parallel wires
of each typeon aceramicsubstrate( -50 milsapart), sandwiching awater
layer between the substrate and a glass slide such that a water path is
established across the electrodes, applying the bias, and observing
dendritic growth as afunction of time. Two electrolytic media were used:
(1) DI water, and (2) dilute H,SO, solution ( - l o 3 M, pH = 3.2), which is

representativeof SO, in a humid environment.
Results show that in DI water the Ag-coated Cu wire grew about 30
dendrites per centimeter of wire in ten minutes starting at 1 1 /2 minute
with the first dendrite, while the Cu-Ag-P wire produced no dendritic
growth even after 90 minutes, at which time the test was terminated.
In the acidic electrolyte, however, the Cu-Ag-Pwire grew an average
of 1 1/2 Cu dendrites per centimeter of wire in ten minutes-about the
same as the Cu wire which grew - 2 dendrites/cm. In any case, all
dendritic growths approached a limit and practically ceased growing after
10 minutes of exposure. It appears that there is a finite number of active
sites that areamenable to dendritic growth and that their exhaustion ends
the process. This appears to be the case even before any one dendrite
shorts the electrodes. Figure 14showstypical dendritesfrom both Cu and
Cu-Ag-Pwires.
Based on the results, it can be stated that the Cu-15Ag-2.5P braze
wire behaves likea Cu wire in termsof metal migration in the system used.
It appears that the metallurgical state of Ag in the quenched wire structure
is not amenable to migration. Considering the Ag dispersed phase in the
Cu matrix and its composition, it appears that, thermodynamically, Cu is
favored. Figure 15 shows the number of dendrites per cm of wire as a
function of time for the three wires in the given electrolytes.
Wire samples were also tested in a standard migration cell, and I-V
curves were measure to determine the Tafel slope during anodic dissolution
in various aqueous media. The results of the migration cell tests in 0, t
H,O and 350 ppm HCL showed no dendrite formation; however, the
anode was roughened (dissolution) and there was Cu in solution. No
silver was detected. The Tafel slope was identical to that of Cu, though
therewas aslight shilft in the noble metal direction. These resultssupport
the observationsin the water-drop experimentthat only Cu dendritesform
and that the Cu-15%Ag-2.5%P wire behaves like pure Cu.
4.0 DISCUSSION
4.1 Model Acceleration and Materials/Process Effects
The time-to-failure model expressed by Eqs. 1 1,15,and 20 are used

to determine an acceleration factor between test conditions T, RH, V,
and field conditions T, RY,V, assumingthat the process is invariant. The
proportionality constant B is, in effect, a function of all the process and
materials variables. The assumption is that the prevailing mechanism at
test conditions is the sameas under field conditions, and projectionof test
Next Page
Previous Page
(1) DI water, and (2) dilute H,SO, solution ( - l o 3 M, pH = 3.2), which is

representativeof SO, in a humid environment.
Results show that in DI water the Ag-coated Cu wire grew about 30
dendrites per centimeter of wire in ten minutes starting at 1 1 /2 minute
with the first dendrite, while the Cu-Ag-P wire produced no dendritic
growth even after 90 minutes, at which time the test was terminated.
In the acidic electrolyte, however, the Cu-Ag-Pwire grew an average
of 1 1/2 Cu dendrites per centimeter of wire in ten minutes-about the
same as the Cu wire which grew - 2 dendrites/cm. In any case, all
dendritic growths approached a limit and practically ceased growing after
10 minutes of exposure. It appears that there is a finite number of active
sites that areamenable to dendritic growth and that their exhaustion ends
the process. This appears to be the case even before any one dendrite
shorts the electrodes. Figure 14showstypical dendritesfrom both Cu and
Cu-Ag-Pwires.
Based on the results, it can be stated that the Cu-15Ag-2.5P braze
wire behaves likea Cu wire in termsof metal migration in the system used.
It appears that the metallurgical state of Ag in the quenched wire structure
is not amenable to migration. Considering the Ag dispersed phase in the
Cu matrix and its composition, it appears that, thermodynamically, Cu is
favored. Figure 15 shows the number of dendrites per cm of wire as a
function of time for the three wires in the given electrolytes.
Wire samples were also tested in a standard migration cell, and I-V
curves were measure to determine the Tafel slope during anodic dissolution
in various aqueous media. The results of the migration cell tests in 0, t
H,O and 350 ppm HCL showed no dendrite formation; however, the
anode was roughened (dissolution) and there was Cu in solution. No
silver was detected. The Tafel slope was identical to that of Cu, though
therewas aslight shilft in the noble metal direction. These resultssupport
the observationsin the water-drop experimentthat only Cu dendritesform
and that the Cu-15%Ag-2.5%P wire behaves like pure Cu.
4.0 DISCUSSION
4.1 Model Acceleration and Materials/Process Effects
The time-to-failure model expressed by Eqs. 1 1,15,and 20 are used

to determine an acceleration factor between test conditions T, RH, V,
and field conditions T, RY,V, assumingthat the process is invariant. The
proportionality constant B is, in effect, a function of all the process and
materials variables. The assumption is that the prevailing mechanism at
test conditions is the sameas under field conditions, and projectionof test
Figure 14- Cu Dendrites Grown From Cu-Ag-P Wires in 10-3 M H2SO4

2 30
i-
6 25
E
0
3 20
.-Q)
ICL
$ 15
:10
.c
0
$ 5
E
i o0 2 4 6 8 10 12 14 16
Time, Min.
Figure 15. Number of Dendrites Per Centimeter of Wire vs. Time

results can be safely made. It must be ascertained that no abrupt change

in the migration behavior is expected outside the experimental range;
process and materialsvariables, on the other hand, are usually kept under
control. These variables include the nature and the concentration of
impurities. If the impurities are characterizable and their effect on dendritic
growth is known, one can often use such species to accelerate or decelerate
the process depending on whether they have a catalytic or poisoning
effect on the dendrites. In any case, any new test variable introduced for
acceleration purposes must not upset the mechanism and the effect on
migration kinetics must be understood and quantified. In addition, the
surfacesof the conductor may be in a different state of oxidation asaresult
of process temperature variation, thus altering the migration kinetics.
Such changes in migration due to subtle process variation cannot be
easily isolated because of the interference of other variables which are
either unknown or not controllable to the desired extent. Thus, the
characteristics which affect the nucleation and growth kinetics of a dendrite
(ionic species, passivation, poisoning, film roughness)are all reflected in
the parameter B which can be termed the process-materialcharacteristic
function. To some extent, this function can be explored by studying the
electrochemical behavior of material systems in aqueous solutions and
water-drop tests with respect to specific contaminants, their concentrations,
and temperature. The model, therefore, has two parts: (1) the acceleration
function which depends on the test conditions (and any process variable
which is well characterized) and provides an acceleration factor for the
given mechanism no matter what the extent or magnitude of migration is,
and (2) the characteristic function, which representsthe reciprocal of the
magnitude, and which is determined experimentally from the fail distribution,
t, under controlled test conditions employing the appropriate model,
e.g., Eq. 11. The objective is to minimize B with the proper choice of
process and materials. Knowingthe characteristic function means having
the tools to design the optimum metallization system for the given
environment.
4.2 Materials Characterization by Water-Drop
Some of the water-drop experiments reported here have illustrated

the usefulness of such data in terms of combating migration under
specific environmental conditions and state of surface oxidation. Other
important factors are process residues, pH, etc., which likewise must be
studied to determinetheir level of tolerance for the given situation. It is no
wonder that 6 , as an experimental parameter, varies from metal to metal
and from process to process. The key is to identify and quantify the
variables responsible for the change.
The metal oxides have shown a considerable slowing of the migration

process which resulted in much longer times to produce a dendrite. The
effect of the metal oxide on migration is dependent on the film oxide
thickness. For instance, a film 50 - 100 A thick will barely produce
detectable changes. One must produce films of the order of 1000 A to
slow down the dendrite growth rate considerably. For instance,a copper
oxide film of this magnitude, grown on a substrate, slowed down the
process by about two orders of magnitude. Thinner films of copper oxide,
as well as Pb oxides, have shown various degrees of effectiveness in
reducingtherateof dendritic growth. Theeffectofoxidesisalsoobserved
in terms of raising the threshold voltage of pure metals by one half and
even one volt for comparable migration times. In addition, the thicker the
oxide film the fewer the number of sites atwhich the dendrite can grow as
it is restricted to grow through defects. That dendrites grow selectively is
illustrated by the fact that at times growths have been observed in a 30 mil
gap in preferenceto a 3 mil gap under the same voltage. The observations
were made in Cu oxide and sulfide films, andonlyafter a substantial delay.
The sites were characterized to be defects.
In addition, the oxide provides a high resistance path through which
a potential drop occurs. This voltage loss reduces the field across the
electrolyte and, therefore, can cause the current to fall below the critical
value.
4.3 Effect of Active Impurities
Another important aspect of migration is the concentration of active

impurities. For one metal ion concentration and one type of impurity at a
given temperature, the critical current density for dendrite growth is (3)
Eq. (23) J, = (Constant) (Cjt/r)’l3
where Ci’ is the concentration of surface active impurities and r is the

dendrite diameter. It is theorized that when a constant current is passed
through a cell, one crystal will form at the cathode and, as it grows, the
current density on its faces decreases. When the current density becomes
less that J,, the crystal will become contaminated and will stop growing;
a new crystal will then be nucleated and it will grow. Occasionally, it will
happen that all but one face of the crystal will become blocked. The last
crystal face will then adjust its size so that the current density is equal to
J, and will grow as a whisker or dendrite.
Measurements of critical current densities in solutions containing
known concentrations of additives are available only for the growth of Ag
whiskers in AgNO, solutions containing gelatin. At 25OC the constant in
/ ~ . known Ci and r, J, was

Eq. 23 is equal to 14 A ~ m ~ ’ ~ / m o I ’From
calculated. The results show that one can quantify the effect of impurities
and incorporate Eq. 23 into the model as a (J - JCr)factor.
Ionic species are also responsible for the enhanced conductivity of
water films. The species can originate from the process residues left on
the substrate (chlorides), from environmental gases dissolved in the
water film, or from the corrosion process providing water soluble compounds
(19-26). The addition of CI, gas, for instance, can promote dendritic
growth in situations where no growth had been possible without the gas.
Sbar and Feinstein (19) found no sign of dendrites after Ti-Pd-Cu-Ni-Au
biased conductors had been exposed to 85’C/85%RH for 1000 hours.
Metal dendrites were observed, however, when the CI, was added. In
many instances chlorine acts as a catalyst by forming intermediate
compounds which are soluble in water. The controlling electrochemical
process usually ends up with lower activation energy. Since the effect of
chlorides may be considerable even for surface concentrations which
escape detection, quantification of the effect is limited by the analytical
capability. When other active species are present, even an accurate
analysis would probably not allow a fair prediction of their effect on metal
migration because the entire mechanism may change. Therefore, one
must know the system well and must know the prevailing mechanism and
its sensitivity to secondary effects from other interacting species. Unless
the metallurgicaland pollutant system is understood, migration quantification
on the basis of species concentration and other electrochemical data is
difficult. For this reason, we can only speak of modelling in terms of
acceleration factor which is based on test variables and is in compliance
with the general theory of migration. Models that go beyond the test
variables by including the effect of specific species and their concentrations
are usuallyjustified on the basis of empirical relationshipsthat quantifythe
systems’ migration behavior with respect to the specific variable.
4.4 Mechanism and Time-to-Failure Results
Relevant works in the electronics packaging industry have dealtwith

silver, copper, cobalt, zinc, lead, iron, tin, nickel, and gold; however, Ag
migration has been the most widely reported (2)(27-35). Barton and
Bockris (2) have measured the rate of Ag dendrite growth from the silver
and silver nitrate system (0.33 to 12%AgN03). The current density during
dendritic growth is superlinear with respect to the overpotential and
proportional to the concentration of silver ions. A critical current density
(Jcr)existsfor the initiation of dendritic growth, which is proportionalto the
silver concentration and increases as the radius of the electrode decreases.
-
The dendritic growth tip is parabolic with a radius of curvature of 1Os’
cm. The current density is controlled by diffusion at all conditions and

obeys laws characteristic of spherical diffusion for small radii. The rate of
dendrite growth is also controlled by the kinetics of the Ag' ion position
on the dendrite tip (charge exchange and spherical diffusion controlled).
For low exchange currents the dendritic growth rate is linear with overvoltage,
but for high current densities, the velocity of growth is proportional to the
square of the applied overvoltage.
Barton and Bockris have, in fact, verified the electrochemical kinetics
with AgNO, solutions within the concentrations mentioned. For diffusion
controlled migration and low current density,
J = (ZFDC/S) (ZF/RT) V
For these conditions:
V = 20mV, applied overvoltage,

C = l o 5 mole/liter,
S = 10-2,diffusion layer,
D = 2 x l o 5 cm2/sec,
the value of J = 1 A/cm2 at room temperature.

In general the current density is expressed as an hyperbolic sine
(6)(7)which reduces to Eq. 24 for low voltage [(ZF/RT) V < < 11.
The superlinear behavior, therefore, is expected at the higher voltage.
Theresults inFig. 4,which isaplotofthereciprocalof$versusthevoltage,
show a threshold of 4 volts in both silicone rubber and epoxy backsealed
modules with a linear fit. Another fit can be obtained by forcing the curve
to pass through the origin, in which case the reciprocal time-to-failure
would depend on V'. If one considers that the voltage-drop at the anode
-
and cathode is 1/2 volt and that the water-drop experiments with Ag
electrodes have shown dendrites growing under 1/2 volt bias, the fit
through the origin is not unreasonable for analytical purpose, to show its
consistency with the theory.
On a statistical basis, a V2dependence is also in line with the fact that
impurities are not homogeneously distributed. As voltages across the
gap raise the current density from below the critical value at the various
sites, having different thresholds, the dependence of dendrite growth on
applied voltage appears stronger than it actually is due to the tail of the J,
distribution dictated by the impurities. This means that there cannot be a
cessation of migration activity at a specific voltage since we are dealing
with statistics. Therefore, a voltage intercept must be more of a gradual
fading than an abrupt ending of the migration process. In any case, the
level of activity is obviously very low below four volts and is inconsequential
from afail rate point of view regardlessof interpretation. Again, it must be

emphasized that since the threshold depends on the level and type of
impurities (among other things), the voltage intercept value must be
considered a characteristic of the particular process.
4.5 Polymer Coating
The leakage current may include extremely low components of

metallic ions derived from the anode, but the predominance is due to
indifferent ions and electrolyzed substances, mainly water through hydrogen
reduction. Gaseous pollutants dissolve into the water film or polymer and
maintain concentration levels proportional to the outside vapor pressure
according to Henry’s law applicable to dilute solutions. Among the
notorious pollutants are SO,, NO,, and CO,, which provide the ionic
species for conduction through electrolysis.
Water electrolysis, as mentioned, is probably a large contributor to
the current. It increases with voltage and, consequently, the migration
component of the metal ions grows with it, as it is proportional to the total
current via the transport number. Despite the initial very low concentration
of metallic ions, the rate of metal deposition is increased when the
electrolysis is carried out with a large current caused by the reduction of
another substance. This phenomenon is called “exaltation of the migration
current” by simultaneouselectrolysis of another compound or substance
(3).
In the case of a polymeric coating, the electrolytic process must
eventually achieve equilibrium with the diffusion process originating at
the polymer surface, which interacts with the outside environment through
adsorption. It is assumed that initially the external pollutants have not
significantly permeated the film and that considerable time would be
required to reach the steady-state concentration, thus satisfying the
diffusion and migration components near the electrodes. During the
transient period, a certain volume of the polymer around the electrodes,
extending to the limits of field interaction, will have to be “charged” with
ions at some rate zFC,,which will vanish at steady state, whatever the time
required. Likewise,the voltage effect diminishes with time, as predicted
by the model and supported by the experimental results (Fig. 10). The
process of achieving steady state is quite slow because the small diffusivity,
-
estimatedat 1O‘g cm2/hrs.and85OC. This is considerably lower than the
diffusivity of water through epoxies. This may be due to solvation and
complexing which can curtail the diffusion process ascribable to ionic
size increase and alteration of the electric interaction. Such an argument
can also be made to explain the relatively high activation energy.
5.0 SUMMARY
The time-to-failure models, derived on the basis of the electrochemistry

of solutions, theory of adsorption and condensation, and transport through
water films and polymers are characterized by the product of four functions:
t, = B(M,P) F(T) G(RH,T) H(E,T)
where B(M,P) is a function of materials and process parametersand must

be studied with respect to materials properties, contaminants, and prevailing
mechanism. It is aconstantfor the same materialsystem under controlled
process. F(T) is the integrated form of the Arrhenius equation. G(RH,T)
is the relative humidity function (which also depends on T) which can
assume the B.E.T. form or the pore-condensation probability form, both
conforming to established physical interactions between liquids and
solids. Other forms such as pure exponential (eBRH), relative humidity to
a power (RH") and other similar fittedfunctions are not sufficientlyflexible
to express water film adsorption throughout the practical RH range, and
their extrapolation to low RH values from test conditions is questionable.
They are not unique and fall within the reach of the two basic functions
above. H(E,T) is the electric field or voltage function, which also depends
on temperature. When a polymer coating is presentwhich adhereswell to
the substrate and electrodes, the function assumes a transient form to
express current leakage as a function of time, which is the result of non
steady-state between migration and diffusion, the latter becoming rate
controlling with time. For short times, the function reduces to the same
form applicable to dendritic growth, however,for long times the leakage
follows an exponential decay growth such that the insulation resistance
approaches a limit. This is the point when migration is completely
dependent on the electrolyzable species diffused from the outside.
The most important variable in metal migration is relative humidity
because, in the absence of a surface water film, no electrolytic process
can take place.
The activation energy for the migration process (leakage) in the
presenceof a polymericfilm is much higher than that for dendritic growths
because in the former the diffusion through the polymer is rate-controlling,
while in dendritic growth the diffusion through the water film controls the
process.
Common metal films used in the electronics and semiconductor
industry show voltage thresholds between 0.5 and 1.6 volts when tested
by water-drop (DI water) with a clean unoxidized surface. On the other
hand, the voltage threshold established through the time-to-failurestatistics
obtainedfrom T/H testing at 85OC/80%RH show higher threshold (e.g. 4
Electrochemica1 Mig ration 29 1
voltsfor silver) by extrapolation from 20, 10, and 5 volts data. It is difficult
to determinethe actual shape of the curve ( t i ’ vs V) below 5 volts because
the probability of producing afail under those conditions isvery low, such
that no fails are possible for the test duration.
The dramatic difference between Ag-migration results in the presence
of a silicone rubber backseat and epoxy backseat indicates that species
that can be leachedoutfrom a polymer play an important role in the failure ,
mechanism, and can introduce a bimodal fail distribution.

A critical current density must be achieved before dendritic growth
can occur. In many instances the critical current has ranged from 0.1 to
1 A/cm2, depending on temperature, type of metal, and impurities.
At critical current density, dendrites begin to form in cathodic regions
which have achievedsharp “macro-rough” geometries,where extremely
high localized currents of the order of 1O3 A/cm2 develop as a result of
parabolic or spherical diffusion to the tip of the dendrite. Dendritic growth
is, therefore, a special case of migration, the case which must satisfy the
requirement of current density and suitable conditions for single crystal
formation.
Migration-induced failures strongly depend on the environmental
atmosphere which modifies the metal migration behavior by altering the
leakage current and even the rate limiting step as a result of surface
passivation or other gating mechanism. In addition, process residues
also play an important role in metal migration through water adsorption
and conductivity enhancement.
Oxidation and other insoluble film formation curtail migration and
even stop it if the films are of sufficient thickness and tenacity. For
instance, the low migration associatedwith Cu-oxide films indicates that
such oxide is a good barrier to ion exchange at the oxide-metal interface
and that the oxide has a very low solubility. Other oxides with such
properties behave similarly.
Ni has the highestvoltage threshold among the various metals tested
bywater-drop test. Its migration is drastically reduced by the formation of
an oxide or hydroxide layer. Introduction of impuritieswhich raise the pH
of the system and favor the formation of Ni(OH), is an important consideration
in the preventionof Ni-migration. At temperaturesgreater than 1OOOC,the
hydroxide impurity layer is notformed, whileat the lower temperatures the
migration process can be practically stopped by such a layer.
The basic model in its general form presents a good base for wide
coverage over the diverse applications. The experimental data on Ag-
migration and leakage kinetics through epoxy coating for acceleration
purpose, servea starting point and set the stage for further evaluationwith
other metals,which are expected to produce different parametersvalues
for H, M, ,V,, F and B. The model’s ability to be expressed as a function
of different materialsandprocesscontamination dependson howwellthe

variables can be controlled and studied and to what extent they affect
metal migration and leakage.
6.0 ACKNOWLEDGEMENT
Stephen Drofitz is acknowledged for performing the water-drop

experiments on thin films and Cheryl Hamiltonfor typing the manuscript.
7.0 REFERENCES
1. Di Giacomo, G., Metal Migration (Ag, Cu, Pb) in Encapsulated

Modules and Time-to-Fail Model as a Function of the Environment
and Package Properties, Proceedings Reliability Physics
Symposium (1982)
2. Barton, J.L. and Bockris, J.O., The Electrolytic Growth of Dendrites

from Ionic Solutions, J . ,Electrochem. SOC.,Vol 268A (1962)
3. Price, 2.B. et al., On the Growth and Properties of Electrolytic

W h i ~ e r s ,Acta Metallurgica, Vol. 6, August (1968)
4. Bockris and Reddy, M o d e r n Electrochemistry, Vol. 1 and 2,

Plenum Press, New York (1970)
5. Adamson, A.W., Physical Chemistry of Surfaces, lnterscience

Publishers, New York (1967)
6. Overbeck, J.Th.G., Colloid and Surface Chemistry, Center for

Advanced ,Engineering Studies, MIT (1971)
7. Bowden, F.P. and Throssell, W.R., Adsorption of Water Vapors on

Solid Surfaces, Proc. Roy. SOC.,Vol. A209, p. 29 (1951)
8. Di Giacomo, G., Current Leakage Kinetics Across Tinned Cr/Cu

Lands Having Epoxy Overlay, / € E € Transaction on Components,
Hybrids, a n d Manufacturing Technology, Vol. CHMT8, No. 4,
December (1985)
9. Chariot, G., Badoz-Lambling,J. and Tremillon, B., €/ectrochemical

Reactions, pp 3-30, Elsevier, New York (1962)
10. Vagramyan, A.T. and Zhamagortsyants, M.A., Electrodeposition

of Metals a n d Inhibiting Adsorption, TT-72-51030, Translated
from Russian (1969)
11. Savchenkov, G.F. and Uvarov, L.A, Study of the Anodic Behavior of
Metals of the Iron Group in a Wide Range of Temperatures; The
Influence of Temperature on the Critical Passivation Current of
Nickel, Zashchita Metallov, 1:636 (1965)
12. Bond,A.P. andUhlig, H.H., J. Electrochem. SOC., 107:488(1960)

13. Okamoto, G. et al, J . Electrochem. SOC.,Japan, 25:199 (1957)
14. Vagramyan, A.T. and Uvarov, L.A., Effect of Inhibition on the

Electrodeposition of Metals, Transactions of institute of Metal
Finishing, Vol. 39 (1962)
15. Ohya, H., A Study of Pore Size Distribution of Modified Ultra Thin
Membranes, J . Applied Polymer S c i . , Vol. 18 (1974)
16. Gregg, S.J. and Sing, K.S.W., Adsorption S u r f a c e Area and

Porosity, Academic Press, 1967
17. Hills, G.J., Kinetic Studies of Ionic Migration, lnternational

Symposium of the Electrochemical Society, 1964
18. Tajima, S. and Ogata, M., Electrocrystallization of Ni and Co

Dendrites From Aqueous Solutions II, Electrochimica Acta, Vol.
15, pp 61-64 (1970)
19. Sbar, N.L. and Feinstein, L.G., / E € € Trans. Parts Hybrids

Packaging, PHP 13,3, 208 (1977)
20. Shumka, A. and Piety, R.R. 73th Annual Proc. Reliability Physics
Symp. /€€,E, p 93 (1975)
21. Hakim, E.B. and Shappirio, J.M., Solid State T e c h n o / . ,p 66, April
(1975)
22. Kozakiewicz, R.P. and Sbar, N.S., Extended Abstracts, Spring

Meet. Electrochem. SOC.,p 123 (1977)
23. Labuda, E.F., Herb, G.K., Ryden, W.D., Fritzinger, L.B. and Szabo,
J.M. Jr., Exfended Abstracts, Spring M e e t . Electrochem.
SOC.,p 195 (1974)
24. Sbar, N.L., /E€€ Trans. Parts Hybrids Packaging, PHP 12, n 3,
p 176 (1 976)
25. Fuss, F.N., Hartwig, C.T.and Morabito, J.M., Thin S o l i d F i l m s , Vol

43, 12, p 189 (1977)
26. Sbar, N.S.andKozakiewicz, R.P., /E€€ Trans. ElecfronDevices,

ED-26, n 1, p 56 (1979)
27. Kohman, G.T., Hermance, H.W. and Downes, G.H.,B e l l Syst.

Techn. J , Vol. 34, p 1115 (1955)
28. Williams and Herrmann, lR€ Trans. Reliability Quality Control,

PGRQC, 11 (1 956)
29. Chaikin, Janney, Church and McClelland, l n d . Eng. Chem, 51,299

(1959)
30. Wirtz, G.P., Proc 22nd Electronic Components Conf, P 840

(1971)
31. Stein, S.J.,Huang, C.,Cang, L. and Schultz, G., Solid State Tech.,
p 25 (1975)
32. Licari, J.J., Perkins, K.L.Caruso,S.V., P r o c . Int. M i c r o e l e c t r o n i c

Symp. (ISHM), p 65 (1975)
33. Rogren, P . E . , P r o c . l n t . M i c r o e l e c t r o n i c Symp. (ISHM), p 267

(1976)
34. Pantanelli, G.P.,and Sedora, E.J., Int. C o n f . o n Manufacturing

and Packaging Technology f o r Hybrid Circuits, Paris, France
(1976)
35. Naguib, H.M. and McLaurin, B . K . , I€€€ Trans. Components

Hybrids M a n u f . T e c h n o / . , CHMT 2 , p 196-207 (1979)
7
ELECTROCHEMICAL CAPACITORS
Ian D. Raistrick
1 .O INTRODUCTION
Electrochemicalcapacitors differ fundamentally from both conventional

capacitors and batteries in their mechanism of charge separation and
energy storage. Conventional types of capacitors store electrostatic
energywhen an electric field is applied across a dielectric. Batteries store
chemical energy in the form of reactants that combine and release the
Gibbs free energy of reaction when an external electronic connection is
made. In the simplest case, an electrochemical capacitor makes use of
the double-layer of charge that forms spontaneously at the interface
between a metallic electrode and an ionically conducting electrolyte.
Depending on the origin of the interaction between the electrode and the
electrolyte, the associated capacitance is usually in the range of tens to
hundreds of p F per cm2of interface area. It is therefore evident that, if an
electrode with a large specific surface area is used, significant energy
storage is possible at this type of interface. Suppose,for example, that a
materialwith a specific surface area of 100 m'/g and a density of 3 g/cm3
is employed in a 10-pm-thick layer. Each electrode would then have an
effective capacitance of about 10 mF/cm'. Such a large capacitance is of
obvious interest in energy storage and power conditioning applications.
The earliest suggestion that practical devices may be constructed on this
principle appears to be due to Becker (1).
A schematic drawing of an electrochemical capacitor, based on
symmetric electrodes is shown in Fig. 1. Here, the active plates are
arranged in a bipolar configuration (a current collector acts as a positive
electrode for one cell and a negative electrode for the neighboring cell).
297
Figure 1. Bipolar arrangement of electrodes in an electrochemical

capacitor.
ElectrochemicalCapacitors 299
The electrolyte compartments are electrically isolated from one another.

This is the simplest and most common arrangementfor this type of device.
It ensures a uniform current density through the structure, thus minimizing
ohmic losses caused by current bunching at collector tabs, and has the
minimum intrinsic inductance (important for pulse applications). A
characteristic of capacitors is that they are usually built with identical
positive and negativeelectrodes. This is not always the case (see Section
4.4 for examples), but, if the electrodes are composed of different materials,
the cells will not typically have zero voltage when uncharged.
In general, the properties of electrochemical capacitors are quite
different from those of other types of capacitors. Like batteries,
electrochemical capacitors are essentially low-voltage devices: their
breakdown occurs by electrolysis of the electrolyte. For aqueous electrolytes,
the thermodynamic potential at which this occurs is about 1.2 volts. For
organic solvent-basedsystems, it may be as high as 3 or 4volts, whereas,
for silver and copper solid electrolytes, it is less than 1 volt. This property
severely limits the amount of stored energy (U) for a given capacitance
because U = 1/2CV2.
In an ideal capacitor, the amount of chargestored, Q , is proportional
to the potential differenceacross the dielectric, Q = CV. Due to the nature
of the interfaces, electrochemical capacitors seldom show this simple
behavior. Usually, the dependence of C on V is complicated by chemical
interactions. Even in the simplest double-layer capacitors (see Section
2.2), the presence of mobile electrolyte ions ensures a nonlinear dependence
of charge on potential. Overall, however, the total quantity of stored
charge is usuallyvery much higher than that availablefroma conventional
capacitor.
Probably more important for many applications than the total energy
storage capability of a device, however, are its ability to deliver current
and the nature of its electrical response, as compared to that of an ideal
capacitor. Electrolytes have an appreciable, and sometimes very large,
resistivity that often dominates the electrical behavior of the device, not
only in the sense that it controls the time constant of the discharge
process, but also in that the current distribution within the electrode
structure is strongly dependenton this quantity. The microstructure of the
high surface-areaelectrodeis often of great importance. Electrochemical
capacitors, therefore, are usually characterized by a relatively large
effective series resistance (esr).
The principal application of electrochemical capacitors to date is as
backup energy for integrated electronics. Here, the long life, small size,
wide temperature range, and low cost of carbon elecVochemical capacitors
give them a distinct advantage over, for example, rechargeable batteries.
This mode of operation, where the capacitor is kept continually charged
by line power, is clearly favorable to capacitors, whose absolute energy

storage capability is not as high as that of batteries (especially primary
batteries). A similar consideration holds for applications such as weapons
fusing, where the capacitor is charged only a short time before being
needed, and until that time has no energy content. Various types of
energy conditioning applications may also prove attractive. In this case,
primary energy is supplied from another source, but peak or pulse power
is provided by the capacitor. Such an application takes advantage of the
high power capability of electrochemical capacitors, compared with
batteries. In contrast, the large capacitance per unit volume of
electrochemical devices, compared with conventional (including electrolytic)
capacitors, offers the possibility of usingthem in compact filter sfor power
supplies. Clearly, the relatively high series resistance and poor a.c.
performance to date are hindrances for this last application.
In the next twosections, the principal mechanismsof charge storage
and the electrical behavior of electrochemicalcapacitorsare discussedin
a generalway. Inthe remainingsections, individualfamiliesof devicesand
materials that have either found, or show promise for, applications are
covered in more detail.
2.0 MECHANISMS OF CHARGE STORAGE AT THE

ELECTROCHEMICAL INTERFACE
In this section, the main types of electrode processes that may be

used for capacitative electrochemical energy storage are briefly described.
Detailed description of specific systems is reserved for later sections. A
formal distinction is made here between electrochemical processes that
are localized at, or near, the electrolyte/electrode boundary and those
that involve bulk phases. The latter type of reactions are more appropriately
classified as belonging to batteries. The distinction is not always clear,
since in some ‘capacitor’ materials, reactions apparently extend over a
significant distance from the surface, and in some batteries, the active
materials may be deployed in very thin films, with the objective of improving
power. Further, in very high surface-area materials such as some transition
metal oxides and the conducting polymers, the active material is present
in such a small particle size and is so completely penetrated by the
electrolyte phase that almost the entire material may be regarded as a
surface species. In the following paragraphs,however,we treat idealized
situations, in which the distinctions are clear.
2.1 Double-Layer Capacitors
In the absence of any chemical interactions between a metal and an

Electrochemical Capacitors 301
electrolyte solution, a double-layer capacitance spontaneously occurs.

This is due to the difference in work function between the two phases,
which causes a small charge to flow across the interface to equalize the
electrochemical potentials.
The excess charge is accommodated in a surface layer on the metal
side of the interface, but on the electrolyte side the charged layer has a
more complex structure (2). Farthest from the interface is the ‘diffuse
region’ in which the excess charge decays to zero in the bulk of the
electrolyte according to Boltzmann statistics.
Eq. (1) P(X) = xz,FC,O exP{ [-zi F 4 (XI1 ART1 I

Here x is the distance from a reference plane, corresponding to the
distance of closest approach of hydrated ions; z. and cio are the charge
number and unperturbed concentration of the it‘ ionic species; 4 is the
local electric potential; and R, F, and T are the gas constant, Faraday’s
constant, and the absolute temperature. The situation is quite analogous
to the junction between a semiconductor and a metal, except that, in an
electrolyte, the density of ionic states close to the interface is large
compared with the electronic density of states, and therefore an ionic
rather than an electronic space charge forms. This perturbation in
concentration extends into the electrolyte for a characteristic distance
known as the Debye length:
Here E is the relative dielectric constant of the electrolyte, and E, is the

permittivity of free space. Electrically, the diffuse part of the double layer
behaves as a potential-dependentcapacitor with a plate separation given
by the Debye length. The capacitance is
C, = [ 2zi2F2E E,c~’
RT ] ’1’
cosh [z]
where 4, is the potential at x = 0 (measured relative to 4 = 0 at x = OD ).
In dilute solutions, the potential drop in the diffuse region is significant,
but in energy storage devices, where high-conductivity situations are
usually encountered, this component is less important than that associated
with the ‘compact’ part of the double layer.
This component is normally described in terms of a ‘molecular
capacitor’ model. According to this description, the electrode is covered
with a monolayer of solvent molecules. The electrolyte ions are essentially
excludedfrom this layer, except when strong chemical interactions result
in desolvation and specific adsorption. The solvent molecules are highly

oriented in the field and, for the case of water, have a much reduced
dielectric constant. A layer of partially oriented water, which also contributes
to the hydration sheath of the first layer of solvated ions, lies outside the
first solvent layer (2). In recent years, an alternative description of the
compact layer has also been considered. This is the ‘jellium’or ‘electronic
capacitor’ model. At the interface, the wave function of the electrons
extends significantly out into the electrolyte, leading to an effective charge
separation. Clearly, the two models differ conceptually to a significant
degree. The experimental evidence which relates to the two models has
recently been reviewed by Martynov and Salem (3).
Fromthe point of view of charge storage devices, the main properties
of the double layer are as follows:
- The double layer behaves as a capacitance in series with the

electrolyte resistance.
- For metallic conductors, the capacitance is on the order of a

few tens of p F per cm‘. In order to achieve significant stored
energy density, very high surface-area electrodes must be
employed. This raises problems of distributed charging
processes, which will be addressed in a later section.
- For a smooth electrode, the double-layercapacitance shows

no frequency dependence up to ionic relaxationfrequencies.
This frequency lies well outside the range of times useful for
energy storage,which are of the order of the RC time constant
or longer.
- The capacitance does show a significant potential dependence

because of non-electrostatic (chemical) factors, as well as
purely coulombic interactions.
2.2 Electrosorption Capacitance
The double-layer charging process described in the previous section

may be regarded as taking place without complete charge-transfer to or
from ionic species either in the solution or adsorbed at the interface.
Evidently, specifically adsorbed species undergo some chemical interaction
with the electrode and there is, therefore, some degree of charge transference
between the metal and the adsorbate. Using surface analytical techniques,
it is possible to assess the degree of covalency of the bond.
Some species, however, can undergo a well-defined Faradaic charge-
Electroc hemical Capacitors 303
transfer process to become adsorbed on the surface. Well-known examples

of this type of reaction involve electrosorption of protons at noble metal
surfaces, e.g.,
Haq+t Pt t l e - + Pt - Had,
This reaction has been suggested as a vehicle for capacitative energy

storage (4) and demonstrated in a solid polymer electrolyte configuration
(5-7). At least some of the transition metal oxide systems, discussed in
greater detail later, also fall intothis category, e.g., thermal RuO, and IrO,.
Because an electron is transferred for every adsorbed proton, much
greater charge densities and capacitances are possible than is the case
for the pure double-layer behavior discussed above. For the example of
H electrosorption on polycrystalline platinum, the charge density associated
with a fully covered surface is 21 0 pC/cm’. The advantage of this is that
less material, i.e., a thinner active layer, is required. Greater charge
densities and capacitances not only reduce the mass and volume of a
device; they also render distributed charging effects less significant. The
differential capacitance associated with electrosorption processes is
often called a ‘pseudocapacitance’ ( 8 ) :
where 0 is the fractional surface coverage by the adsorbed species and

q, is the charge associated with the adsorption of a monolayer ( 9 = 1).
Because the proton adsorbs at a specific site on the surface, with a
well-defined enthalpy of adsorption, the dependence of differential
capacitance on electrode potential may be quite strong. On platinum, for
example, several different site energies may be resolved in the underpotential
deposition region.
Each adsorption process may be described in terms of an adsorption
isotherm, which quantifies the relationship between the degree of coverage,
9 , and the electrode potential. The simplest example is the Langmuir
isotherm (9), which assumes no interaction between the adsorbed species
(i.e., a constant free energy of adsorption, AGO). The coverage is just the
balance between an enthalpic term (driving the adsorption process) and
the entropic term, giving the isotherm:
-A Go -zi EF
-9 - - ci exp - exp -
1-9 RT RT
where ci is the bulk concentration of i.

More sophisticated treatments of electrosorption equilibria lead to

other formulationsof the isotherms (1 0)( 1 1). Terms that may be included
are those due to surface heterogeneity and lateral interactions between
the adsorbed species. In the Temkin isotherm, these terms appear as a
parameter, r, which reduces the free energy of adsorption :
6
-- - ci exp
-
-AGO re -zi EF
1-8 RT RT
In Fig. 2(a), this isotherm is plotted for different values of the interaction
parameter, r. For the case of r = 0, the isotherm corresponds to that of
Langmuir. From the coverageequations, it is straightforwardto derivethe
equations for the differential capacitance.
CadsLangmui=
r
90
zF e( i - e )
Cad?mkin
= q, -
RT 1 +rB(1 - e )
The potential dependence of Cadsis shown graphically in Fig. 2(b),
corresponding to the r values in Fig. 2(a). Dolin and Ershler (1 2) were the
first to investigate pseudocapacitance experimentally. Using a.c. methods
they found Cads'sas high as 2 mF/cm2 for H on platinum. For a more
complete discussion of interaction terms and pseudocapacitance, see
Conway et at. (1 3).
The strong dependence of charge on electrode potential exhibited
by a system such as H/Pt may be a disadvantage in capacitor applications.
A symmetrical device would operate over a rather narrow potential range.
Some conducting metal oxides of the transition and noble metals, however,
show very broad capacitance/potentiaI behavior. This implies that
adsorption siteswith averywide distribution of free energiesare available
for the protons. This behavior may be formally described in terms of
several broad, overlapping Temkin isotherms. It is likely that the origin of
this behavior is to be found in the following effects:
- a broad distribution of enthalpies of ionic interactions, and
- a strong dependence of Fermi level on amount of injected

electronic charge, i.e.,a rather small density of states function
over the potential range.
In contrast to the true double-layer capacitor situation, we may also

(a) POTENTIAL (mV)
2.0 I I I I I I I I I I I 1
t
A
N
.
E
0
1.8
2
i3
1.6
1.4
1.2
c
t
I- I / .
I \
R 1
-I
-I I
U
1.0
0
.J 0.8
4
4 0.6
W
E 0.4
LL
0.2
0
90 120 150 180 210 240 270 300 330 360
(b) POTENTIAL (mV)
Figure 2. (a) Plot of Temkin isotherm for various values of the interaction
parameter, r. (b) Differential capacitance corresponding to the same r-
values.
have to consider the kinetic aspects of the charge-storage reaction. If a

heterogeneous electron-transfer reaction is involved, then this reaction
proceeds at a finite rate, which may be slow enough to appear as an
additional resistancein the charging process. An equivalent circuit which
describes the small-signal response of an electrosorption capacitor
(C,) with finite charging kinetics, as well as finite series electrolyte resistance,
R,, is shown in Fig. 3. The true double-layer capacitance, corresponding
to the effects discussed in the previous section is also includedas C,. The
charge-transferresistance, R, is related to the exchange current density
by the equation
R, = RT/zi Fi,
The exchange current may also vary with the electrode potential through
the dependence of the surface coverage on potential. It should be noted,
however, that the kinetics will not be of the usual Butler-Volmer type,
which describes steady-state behavior, because, in the type of reaction
we are considering, there is no true steady-state current, and a continuous
series of open circuit potentials occurs depending on the amount of
charge that has been passed to or from the electrode. The kinetics are
most conveniently discussed in terms of small-signal behavior,where the
kinetic parameters can be assumed fixed by the d.c. potential and then
investigated by a small-signal perturbation around the d.c. level. In this
way it provespossible, in principle, to study the d.c. potential dependence
of R, and C,. There appear to have been rather few attempts to make
electrochemical studies of this type. For the case of platinum, the kinetic
parameters governing the electrosorption of hydrogen have been determined
(Volmer reaction) (1 4).
In principle, of course, mass transport limitations can also occur in
the electrolyte phase. However, this would usually occur (at least for
protons) only at very low concentrations. Under these circumstances,
diffusion as well as migration would play a role in limiting the interface
charging rate, and the electrolyte could no longer be represented as a
pure resistance in the circuit above.
2.3 Surface Redox Processes
In the previous section, the electrosorption of mobile ionic species

onto an electrode surfacewas discussed. It is also possible to imagine the
oxidation and reduction (with charge transfer) of immobile surface-bonded
species.
S - O x t e- + C++ S-Red
Figure 3. Equivalent circuit for an electrosorption process: R, is the

bulk solution resistance, R, is the charge transfer (kinetic) resistance, C,
is the double-layercapacitance,and C, is the adsorption pseudocapacitance.
Here, S represents the electrode surface. As indicated in this equation,

counterion movement, C', in or out of the interface region is necessary to
restore local electroneutrality. Note that either anions or cations or both
may be involved. The extent to which the charge is localized on the
surface and the detailed involvement of the counterion must be determined
for each individual case. The clearest example of this type of behavior
might be that found in chemically modified electrodes, in which a redox
species is bound to an electrode surface by means of an electrochemically
inactive polymer (1 5)( 16). An example of the modification of a surface by
an adsorbed redox species is to be found in a patent by Calahan et at. (1 7),
where metallorganic molecules are adsorbed onto the surface of carbon,
thus increasing the charge-storage capability of the surface. In fact, it is
likely that many carbons derive a significant fraction of their charge-
storage capacity from the oxidation and reduction of surface functional
groups, e.g.,the quinone-hydroquinone type of reaction (18):
Q t 2H' t 2e-+ H,Q
As in the previous section, electrode kinetics may be important in determining

the total series resistance and hence the charge and discharge rate. As
mentioned above, movement of the counterion may also be important.
This aspect may be easily overlooked, because the counterion may not be
directly involved in a formal description of the charge-transferreaction. It
is probably important in determining the rates of oxidation and reduction
at immobilized electroactive layers (e.g., in Langmuir-Blodgettfilms) and
in conducting polymers (e.g., polyacetylene).
2.4 Thin-Film Bulk Reaction
Dependingon the system, it could certainly be argued that reactions

involving charge storage in conducting polymer electrodes and in some
thin-film oxide systems (e.g.,anodically grown IrO, and nickel oxide) are
actually bulk reactionswhich extend throughout the layer of electroactive
material. Charge is injected into a x electron system in the polymer and
an electrolyte cation or anion moves in or out of the polymer as the charge-
compensating species. Here, the polymer macromolecule behaves as a
giant redox system. The conceptual distinction between polymers and
some metal oxide systems (especially highly hydrated anodic films) may
not be large. The bands of the inorganic crystal act in the same way as the
extended 'IC electron systems in the polymers, and the transition metal
ions undergo redox processes. For a thin film, the electrical responsewill
appear to be capacitative, as long as solid-state diffusion kinetics are
rapid enough to allow equilibration of the concentration profiles throughout

the film at the times of interest (see Section 3.5).
In addition to considerations of solid-state diffusion coefficients and
the intrinsic kinetics of the interfacial electrochemistry,we may also have
to consider nucleation and growth kinetics of new phases that are produced
or removed during the reaction.
As was mentioned in the introduction, this type of reaction is more
usually associated with secondary batteries than with capacitors. Most
battery systems are designed to produce a relatively flat potential vs.
charge relationship during discharge (Fig. 4a), in contrast to the linear fall
in potential expected from a capacitor (Fig. 4b). In a chemical reaction
that involvesthe coexistence of multiple phases, the equilibrium potential
is, in fact, independent of the degree of discharge throughout the multiphase
region. If such electrodes are deployed in thin films, then the entirecharge
will be delivered in a narrow rangeof voltage,followed by a precipitous fall
after the reactants have been consumed. The cyclic voltammogramsfor
a single electrode correspondingto the ‘battery’ and ‘capacitor’ situations,
measured with respect to an arbitrary reference electrode, are shown
schematically in Figs. 4c and 4d. The behavior of a ‘capacitor’ constructed
from symmetric electrodes of these two systems is shown in Figs. 4e and
4f. Dependingon the application, these alternative characteristicsmay be
an advantage or a disadvantage. The possibility also exists of combining
two chemically different electrodes into a single device. This would have
the effect of shifting the origins of Figs. 4e and 4f.
Solid-state reactions leading to useful energy storage or conditioning
processes have been intensively investigatedduring the last two decades
from a battery development point of view (19-21).
The amount of charge stored in an electrode is a function of the
potential, and is related to the amount of stored energy:
We can therefore define an equivalent integral capacitance per electrode

Eq. (1 1) C = 2U/AE2
where A E is the potential range over which the device is to be used. We

can also define a differential capacitancedQ/dE, which will be useful later
when discussing the kinetics of these types of devices.
Eq. (12) C = f(dQ/dE)dE

Figure 4. Schematics of discharge curves for (a) ‘battery-like’ and (b)

‘capacitor-like’electrodes; (c) and (d) are correspondingsingle-electrode
cyclic voltammograms; (e) and (f) are corresponding voltammogramsfor
a symmetrical electrochemical capacitor.
Electrochemical Capacitors 31 1
When E vs. Q is as shown in Fig. 4a, then clearly the differential capacitance
can be very large. If operation over a narrow potential range is required,
then very high capacitances can be obtained.
3.0 DYNAMIC BEHAVIOR OF ELECTROCHEMICAL

CAPACITORS
3.1 General Considerations
In this section, the dynamic electrical behavior of electrochemical

capacitors will be considered. Details of specific systems will be discussed
later, as appropriate.
The traditional approach to understanding both the steady-state and
transient behavior of battery systems is based on the porous electrode
models of Newman and Tobias (22), and Newman and Tiedermann (23).
This is amacroscopicapproach, in that noattempt is made to describe the
microscopic details of the geometry. Volume-averaged properties are
used to describe the electrode kinetics, species concentrations, etc.
One-dimensional expressions are written for the fluxes of electroactive
species in terms of concentration gradients, preferably using the
concentrated solution theory of Newman (24). Expressions are also
written for the species continuity conditions, which relate the time
dependence of concentrations to interfacial current density and the
spatial variation of the flux. These equations are combined with expressions
for the interfacial current density (heterogeneous rate equation),
electroneutrality condition, potential drop in the electrode, and potential
drop in the electrolyte (which includes spatial variation of the electrolyte
concentration). These coupled equations are linearized using finite-
difference techniques and then solved numerically.
This approach has been used to model the behavior of several
different types of battery systems, including a recent study by Lafollette
(25) of a very high-rate lead-acid battery. An important conclusion of this
work was that for thin electrodes, operated at very high rates, electrolyte
depletion at the pore wall is very significant, a factor not usually considered
in more conventional battery situations in which the actual interfacial
current densities are not particularly high.
Although this approach has not been explicitly applied to
electrochemical capacitor modelling, some variation will probably be
necessary when large-signal pulse behavior of, for example, the metal
oxide systems is studied in depth.
The second approach to modelling the behavior of high surface-area
electrodes, is to consider explicitly the geometry of the microstructure
and to solve the linearized response. The simplest model is that of the
uniform cylindrical pore (or rectangular slot, in one dimension), which is

discussed in the next section. An equivalent circuit approach is usually
adopted which makes use of the concepts of distributed impedance and
transmission lines. Variations of this type of model allow for nonuniform
pores, curvature of potential surfaces (in the case of shallow pores or a
rough surface), and a biporous geometry. They are briefly considered
here.
Solid-state mass transport and its effect on the electrical response
are considered in Section 3.5. Again, the discussion is given in terms of
the small-signal response, i.e., the linear behavior of the system. In the
final part of this section, consideration is given to some aspects of the
large-signal response, which are important for real applications, and
which may be difficult to predict on the basis of linear-response
considerations.
3.2 Uniform Transmission Line Model of the Response

of Porous Electrodes
The simplest approach to understanding interfacial charging processes

in porous electrodes is through the use of uniform transmission line
models (26), such as that shown in Fig. 5. Here, the electrode is supposed
to consist of a set of uniform cylindrical pores, each of length I. The
electrolyte resistance per unit length is r (ohm/cm), and the capacitance
per unit length is c (F/cm). The differential equations describing the
current and potential variation with distance are
Differentiating the first equation again with respect to x and substituting

into the second, we obtain
The impedance of this line, for the case in which it has an open
termination ( [ad/&] I,=, = 0 ), is given by the expression:
Eq. (16) Z(j o ) = [r/( j o ~ ) ] ’ /coth

~ I ( jwrc)1/2
This impedance is plotted in the complex plane representationin Fig.

6. Qualitatively, the impedance appears to be a pure capacitance at low
. higher frequencies,
frequencies, where the phase angle tends toward ~ / 2 At
Figure 5. Uniform transmission line equivalent circuit.
REAL Z
Figure 6. Impedance plot corresponding to a uniform, finite-length rc

transmission line with an open termination.
the differential phase isx/4. The critical point to note is that there isa r e a l
partassociatedwith the impedancewhich arisesfrom the distributed pore
resistance, through which the low-frequencycapacitance must be charged.
This is, of course, in addition to any high frequency resistance associated
with the electrolyte outside the porous structure. In a porous structure
it is therefore not possible to charge the interface as quickly as would be
expected on the basis of a single RC time constant, where the value for R
is calculated solely from the electrode plate separation. The additional
distributed resistance can be shown to be equal to rl/3.
This, therefore is the fundamental problem with high surface-area
electrochemical capacitors. In order to achieve the desired high energy
densities, large surface-area materials must be used in a layer of finite
thickness. This automatically ensures that an additional resistance
associated with current penetration of the porous structure will be present.
Whether this additional resistance interferes with a proposed application
isaseparate question. Obviously, since r isequal top/aa2, where p is the
specificelectrolyte resistanceand a is the pore radius, then the greater the
values of p and I and the smaller the pore radius, the more severe is the
distributed resistance problem.
In this simple explanation of the fundamental nature of double-layer
charging in porous materials, the following assumptions were made:
1. Curvature of the potential surfaces was neglected. This is a

good approximation for pores that are long compared with
their diameter. For shallow pores and ‘rough’ electrodes, this
is no longer a good assumption, and Laplace’sequation must
be solved to give the current distribution. This topic is addressed
below.
2. Electrode resistance has been neglected in this treatment. It

may be included in the equivalent circuit of Fig. 5 by including
distributed resistance in the ground line of the circuit.
3. Concentration polarization within the electrolyte has been

neglected.
4. The model is strictly valid only for an identical set of uniform

(constant diameter) pores. As will be seen below, however,
this simple model provides a good description of the electrical
response of carbon electrochemical capacitors. The model
is evidently qualitatively descriptive of the average properties
of some porous materials. It is expected that the model will
apply to a material composed of a set of pores in which r and
c do not vary with position. It is therefore not useful if these

quantities vary with depth of current penetration, as in a
material with a well-defined biporosity, or one in which all
length scales are present, as is found for the case of a fractal
geometry.
3.3 Rough Electrode Surfaces
Under some conditions, it is necessary to calculate the current

distribution in the system from first principles, rather than by merely
assuming a transmission-line-like behavior. Thiswill be the casewhen the
pores are not long compared with their radius. This type of electrode has
been called a rough electrode by de Levie (27). The equipotential surfaces
cannot be assumed to be perpendicular to the axis of the pores. Here, it
is necessary to solve Laplace’s equation
throughout theelectrolyte for the particular geometry and condition at the

boundary. Although numerous methods exist for the time independent
solution to this problem, little attention has been given to the a.c. or
transient solution. This problem has recently been approached numerically
using a boundary integral method (28). This method, in which the
potential in the bulk of the electrolyte is calculated by an integration
around the surface of the region, is particularly suited to electrochemical
problems in that the boundary conditions can be dealt with in a manner
appropriate to the problem. At any frequency and potential, an
electrochemical interface can be representedasa parallel combination of
aresistor and acapacitor. This means that the boundary condition can be
expressed as a linear combination of a Dirichlet and a von Neumann
condition. An example of the results of a calculation for a sawtooth rough
electrode is shown in Fig. 7. In Fig. 7a, the equipotential lines are shown
at a low and a high frequency. In Fig. 7b, the computed complex
impedance plot for the whole frequency range is shown for several aspect
ratios (amplitude of sawtooth divided by peak-to-peak spacing). Several
features are noteworthy. First, the numerical calculation shows some
strong similarities to the transmission line model, in which a region of
lower phase angle is succeeded by purely capacitative behavior at lower
frequencies. The high-frequency phase angle is not n / 4 however, as
would be true for a uniform pore. Rather, it is greater than x/4 to an extent
determined by the aspect ratio of the pore. In addition, the effect of
curvature of the equipotential lines is to lead to a higher effective resistance
at very high frequencies. Instead of asymptotically approaching the real
(4
D: 1 2 5 10
0 1 2 3 4
REAL
(b)
Figure 7. (a) Equipotential lines in the eleclrolyte for a sawtooth electrode

surface in the high and low frequency limits,(b) complex impedance plots
corresponding to (a) for different aspect ratios (height to half-widthratio)
of the groove .
axis at x/4, the locus of the impedance approaches the real axis at x / 2
and at a higher value of resistance than would be expected for a smooth
electrode at the same distance from the reference potential plane. The
cause for this is that the current lines concentrate at the points of the
sawteeth. One conclusion is therefore that for a very thin, high resistivity
electrolyte, the theoretical high frequency resistancecannot be achieved
for a rough front surface.
3.4 Nonuniform Pores
An example of the electrical responseof a nonuniform pore structure

was given in the previous section. An alternative approach is to use the
transmission line approximation discussed in Section 3.2. The model is
modified by allowing r and c to be functions of x. Equation (15) is then
replaced by:
which can be solved analytically for certain functionalformsfor r and c, or,

in general, may be solved numerically. This approach has been adopted
by Keiser et al. (29),who examined the frequency response of various
assumed pore geometries,ranging from the sawtooth, discussed above,
to a very occluded pore (ink-bottle pore). A qualitative representation of
their results, plotted in the complex impedance plane, is shown in Fig. 8.
Sometimes, a particular geometry can be approximated by a combination
of transmission line circuits. An example is the occluded pore (shown in
Fig. 8), which may be modelled by two transmission lines, the first being
terminated by the second. The result of such a calculation is shown in Fig.
9. Note that all geometries have one thing in common. At low enough
frequencies, the result tends to a purely capacitative behavior. It is at
intermediate and high frequencies that the geometry affects the form of
the impedance.
One type of behavior is observed experimentally to be particularly
widespread. It is described by the constant-phase element (CPE):
In the complex plane,this element appearsas a straight line inclined at the

angle Q n / 2 to the real axis. Macdonald has shown that, for physical
situations in which a relaxation time description is appropriate, CPE
behavior may arise from anexponential distribution of activation energies
for the relaxationprocess (30). For porous electrodes, such a description
REAL Z
Figure 8. Results of the calculationsof Keiseret al. (29) of the influence

of pore geometryon impedance. (Adapted w i t h permission from Ref.
29).
ELECTRO
N
*a
rc7a
a
I
REAL Z
Figure 9. Simple two-transmission line approach to the calculation of

the impedance of a pore with two characteristic radii. The different impedance
plots correspond to different relative sizes of the two regions.
is not necessarily helpful, and other explanations should be considered.

If we assume that some geometrical aspect of the porosity is responsible
for this behavior,then an obvious extension of Macdonald’s approach is
toinquirewhetheraparticular distribution functionof poresizesor shapes
might lead to this behavior.
The transmission line response itself, in the region where the current
penetration is not complete, is an example of CPE behavior. It should
perhaps be emphasized here that CPE behavior cannot continue indefinitely
to either infinitely high or low frequencies. At low frequencies, the behavior
will ultimately become purely capacitative and the phase angle will approach
x / 2 . At high frequencies, the current will ultimately withdraw from even
the shallowest pores,and the phase angle willapproach n/4 in the simple
transmissionline case or x / 2 , according to the more realistic calculations
based on Laplace’s equation described above.
It has been suggested that fractal pore structures lead to CPE
behavior during double-layer charging (31)(32). In afractal structure, all
length scales are present, and for a process, such as diffusion, that
contains a well-defined size scale (= [D/u]’~*),it seems to be true that a
CPE with an exponent closely related to the fractal dimension occurs in
the response (33). For the case of double-layer charging, however, the
situation is not so clear because there is no similar characteristic distance
involved (34)(35). It appears that the fractal blocking interface does lead
to CPE behavior, but the exponent is not simply related to the fractal
dimension.
3.5 Mass Transport into a Thin Film
Electrochemical charge storage in a bulk film of material consists of

the movement of two types of species, into or out of the film. These may
be, for example, a cation and an electron. The requirement of charge
neutrality stipulates that the two species move essentially together. We
may therefore solve the coupled transport equations to give an equation
that describes the flux of the neutral combination of species, in terms of
the gradient in concentration (36). This allows us to define the chemical
d i f f u s i o n c o e f f i c i e n t , which describes the motion of a neutral species
in a non-ideal solid solution.
Here D is the chemical diffusion coefficient, and D is the component

diffusion coefficient (which is proportionalto individual particle mobilities).

The asterisk indicates a neutral combination of species, e.g., the ion plus
electron. This equation is valid for general transference numbers and
thermodynamic conditions. From the point of view of overall electrical
response,the most important point is that this equation is still of the Fick’s
law form, even though thermodynamicallythe solution may be non-ideal.
The a.c. response has been calculated for such an electode, subject
to a change in the applied potential or current. The complex impedance
is of the form (37)
1 dE cothl [jW/6I1l2
Z(jW) =- -
zF dc* (
where (1/zF)dE/dc is the thermodynamicenhancementfactor. This term
may be recognized as being the reciprocal of the differential capacitance,
defined in Eq. (1 2). It specifies the relationship between the charge and
the potential in the electrode. The small-signal response to a voltage or
current step has also been calculated (36)(38-40). Diffusion coefficients
have been measured, based on such experiments, for proton and alkali
cationdiffusionintoanumberof different materialsin theformofthinfilms.
Particular attention has been given to tungsten trioxide (37) and iridium
oxide (41), from the point of view of electrochromic applications.
The same equations are also applicable to the kinetics of charge
diffusion in redox systems. For example, Hunter et al. (42) have studied
the charging behavior of polyvinylferrocenefilmsusing this methodology.
It is very important to note the identical forms of Eqs. (22) and (16).
Electrochemically, they correspond to very different processes, even
though the essential physics, that of a distributed storage/dissipation
process is the same. The mathematicalsimilarity is a source of ambiguity
in the interpretation of ax. and transient data. The presence of a ‘Warburg’
impedance (W ’/* dependenceof the impedance) is often taken as proof of
the existence of a diffusion process, as in Eq. (22). It may often, however,
be caused by the presenceof porosity or even roughness in the electrode.
3.6 Large-Signal Response
Most of the discussion of the transient and frequency-dependent

response has been given in terms of the small-signal or linear behavior.
This is obviously a significant simplification, which allows, for example,
analysis of processes in terms of a perturbation around a well-defined
thermodynamic state. However, when larger signals are involvedthis will
not be a good description of the system. This observation is particularly
important for pulse applications of electrochemicaldevices OT in conventional
time-domain studies, where large-amplitude cyclic voltammetry (CV) is

the usual experimental technique.
First, in a porous electrode, an ionic species may be removed during
an interfacialor a bulk discharge process. This may deplete the concentration
of the species in the electrolyte phase within the pore. The fall in concentration
may only be replenished by diffusion into the pore. This situation may be
contrasted with the small-signal case in which the concentration within
the pore may be regarded as constant in the axial direction and only
gradients in a radial direction have to be considered. Depending on the
magnitude of the perturbation, a strong nonlinear response may be
expected. As mentioned in Section 3.1, Lafollette (25) found this to be
important in thin electrodes operated at high current density. A second
class of examples is associatedwith situations in which a critical property
of the electrode material depends very strongly on potential or injected
charge. Gottesfeld et al. (43) have theoretically examined the cyclic
voltammetric behavior of films (such as sputtered IrO, and polyaniline) in
which the film resistance isa strongly nonlinear function of Q. These types
of films often show a ‘prepeak’ in the anodic sweep, which comes before
the main film oxidation peak. No corresponding cathodic peak is seen.
The characteristic of these materials is that, in the reduced state, the film
resistance is very much greater than it is in the oxidized state. Gottesfeld
et al. (43) found, that under these circumstances,the onset of oxidation is
delayed because of IR losses, until sufficient charge has entered the film
to make it conducting. At this point, the thermodynamically required
charge will then flow into the film. The prepeak corresponds to this
required charge. Once the film is conducting, subsequent charge movement
is reversible. From an applications point of view, it is clear that such
behavior is of great importance for large-signal pulse applications, where
the kinetics of charging cannot easily be predictedfrom asmall-amplitude
measurement. A closely related example (the nickel oxide electrode) has
been analyzed experimentally (44).
4.0 CARBON ELECTROCHEMICAL CAPACITORS
4.1 Introduction
In the last two sections, we have considered in a general way the

mechanism of charge storage and the electrical response we might
expect for different types of systems. In the next three sections, we will
consider the properties of specific materials systems and try to relate
those properties to the earlier discussion.
Electrochemical capacitors, based on high surface-area carbon
electrode materials and a variety of electrolytes, are the most highly

developed of the systems discussed in this chapter. Capacitors using
sulphuric acid as the electrolyte are commercially availableand are used
as backup power for memory. They have also found a number of military
applications, includingfusing. After the initial patent of Becker (l), much
developmentwork has been carried out at Sohio (45)(46).Much of the
recent work on carbon systems (especially involving the developmentof
nonaqueous electrolyte systems) is to be found in the Japanese patent
literature.
Carbon is availablewith a wide variety of morphologies and physical
and chemical properties. The subject is comprehensivelyreviewed in a
recent book by Kinoshita (18).To a large extent, the actual carbon used
in commercial electrochemical capacitors has remained proprietary, but
most would appear to be carbon blacks (47) and very high-surface-area
active carbons. The latter materials are prepared from carbonaceous
precursor materials by steam activation. Recent Japanese patents indicate
that a carbon obtained by activation of a carbonized phenolic resin is
particularly suitable, especially for organic solvent-basedsystems. This
material may have a specific surface area up to 2500 m'/g.
Some data on the capacitance of different activated carbons has
been given by Currie (46).In addition, the double-layer capacitance of
carbon electrodes has been studied by a variety of techniques. These
data have been collected by Kinoshita (1 8)and Randin (48).Both carbon
blacks and activated carbons appear to have double-layer capacitances
of about 10 pF/cm2. This is somewhat less than the value expected for a
good metal and is ascribed to the semi-metallic nature of carbon, which
allows a semiconducting space charge to form on the carbon side of the
interface. This space-charge capacitance is significantly less than the
electrolyte dou ble-layer capacitor and therefore dominatesthe behavior.
The values for the capacitance are typically obtained by dividing the
measured capacitance of a flooded electrode by the BET surface area.
Some reservation must be associated with this type of measurement, as
very small pores ( < 20 angstroms in diameter) may not easily be wetted.
The contribution of surface active groups to the interface capacitance
is increased by oxidation, and decreased by a heat treatment in an inert or
reducing atmosphere. Kinoshita and Bett (49)have studied the oxidation
of carbon black (Vulcan XC72) in chromic acid and found that a reversible
-
peak develops in the cyclic voltammogram at 0.5 0.6 volt vs. SHE) with
an increasing level of oxidation. This type of reaction is associated with
oxidation and reduction of carbonyl/quinone groups. Numerous other
studies of the electrochemicalactivity of species on carbon surfaces have
been reported and are collected and summarized by Kinoshita (1 8).
324 Electrochemistry af Semiconductors and Electronics
4.2 Systems with Aqueous Electrolytes
By far the most commonly used electrolyte in electrochemical

capacitors is aqueous sulphuric acid, largely because of its high conductivity,
wide temperature range of applicability, and low cost. If a concentration
of 35% - 40% H,SO, is used, performance remains good down to a least
-55"C,although freezing caused by concentration polarization during
discharge of the electrode would seem to be a possibility at low temperature.
Performancegenerally increases up to at least 1OOOC, the reason presumably
being improvedelectrolyte conductivitythat gives better access to charge
stored in pores at the end of long resistive paths.
This type of capacitor is commercially available with an energy
density of about 1.5 J/cm3 (46)(50). This may be compared with a Ni-Cd
rechargeable battery (450 J/cm3) and a typical electrolytic capacitor
(0.02 J/cm3). Volumetric capacitance density scales as 1/V2 for
electrochemicalcapacitors; electrolytic capacitors decrease less quickly,
and thus the biggest advantage of the electrochemical devices is to be
found at low voltage. Cycle life is excellent ( > 105)and shelf life is also very
good. These devices can be manufacturedfrom very low-cost components.
Typicalfailure modesare solventevaporation,which can be eliminated by
incorporation of a hermetic seal, and container corrosion leading to high
contact resistances and increased leakage currents.
As mentioned earlier, leakage currents are an important consideration
in memory backup applications, where they may represent a significant
fraction of the needed current to maintain memory voltage. Typical
leakage currents for this type of device are about 4 pA/cm2, and are
strongly temperature dependent. Currie (46) has reported an activation
energy of 14 kJ/g for leakage currents in H,SO,. The current is caused by
electrochemical reactions involving active species dissolved in the
electrolyte. Various attempts have been made to reduce leakage currents,
including addition of inert material, such as SO,, A120,, and SIC, to the
electrolyte (51) and saturation of the electrode compartments with
nonproton-containing salts, such as Li2S0, (52). Currie (46) has reported
that the leakage current of a C capacitor using aK,SO,electrolyte is much
reduced compared with a H2S0, electrolyte. A comparison of the rate of
voltage decay for a 5 volt capacitor using these two electrolytes is shown
in Fig. 10. Leakage current decreases with decreasing specific surface
area of the carbon.
A further problem with leakage currents is that they lead to cell
balancing problems. Individual cells with high leakage rates do not
charge to the same voltage as low-leakagecells, leading to failure of the
higher voltage cells. This problem may beaddressed by (1) de-rating the
overall voltage, (2) incorporating redox couples with an appropriate
io5 . . , I
7ESTVALUE
104
I-
10’2l-
F
4‘1 I I I I
-I
I 1
lo-’ 1 10 lo2 lo3
DISCHARGE CURRENT (FA)
Figure 10. Rate of self-dischargeof carbon double-layer capacitorswith

H,SO, and K,SO, electrolytes, expressed as back-up time (time for a
capacitor to discharge from 5.0 to 2.0 volts) at a particular discharge
current. (Adapted with permission from Ref. 4 6 . )
potential to buffer the overcharge, or (3) electronic cell balancing using

shunt resistors or zener diodes.
A great variety of carbons have been used for electrode materials,
some of them exhibiting surface areas significantly in excess of 2500 m2/
g, although values around 1100 m2/g are more normal. Important
considerations in the choice of carbon will include the way the surface
area is distributed (micropores vs. mesopores, for example) and the
distribution function of pore sizes. An interesting recent concept is the
addition of an inorganic redox substance to the surface of an active
carbon, e.g., H,PMo,,VO,,, to add storage capacity (17). This idea could
undoubtedly be extended to other substrates and redox couples.
Small-signalimpedancedata for a carbon capacitor are presentedin
complex impedance-planeform in Fig. 11. The best fit to Eq. (1 6), which
describes the behavior of a uniform transmission line, is also plotted. It is
clear that, to a first approximation, the transmission line model is quite
good, and can certainly be used for predicting the small-signal behavior
of the device, once the transfer-function parametershave been obtained.
It is also obvious that, in order to access the low frequency capacitance,
the devices have a high esr. As was discussed in an earlier section, this is
due to the contribution of the electrolyte resistancewithin the poresof the
electrode. Currie (46) has also examined the effect of carbon type on the
electrical response. Figure 12 shows some of his results on the frequency
dispersion of the phase angle for several types of carbon. The frequency
response of devices with 60 b diameter pores is significantly better than
that of an activated carbon with 20 b pores, and is improved even more if
a carbon with a controlled 170 b pore size is used.
Another consequence of the delay associatedwith the transmission
line behavior is voltage recovery after termination of a constant current
discharge. This is caused by a redistribution of the remaining charge in
parts of the electrode with a longer time constant.
4.3. Nonaqueous Electrolyte Systems
Much recent development work has focused on the use of organic

solvent-based electrolyte systems for use with carbon electrodes. The
principal advantage of this approach is the removal of the limitation on cell
voltage imposed by the electrolysis of water and the consequent increase
in energy density, which is proportional t o v (53). Because cell voltages
of 3 volts or more are easily attained, an order of magnitude increase in
energy density is possible if the stored charge density were the same as
in aqueous systems. Several Japanese patents claim cell voltages in the
range of 4 to 5.5 volts and specific capacitances greater than 100 F/g. A
list of the more common electrolytes and solvents is given in Table 1.
0
1 2 3 4
REAL Z (Q)
Figure 1 1. Impedance plot (0) for a commercial carbon/H,SO,double-

layer capacitor at room temperature.Calculateddata points ( A ) , based on
the fit to Eq. (1 6 ) are also shown.
Figure 12. Effect of pore size and distribution on the electrical behavior
of various types of carbon. Dependenceof phase angle on frequency for:
( 0 )76-angstrom controlled porosity carbon; ( 0 ) 166-angstrom controlled
porosity carbon; ( x ) C black (Black-Pearls); ( +) high surface-area
activated carbon. (Adapted with permission from R e f . 4 6 . )
Table 1
Nonaqueous Solvents and Solutes Used
in Carbon-Based Electrochemical Capacitors
Solvents Electrolytes
propylene carbonate LiCIO,
7-butyrolactone R,N+A
where R = Et, Pr, Bu etc.
A = BF,, CIO,, CF,SO,
sulpholane RSO,M
where M = alkali, NH4+,NR,+
These are essentially the same electrolyte solutions that have been
found useful for nonaqueous lithium batteries, and their properties relevant
to that application have been reviewed (54). Other advantages may
include low leakage currents and absence of gas evolution on overcharge.
The conductivities of these electrolytes, however, are much lower than
those of aqueous solutions and an increase in series resistance is unavoidably
associated with their use. Although detailed transient or impedance
studies appear not to have been published, it is safe to assume that the
increase in electrolyte resistivity is also detrimental to dynamic response,
because of increased transmission line effects in the porous electrode
structures.
A partial description of the Matsushita ‘Gold’ capacitor has been
given (55). The carbon is based on a carbonized and activated (800OC -
-
1 100°C for 1 2 hrs. in water vapor carried in nitrogen) phenolic resin and
hasasurfaceareaof upto2500m2/g. Thistypeofcarbonalsohasarather
high electrical conductivity, of about 60n-lcm-l.It can be made in various
forms, such as felts, fiber, cloth, and paper, and is thus quite suitable for
electrode fabrication. Specific capacitance vs. specific surface area was
found to be linear at room temperature. In order to get good performance
at low temperatures (-25OC), it was found necessary to have more than
50% of the pore volume in pores with a diameter of 200 h or greater.
Plasma-sprayed aluminum was used for making contacts.
Nawa et al. (56)have also described a capacitor based on the same
type of carbon using LiCl0,and rbutyrolactone as the electrolyte. These
authors believe that cell capacity is due to a superimposition of a double-
layer charging process and an electrolyte intercalation process (galvanic
cell mode). The intercalated species are Li+ and CIO,-. Evidencefor the
Faradaic processes is found in the dependence of charge recovered on
cell voltage. It was found that more charge was stored in the cell than
would be expectedfrom the linear dependenceof Q on V associatedwith
double-layer charging. Intercalation in this type of carbon is not unlikely,
as its high conductivity implies a well-developed graphitic nature. Specific
capacitance of the system was 36.5 F/g of carbon, and specific energy
and power were 130 kJ/kg and 11.1 kW/kg, respectively, based on
carbon and necessary electrolyte. Self-dischargewasreportedto be 40%
- 50% after one day.
Yashima et al. (57) have investigatedthe self-discharge in activated
-
carbon fiber LiCIO,/PC cells. Water and oxygen, dissolved and adsorbed,
were found to be the principal sources of self-discharge. After removalof
these impurities, residual reactionswere probably the reduction of propylene
carbonateat the negativeelectrode,irreversible dopant reactionswiththe
carbon and the presence of surface functional groups, e.g., -OH, -COOH
and -CHO.About 70% of the self discharge apparently occurred at the
negative electrode.
Nogami et al. (58) have described a symmetrical activated carbon
fiber cell using Bu,NCIO, as the electrolyte.
4.4 Solid Electrolyte Systems
From the point of view of lifetime, mechanical ruggedness, and,

possibly, ease of manufacture, solid electrolyte systems are very attractive.
Two main difficulties are, however, apparent. The first of these is the
difficulty in makingan intimate high-surface-areasolid-solidinterface and
maintaining it over a wide temperature range. Even if the electrolyte can
be introduced into the pores as a liquid, it may prove difficult to maintain
contact on solidification by cooling or solvent evaporation. The second
problem is that most solid electrolytes are distinctly inferior to liquid
electrolytes in terms of their resistivity and the temperature coefficient of
that resistivity.
An early solid-state device, based on the solid electrolyte AgBr, was
patented by Raleigh (59). The solid Ag’ conductor RbAg,l,, however, has
a much higher room temperature ionic conductivity (= 0.27n-’cm-l)(60).
Oxley (61)(62) studied an electrochemical device that made use of this
electrolyte with an asymmetric arrangement of electrodes, Ag 1 RbAg,I, IC.
Within the potential range of 0 to t 5 0 0 mV (vs. Ag+/Ag), the device
behaved almost perfectly. Interface areas up to 10 m2/gwere obtained,
leading to packaged energy densities of 1 to 1.7 J/cm3. The device had
excellent charge retention properties, because of the very low electronic
conductivity of this electrolyte, and also an attractively wide temperature
range. At -55OC, the conductivity only falls to 0.09Q-’cm-’, and at t 75OC

it is 0.39 Q %n-’. Other silver ion conductors have also been proposed
(63).
Although they have lower ionic conductivities, analogous and related
copper compounds have also been used (64). They appear to have about
half of the specific energy achievedwith the Ag’ conductor. The best of
the Cut electrolytes is Rb2Cu,13CI, which has a 25OC conductivity of
1 Q-Icm-’. The copper electrolytes seem to show considerably
6 . 4 ~0‘3
more self-dischargethrough electronic conductivity than does RbAg,15.
Other inorganic solid electrolytes have also been employed in
electrochemical capacitor applications, including hydrogen uranyl
phosphate, (H,OUO,PO,) and Zr(HPO,),-xH,O (proton conductors) (65),
Li and Na /3-alumina (66), LiNaSO, (67). Clearly, these systems will only
be satisfactory for relatively slow applications.
Devices basedon ionically-conducting polymers appear perhaps to
be more attractive than the inorganic solid electrolytes. U.S. patents
describe a capacitor based on the use of polyethylene oxide (PEO)
containing dispersed lithium salts as an electrolyte (68)(69). The PEO
acts as a solvent for the electrolyte salt, and can be applied as a liquid (in
a solvent such as acetonitrile) to the high surface-area electrode, ensuring
good penetration of the pores. Much research has been carried out on
polymeric lithium ion conductors for lithium-battery applications (70),
and PEO is no longer considered one of the best (Le., highest conductivity)
systems (71). We can presume that improved capacitor performance
could also be obtained if some of the more recently developed systems
were tried, particularly close to room temperature.
5.0 TRANSITION AND NOBLE METAL OXIDE CAPACITORS
5.1 Introduction
In this section, we will consider the properties of perhaps the second

most well developed of the electrochemical capacitor systems: those
based on the oxides of the transition and noble metals. In general, a
distinction is made between the metal oxide systems and the carbon
systems on the basis that the former are redox active materials, Le.,
charge is stored by means of changes in the oxidation state of the metal
when an ionic species (e.g., a proton) reacts with the surface (or bulk) of
the oxide, and the ‘chemisorption’ is accompanied by the simultaneous
injection of an electron into the oxide. The carbon capacitors are supposed
to be more purely of the double-layeror space-chargevariety. In fact, as
was pointed out earlier, the distinction is less clear. In the case of carbon,
surface functional groups may be oxidized and reduced in very much the
Next Page
Previous Page
range. At -55OC, the conductivity only falls to 0.09Q-’cm-’, and at t 75OC

it is 0.39 Q %n-’. Other silver ion conductors have also been proposed
(63).
Although they have lower ionic conductivities, analogous and related
copper compounds have also been used (64). They appear to have about
half of the specific energy achievedwith the Ag’ conductor. The best of
the Cut electrolytes is Rb2Cu,13CI, which has a 25OC conductivity of
1 Q-Icm-’. The copper electrolytes seem to show considerably
6 . 4 ~0‘3
more self-dischargethrough electronic conductivity than does RbAg,15.
Other inorganic solid electrolytes have also been employed in
electrochemical capacitor applications, including hydrogen uranyl
phosphate, (H,OUO,PO,) and Zr(HPO,),-xH,O (proton conductors) (65),
Li and Na /3-alumina (66), LiNaSO, (67). Clearly, these systems will only
be satisfactory for relatively slow applications.
Devices basedon ionically-conducting polymers appear perhaps to
be more attractive than the inorganic solid electrolytes. U.S. patents
describe a capacitor based on the use of polyethylene oxide (PEO)
containing dispersed lithium salts as an electrolyte (68)(69). The PEO
acts as a solvent for the electrolyte salt, and can be applied as a liquid (in
a solvent such as acetonitrile) to the high surface-area electrode, ensuring
good penetration of the pores. Much research has been carried out on
polymeric lithium ion conductors for lithium-battery applications (70),
and PEO is no longer considered one of the best (Le., highest conductivity)
systems (71). We can presume that improved capacitor performance
could also be obtained if some of the more recently developed systems
were tried, particularly close to room temperature.
5.0 TRANSITION AND NOBLE METAL OXIDE CAPACITORS
5.1 Introduction
In this section, we will consider the properties of perhaps the second

most well developed of the electrochemical capacitor systems: those
based on the oxides of the transition and noble metals. In general, a
distinction is made between the metal oxide systems and the carbon
systems on the basis that the former are redox active materials, Le.,
charge is stored by means of changes in the oxidation state of the metal
when an ionic species (e.g., a proton) reacts with the surface (or bulk) of
the oxide, and the ‘chemisorption’ is accompanied by the simultaneous
injection of an electron into the oxide. The carbon capacitors are supposed
to be more purely of the double-layeror space-chargevariety. In fact, as
was pointed out earlier, the distinction is less clear. In the case of carbon,
surface functional groups may be oxidized and reduced in very much the
same way as surface oxide and hydroxide groups on oxides, and intercalation
reactions are a real possibility. Also, the term redox is perhaps misleading,
in that most of the oxides are good electronic conductors and it is to be
expected that crystallographically-equivalent individual metal ions will
have the same oxidation state, which is intermediatebetween those of the
formal redox states. This is not necessarily always the case, since
electrons may be localized in surface states, or other defects, in which
case specific ions may be said to have different oxidation states. This is
also true of some transition metal oxides that are intrinsically semiconducting
(e.g., the vanadium oxides) and will, in general, require electronically
conducting additives to function well as battery or capacitor electrodes.
Other featuresalso blur the distinctions. Aswill be described below,
the evidence suggests that, in the case of thermal RuO,, the ionic processes
involvedare largely restricted to the surfaces of individual oxide grains. In
the caseof nickel oxide, however,it is much more likely that the bulk of the
lattice is readily accessed by protons. For IrO,, it seems that anodically
grown films behave rather like nickel oxide, whereas thermal films behave
almost identically to RuO,. There is thus a continuous spectrum of
behavior which, in principle, may be defined in terms of bulk vs. surface
ionic reactions and delocalized vs. localized electronic behavior. From
the point of view of the energetics of the charging process, both the ionic
and electronic components may formally be said to contribute to the total
free energy of the interaction. Clearly the availability of empty electronic
states (either bulk or surface) is a prerequisitefor any type of interaction,
and thedensityofsuchstatesisprobablyadominantfeatureinthevoltage
vs. charge characteristic of many electrodes. However, the ionic contribution
to the enthalpyis also important and it controls the ion exchangeandacid-
base behavior of oxide interfaces. The ionic interactions may in turn
modify the electronic states and, therefore, the assumption of rigid bands
may not be a good basis for predicting the energetics of the charging
process. Unfortunately,the degree to which these features are important
for individual oxides has scarcely been addressed, and conflicting opinions
are widely found in the literature.
5.2 Thermally Prepared Oxide Films
Because of its importance as a dimensionally stable anode material

(DSA) for the evolution of chlorine and oxygen, RuO, in pure form, or
mixed with the oxides of other metals, including Ta, Ti and Ir (72),has
received considerableattention. IrO, has also been investigated, but to a
lesser extent, presumably due to its greater cost. Several books and
review articles have been devoted to the physical, chemical, and
electrochemical properties of these and related oxides (73).
Films are prepared by thermal decomposition and oxidation of

soluble precursor materials. For RuO,, the most common precursor is
RuCI,.xH,O, dissolved in water or an alcohol, such as isopropanol. The
solution is applied to a valve-metal substrate, the most common material
beingtitanium, by painting, dipping, or spraying. Thechloridecan then be
decomposed and oxidized in air or oxygen to RuO, at relatively low
temperatures (3OOOC - 5OOOC). The properties of the oxide, e.g., surface
area, depend significantly on the preparation temperature.
The highly capacitive nature of thin films of these oxides, when
immersed in an aqueous electrolyte, is well known. Cyclic voltammograms
of RuO, and IrO, are shown in Figs.13a and b. The specific capacitance
of the RuO,, determined by Rutherford backscattering and cyclic
voltammetry is about 350 F/g (74). This may be comparedwith about 100
C/g for carbon, achieved on much higher surface-area material. A
striking feature of these data is the broad range of potential over which
charge is stored. This is actually an important distinguishing feature
between ‘chemisorption’ on these oxides and the underpotential deposition
of, for example, hydrogen on platinum, in which case adsorption takes
place over a rather narrow potential range.
In situ quartz crystal microbalance studies (75)(76) have confirmed
the cationic nature of the charging process: the film mass decreases
proportionately to charge as the potential is made more positive (Fig. 14).
Attempts have been made to determine the real charge density on RuO,
by comparing BET surface areas and electrochemically measured
capacitance. Burke et al. (77) found a value of about 160 pC/cm2, and
Raistrick and Sherman (74) found somewhat higher values, in the range
of 200 - 300 pC/cm2, on oxide optimized for capacitor applications. The
specific surface area of the RuO, was about 120 m2/g, a value in line with
transmission electron micrographs that indicated a microcrystalline structure
with a particle diameter of about 200 angstroms. A straightforward
calculation of the densityof -OH groups on the surface of a rutile-structure
oxide (78) indicates that a charge density of 200 pC/cm2 is expected if
each surface group is singly ionizable, Le.,
M-OH + H+ t e - = M-OH,
Obviously, if multiple ionization processes are possible, then more charge

can be accommodated on the surface. It is therefore unnecessary to
postulate bulk reaction of protons with RuO,. Other evidence that points
to the same conclusion may briefly be summarized as follows:
- Although the a x . impedance behavior of theelectrodes is not

well explained by solid-state bulk diffusion, it is understandable
- 8 I I I 1 I
4
N
-
5 2
h
v
E
I - 0
z
W
a
a -2
3
0
-4
I I I I I I I I I
12
10
8
-
N
6
5 4
4
E
v 2
I - 0
z
w -2
oc
5 -4
-6
-8
-10
-12 I I I I I I I I I I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
POTENTIAL (Volts vs SCE)
Figure 13. Cyclicvoltammograms in 0.5 M H,SO,at 50 mV/ssweeprate,

for (a) thermal RuO, and (b) thermal IrO,.
I I
RuO,,
-
50 mV,s
I I
/
-
-0.3 -- e
200 -
2 100 -
-1
0 0.2 0.4 0.6 0.8 1.o

V (SCE)
Figure 14. Dependence of frequency of Ru0,-coated-quartz crystal

microbalance electrode on potential in 0.5 M H,SO, shown below the
corresponding cyclic voltammogram.
in terms of the charging of a porous medium (see below).
- Other ions (e.g., the alkali metals) also interact strongly with
these oxides, and, although the charge density is somewhat
lower, the kinetics of charging are essentially identical to
those found with protons, if the differing conductivities of the
electrolytes are taken into account. This result is expectedfor
a fast surface electrosorption process, but is quite incompatible
with solid-state diffusion.
The kinetics of the charging process have been studied in some

detail using a.c. impedance spectroscopy (74)(79)(80). The nature of the
responsedepends on the methodof film preparation, but usuallytwo time
constants are clearly seen in the impedancedata, a result which suggests
an approximate equivalent circuit of the type shown in Fig. 3. Various
authors have proposed differing explanationsfor theorigin of the two time
constants. Rishpon and Gottesfeld (79) have suggested that two mass-
transport processes are involved, the faster being associated with the
surface of the grains and the second with diffusion into the bulk. The
possibility that the high-frequency time constant was associated with
finite interfacial kinetic processes (Le., a charge transfer resistance in
parallelwith a double-layercapacitance)was suggested by Abraham and
Rajeshwar (80), who also attribute the low-frequency behavior to a diffusion
process. Raistrick et al. (74) have examined the possibility of charge-
transfer kinetic limitations by varying the proton concentration and the
conductivity of the electrolyte separately. ltwas found that both R, and R,
were proportional to the electrolyte conductivity, but independent of the
proton concentration. This suggests that the form of the response is
determined by charging in a porous medium, but with an effective geometry
which is significantly different from that encountered in the case of the
carbon electrodes.
Films of oxides may also be prepared by sputtering. The most
intensively studied material is IrO, because of its possible application as
an electrochromic material (81-83). Films prepared by this method seem
to be more similar to films prepared thermally, than to those prepared by
anodic oxidation (see next section). Aurian-Blajeni et al. (81) have
interpretedthe two time constants in the impedance spectrum as caused
by a two-layer structure in the oxide.
5.3 Anodically Prepared Films
Many transition and noble metaloxides can be anodically oxidized in

aqueous electrolyte solutions to form oxide coatings. In some cases,
these special electrochemical treatments, e.g., cycling or pulsing, allow

very thick layers to be built up, which have significant charge-storage
capacity. Besides Ir, other metals that show this behavior are Ru (84), Ni
(44)(85), Rh (86), and Mn (87).
In some cases, the behavior of the anodic films is significantly
different from the thermally prepared materials. Hepel et ai. (88) found
that the energy of interaction of protons with IrO, was about the same for
sputtered films and for single crystals, but differed significantly from that
associated with anodic films.
The growth of an iridium oxide film in acid solution is shown in Fig. 15.
(Note the small anodic pre-peak discussed in section 3.6.) The charge
increases steadily with number of growth cycles and approaches values
comparable to those found in thick thermal films. A comparison with the
cyclic voltammogram of a thermal film of IrO, in the same electrolyte (Fig.
13b) illustrates some of the differences. At cathodic potentials, the
capacity of the anodic films is rather low, but it shows a strong peak at
about 0.7 volt (SCE) and a broad highly capacitive response at more
positive potentials. Small-amplitudeimpedance data for such a film are
presented in Fig. 16 for several different potentials which cover the
complete range of behavior. At anodic potentials, the data are well
explained in terms of a single, almost ideal capacitor in series with the
electrolyte resistance. As the potential is made more negative, it is clear
thatasecondresistanceentersintothe descriptionof theresponse. Atthe
lowest potentials, this resistance dominates the behavior and allows only
avery limited degree of charging of the film. This resistance may safely be
ascribed to the film itself and presumably corresponds to a dramatic
change in the electronic properties. From the point of view of applications,
this change in the capacitance and conductance of the film is a disadvantage.
Equally striking, however, is the almost perfectly capacitive behavior of
the film at positive potentials. Data at 0.9 volts show no evidence of
distributed charging behavior. This behavior is intriguing in that the
electrode is behaving as if it were nonporous (Le., no frequency dispersion)
and yet shows very high charge-storagecapacity. The likely explanation
is that the oxide is, in fact, exceedingly porous, with a structure similar to
that of a polymer, inwhich chainsor small numbers of atomsor molecules
are completely permeated by and accessible to electrolyte species. A
hexagonal structure has been proposed for anodic IrO, (89).
Confirmationof cationic (H,O+) involvementin the charging process
for anodic IrO, was obtained by Mclntyre et ai. (90) who examined the
composition of oxidized and reducedfilms using RBS and nuclear reaction
analysisof deuterated specimens. The average D:lr ratiowas found to be
1.51 in oxidized and 2.28 in bleachedfilms. The composition of the films
was also found to be nonuniform. The interior of the films becomes
14
12
10
8
6
a
-
E 4
0
$ 2
ZO
-2
-4
-6
-8
-0.2 0 0.2 0.4 0.6 0.8 1.o 1.2
v (SCE)
Figure 15. Growth of anodic IrO, on an Ir wire in 0.5 M H,SO,. The
number of cycles is indicated.
c
I I I
-:- -1 0.2 v
-
-+-
i
-
---
0.4 V
4
-\-
1
%
-- 0.5 V
h -3---=- 0.6 V
C
v -p 0.7 V
w c-- 0.8V
2 3 -’*c,-
0.9 v
3 ---- 1.0 v
w
n
2 2
c7
s
1
0 I I
-1 1 3 5
LOG OMEGA (radiansk)
Figure 16. Impedance magnitude data for an anodic IrO, film measured
at different electrode potentials (vs. SCE).
increasingly anhydrous with almost no water being found at the metal/

oxide interface. The composition of the outer layer was lr0,-2D20 in
oxidizedand IrO,. De 3D,O in reducedfilms. Thesameauthorsfoundthat
in aprotic solvents, lithium and sodium ions could be inserted into the
structure, to a composition of about LilrO,; however, K’ could not be
inserted.
Using chemical analysis, Pickup and Birss (91) found evidence of
non proton cation involvement in the redox processes in aqueous solutions.
5.4 Other Oxides
Many transition metal oxides are electrochemically active and can

therefore be used as the electrode material in an electrochemical capacitor.
Most of these oxides, however,are more justifiably considered as battery
materials because extensive bulk involvement is normal. Well-known
examples are the oxides of manganese, nickel, vanadium, tungsten,
chromium, copper, cobalt, and molybdenum. Some of these materials
are among the most important components of well-established secondary
battery systems and have been extensively studied and the results reviewed
(19)(20)(80). Many of them are usedwith both aqueous and nonaqueous
electrolyte systems. In the table below, the charge densities of some
oxide electrode systems are given. These numbers give some indication
of the energy-storage capability of a ‘capacitor’ constructed from these
materials. The corresponding charge density for RuO, in H,SO, is about
350 C/g.
Table 2
Rechargeable Oxide Electrodes and Charge Densities.
Oxide Charge Density

(C/g)
LixV,O, = 650
LiXV8013 = 1400
LixCr20, = 972
MnO, 1120
LixCoO, 670
NiOOH 1040
Ago 1570
P bo, 810
These oxides will not be considered further here as little work has
been carried out with the express purposeof optimizingthem in capacitor-
like configurations. Exceptions are the patent of Elliot and Huff (92) and
the recent work of Rauh (93) (see below).
5.5 Metal Oxide Capacitors Utilizing Solid and Polymeric

Electrolytes
As is the case for carbon-baseddevices, polymeric electrolytes have

been usedwith transition metal oxides. McHardy (5-7) has used IrO, (and
also platinum metal) as electrodes deposited in and on the surface of
Nafion. (Nafion isaregistered trademark of E. I.duPont de Nemours, Inc.)
The electrode fabrication is achieved by ion-exchanging hydrazine for
polymer protons and then immersing the structure in a salt of the noble
metal. Charge storage capacities in excess of 10 mC/cm2were obtained.
The best performancewasobtainedwith very thin (1 mil) membranes, and
excellent discharge characteristics were obtained on millisecond time
scales. Between 0.5 and 1.1 volts the charge was proportional to the
voltage.
A similar concept has been explored by Rauh (93) who prepared
highly dispersed,very thin electrodes of a number of oxides and organic
polymers, includingVOX,IrO,, RuO,, and NiO, in a capacitor configuration.
Vacuum and electrochemical depositiontechniqueswere used to deposit
very thin layers of metals and oxides on ion-exchangeor other polymeric
electrolytes and separators. Clearly, these types of configurations show
much promise for very high rate applications.
6.0 CONDUCTING POLYMERS
6.1 Introduction
Although little direct developmental work appears to have been

carried out, the possibility exists of using conducting organic polymers as
active materials in electrochemical capacitors. They have certainly been
actively investigatedas battery electrode materials (94), and many of the
same requirements hold for the capacitor case. Superficially at least,
conducting organic polymers resemblethe inorganic oxides discussed in
the previous section, especially low-density anodic oxide films, such as
anodic IrO,. Although thepossibilityof an all-polymercapacitor or battery
is very attractive (using for example a PEO-typeof electrolyte) (95)(96),
the conducting polymeric materials are not particularly processable,
being neither meltable nor soluble (although poly 3-alkyl thiophenes have
been made soluble) (97).
6.2 Charge-Storage Mechanism
The main classes of organic materialsare as follows: polyacetylene,

polypyrrole, polyaniline, polyphenylene,and polythiophene. These materials
share the characteristic that, on ionic doping, electronic carriers are
introduced into a delocalized ‘IC system, rendering them very good conductors
(98). As in the case of the oxides, this is a double-injection process, with
counterions entering the polymer structure. Several of the polymers can
be produced by in situ electrochemical processes. Their microstructure
is typically very dependent on preparation conditions.
Some of the polymers, e.g., polyacetylene, can be either p- or n-
doped. For example, the empirical equations describing potential vs. a
lithium electrode, as a function of doping for this material are (99):
v, = 3.43 t 0.1 4 log Q , and
V, = 1.5-(0.13 t 0.02Q)lOgQ
where Q is the %oxidation. There is thusa gap in the densityof states. For
a symmetrical cell with an electrolyte such as n-Bu,N+PF,- in propylene
carbonate or sulpholane, there is a potential difference of about 2.5 volts
at 7% doping. A capacitor based on this arrangement would not have a
linear Q-V behavior; the potential would fall rapidly to zero once the gap
potential difference is reached on discharge. The behavior is more
‘battery-like’than ‘capacitor-like’. Polyacetyleneprepared by the Shirakawa
method (100) is about 33% dense and consists of a mat of interconnected
fibrils with a diameter of about 200 angstroms. The surfacearea is 60 - 100
m*/g. Both the n- and p-doped materials are very sensitive to air and
water. Recently, new preparative routes have led to polyacetylene with
conductivity comparable to that of metallic copper (101).
Polyaniline can be prepared electrochemically, and its microstructure
depends on the preparative conditions (102). The charge capacity is
comparable to the inorganic oxides: Gottesfeld et al. (103) report a
capacity of 800 C/cm3 in an aqueous acidic electrolyte and Genies et al.
(102) a value of 450 C/g in propylene carbonate/LiCIO,.
The use of aqueous electrolytes is certainly an advantage for high-
rate applications, and polyaniline can be cycled in water over a limited
potential range. Polyacetylene,however, is very unstable, in both n- and
p-dopedforms, to traces of water and oxygen. At more positive potentials,
polyaniline is unstable to oxidation and undergoes fragmentation and
dissolution (104). Similarly, polypyrrole is fatally oxidized above 1 volt vs.
SCE in aqueous solutions (105).
6.3 Electrical Response
A cyclic voltammogram for electrochemically grown polyaniline is

shown in Fig. 17. The electrical responseof the conducting polymers, like
that of the anodic oxide films, is not understood in great detail (106). The
film resistancestypically vary strongly with potential through the doping
process, and there appear to be two time constants associated with
charge storage. One explanation is that a fast double-layer charging
process occurs on the surface of the polymer fibrils, and then a slow
diffusive process involving penetration of the fibrils by ions takes place.
An alternative explanation may be that there are two phases which are
dynamically distinguishable. Rubinstein(107) has pointed out, however,
that the fact that the two capacitances are potential dependent argues
against this hypothesis. Clearly, charging kinetics may depend profoundly
on microstructure, and a wide range of size scales and geometries have
been found in all these materials, depending on the conditions of preparation.
It is to be expected that charging kinetics will depend specifically on the
ion size, charge etc.
Jow and Shacklette (108) have interpreted their results on n-doped
polyacetylene in terms of a charge-transfer resistance and a diffusion
process. (See Section 3.5). The measured double-layer capacitance,
however, does not scalewith thicknessof the films, and isfar smaller than
would be expectedfor a materialwith the observed surface area (50 - 150
m'/g). This suggests, perhaps, that the observed high-frequency process
is controlled by the porosity rather than a charge-transferreaction, as in
the case of RuO,. Specific capacitances for n-doping are voltage dependent
and range from 95 F/g at 1.1 volts vs. Na (NaB4, in THF) to 320 F/g at 0.6
volts.
7.0 CONCLUSIONS
In this section, the performance characteristics of electrochemical

capacitors, both realized and projected, are discussed. The limited data
available are presented in the form of a specific energy vs. specific power
plot (Fig. 18). Although this type of plot is not particularly useful or
significant for assessment of conventional capacitors, where other
considerations, such as average power capability at some working
frequency, prevail, it provides a useful method of separating different
characteristics of energy storage devices.
The power capability of a capacitor-like device may be quantified in
several different ways. The power delivered into a load resistance, R,, by
a capacitor, C, with an internal resistance, R,, is given by the equation:
Om41
0.3 1
I I I
I I I
0.2 0.3 0.4

V (SCE)
I -0.3 4
Figure 17 . Cyclic voltammogram of electrochemically prepared polyaniline

on a platinum electrode. 50mV/s in 0.1 M HCI.
!- I
i
100
F BATTERIES
I
. I”.
CONVENTIONAL CAPACITORS
1 1
Figure 18. Specific energy vs. specific power for a number of batteries
and capacitors. 1 = electrolytic capacitors (maximum power); 2 =
Maxwell polycarbonate,with 2a = average energy at 1 kHzrepetition rate,
2b = single-shot energy density; 3 = projected Ru0,-based devices (see
text); 4 = Pinnacle Ultracapacitor (1 lo), 4a = maximum power, 4b =
power at 1 msec (=T /25); 5 = Sohio Maxcap (46)(50) 5a = maximum
power, 5b power at T . Batteries (from Ref. 19): a = Li/MnO,; b = Na/S; c
= Ag/Zn; d = alkaline; e = Li/V,O,; f = NijZn, g = Ni/Cd; h = Ni/H; i =
P b/H,SO,.
In the absence of any information besides V and R,, it is therefore possible

to define a maximum power at zero time equal toV/R,. Clearly, however,
the time and load requirementsshould be matched to the propertiesof the
- -
capacitor (time r , where r = R,C, RL R,) for efficient power transfer.
V2/R may be very large for a small capacitor, but if t > > r then little useful
power is available at times of interest. If weight and size are not critical
then more power can be obtained by making t < r , Le., by using a larger
capacitor.
The average power delivered to timet
-
P(t) = 1/tS'p(t' ) dt '
0
is equal to
Significant quantities can be definedfor P and Pinto matched loads (R, =

RL),and at the time constant. In Fig. 18, the method for calculating the
power is indicated where appropriate.
The specific energies and powers of a number of secondary battery
systems are collected in the same figure. Their relatively large energy
storage capability is a consequence of the fact that most batteries utilize
bulk electrochemical reactions as their storage mode. Relatively few
attempts have been made to optimize conventional batteries for power
output, rather than energystorage. It is clear that by developing multicell,
thin-layer, bipolar stacks, great improvementsin battery power would be
possible. The recent work of Lafollette (25) and Gibbard (109) illustrates
the potential of this approach. Lafollette has reported power densities up
to800 kW/kg, for a bipolar lead-acid battery, and Gibbard has reported 75
kW/kg at 10 msec for a high temperature Li-Si/FeS2 stack. Battery
current density is, in some cases, limited by kinetic considerations (electrode
kinetics, solid-state diffusion, or phase nucleation) and, in others, by
ohmic losses caused by low electrolyte conductivity.
Conventional capacitors, on the other hand, have very low specific
capacitances, and hence rely on a high cell potential difference to give
them appreciable energy-storage capability. They are characterized by
very high power output for short times. Their energy density is limited by
dielectric breakdown. It seems that, even in systems specifically engineered
for highenergycontent, avalueof 0.5 J/cm3 (or J/g) is difficulttoachieve,
and values several times less than this are usual upper limits on energy
density.
The propertiesof electrochemical capacitors should be intermediate
between those of conventional capacitors and of batteries, both from an
energy and power density point of view. In battery terms, bulk mass
transport in the electrode structure is avoided by the confinement of
chemical reactivity to a surface region. Surface electrosorption processes
reduce the importance of electrode kinetics. (In the case of a pure double-
layer process, of course, there is no kinetic component.) From a conventional
capacitor perspective, electrochemical capacitors achieve high charge
densities by using the equivalent of very large area plates. This last
advantage is offset considerably by low operating voltages and high
effective series resistances.
As may be seen from Fig. 18, commercial carbon-based
electrochemical capacitors exhibit a relatively low energy density,
approximately 1 to 2 J/cm3 and less than 0.5 J/g. It should be remembered,
however,that this is still several times greater than the energy density of
most electrolytic capacitors. Further, the complications caused by
distributed charging ensure a relatively low power density. (The value
shown in the figure is the maximum power based on an esr measured at
1 kHz. In fact, as has been seen, most of the capacitance is accessed
through a somewhat greater esr.) Carbon electrochemical capacitors,
therefore, owe their commercial success, not to superior energy and
power capability so much as to low cost, extremely high cyclability, and
wide temperature range of operation. Because of their much higher
operating cell voltage, the organic electrolyte-based carbon capacitors
offer significantly higher energy densities.
In many respects, the use of metal oxides as active materials for
electrochemical capacitors promises much improved performance over
that of the carbon type. Particularly for applications in which fast response
is important (e.g., pulse power, filtering), the high charge density in these
materials means that thinner active layers are possible for the same
overall capacitance. The main advantage of this is not in the associated
improvementsin energy density, but in the improved electrical response.
As we saw in an earlier section, the effects of distributed charging are not
completely eliminated in metaloxide capacitors, but they are significantly
reduced compared with those in carbon capacitors.
Based on the small-signalresponseof RuO, capacitors, it is possible
to estimate their performance,assuming that no large-signal problems
arise (such asconcentration polarization). In Fig. 18,aregion of expected
performance is shown for devices based on RuO,. The calculation
included measured pore resistance (based on the data in Ref. 74) and the
weight of all materials except for the package. A factor of 2 - 3 reduction
in specific quantities might be expected if the packaging were included.

The power densities are calculated at the capacitor time constant, for
discharge into a matched load. It was assumed that bipolar plates would
be constructedfrom 1 mil-thick Ti, and that the separator would also be 1
mil thick. A reduction in sulphuric acid conductivity of 50% was assumed
to account for separator tortuosity.
Pinnacle Research Inc. has published performance data on metal
oxide capacitors (of undisclosed composition) in which H,SO, was used
as the electrolyte (1 10). A 1O-volt bipolar device,with an active electrode
area of 2 cm2,gave a volumetric energy density of 0.9 J/cm3. Although
this is comparable to presently available carbon electrochemical capacitors,
the esr is significantly better at 30 m n per repeat, and the energy density
of the active region (23 J/cm3) indicates that the packaging of the device
is not optimized. A prototype 20-kW device demonstrated some of the
potential of this technology. It was an 84-volt package with 0.49 F
capacity. The esr (at 1 kHz) was 50 m i l . The energy density was 6.4 J/
g and 20 J/cm3. This result is quite close to estimates of projected metal
oxide performance. The power densities quoted by Pinnacle are for a 1
msec pulse, whereas the actual time constant of the capacitor was closer
to 25 msec. The power at the time constant would be significantly less
than the 1 msec power level, as discussed above.
Conducting polymers offer a charge density similar to that of metal
oxides. In addition, it seems possible that all-plastic, electrochemical
capacitors may become a reality. A more detailed understanding of the
kinetics of charging in both polymersand metal oxides is clearly desirable
if their potential is to be realized.
From the point of view of future applications, it seems that, at the
moment, the limiting feature of electrochemical capacitors is their poor
frequency response. Good performance at 60 Hz would be an important
achievement, opening up the area of compact power supply application
for these devices.
8.0 ACKNOWLEDGEMENTS
The author would like to thank Thomas Springer (Los Alamos) for
many contributions to the mathematical modelling aspects mentioned
here, Ruth Sherman (Los Alamos)who gathered much experimentaldata
on RuO,, Judith Rishpon (University of Tel Aviv) who made the quartz
crystal microbalance measurements of Fig. 14, and Bernard Boukamp
(University of Twente) who has allowed the author to use his non-linear
least-squarederror fitting routines for impedance data analysis.
9.0 REFERENCES
1. Becker, H. I., U.S. Patent 2,800,616, July 23, (1957)
2. (a) Bockris, J. O’M., Devanathan, M. A. V. and Muller, K., Proc.

Roy. SOC.,Ser. A 274, 55 (1963)
(b) Bockris, J. O’M. and Reddy, A. K. N., Modern Electrochemistry,
Plenum, New York (1970)
3. Martynov, G. A. and Salem, R. R., Adv. Coll. Interface Sci.,

22:229 (1985)
4. Craig, D. R., European PatentApplication No. 78,404A2, September

30 (1983)
5. McHardy, J., Proc. l€C€C. 22:306(1987)
6. McHardy, J., Ludwig, F. A., Higley, L. R., Kindler, A. andTownsend,

C. W., U. S. Patent 4,766,522, August 23 (1 988)
7. McHardy, J., in: Proc. U . S. Army Workshop o n Capacitors and

Batteries for Pulse Power Applications, Laboratory of the U. S.
Army, LABCOM, Asbury Park, NJ, Nov. 17-18 (1987)
8. Grahame, D. C., J . Electrochem. SOC., 99:370c (1952)
9. Langmuir, I., J . Am. Chem. SOC., 40:1361 (1918)
10. Frumkin, A. N., Z. Physik, 35:792 (1926)
11. Temkin, M. I., Zh. Fiz. Khim., 15:296 (1941)
12. Dolin, P. andErshler, B., Acta Physicochim., USSR 13:747(1940)
13. Conway, 8. E., Tellefsen, K. and Marshall, S . , in: The Chemistry

and Physics of Electrocatalysis, (J. D. E. Mclntyre, M. J. Weaver,
and E. B. Yaeger, eds.), Proceedings Vol. 84-12, p. 15. The
Electrochemical Society, Pennington, NJ (1 984)
14. Breiter, M., in: Trans. Symp. on Electrode Processes, (E.

Yaeger, ed.), p. 307, John Wiley and Sons, New York (1961)
15. lwakura, C., Kawai, T., Nojima, M. andyoneyama, H., J. Electrochem.

SOC., 134:791 (1987)
16. (a) Neff, V. D., J. Electrochem. SOC., 132:1382(1985)

(b) Honda, K. and Hayashi, H., J. Electrochem. SOC., 134:1330
(1 987)
17. Calahan, J. L., Cuellar, E. A., Desmond, M. J. and Currie, J. C., U.S.
Patent 4,633,372, December 30 (1 986)
18. Kinoshita, K., Carbon, €lectrochemica/ and Physicochemical

Properties, John Wiley & Sons, New York (1988)
19. Desilvstro, J. and Haas, O., J. Electrochem. SOC., 137:5c(1 990)
20. Eichinger, G. and Besenhard, J. O., J. Electroanal. Chem., 72:l

(1 976)
21. Abraham, K. M., J. Power Sources, 7:l (1981)
22. Newman, J. andTobias, C. W., J. €lectrochem. SOC., 109:1183

(1 962)
23. Newman, J. and Tiedermann, W., AlChE Journal, 21 :25(1975)
24. Newman, J., €lectrochemical Systems, Prentice-Hall, Princeton,

NJ (1973)
25. Lafollette, R. M., Design Fundamentals of High Power Density,

Pulsed Discharged, L e a d Acid Batteries, Ph. D. dissertation,
Brigham Young University, Provo, Utah (1 988)
26. de Levie, R., in: Advances in Electrochemistry a n d Electro-

chemical Engineering, (P. Delahayand C. W. Tobias, eds.), Vol.
6,p. 329,Interscience, New York (1967)
27. deLevie, R., Electrochim. Acta, 10:113(1965)
28. Springer, T. E. and Raistrick, I. D., To be published
29. Keiser, H., Beccu, K. D. and Gutjahr, M. A., €lectrochim. Acta,

21 539 (1 976)
30. Macdonald, J. R., J. Appl. Phys., 61:700(1987)
31. Liu, S.H., Phys. Rev. Lett., 55529 (1985)

32. Nyikos, L. and Pajkossi, T., Electrochim. Acta., 30:1533 (1985)
33. Nyikos, L. and Pajkossi, T. Electrochim. Acta., 31:1347 (1986)
34. Wang, J. C., Electrochim. Acta., 34:987 (1989)
35. Keddam, M. andTakenouti, H., Electrochim. Acta, 33:445 (1988)
36. Weppner,W.andHuggins, R.A., J . Electrochem. SOC.,124:1569

(1977)
37. Ho, C., Raistrick, I. D. and Huggins, R. A., J . Electrochem. SOC.,

127:343 (1980)
38. Weppner, W. and Huggins, R . A., in Annual Review of Materials

Science, (R. A. Huggins, ed.), Vol. 8, p. 269,Annual Reviews Inc.,
Palo Alto (1978)
39. Wen, C. J., Ho, C., Boukamp, B. A., Raistrick, I. D., Weppner, W. and
Huggins, R. A., lnt. Met. Rev., 5:253 (1981)
40. Raistrick, I. D. and Huggins, R . A., Solid State lonics, 7:213

(1982)
41. Glarum,S. H.and Marshall, J. H., J . Electrochem. SOC.,127:1467

(1980)
42. Hunter, T. B., Tyler, P. S., Smyrl, W. H. and White, H. S., J .

Electrochem. SOC.,134:2198 (1987)
43. Gottesfeld, S., Redondo, A., Rubinstein, I. and Feldberg, S. W., J .

Electroanal. Chem., 265:15 (1989)
44. Glarum, S.H. and Marshall, J. H., J . Electrochem. SOC.,129535

(1982)
45. (a) Boos, D. L., U. S.Patent 3,536,963, Oct. 27, 1970

(b) Boos, D. L., Adams, H. A., Hacha, T. H. and Metcalfe, J. E., in:
2 Ist Electronics Component Conference, p. 338,
Washington, DC (1971)
46. (a) Currie, J. C., in: Proc. U . S . Army Workshop on Capacitors

and Batteries for Pulse Power Applications, Laboratoryof
Electrochemical Capacitors 3s1
the U. S. Army, LABCOM, Asbury Park, NJ, November 17-18

(1 987)
(b) Currie, J. C., Boos, D. L., Gaylor, V. F.and Bennet, P. D., in: The
Chemistry and Physics of Composite Media, (M. Tomkiewicz
and P. N. Sen, eds.), Proceedings Vol. 85-8, The Electrochemical
Society, Pennington, NJ (1 985)
47. Watanabe, A., Mori, K., Ishikawa, H. and Nakamura, Y., J.

Electrochem. SOC.,134:1318(1987)
48. Randin, J. P., in: Encyclopedia of the Electrochemistry of the

Elements, Vol VII, (A. J. Bard, ed.), p1, Dekker, New York (1976)
49. Kinoshita, K., and Bett, J. A. S., Carbon, 1 1 :403(1 973)
50. Maxcap Double Layer Capacitor, Characteristics, Operation

and Use, Sohio EngineeredMaterials Company, Niagara Falls, NY
(1985)
51. Marse, A. V., Bennett, P. D. and Boos, D. L., U.S. Patent 4,605,989,
August 12 (1 986)
52. Bennett, P. D. and Currie, J. C., Eur. Pat. Appl. 200,327,

December
10 (1986)
53. Butherus,A. D. and Newby,K. R., U. S. Patent 3,700,975,

October 24
(1 972)
54. Besenhard, J. 0.and Eichinger, G., J. Electroanal. Chem., 68:l

(1 976)
55. Yoshida, A., Tanahashi, I., Takeuchi, Y. and Nishino, A., / € E €

Trans. Components, Hybrids and Manufacturing Tech., 1O:lOO
(1 987)
56. Nawa, M., Nogami, T. and Mikawa, H., J . ,Electrochem. SOC.,

131:1457(1984)
57. Yashima, H., Nogami, T. and Mikawa, J., Synth. Metals, 10:229
(1985)
58. Nogami, T., Mikawa, H. and Nawa, M., J. Phys. (Paris), 44:C3
(1 983)
59. Raleigh, D. O., U.S.Patent 3,419,760, December 31 (1968)

60. (a) Bradley, J. N. and Greene, P. D., Trans. Farad. SOC.,62:2069
(1966)
(b) Ibid. 63:424 (1967)
(c) Owens, B. B., and Argue, G. R., Science, 157:308 (1967)
61. Oxley, J. E., in: Proc. 24th. Power Sources Conference, p. 20

(1970)
62. Oxley, J. E., in: Proc. U . S. Army Workshop on Capacitors and

Army, LABCOM, Asbury Park, NJ, November 17-18 (1987)
63. Portier, J., Reau, J. M., Sun, H. W.,Tanguy, B., Astier, R., Cambie, C.,
Maurin, M., Pradel, A. and Ribes, M., F r . Demande, 25941 16:14
August (1987)
64. (a) Sekido, S. and Ninomiya, Y., Solid State lonics, 3/4:153
(1981)
(b) Sekido, S., ibid. 9/10:777 (1983)
(c) J. Electrochem. SOC.,126:1654 (1979)
65. Velasco, G., Colomban, P., Adet, P. and Pham, T. M., Eur. Pat.
Appl., 208:589, January 14 (1987)
66. (a) Xue, R. and Chen, L., Solid State lonics. 18-19: 1 134 (1986)
(b) Matsushita Electric Industrial Co., Ltd. Japanese Patent 57/
1 1139 84, March 2 (1982)
67. Matsushita Electric Industrial Co., Ltd. Japanese Patent 57/14019

B4, March 20 (1982)
68. Lundsgaard, J. S., U. S. Patent 4,638,407, January 20 (1987)
69. Currie, J. C., Difranco, L. F. and Bennett, P. D., U.S. Patent 4,730,239,
March 8 (1988)
70. Armand, M., in: Ann. Rev. Mat. Sci., (R.A. Huggins,ed.),Vol. 16,
p. 245, Annual Reviews Inc., Palo Alto (1986)
71. Scrosati, B., J . Electrochem. Soc., 136:2774 (1989)

72. See, for example:

Beer, H. B., U. S. Patent 3,632,498, January 4 (1972)
73. (a) Trasatti, S., in: Electrodes of Conductive Metal Oxides

(Part B) (S. Trasatti, ed.), Elsevier/North Holland, New York
(1981)
(b) Daghetti,A..Lodi,G.andTrasatti,S.,Mat. Chem. Physics, 8:l
(1 983)
(c) Trasatti, S., and O’Grady, W. E., in: Advances in Electro-
chemistry and Electrochemical Engineering, (H. Gerischer
and C. W. Tobias, eds.), 12, p177, Wiley, New York (1981)
74. (a) Raistrick, I. D.and Sherman, R. J., in: Proceedings of the

Symposium on Materials and Processes for Energy
Conversion and Storage, (S.Srinivasan, S.Wagner and H.
Wroblowa, eds.), Proc.Vol. 87-12, The ElectrochemicalSociety,
Pennington, NJ (1987)
(b) Raistrick, I. D.,Beery, J. G. and Sherman, R. J., in: Proceedings
of the Symposium o n Electroceramics and Solid-State
lonics, (H. L.Tuller and D.M. Smyth, eds.), Proc. Vol. 88-3, The
Electrochemical Society, Pennington, NJ (1988)
75. Buttry, D.and Gottesfeld,S., Electrochem. SOC.Ext. Abstracts,

Abstract No. 1404, Honolulu Mtg (1987)
76. Raistrick, I. D.and Rishpon, J., unpublished work
77. Burke, L. D.,Murphy, 0. J., O’Neill, J. F. and Venkatesan, S., J .

Chem. SOC.Faraday Trans. , 1 :73 (1977)
78. Furlong, D. N., Yates, D. E. and Healy, T. W., in: Electrodes of

Conductive Mefal Oxides (Part B), (S. Trasatti, ed.), Chapter 8,
Elsevier/North Holland, New York (1981)
79. Rishpon, J. and Gottesfeld, S., J. Electrochem. SOC., 131:1960

(1984)
80. Abraham, P. and Rajeshwar, K. (To be published.)
81. Aurian-Blajeni,B., Boucher, M. M.,Kimball, A. G. and Roblee, L. S.,

J . Mat. Res., 4:440 (1989)
82. Gottesfeld, S., J . Electrochem. SOC., 126:742 (1979)

83. Kang, K. S. and Shay, J. L., J . Electrochem. S O C . ,130:766 (1983)
84. Burke, L. D. and Whelan, D. P,, J . Electroanal. Chem., 103:179

(1979)
85. Burke, L. D. and Whelan, D. P., J. Electroanal. Chem., 109:385

(1 980)
86. Burke, L. D. and O’Sullivan, E. J. M., J. Electroanal. Chem., 93:ll

(1 978)
87. Burke, L. D. and Murphy, 0. J., J. Electroanal. C h e m . , 109:373

(1980)
88. Hepel,T., Pollack, F. H. and O’Grady, W. E., J. Electrochem. SOC.,

132:2385 (1985)
89. Michell, D., Rand, D. A. J. and Woods, R., J . €lectroanal. Chem.,

84:117(1977)
90. Mclntyre, J. D. E.,Basu, S., Peck, W. F., Brown, W. L. and Augustyniak,

W. M., Phys. Rev., B25:7242 (1982)
91. Pickup, R. and Birss, V., J . Electroanal. C h e m . , 240:171 (1988)
92. Elliot, W. E. and Huff, J. R., U. S. Patent 3,317,349,May 2 (1967)
93. Rauh, D., in: Proc. U . S. Army Workshop on Capacitors and

Army, LABCOM, Asbury Park, NJ, November 17-18 (1987)

(a) Synth. Metals 28 (1 989)
(b) Proceedings of the lnternational Conference on Sci.
and Techn. of Synthetic Metals, Santa Fe,NM, June 26-
July 2 (1988)
95. Chiang, C. K., Polym. Comm., 22:1454 (1981)
96. Nagatomo, T., Ichikawa, C.,and Omoto, O., J. €lectrochem.

S O C . ,134:305 (1987)
97. Feldhues, M., Karnpf, G., Litterer, H., Mecklenberg, T. and Wegener,
P., Synth. M e t . , 28:C487 (1989)
Electrochemica Capacitors 355
98 Electronic Properties of Conjugated Polymers, Springer Ser.

I.
in Solid State Sciences, (H. Kuzmany, M. Mehring,and S. Roth,eds.),

Springer Verlag, Berlin (1 987)
99. MacDairmid,A. G.andKaner, R. B., in: Handbook of Conducting

Polymers, (T. A. Skotheim, ed.), Volume 1 , p. 689,Marcel Dekker,
New York (1 986)
100.Shirakawa, H. and Ikeda, S., J. Polym., 2:231 (1971)
101.Naarman, H. andTheolon, N., Synth. Met., 22:l (1987)
102.Genies, E., Hany, P. and Santier, C., Synth. Met., 28:C647(1989)
103.Gottesfeld, S.,Redondo, A., and Feldberg, S. W., €/ectrochemical

SOC.Extended Abstracts, No. 507,San Diego, CA (1986)
104.Glarum, S. and Marshall, J., J. Electrochem. SOC., 134:2160

(1 987)
105.Asavapiriyanony, S.,Chandler, G. K., Gunawardena, G. A. and Pletcher,

D., J. €lectroanal. Chem., 177:229(1984)
106.Feldberg, S.,J. Amer. Chem. SOC., 106:4671(1984)
107.Rubinstein, I., J . Electrochem. SOC.,134:3078(1987)
108.Jow, T. R., and Shacklette, L. W., J. €lectrochem. SOC., 135541

(1 988)
109.Gibbard, F., in: Proceedings of the Symposium on Electro-

chemical and Thermal Modelling of Battery, Fuel Cell and
Photoenergy Conversion Systems, (J. R. Selman and H. C.
Maru, eds.), Proc. Vol. 86-12,
p. 193,The Electrochemical Society,
Pennington, NJ (1986)
110.Tong, R. R., Mason, G. E., Lee, H. L., Bullard, G. L., in: Proc. 34th.
Power Sources Symposium (1 988)
INDEX
activation energy 57-59, 93, 104, 115, chemical potential 55, 56

260, 269,281, 287,289, chemical vapor deposition see CVD
290,324 codeposition 5, 6, 18, 22
activation overpotential 64,65 compound semiconductor 4, 116
activity coefficient 3 concentration ovelpotential 64
admittance 227 condensation 255, 257, 258, 263, 264,
adsorption 55, 56, 72, 76, 77, 81, 82, 290
85, 104, 105, 107, 224, conduction band 6567, 131,178, 187,
257-260, 262, 269, 204, 208, 220, 223
289-291, 302-304, 333 contact displacement 85, 91
amorphous silicon 15 corrosion 150, 182, 183, 185, 187,
anisotropic etching 105 190-192, 199, 217, 220,
anode 4, 10, 14, 15, 34, 35, 39, 45, 62, 243,245,287,324
64,69, 100, 259, 278, critical current 6, 256, 257, 259, 269,
282,288,289,332 276,278,281, 286,287,
anodic oxidation 127, 137, 139, 141, 29 1
144,149,150,165,166, cryolite 9, 10
168,336 crystal growth 2, 72, 139, 166, 256
Arrhenius equation 93, 290 current leakage 290
arsenide 44 CVD 47, 101,139,168,186
autocatalytic reaction 68 Czochralski 139
BaTiO, 182 dark current 137, 206,220, 225
BET isotherm 257, 262, 269, 290,323, Debye length 301
333 dendrite 255-257, 262, 263, 266, 276,
Bi,S, 16 278, 281, 282, 285-288,
bipolar electrode 297, 345, 347 291
blazed structure 196 deposition potential 2, 6, 18, 22, 24, 25,
block patterns 101 27, 32, 35, 38
borides 2 desorption 56, 104, 148
Burgers vector 76, 79 dielectric constant 88, 144, 223, 224,
butyrolactone 329 234,238,301,302
capacitor 127, 128, 131, 150, 225, 297, differential capacitance 303, 304, 309,
299-302, 304, 306, 309, 311, 321
31 1,315,323,324,326, diffraction grating 243
329, 331-333, 337, diffusion coefficient 22, 57, 75, 159,
339-342, 345-347 227,320,321
carbides 2,44 diffusion length 189, 190, 227, 240
carrier concentration 100, 128, 131, diffusion overpotential 63, 65
137, 243 dislocation 75-77, 79-82, 85, 88, 89,
cathode 2, 4, 9, 10, 14, 18, 35, 38, 39, 104, 107, 110, 112, 218
41, 45, 62-65, 70, 100, distributed impedance 312
260,278,281, 286, 288 DMSO 24,25, 27, 30
CdS 3, 8, 16, 18, 24, 25, 27, 30, 39, donor concentration 16
41, 141, 166, 182, 243 doping 2, 15, 34, 141, 190, 192, 220,
CdSe 16,22,141,182,228 223,228, 234, 235,245,
CdTe 3, 4, 18, 19, 22, 24, 25, 27, 139, 341,342
168,182 double layer 25, 55, 56, 59, 60, 62, 63,
charge storage 300, 302, 308, 320, 69,85, 242,301,302
322,340,342 drift velocity 261
chemical diffusion coefficient 320 ECE mechanism 161, 163
chemical etching 53, 54,58, 96, 101, edge dislocation 76, 80
112,139,191,192 EDTA 22
356
Index 357
effective series resistance see ESR 195, 196, 199,

electrochemical cell 5, 35, 62, 63, 203, 203, 204, 206,
232 208, 209, 211,
electrochemical potential 56 217, 218, 222,
electrocrystallization 44, 255 238, 245
electrodeposition 2, 3, 6, 6-10, 15, 16, Galvani potential 64
19, 24, 27, 32, 34, 35, galvanic cell 62
38, 39, 41, 45, 47, 141, GaP 2, 3, 24, 32, 34, 35, 38, 39, 128,
204, 263 139, 182, 220, 222, 232,
electrodissolution 100 234,240,242,243,257,
electroless deposition 24 262-264, 266, 273, 278,
electroluminescence 70, 223, 243 286, 288, 341
electrolytic cell 62, 98 Ge 30, 41, 178, 182, 222
electrolytic etching 58, 64, 98, 100 Gibbs free energy 54,297
eleclron-hole pair 223 Gouy-Chapman layer 55
electrons 3, 8, 62, 65, 67, 68, 96, 98, grain boundaries 22, 110
133,159,177, 178,183, heat treatment 15, 22, 30, 209, 323
ia7,204,208,223,23a, Helmhotlz layer 234
278,281,302,332 HgTe 143, 152
electropolishing 98, 100, 139, 150 hole drilling 196, 199, 201
electroreflectance 219, 222, 224, 232, holes 65,67, 68, 70, 96, 98, 100, 133,
234, 235,245 177, 178, 182, 185-187,
electrosorption 302-304, 306, 336, 346 190, 192, 199, 203, 204,
electrowinning 9, 14 208,236
elemental semiconductor 4, 8 holography 189
energy conversion 16, 218 humidity 257-259, 262-264, 266, 269,
energy storage 297, 299-303, 309, 342, 273, 290
345 Il-VI Compounds 3, 16, 24, 55, 127,
entropy 54 141, 166, 178, 190
epitaxial growlh 34, 35, 38, 211 Ill-V CompOUnds 3, 16, 34, 55, 178,
epitaxy 47, 101, 139 187,190, 192, 196, 209
equivalent circuit 163, 225, 228, 242, impedance 161, 163, 219, 224, 225,
306,312, 314, 336 228, 232, 240-242, 312,
ESR 299, 326, 346,347 314,315,317, 321, 326,
etalon 189 329,333,336,337, 347
elch figures 53, 76, 110, 112 impedance spectroscopy 219,225,228,
etch pit 88 232,336
facet formation 101 impurity atoms 256
Faraday constant 56 infrared spectroscopy 240
Fermi level 131, 209, 218, 220, 228, InP 2, 3, 39, 41, 44, 100, 107, 112,
234, 235, 304 115, 116, 182, 186, 190,
fiber optics 177, 186, 199 191,201,203,204,218,
Fick's laws 57, 63, 321 243
fixed charge 137, 150, 165, 166, 168 interfacial tension 88
flat band 128, 183, 204, 220, 223, 228, intermediates 69, 165, 287, 317, 332,
232,234,235,243 346
fluorosilicate 14 ion beam 143, 186,196, 201,219
free energy 3, 54, 72, 73, 79-81, 143, ionic transporl 258, 262, 265
256,261,297,303, 304, IrO, 303,308, 322,332, 333, 336,337,
332 339,340
fused saii 34, 35, 44 isotropic etching 105, 116
GaAs 2-4, 44, 69, 70, 91, 93, 96, 98, IV-IV compounds 16,44
112, 116, 182, 183, 185, jellium 302
186, 191, 192, kink site 75,85
Langmuir isotherm 303 passivation 98, 100, 127, 128, 137,

Langmuir-Blodgenfilms 308 139,141,144,148,150,
laser 139, 143, 148, 178, 187, 189, 152,161,165, 166, 168,
191,196,199,201,203, 260, 276, 281, 285, 291
211, 242 passive film 100
lateral diflusion 190, 192 passive layer 259
leakage current 289, 291, 324 PBC 105, 107
LEDs 38, 44, 177, 186, 201, 203, 206, PC 15,25, 27, 30,329, 330, 341
34 1 PEO 331,340
light emitting diodes see LEDs periodic bond chain see PBC
light guides 199 periodic structures 187, 192, 21 1
limiting current 6, 19, 63, 72 phase diagram 142, 143
linear defects 110 phosphide 34, 39
liquid junction 218-220, 222-224, 228, photocapacitance 219, 232
232, 234, 240, 241 photocapacitor 133
luminescence 70, 222-224 photocarriers 96
macrosmoothing 100 photochemical process 19, 22, 24, 30,
mass transfer 183, 220 168, 209
MBE 139 photoconductor 127, 128
mechanical polishing 127, 139, 149 photoelectrochemical deposition 203
mercury depletion 143, 148, 149 photoelectrolysis 218
metal-insulator-semiconductor see MIS photoelectrolytic etching 58, 100
metal-oxide-semiconductor see MOS photoetching 58, 96, 187, 211, 218
microsmoothing 100 photoluminescence 30, 209, 222, 223,
migration 16, 72, 115, 255, 258, 259, 241,243
262,263,265,266,269, photon flux 180, 183
273,276,278,281,282, photon limited region 183, 195
285-292, 306 photons 98, 180, 183, 185, 190, 195,
MIS 127,128,133, 137 222,242
MOCVD 38,139 photoreflectance 222, 232, 235, 238,
molten salt 2, 15, 34,39, 44 240
MOS 128, 150,209,218,228 photovoltage 96, 232, 235, 240
Mott-Schottky equation 228, 232, 242 photovoltaic 127, 133, 137, 218, 244
n-type 18, 22, 24, 32, 35, 44,68,96, piezoelectric 25, 244
98, 100, 131, 133, 143, pinning 218, 228, 235
144, 161, 178, 183, 187, pits 53, 76, 77, 82, 85, 93, 105, 107,
206,220,223 199
non aqueous electrolytes 24, 243 pitting 98, 100
nucleation 25, 38, 72, 73, 76, 77,79, planar defects 110
80, 206, 208, 263, 285, plasma oxidation 168
309,345 Pofonak 39
nucleophilic attack 70 polishing 88, 89, 91, 98, 100, 101, 110,
ohmic contacts 180, 187 115,127, 128, 139, 149,
optoelectronics 16, 44 161, 196
organic electrolyte 346 polyacetylene 308, 341,342
outer potential 55 polyaniline 322, 341, 342
overpotential 6, 18, 39, 63-65, 152, polyethylene oxide see PEO
178,260,281,287, 288 polymer 180, 206, 258, 265, 276,
overvoltage see overpotential 289-291, 303, 308, 337,
psi 204, 208 340-342
ptype 15, 18, 27, 32, 68, 70, 96, 98, polyphenylene 341
131,133,137,163,165, polypyrrole 341
166,178,187,203, 204, polythiophene 341
209 pore distribution 257, 263, 264
index 359
pore size 269, 326 301, 323

porous electrode 311, 322, 329 spherical diffusion 255, 288, 291
potenticdynamic 143, 150, 152, 159 step motion 104
Pouaaix diagram 143 sulfide 8, 16, 18, 19, 24, 30, 32, 34,
pre-electrolysis 18, 22 141, 166,209,286
propylene carbonate see PC sulfidization 139, 165, 166
pseudocapacitance 303, 304 sulpholane 329, 341
pulsed electrolysis 15 surface adsorption 72,76
quantum efficiency 183, 185, 223, 234, surface depletion 183
244 surface diflusion 72, 75, 77, 190
quantum yield 183, 185, 190, 192,201 surface potential 55, 128, 131, 133,
rate-limiting step 73, 93, 104 137,234,235, 240
recombination 96, 98, 100, 112, 131, surface preparation 133, 139, 141, 149,
133, 137, 178, 183, 190, 152, 166, 211
196,223,241, 243 surface reaction 70, 104
rectifying contacts 204 surface roughness 100
redox process 24, 89, 91, 107, 199, surface state 168, 177, 209, 228
242,306,308,321,324, Temkin isotherm 304
326,331,332,339 ternary compounds 2, 30, 32
reference electrode 6, 16, 25, 180, 182, thin film 2, 32, 308, 320
220,309 threshold volage 278, 281, 286
refractive index 144, 189, 244 TiO, 182, 206, 218, 228, 242
relative humidity 228, 257-269, 273, transfer coelficient 59, 65
276, 278, 290, 304, 337 transkion time 259, 260
relaxation time 317 transmission line 312, 315, 317, 320,
reverse bias 220 326,329
ring-disk 70, 152, 159, 163 twin boundaries 110
roughness 100,228, 285,321 undercutting 112, 115, 192
RUO, 303, 332, 333, 339, 342, 346, underpotentialdeposition 303, 333
347 valence band 65,67, 70, 178, 182, 199,
S-pits 107 204, 223
saturated calomel electrode see SCE vitreous carbon 44,45
scanning tunneling microscope 203 Vola potential 55
SCE 18, 22, 32, 100, 163, 180, 185, volammetry 5, 322, 333
186, 192, 195, 196, 203, Warburg impedance 163,321
206,208,337,341 water-drop test 259, 260, 276, 282,
Schottky Barrier 177, 206, 209, 285. 288, 290-292
217-219, 228, 232, 238, zeta potential 163
240, 242, 243 ZnO 34,58, 182,206, 243
screw dislocation 79 ZnSe 30
selective etching 53, 76, 82, 89, 115
sensitization 242
sic 2, 3, 44, 45, 182, 186, 211, 324
silicate 9, 10
silicide 9
silicon 4, 6-10, 14, 15, 30, 34, 38, 44,
45,47,67,92, 101, 104,
127,133,137,168,209,
240
single crystal 34, 35, 45, 291
solid electrolyte 330
solvation 27, 58, 289
space charge 131, 187, 189, 190, 201,
223,225,228, 238, 242,

Electrochemistry of Semiconductors and Electronics Processes and Devices - John Mchardy

Uploaded by

Electrochemistry of Semiconductors and Electronics Processes and Devices - John Mchardy

Uploaded by

ELECTROCHEMISTRY OF

John McHardy and Frank Ludwig

Published in the United States of America by

Library of Congress Cataloging-in-Publication Data

Electrochemistry of semiconductors and electronics : processes and

Electronic Materials and Process Technology

Ceramic and Other Materials-Processing and Technology

Giulio Di Giacomo R. David Rauh

Robert C. DeMattei Keshra Sangwal

Ian D. Raistrick Micha Tomkiewicz

Final determination of the suitability of any information or

In Chapter 1, R.C. De Mattei and R.S. Feigelson review electrochemical

The next three chapters deal with electrochemical aspects of

is well established, the electrochemical viewpoint of this article provides

In Chapter 3, R.T. Talasek reviews the anodic passivation of Il-VI

In Chapter 4, R.D. Rauh introduces the relatively new subject of

In Chapter 5, Micha Tomkiewicz reviews photoelectrochemical methods

August, 1991 John McHardy

2 . CHEMICAL ETCHING: PRINCIPLES AND APPLICATIONS . . 53

3 . ELECTROCHEMICAL PASSIVATION OF (Hg.Cd)Te . . . . .. 127

1.1 Types of Infrared Detectors . . . . . . . . . . . . . 127

5 . PHOTOELECTROCHEMICAL CHARACTERIZATION . . . . . 217

6. ELECTROCHEMICAL MIGRATION . , , . . , , . . . . . . , , , , , , 255

Ag, Pb, and Cu Films . . , , , . . . . . . . . . . 276

4.4 Mechanism and Time-to-Failure Results . . .. 287

7. ELECTROCHEMICAL CAPACITORS , . . , , , , . . . , . . . , . . 297

INDEX ....................................... .... 356

solutions, but organic electrolytes are sometimes used), and 2) high

The electrodeposition of semiconductor compounds, like any other

The change in free energy is given by (4)(5):

Eq. (2) A G = AGO + RTIn(a,/a,,,)

where R is the gas constant, T is the absolute temperature and ai is the

A more complete discussion of activity and activity coefficients can be

Eq. (4) AG = AGO + RTIn( 1 / [ M + " ] )

It can be shown that (5)

where n is the number of moles of electrons involved in the reaction, F is

Eq. (7) bAfa + a B b = bA + aB

Eq. (8b) B-b B + b e- (oxidation at anode)

The cell potential is given by

The reactions described in Eqs. 7 thru 9 are typical of those involved

Eq. (1Oa) M+, + m e- 4 M

Eq. (lob) NOin+(2y-n)e-- N+YO-~

Eq. (1Oc) 0-24 0 . 5 0 2+ 2e

yielding an overall reaction:

Eq. (1Od) 2Mfmt 2NOy-"t (m - n)02-2MN t (y t (m - n)/2) 0,

Eq. (12a) 2M'"' t m0-24 2M t 0.5m0,

Eq. (12b) 2NOy-" 4 2N t (y - n/2) O2 t no'*

with cell potentials given by:

deposition potential is exceeded, as shown in Fig. 1 . In most practical

If the values are available in the literature or if an investigator wishes to

Figure 1. Theoretical current vs. voltage (I-V) plot showing deposition

Figure 2. I-V (voltammogram) plot of 1 m/o K,SiF, in Flinak at 75OoC

where M is the molecular weight of the material and c is the deposition

Eq. (18) v = (MEq) / (pnF) = Ay

where p is the density of the material, A is the area of deposition and y is

3.0 ELEMENTAL SEMICONDUCTORS

The most important elemental semiconductor materialfor industrial

Introduction. The first attempts to produce silicon electrolytically

cathode. The anode in thiscellwas graphiteand the measureddeposition

Figure 3. Cell for silicon production in which the electrolyte is contained

Figure 4. Furnace used for electrodeposition of silicon and silicon carbide.

Figure5. Silicon lump (1.6 gm) produced by electrodepositionabove the

commercial applications. The silicon produced by this method had a

grade silicon. In the same vein, Sharma and Mukherjee (42)investigated

16 Electrochemistry of Semiconductors and Electronics

4.0 COMPOUND SEMICONDUCTORS

Compound semiconductors can be chosen to match their

4.1 Il-VI Compounds

Aqueous Solvents. The interest in the electrodeposition of Il-VI

Figure 6. Linear sweep voltammogram. Polycrystalline cadmium electrode

porous or polycrystalline. The process ended when the layer began to

Eq. (20) v, = -0.403 t -

at the CdTe/Cd phase boundary and by:

Eq. (21) V, = 0.143 t 0.295 aCd2+

at the CdTe/Te phase boundary. AV is the overpotential and is given by

where q is the discharge overpotential, i the current density and R the