Electrochemical and Optical Techniques

for the Study and Monitoring

of Metallic Corrosion
NATO ASI Series
Advanced Science Institutes Series

A Series presenting the results of activities sponsored by the NA TO Science Committee,

which aims at the dissemination of advanced scientific and technological knowledge,
with a view to strengthening links between scientific communities.

The Series is published by an international board of publishers in conjunction with the

NATO Scientific Affairs Division

A Life Sciences Plenum Publishing Corporation

B Physics London and New York

C Mathematical Kluwer Academic Publishers

and Physical Sciences Dordrecht, Boston and London
D Behavioural and Social Sciences
E Applied Sciences

F Computer and Systems Sciences Springer-Verlag

G Ecological Sciences Berlin, Heidelberg, New York, London,
H Cell Biology Paris and Tokyo
I Global Environmental Change

NATO·PCa-DATA BASE

The electronic index to the NATO ASI Series provides full bibliographical references
(with keywords and/or abstracts) to more than 30000 contributions from international
scientists published in all sections of the NATO ASI Series.
Access to the NATO-PCO-DATA BASE is possible in two ways:
- via online FILE 128 (NATO-PCO-DATA BASE) hosted by ESRIN,
Via Galileo Galilei, 1-00044 Frascati, Italy.
- via CD-ROM "NATO-PCO-DATA BASE" with user-friendly retrieval software in
English, French and German (© WTV GmbH and DATAWARE Technologies Inc.
1989).

The CD-ROM can be ordered through any member of the Board of Publishers or
through NATO-PCO, Overijse, Belgium.

Series E: Applied Sciences - Vol. 203

Electrochemical and
Optical Techniques
for the Study and Monitoring
of Metallic Corrosion
edited by

Maria G. S. Ferreira
Department of Chemical Engineering,
Technical University of Lisbon, Portugal
and

Carlos A. Melendres
Materials Science/Chemical Technology Divsion,
Argonne National Laboratory,
Illinois, U.S.A.

..
Springer Science+Business Media, B.V.
Proceedings of the NATO Advanced Study Institute on
Electrochemical and Optical Techniques for the Study and Monitoring of
Metallic Corrosion
Vienna do Castelo, Portugal
9-21 July 1989

Library of Congress Cataloging-in-Publication Data

NArD Advanced Stuey InstItute on Electrochemlcal and Optlca1
Technlques for the Study and Monltorlng of Metal!1!: COrrQS10n (1989
Vlana do Castelo, Portuga!)
Electrochenlcal and optIca! teChnlques far the study and
lIonltol"lng of metal1,c corraSlon proceedlngs of the NATD Advanced
Study Institute on Electrochelilcai and Optlca1 TeCnnTQUeS for the
Study and Monltorlng of Metal11c Corrosl0n, Vuna do Castelo.
Partugal • .July 9-21, 1989 I edlted by Mărie O.S. Ferrelra and Carlas
A. Melendres.
p. cm. -- <NATD ASI serles. Serles E. Applied Sela"CeS ; voI.
203)
"PublJshed In cooperat Ion wlth NATD SClentlf,C Affa,I"S Dlvlslon.
Inc 1udes 1ndex.
ISBN 978-94-010-5614-4 ISBN 978-94-011-3636-5 (eBook)
DOI 10.1007/978-94-011-3636-5
, _ Corros 10n and ant1-c:orros lves--Test tng--Congresses.
2. Elec:trolyt1c: c:orroslon--Testlng--Congresses. 1. Ferreu'a, Mârl0
G. S. II. Melendres. C. (Carlos). 1939- III. Nortn AtlantlC:
Treaty Organ1zatl0n. SClentlfic Affalrs D1111510n. IV. Title.
V. Serles: NAro ASI serleS. Serles E. Appl1ea SClences ; no. 203.
TA462,N34 1989
620,1'623--dC20 91-612

ISBN 978-94-010-5614-4

Printed on acid-free paper

AII Rights Reserved

© 1991 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 1991
Softcover reprint of the hardcover 1st edition 1991
No part of the material protected by this copyright notice may be reproduced or
utilized in any ferm er by any means, electronic or mechanical, including photo-
copying, recording or by any information storage and retrieval system, without written
permission from the copyright owner.
 CONTENTS
Preface IX

Organization xi
List of Participants xiii
List of Contributors xix

Marcel Pourbaix
Thermodynamics and Corrosion 1

D.D. Macdonald
Electrochemical Impedance Spectroscopy 31

M.M. Lohrengel
Principles and Applications of Pulse Techniques
to Corrosion Study 69

R.A. Cottis and S. Turgoose

Linear Electrochemical Techniques - Are
They All The Same? 123

C. Gabrielli, F. Huet and M. Keddam

Investigation of Metallic Corrosion by
Electrochemical Noise Techniques 135

M. da Cunha Bela
Photoelectrochemistry: Theoretical Basis 191

M. da Cunha Belo
Photoelectrochemistry: Application to the
Study of Passive Films 219

R.E. Hummel
Differential Reflectometry and Other optical
Techniques in Corrosion Research 241
vi

Z. Szklarska-Smialowska and R. Krishnakumar

Ellipsometry in Studies of Metallic
Corrosion and Oxidation 285

C.A. Melendres
Laser Raman Spectroscopy: Principles and
Applications to Corrosion Studies 355

T.M. Devine
Use of Surface-Enhanced Raman Spectroscopy
in Studies of Electrode-Electrolyte
Interfaces 389

T.M. Devine
Infrared Spectroscopy at Electrode-Electrolyte
Interfaces 439

M.G.S. Ferreira and A.M.P. Simoes

Passivation and Localized Corrosion 485

F. Mansfeld
Inhibitors and Coatings 521

Z. Szklarska-Smialowska
Methods for Studying Corrosion
Inhibition 545

F. Mansfeld
Atmospheric Corrosion 571

R.C. Newman
Stress-Corrosion Cracking and Corrosion Fatigue:
Application of Electrochemical Techniques 585
 vii

D.D. Macdonald
Material Corrosion in High Temperature Aqueous
Solutions in Power Generating Systems 611

M.A. Barbosa and R.A. Silva

Developments in the Study of Biomaterials
Degradation 651

R.G. Kelly and R.C. Newman

Advanced Materials: Corrosion and
Electrochemistry 665

Index 697
 PREFACE
In spite of considerable efforts over the years to
understand and combat materials degradation via corrosion
processes, many challenges still remain both in the
theoretical understanding of the phenomena and in seeking
pratical solutions to the perennial problem. Progress has
been slow due to the complexity of the processes and the
systems involved. Fortunately, in recent years there has been
a renaissance in the development of new electrochemical and
optical techniques, as well as advances in instrumentation,
which have greatly aided our quest to gain insight into the
complex mechanisms involved in metallic corrosion and
passivation. Numerous scientific meetings, symposia, and
workshops have been held allover the world which attest to
the frenzy of activities in corrosion science and technology.
However, most of these conferences have dealt mainly with
recent research results. There appeared to be a need to
assess and disseminate our present state of knowledge in the
field as regards measurement techniques, theory, and
instrumentation. The present NATO Advanced Study Institute
was therefore held in Viana do Castelo, Portugal from July 9
to 21, 1989. The Institute consisted of a series of tutorial
lectures, poster sessions, and round-table discussions
interspersed evenly over the two-week period. It was attended
by 75 participants from several countries representing
industry, government and university laboratories.
In the present volume, most of the lectures given at the
Institute have been collected. They cover the principles and
application of electrochemical and optical techniques to
different. aspects of metallic corrosion and passivation as
well as corrosion monitoring and protection. A purpose of
this book is to help those engaged in corrosion studies to be
more aware of the techniques presently available to
facilitate their work.

ix
x

We thank all the lecturers who contributed to this book

even those who have been impeded and too passivated to be
unable to prepare their manuscripts for printing. We deeply
appreciate the time they spent in preparing their lectures
and after-hours discussions with the participants. We also
thank all the participants for coming and making this ASI a
success.
Finally we thank the Scientific Affairs Division of NATO
(Belgium) , the portuguese councils, Junta Nacional de
Investiga9ao Cientifica e Tecnologica (JNICT) and Insti tuto
Nacional de Investiga9ao Cientifica (INIC), Funda9ao Calouste
Gulbenkian (Portugal), Haynes International (USA), Argonne
National Laboratory and the U. S. Department of Energy, as
well as the local authorities at Viana do Castelo for their
financial and logistic support.

Mario G.S. Ferreira

Carlos A. Melendres
ORGANIZATION

Director:
Mario G,S, Ferreira
Department of Chemical Engineering
Instituto Superior T~cnico

1096 Lisboa Codex, Portugal

Co-Director:
carlos A, Melendres
Materials Science/Chemical Technology Division
Argonne National Laboratory
Argonne, IL 60439, U,S,A,

Organizing Committee:
M, da Cunha Belo, CNRS, Vitry-Sur-Seine, France
],L, Dawson, UMIST, Manchester, U,K,
M,G,S, Ferreira, Instituto Superior T~cnico, Lisbon, Portugal
C. Gabrielli, CNRS, Paris, France
D,D, MacDonald, SRI, Menlo Park, U,S,A,
C,A, Melendres, Argonne National Laboratory, Argonne, U,S,A,
],W, schultze, University of Dusseldorf, Dusseldorf, Germany
Z, Szklarska-Smialowska, Ohio State Univ" Columbus, U,S,A,
G, Trabanelli, University of Ferrara, Ferrara, Italy

xi
 LIST OF PARTICIPANTS
ACHA, MIGUEL Instituto Eduardo Torroja de la
Construccion y del Cemento, C/Serrano
Galvache sin, 28033 Madrid, Espana

AKID, ROBERT University of Sheffield, Department of

Mechanical & Process Engineering, Mappin
St., Sheffield Sl 3JD, U.K.

ALMEIDA, ELISABETE L.N.E.T.I./S.C.P.M., Azinhaga dos

Lameiros a Estrada Pa90 do Lumiar, 1699
Lisboa Codex, Portugal

ALVES, HELENA Instituto Superior Tecnico, Departamento

de Engenharia Quimica, Laborat6rio de
Electroquimica, Av. Rovisco Pais, 1096
Lisboa Codex, Portugal
ARNVIG, POUL-ERIK The Danish Corrosion Centre, Park Alle
345, DK-2065 Brondly, Denmark

BARBOSA, MARIO Department of Metallurgy, Faculty of

Engineering, University of Oporto,
Rua dos Bragas, 4099 Porto Codex,
Portugal

BADE, KLAUS Institut fUr Physikalishe Chemie Und

Elektrochemie II, Heinrich-Heine-
Universitat, Universitatsstr. 1, D-4000
DUsseldorf 1, F.R.G.

BEREKET, GOZEN Faculty of Arts and Science, Chemistry

Department, Univerity of Anatolia,
Eskisehir, Turkey

BOISSEL, J. Souriau et Cie., BP 50, 72402 Ferte-

Bernard, France

BRETHERTON, NEIL Corrosion and Protection Centre, UMIST,

P.O. BOX 88, Manchester, M60 1QD, U.K.
BROOKES, HUGH Chemistry Department, University of
Natal, Durban 4001, South Africa
BUI, NAM Laboratoire de Metallurgie Physique,
E.N.S. Chimie, 118 Route de Narbonne,
31077 Toulouse, France

xiii
CAHAN, BORIS Case Western Reserve University,
Cleveland, Ohio, U.S.A.

CARVALHAL, VERONICA E.D.P. - Electricidade de.Portugal, Av.

Mouzinho da Silveira, 10, 1200 Lisboa
Codex, Portugal

CAVALCANTI, EDUARDO I.N.T., Corrosion Programme, Av.

Venezuela, 82/614, 20081 Rio de Janeiro,
RJ, Brasil

CETIN, INCI Yildiz University, Faculty of Science and

Art, Chemistry Department, 80270 Sisli,
Istambul, Turkey

COLINO, JOAQUIM J. Oficinas Gerais de Material Aeronautico,

Gabinete de Engenharia da 2g Divisao,
2615 Alverca, Portugal

COTTIS, R.A. Corrosion and Protection Centre, UMIST,

P.O. BOX 88, Manchester, M60 1QD, U.K.

DA CUNHA BELO, M. CNRS - CEMS, 15 rue Georges Urbain,

94400 Vitry-Sur-Seine, France

DARBY Jr., J.B. ER-131, G236, GTN, U.S. Department of

Energy, Washington/DC 20545, U.S.A.

DAWSON, JOHN L. Corrosion and Protection Centre

Industrial Services (CAPCIS), UMIST,
Barinbridge House, Manchester M60 lQD,
U.K.

DEVINE, THOMAS M. Department of Materials Science and

Mineral Engineering, University of
California, Berkeley, CA 94720,
U.S.A.

EBLING, DIRK Institut flir Physikalishe Chemie lind

Electrochemie, Universitats St. 1, 4000
Dlisseldorf, F.R.G.

FELIU, SEBASTIAN Jr. CENIM (CSIC), Av. Gregorio del Arno 8,

Ciudade Universitaria , Madrid 28040,
Espana

FERNANDES, JOAO Instituto Superior Tecnico, Departamento

de Engenharia Quimica, Laboratorio de
Electroquimica, Av. Rovisco Pais, 1096
Lisboa Codex, Portugal
 xv

FERREIRA, MARIO Instituto Superior Tecnico, Departamento

de Engenharia Quimica, Laborat6rio de
Electroquimica, Av. Rovisco Pais, 1096
Lisboa Codex, Portugal

FONSECA, ANT6NIO PETROGAL - Refinaria de Cabo Ruivo, Av.

Infante D. Henrique, 1900 Lisboa Codex
Portugal

FONSECA, INES Departamento de Quimica, Faculdade de

Ciencias, Edificio C1, Cidade
Universitaria, Lisboa, Portugal

GABRIELLI, C. LP 15 du CNRS, Physiques des Liguides et

Electrochimie, Laboratoire de
L'Universite Pierre et Marie Curie, Tour
22, 4 Place Jussieu, 75257 Paris Cedex,
France

GAGO, ISABEL M.M. Instituto Superior Tecnico, Departamento

de Engenharia Quimica, Laborat6rio de
Electroquimica, Av. Rovisco Pais, 1096
Lisboa Codex, Portugal

GARCIA-TELLAECHE, PILAR INASMET, Barrio de Igara sin Camino

de Portueche, 20009 San Sebastian,
Espana

GEENEN, F.H. Division of Corrosion Technology and

Electrochemistry, Lab. for Materials
Science and Technology, Delft University
of Technology, P.O. BOX 5052, 2600 GA
Delft, The Netherlands

GURBUZ, RIZA Metallurgical Engineering Department,

Middle East Technical University, 06531
Ankara, Turkey

HUMMEL, R.E. Department of Materials Science

Engineering, University of Florida,
Gainesville, Florida, Flo. 32611, U.S.A.

IMER, FILIZ Yildiz University, Department of

Chemistry, Faculty of Science and Art,
80270 Sisli, Istambul, Turkey

IZQUIERDO, P.M. Escuela Tecnica Superior de Ingenieros

Industriales, Apartado 62, 36200 Vigo,
Espana

JAFAR, M.I. CAPCIS/UMIST, P.O.BOX 88, Manchester,

M60 1QD, U.K.
JOHNSON, C.S. Department of Chemistry, Northwestern
University, 2145 Sheridan Road, Evanston,
IL 60208, U.S.A.

JONES, R.D. University of Wales, Cardif, U.K.

JUSTO, M.JULIA LNETI/DTM, Az. dos Lameiros a Estrada

Pa90 do Lumiar, 1699 Lisboa Codex,
Portugal

KING, FRASER Atomic Energy of Canada, Whiteshell

Nuclear Research Establishment, Pinawa,
Manitoba, Canada ROE lLO, Canada

KISSEL, JACQUES CEBELCOR, Av. Paul Herger, Grille 2,

1050 Bruxelles, Belgique

KIRIAKIDIS, N. Agricultural College of Athens, Athens,

Greece (absent)

KRISHANAKUMAR, R. Fontana Corrosion Center, Department of

Materials Science, The Ohio State
University, Columbus, Ohio 43210, U.S.A.

LAWSON, KEVIN Corrosion and Protection Centre, UMIST,

Sacville St., Manchester M60 1QD, U.K.

LEITAO, EUGENIA LNETI/PORTO, Rua d~ Salazares, 842, 4100

Porto, Portugal

LOHRENGEL, MANUEL Institut fUr Physikalische Chemie,

Universitat DUsseldorf, Universitatsstr.
1, GebaUde 26.32, 4000 DUsseldorf, F.R.G.

MACDONALD, DIGBY D. SRI International, 333 Ravenswood Avenue,

Menlo Park, CA 94025, U.S.A.

MACDONALD, MYRNA SRI International, 333 Ravenswood Avenue,

Menlo Park, CA 94025, U.S.A.

MANSFELD, FLORIAN Corrosion and Environment Effects

Laboratory, Department of Materials
Science, University of Southern
California, Los Angeles, CA 90089-0241,
U.S.A.

MELENDRES, CARLOS Materials Science and Chemical Technology

Divisions, Argonne National Laboratory,
Argonne, Illinois 60439, U.S.A.

MORI, ERIK Materials Science and Chemical Technology

Division, Argonne National Laboratory,
Argonne, Illinois 60439, U.S.A.
 xvii

NEWMAN, ROGER C. Corrosion and Protection Centre, UMIST,

P.O. BOX 88, Manchester M60 1QD, U.K.

PENA, U.G. Escuela Tecnica Superior de Ingenieros

Industriales, Dept. Chemical Engineering
Apartado 62, 36200 Vigo, Espana

PEREIRA, DULCINEA LNETI, Az. dos Lameiros a Estrada do Pa90

do Lumiar, 1699 Lisboa Codex, Portugal

PONTINHA, MANUELA ISEL - Instituto Superior de Engenharia

de Lisboa, Laboratorio de Corrosao, Rua
Conselheiro Emidio Navarro, 1800 Lisboa
codex, Portugal

POURBAIX, MARCEL Belgium Centre for Corrosion Study,

CEBELCOR, Av. Paul Heger, Grille 2, 1050
Brussels, Belgium

RANGEL, CARMEN M. LNETI/DTM, Az. dos Lameiros a Estrada do

Pa90 do Lumiar, 22, 1699 Lisboa Codex,
Portugal

ROBERGE, REAL IREQ - Institut de Recherche d'Hydro-

Quebec, 1800 Montee Ste-Julie, Varennes
(Quebec) JOL 2PO, Canada

SALTA, MANUELA R. LNEC - Laboratorio Nacional de Engenharia

Civil, Av do Brasil, 101, 1799 Lisboa
Codex, Portugal

SARA~, A.S. Technical University of Istambul,

Department of Chemistry, ITU Fer-Eded.
Fak. Kimya Bolunu, Maslak 80626,
Istambul, Turkey

SCANTLEBURY, DAVID Corrosion and Protection Centre UMIST,

P.O. BOX 88, Manchester M60 1QD, U.K.

SILVA, FERNANDO Faculdade de Ciencias da Universidade do

Porto, Porto, Portugal

SILVA, TERESA O.M. Instituto Superior Tecnico, Laboratorio

de Electroquimica, Av. Rovisco Pais, 1096
Lisboa Codex, Portugal

SIMOES, ALDA M.P. Departamento de Engenharia Quimica,

Instituto Superior Tecnico, 1096
Lisboa Codex, Portugal

SIMPLICIO, H. LNETI/DTM, Az. dos Lameiros a Estrada do

Pa90 do Lumiar, 1699 Lisboa Codex,
Portugal
xviii

SOUSA, SUSANA M. FEUP, Departamento de Metalurgia, Rua

dos Bragas, 4099 Porto Codex, Portugal

STATHEROPOULOS, M. Laboratory of Inorganic and Anal.

Chemistry, Department of Chemical
Engineering, National Technical Uni.
of Athens, 42 Patissiov Str., 10682,
Athens, Greece

SZKLARSKA-SMIALOWSKA, Z. Fontana Corrosion Center, Department

of Materials Science, The Ohio State
University, Columbus, Ohio, U.S.A.

TASSINARI, MICHELA Centro Studi Corrosione "A. Dacco",

Chemistry Department, University of
Ferrara, Via L. Borsari, 46/44100
Ferrara, Italy

TRUJILLO, M. CNRS-CEMS, 15 rue Georges Urbain,94400

Vitry-Sur-Seine, France

TRAVASSOS, MARIA LNETI, Az. dos Lameiros a Estrada Pa90 do

Lumiar, 1699 Lisboa Codex, Portugal

WATKINS, KEN G. Department of Materials Science and Eng.,

The University of Liverpool, P.O. BOX 147
Liverpool, L69 3BX, U.K.

WEBSTER, BARBARA Corrosion and Protection Centre, UMIST,

P.O. BOX 88, Manchester M60 1QD, U.K.

WEST, ALAN Department of Chemical Engineering,

University of California, Berkeley, CA
94709, U.S.A.

ZDUNEK, ALAN Illinois Institute of Technology, 10 W.

33 rd St., hall room 105, Perlstein,
Chicago,
List of Contributors
BARBOSA, M.A. Department of Metallurgy, Faculty of
Engineering, University of Oporto,
Rua dos Bragas, 4099 Porto Codex,
Portugal

COTTIS, R.A. Corrosion and Protection Centre,

UMIST, Manchester M60 1QD, U.K.

DA CUNHA BELO, M. CNRS - CECM, 15 rue Georges Urbain,

94400 Vitry-Sur-Seine, France

DEVINE, T.M. Department of Materials Science and

Mineral Engineering, University of
California, Berkeley, CA 94720,
U.S.A.

FERREIRA, M.G.S. Departamento de Engenharia Quimica,

Instituto Superior Tecnico, 1096
Lisboa Codex, Portugal

GABRIELLI, C. LP 15 du CNRS, Physiques des

Liquides et Electrochimie,
Laboratoire de L'Universite Pierre
et Marie Curie, Tour 22, 4 Place
Jussieu, 75257 Paris Cedex 05,
France

HUET, F LP 15 du CNRS, Physiques des

Liquides et Electrochimie,
Laboratoire de L'Universite Pierre
et Marie Curie, Tour 22, 4 Place
Jussieu, 75257 Paris Cedex 05,
France

HUMMEL, R.E. Department of Materials Science

Engineering, University of Florida,
GAINESVILLE, Florida, Flo. 32611,
U.S.A.

KEDDAM, M. LP 15 du CNRS, Physiques des

Liquides et Electrochimie,
Laboratoire de L'Universite Pierre
et Marie Curie, Tour 22, 4 Place
Jussieu, 75257 Paris Cedex 05,
France

xix
xx

KELLY, R.G. Corrosion and Protection Centre,

UMIST, P.O. BOX 88, Sackville St.,
Manchester M60 lQD, U.K.

KRISHANAKUMAR, R. Fontana Corrosion Center, Department

of Materials Science, The Ohio State
University, Columbus, Ohio 43210,
U.S.A.

LOHRENGEL, M.M. Institut flir Physikalische Chemie

und Elektrochemie, Heinrich-Heine
Universitat DUsseldorf,
Universitatsstr. 1, D-4000
DUsseldorf 1, Germany

MACDONALD, D.D. SRI International, 333 Ravenswood

Avenue, Menlo Park, CA 94025, U.S.A.

MANSFELD, F. Corrosion and Environmental Effects

Laboratory, Department of Materials
Science, University of Southern
California, Los Angeles, CA 90089 -
0241, U.S.A.

MELENDRES, C.A. Materials Science and Chemical

Technology Divisions, Argonne
National Laboratory, Argonne,
Illinois 60439, U.S.A.

NEWMAN, R.C. Corrosion and Protection Centre,

UMIST, P.O. BOX 88, Manchester M60
lQD, U.K.

POURBAIX, M. Belgium Centre for Corrosion Study,

CEBELCOR, Av. Paul Heger, Grille 2,
1050 Brussels, Belgium

SILVA, R.A. Department of Metallurgy, Faculty of

Engineering, University of Oporto,
Rua dos Bragas, 4099 Porto Codex,
Portugal

SIMOES, A.M.P. Departamento de Engenharia Quimica,

Instituto Superior Tecnico, 1096
Lisboa Codex, Portugal

SZKLARSKA-SMIALOWSKA, Z. Fontana Corrosion Center, Department

of Materials Science, The Ohio State
University, Columbus, Ohio, U.S.A.

TURGOOSE, M.A. Corrosion and Protection Centre,

UMIST, Manchester M60 lQD, U.K.
THERMODYNAMICS AND CORROSION

f1i\RCEL POURBAIX
Belgium Centre for Corrosion Study
CEBELCOR
~v. Paul Heger, Grille 2
1050 Brussels
Belgium

ABSTRACT. After having paid tribute to the pionner ',-lorks

of Fritz Haber, Walther Nernst, Ulick Evans and Carl
\~agner , the author presents graphical methods in
elect u)chcmical thermodynamics - electrode potential/pH
and electrode potential/temperature diagrams and in
electrochemical kinetics (polarization curves in stirred
solution). Examples of results obtained by these methods
are given concerning corrosion and some other fields
related to electrochemistry.

1. Introduction

As said in 1948 by Ulick Evans (1), who was the father of

the scientific study of metallic corrosion in the presence
of aqueous solutions, the researchers who have whished to
apply chemical thermodynamics to the study of these
phenomena (which are essentially electrochemical, and not
chemical) felt themselves on "quicksands": the basis of
their studies was real, but insufficient, and the
researcher was sucked down without being able to progress
and, despite the classical works of Walther Nernst in
Germany and G.N. Lewis in the U.S.A., one has for many
years believed that thermodynamics could only be of little

M. G. S. Ferreira and C. A. Melendres (eds.).

Electrochemical and Optical Techniquesfor the Study and Monitoring of Metallic Corrosion. 1-30.
© 1991 KluwerAcademic Publishers.
2

help in this area. It has been said that the often complex
phenomena which occur in metallic 'corrosion do not follow
the principles of thermodynamics. But, if
thermodynamicians are sometimes wrong, thermodynamics
itself is never wrong.
Although the first applications of electrochemical
thermodynamics may be found in the works of Fritz Haber
(who mentioned insistently already in 1898 [2) and in 1899
[3] the importance of the concept of "electrode
potential") and of Walther Nernst [4] who published
shortly later his famous "scale of dissolution tensions",
one had to wait a long time before the value of this
concept could be fully appreciated: this appreciation was
thwarted by the existence of two opposite conventions
concerning the sign of these electrode tensions (or
potentials), which was, for the non "noble" metals, as
zinc for example, sometimes considered as negative ("east
c:oast convention" in the U.S.A.) and sometimes considered
as positive ("west coast convention"), and this until
1953, when the "Stockholm convention" of IUPAC
(International Union of Pure and Applied Chemistry)
prescribed the sign "minus" for the electrode potential of
zinc [5]. In the meantime Georges Chaudron once said to
me: "Electrochemistry is not science. I do not know two
electrochemists who agree".

2. The Affinity. The First and Second Principles of

Thermodynamics

According to concepts introduced by Theophile De

Donder (6), the affinity A of a chemical reaction I:"M 0
may be expressed by the relation A = I:vl-L (where the l-L are
the "molar chemical potentials" of the reacting species
M). The 1st principle of thermodynamics prescribes that
the equilibriwn condition of such a reaction, which
 3

corresponds to an affinity equal to zero, may be expressed

by the relation l: v U O. The 2nd principle of
thermodynamics, expressed by the De Donder's relation
AxV ~ 0 (which its author familiarly called "Ave Maria")
prescribes that the speed of a chemical reaction is
always, either zero, or of the same sign of its affinity.
I have been concerned with the extrapolation of
these concepts of De Donder to electrochemical reactions
l:vM + ne0 involving, besides chemical species M, free
=
negative electrons e . This has led me to consider that,
according to the 1st principle of thermodynamics, there
exists, for such'reactions, an "electrical equilibrium
potential" EO measured vs a given reversible reference
electrode (for instance the standard hydrogen
electrode SHE) whose value is given by the relation
EO (l: v 1l/23060n) V (SHE), where the molar chemical
potentials II are expressed in calories.
The affinity A of an electrochemical reaction may be
expressed, in V, by the relation A = (E - EO) where
(E-EO) is the difference between the "reaction potential"
E and the "equilibrium potential" EO of the reaction,
i.e., the "overpotential" (or "overvoltage") under which
the reaction occurs. The 2nd principle of thermodynamics,
expressed by the reaction (E - EO)xi~ 0 prescribes that
the velocity of an electrochemical reaction (measured by
the intensity of the "reaction current" i) considered as
positive in case of an oxidation and as negative in case
of a reduction, is always, either zero, or of the same
sign as its overpotential:
- if E > EO, i ~ 0, i.e., if the reaction potential
E is greater (more positive) than the equilibrium
potential EO of a reaction, this reaction can occur only
,In h
t e d' ' f
lrectlon 0 'd ' F
an OXl atlon, e.g. e _ F e 2+ +
2e -, or (H2) - 2H+ + 2e -. A reduction is impossible.
- if E < EO, i ~ 0, i.e., if the reaction potential
is lower (less positive, or more negative) than the
equilibrium potential EO of a reaction, this reaction can
4

occur only in the direction of a reduction, e.g. Fe - Fe 2 +

+ 2e ,or (H2)-2H+ + 2e-. An oxidation is impossible.
When around 1936-1939, Theophile De Donder became
aware of these formulae EO = (~v ~/23060n) and (E - EO)
xi ? 0 and knew that I had made some experiments which
proved that these formulae were correct and useful, he
jumped for joy (for, despite his 64 years, he had kept the
enthusiasm of a child) and he exclaimed "The good Lord is
the simplest being in the universe! ... " and I had the
surprise of receiving, from the society "La Chimie
Mathematique" of which he was the director, a cheque for a
thousand francs.
In conclusion, if one wishes to study phenomena
involving electrochemical reactions, it is rational:
- on the one hand, to determine, by calculation, the
equilibrium conditions of all possible chemical and
electrochemical reactions. This may be easily done with
the formulae given in Table 1 if one knows the values of
the standard chemical potentials ~ of all reacting
chemical species.
- on the other hand to determine experimentally, for
everyone of the electrochemical reactions r of pratical
interest, the influence of the "reaction potential" E on
the direction and intensity of the "reaction current" i r .
This may be done by the determination of "polarization
curves", of type E = f(i r ) under conditions where the
reaction potential E and the equilibrium potentials EO,r
are perfectely known. This generally implies that these
experiments are performed in stirred solution and that the
electrode potentials be measured vs a well known
reeversible reference electrode, using a junction-siphon
of type Haber-Luggin or Piontelli emerging in the solution
at immediate vicinity of the metallic surface on which the
reactions to be studied are taking plece (Figure 1).
When these two conditons are fulfilled, the
polarization curves show, in fact, the relationship which
exists between the velocities ir of the reactions rand
 5

their affinities (E EO,r), which is a perfectly

Cartesian way to study a phenomenon. According to the
method of representation used at the bottom of Table 2,
all polarization curves are rising from left to right and
cut the ordinate axis at the point where the electrode
potential is equal to the equilibrium potential EO,r.

Jet../

t--
-I--'.so/ufian
H2 0, 0" II.

'\ ft: (OH}2

r. J 01

"\
Fr,O,
11,02
, '

Figure 1. Electrode potential measurement.

In the frequent case where several electrochemical

reactions may occur simultaneously on a same interface
metal/solution, the "principle of superposition of partial
tension-current curves" established in 1938 by Carl Wagner
and Wilhelm Traud [7) may powerfully lead to the perfect
elucidation of the studied phenomena.
Due to the frequent complexity of the
electrochemical phenomena of corrosion (and others) in
which take part many substances involved in many
reactions, it is often useful to make use of graphical
methods. They represent the whole of the thermodynamic
and/or kinetic data relating to these problems (for
instance equilibrium diagrams and polarization curves set
up as a function of an electrode potential). The study of
these problems then is similar to a picture-game which is
6

particularly simple for the players who are used to it.

TABLE 1. Calculation of the equilibrium conditions of

reactions as a function of the standard chemical
potentials ~o (Calories)
----------------------------------------------------------
Chenzica/ reactions

L vM = 0

LVI' = 0 where I' = I'" + RTln (M)

= 1<" + 4.5756Tlog (M) (calories per mole)

LVI''' + 4.5756TL l' log (M) = ()

L V log (M) = log K

'\' (J

where VI' (equilibrium constant)

log K = - 4~756T

for t = 2S.0"C (T=298.1 K): Lvlog(M) = logK

'\' Villi
where lonK= -----
" 136~T

Electrochemical reactions

Eo = E "0 + O.OOOI984T,)
... l' Iog (M) (equilihrium potential)
II

t.:: = 2306011
\.' 0
where ~[V(SHE)l (standard equilihrium potential)

for t = 2S.0°C (T = 298.1 K): E" = E:: + 0.0591 :s v log (M)

II

\' fl
where E"=~
" 2306011
 7

TABLE 2. Affinity and Direction of Chemical and

Electrochemical Reactions: Polarization Curves.

ChetlllC(l/I"CllCliol!.\ Electrochemical rcucliuns

ReaClirm ~l'iH =- () 2: 1'1\1 + tiC - = ()

Fe + 2H+ = Fe" + (Hol Fe = Fe=:!+ + 2c- (c)

2H+ + 2e- = (H,) (a)

Affil1ity A..::: -:2: )',Il (calories) ,4=11(£-£,,) (V)

Equilibrium L 1'.11 = ()
Direerion of [he reller/on

(£-£,,)i30

if E> Eo, i ~ ()

oxidation or- no reaction

FC-----40 Fc:::!+ + 2c- (c)
(H,)--> 2H+ + 2c- (a)
Oxidation of gaseous
hydrogen reduction is impossible
Evolution of gaseous
hydrogen if E< Eo,; 0::; 0
reduction or no reaction

Fe +--- Fc~+ + 2c- (c)

Electrodeposltion (H,) ~2H+ + 2c- (a)
of iron
Dissolution 01 iron oxidation is impossible

Reaction currents

Reductions o Oxidations

Fe: Fez~+ 2e- (c)

IH),' 2M' + 2.- (0)

3. Diagrams of electochemical equilibria

3.1. The potential-pH diagrams at 25°C, and their

applications to the protection against corrosion [8-11)

A powerful support from the UNION MINIERE, from the

IRSIA (Institut de Recherches Scientifiques dans
 8

L'Industrie & l'Agriculture) and from U.S. Air Force led

to the setting-up and to the publication by Robert
Maxwell, in French and English, of an "Atlas of
Electrochemical Equilibria in Aqueous Solutions" [9J which
includes "Potential-pH Equilibrium Diagrams" for all
chemical elements in the presence of water, at the
temperature of 25°C.
Here are some of these diagrams:
Figure 2(al, which relates to pure water, shows that
water is thermodynamically stable under a pressure of 1
atmosphere only in a well determined "area", below which
it may be reduced with evolution of hydrogen, and above
which it may be oxidized with evolution of oxygen - Figure
2(b) shows the acid, alkaline, oxidizing and reducing
media.

-2 0 2 + 6 e 10 12 14 16 -2 0 2 + 6 e 10 12 14 16
2 '2 2
E{V) degagement d oxygene milieux milieux
et ac,dd,caf,on 1,6 oxydenlS ef aCldes oxydants et alcalms
1,6 1.6
1,2
-®---""'--- 1,2
0,8
- --- 0,8

---
.......

0,4

0
-0-...
--------
- 0,4
- 0,8
-0,4
-0,8
---- --- -- -0,4
-0,8
milieux mdieux
degagement dnydrogene
-1,2 et alcabmsatlon -1,2 redLideLirs el aodes redLicteLirs et alcalms -1,2
-1,6 -1,6 -1,6
-2 a z ... 6 8 10 12 14 16 -2 0 2 ... 6 8 10 12 14 16 pH

Figure 2. (a) Domain of thermodynamic stability of water

under 1 atmosphere pressure; (b) Acid, alkaline, oxidizing
and reducing media.

Figure 3, related to iron in the presence of water

shows notably, besides the lines (al and (b) relating to
 9

water, the circumstances where thermodynamically stable

metallic iron Fe, its solid oxides Fe304 (magnetite) and
Fe203 (haematite), and its dissolved derivates Fe 2 +
3
(ferrous), Fe + (ferric) and Fe02H- (bihypoferrite) exist;
from which may be easily deduced Figure 4 which shows,
according to some hypotheses, the theoretical areas of
iron corrosion (where the metal may be dissolved), of
immunity (where the metal is thermodynamically stable, and
thus uncorrodible) and of passivation (where the metal may
be covered by a more or less protective oxide film.

o 1 2 3 4 5 5 7 8 9 10 11 12 13 14 ,5 15
~ ~.2
1a
1,6

'.4
-$~'@, 1.2.

I -~
: FeOH1'''' ,~ ..._
$ '<":',;:- 0.8

--
-- - .. ~'--'5'
&
\::)-~-
'... ' ... 0,6
0.4 "
......... Ee{OH1! ......... --_ 0.4
0.2 \~\ ""J"e 2 0 • - -. 0.2
"-"-" -0ir>---,-
o
-0.2
'®-_-..f, ••
-- -,' "'@-'- -0.<
-0.4 -0.4
-0.6~~~~ -0.6
-0.8 __--''''"''''-:-:0 -0.8
-1 -1
-1.2 Fe -1,2

-1,4 -1,4

~ ~

-1'~2~_:-1--;;-0--:-1---::2"'--:3:--40--7-
5 --::-6---':7---:8'--'9:--"', 4 """1""5~16·1,8
0';"'7:'11-:':12'--';13".--;,7
pH

Figure 3. Equilibrium diagrams for the system iron-water,

at 25°C.

Happily this theoretical diagram fits very well into

the experimental facts. It has enable us to explain
strange facts, which were hitherto not understandable: the
absence of iron corrosion in certain solutions of nitric
10

acid (which raise the electrode potential of the metal in

t.he "passivation area" with coverage by an invisible
oxide film); the conditions of success of the famous and
then much debated "Evans experiment" relating to the
"differential aeration of iron" (Figure 5), according to
which local aeration of an iron surface sometimes leads to
corrosion of the non-aerated spots. From a technical point
of view, such figures very much helped to set up perfectly
safe criteria for the corrosion protection of the
thousands of miles of buried pipelines used for the
transport of oil and gas. The search for such a criterion
had been the main reason for the creation, in the U.S.A.,
of the National Association of Corrosion Engineers.

2
-2 a 2 .. 6 B 10 12 14 16
2
E(V)
1,6 1,6

-- -
1,2 -(§).. 1,2

O,B 0,8
0.4

0
-~OrroSi!!'!..
-0,4

-0,8
--- -- ,8
corrosion
-1,2
immunife
-1,6
-2 0 Z 4 6 8 10 12 14 16 pH

Figure 4. Theoretical conditions of corrosion, immunity

and passivation of iron, at 25°C.

Thisrapid success came to us as an agreeable

surprise, more than the first of the potential-pH
diagrams, which related to coper (12), had been set up,
not for a corrosion study, but for a study of catalysis.
And this encouraged us very much to pursue the setting up
 II

li'cbi".1-
_-VJ,,,,,V-=-d .J"oiu.

o
rnA
o
SoIuholl "" ch/o,,,,~ d. ,.faUirJ11'
11,1.,/

Figure 5. Production of a current by differGntial aeration

(Evans experiment).

of such diagrams as well as experimental studies related

to their scientific and technological applications.
Figure 6 [13] which was stablished as a result of
the "Washington Corrosion Dialogue" held in March 1966,
helped the setting-up of methods of quality control of
"weathering steels" such as Cor ten and Mayari-R, as well
as of recommendations intended to avoid disappointment in
the industrial application of such steels.
Alex Pirson, who was the first technical secretary
of CEBELCOR has, with much talent and humor, popularized
this iron diagram (Figure 7).
12

1,6-+-_.-_ _ _ _ _ _ _ _ _ _-+1,6
E('ns ty )
1,2 1,2

0,8

0,4 ,4

o
-0,4

-0,8

-1. -1.2
Irnmunttd:

-1,S -1.6

-2 ~--._-.---r--._--r__.--_r-2
0 2 6 8 10 12 14
pH

Figure 6. Conditions of immunity general corrosion and

perfect passivation of iron in chloride-free solution;
line 1: in the absence of oxygen; line 2: in the presence
of oxygen.

There exists, of every metal, a potential-pH diagram

which is a specific to him as are, for everyone of us, our
fingerprints and the lines in our hands.
Figure 8 [14) gives such theoretical "corrosion-
immunity-passivation diagrams" for 43 metals and
metalloids.
One sees from this figure that:
Gold (no.l), whose area of immunity completely
covers the area of stability of water is the least
corrodible of all metals.
The frequent excellent resistance to corrosion of
aluminium (no.39) is due to the formation of a protective
atmosphere one has, however, to avoid any contact with
substances which are acid (such as sulphuric soots) or
alkaline (such as fresh cements). Aluminium frames should
be kept with the same care as glass windows.
 13

Figure 7. Corrosion, immunity and passivation diagram of

oxide film. In the case of exposure to the action of an
iron at 25°C (by Alex Pirson).

Lead (no.B) will be corroded by most soft waters,

and this may, in case of water distribution through lead
pipes, cause poisoning to the drinkers and death from
"saturnism".
Incidentally, according to Figure 9, which shows the
equilibria of the system Pb-S042+-H20, in a lead batery
14

containing sulphuric acid, the open circuit potentials are

respectively -0.31 V(SHE) for the negative electrode (in
lead) and +1,64 V(SHE) for the positive electrode (in lead
covered with lead peroxide PB02), which corresponds to an
electromotive force of 1.95 V. When using the battery for
starting a car, the representative points of the two
electrodes move closer together: lead of the negative
electrode dissolves by oxidation, and lead peroxide of the
positive electrode dissolves by reduction. After starting
the car, the dynamo induces a reversal of the current with
recharging of the battery: redeposition of metallic lead
with hydrogen evolution on the negative electrode, and
reformation of Pb02 with evolution of oxygen on the
positive electrode. As point 2 of Figure 9 is situated
above the "oxygen line" (b) the Pb02 layer which covers
the positive electrode is thermodynamically unstable when
at zero current, and thus tends to dissolve with evolution
of oxygen and diminution of the acidity according to the
reactions.
+ - ?+
2Pb02 + 8H + 2e - 2Pb- + 4H20 (52)
2H20 - (02) m+ 4H
+ + 4e (b)

This is the reason for the well known phenomenon of

destruction of lead batteries out of service, when the
Pb02 layer of the positive electrode is not regenerated by
recharging on time, before it becomes too late [15).

3.2. The pitting corrosion of copper pipes for cold town

water distribution [16).

In 1961, the "International Copper Research

Association", INCRA, asked CEBELCOR to try to elucidate
some electrochemical aspects of the pitting corrosion
 15

which affected copper pipes for water distribution, and

was then a severe problem in several countries, such as
Belgium, Germany, Netherlands, U.K. and U.S.A., but not in
France.
Figure 10 shows a cross-section through a copper pit
in Brussels drinking water. At the bottom of the pit white
cuprous chloride CuCl and a large quantity of red cuprous
oxide cU20 coexist: the pit is covered by a layer of green
malachite Cuco3, CU(OH)2 (and by a small quantity of
calcium carbonate CaC03). It is obvious that the solution
existing in the cavity does not have the same composition
as the bulk of the water" but is saturated altogether with
CuCl, with CU20 and with metallic copper Cu. Figure 11
shows the cavity after removal of the non-adherent CU20.
Figure 12 is the equilibrium diagram of the ternary
-2
system Cu-CI-H20 for solutions 10 M in Cl ions (355 mg
Cl 1-1), which is approximately the chloride content of
solutions saturated in cuprous chloride. The only point of
this Fig. 12 where the three solid substances Cu, CU20 and
CuCl which coexist in the pit are simultaneously stable is
the crossing point of the three lines no. 12, 51 and 55.
This point, whose coordinates are E = +270 mV(SHE) and pH
3.5 thus represents the conditions of thermodynamic
equilibrium in the pit. As, under these pH conditions, the
corrosion reaction of copper is reversible, it is easy to
understand that any similar pit will grow or not grow
according to the electrode potential of the metal inside
the pit will be higher or lower than +270 mV(SHE). If, for
any reason, the electrode potential inside the pit is
higher than +270 mV(SHE), the copper will corrode there
according to a reaction such as cu~cu2+ + 2e-; if on the
contary this potential is lower than +270 mV(SHE),
copper will deposit there according to a reaction such as
cu 2 + + 2e-~Cu, and the corrosion will be stifled. And
this extremely simple criterion has allowed to avoid any
corrosion of copper in cold drinking water.
16

~",
o~.,,~., "z
0·" 1

, z
a~l.plt

II"1..a._I'4,.

CCImMion 0, ~

"--"'-
~o,CllGor~."...

~-",-­
[===:J I ......."

Figure B. Theoretical corrosion, immunity and passivation

domains of 43 metals and metalloids, ordered according to
their thermodynamic nobili ty.
 17

-- 2----------------<'

t decnarg Pb02.
Pbo,--

PbSO.,

PbO
~Pb02 r 0

-®
~-1
6
, ,
7 8 9 10 \'\ 12 13 IY ,5
pfi

Figure 9. Operating conditions of lead accumulator (with

1M H2S04, pH 0.2).

CuCO)-<:U(OHi 1 (~eeni

-----
/' IooaCu,O (red)

Figure 10. Copper pit in Brussels water. Cross section

showing the presence of red CU20 and cuCl beneath a green
mushroom of malachite.
18

Figure 11. Copper pit in Brussels water. Cross section of

the pit shown at Fig. 10 after removal of the loose
deposit of CU20.

But, why did one not observe in France copper

failures similar to those observed in other countries?
Simply because, in these other countries, soft annealed
copper tubes were generally used. The tubes were annealed
in a reducing atmosphere which led to the superficial
formation of carbon which increased the electrode
potential of the metal above the critical value; this was
not the pratice in France where cold drawn hard tubes, not
annealed, were prefered, which were free from superficial
carbon. This criterion has also prevented any pitting
corrosion of copper and of coper alloys in other cases,
notably in the protection of condensers in copper alloys
by addition of ferrous salts to sea water [17-19]. It has
also allowed under the conditions which sculptures in
gilded bronze resist or do not resist to the corrosive
action of an atmosphere to be elucidated [20,21].
 19

:: "~:.'
1.8
'';'': ~

..

1.4
.(J)
;;
::1: 1 <.-
0.8

-0.4

·0.8 Cu
·1

·1.4

·18 .':-4--0~~::--~4--6----'-O --'2--'-'-"""6

01-<

Figure 12. Potential-pH equilibrium diagrams for the

three-component system Cu-Cl-H20, for solutions containing
355 ppm Cl- (10- 2 g ion 1- 1 ).

3.3. Other applications of potential-pH diagrams.

The potential-pH diagrams which have just been

presented are only a few examples among many, set up by
many searchers, generally in the frame of the
collaboration initiated in 1949 by CITCE, and this, not
only in corrosion, but also in other areas, like the four
last mentioned below which had not been originally
foreseen:

- analytical chemistry Gaston Charlot, France;

- general chemistry Gabriel Valensi, France;
- electrodeposition Thomas P. Hoar, England;
- electrometallurgy Paul Duby, U.S.A.;
- batteries and fuel cells Jean Brenet, France.

In this area of batteries and fuel cells, a synthesis-

20

statement about the thermodynamic and kinetic aspects of

the electrode reactions would be particularly welcome.
Geology: Robert M. Garrels, U.S.A. [22,23J who
pioneered thermodynamic study of sedimentary deposits and
established such diagrams for the Colorado minerals of
uranium and vanadium, and stayed in 1963 at U.L.B., where
he was appointed Dr. Honoris Causa. His decease, which
occured on 8 March 1988, has sadly affected his many
friends.
Flotation: William T. Denholm, Australia [24J who,
pioneer of the electrochemical study of flotation, applied
such diagrams, with great success, to the treatment of
sulfide minerals.
Agriculture: F.N. Ponnamperuma, Philippines [25-27J
who, when visiting me around 1975, told me that these
diagrams (mainly those related to iron and to manganese)
had helped the "Rice Research Institute R.R.I." to double
the yield of rice fields in south-east Asia (optimal
conditions about pH 6.6 and E +0.03 to 0.14 V(SHE».
Ecology: Mitsuo Kurachi [28J, Japan who, searching
for a remedy to very severe damages (illness known as
ital-ital, and contamination of rice fields) provoked in
Kamioka by effluents of the Mitsui Mining and Smelting Cy
containing a too high amount of cadmium, set up, with the
help of such diagrams, a treatment which decreased the
cadmium content of these effluents below 1 ppb.
Moving to medicine and coming back to corrosion,
Fig. 13 shows the behaviour of iron and of stainless
steels in the presence of chloride containing solutions,
somewhat similar to sea water and to the fluids which
impregnate the human body. These diagrams have been set up
for a study about the behaviour of surgical implants [14J.
The hatched areas show the conditions to be avoided, where
these implants may corrode with dissolution of alloying
elements which may be very damaging for the health. The
triangular hatched regions which appear in the lower left
part of the diagrams relating to stainless steels may be
 21

particularly dangerous, for they correspond to localized

corrosions occuring inside interstices which have become
acid (for instance under a screw) and which, as shown by
the experience of F.L. Laque (Fig. 14) may allow a rubber
band to shear a plate of stainless steel.
One of the consequences of the diagrams and
experiences shown in Figs. 13 and 14 is that, in the case
of young persons, who still have a long life expectation,
it is generally advisable to remove the implants after
healing of the fracture.In the case of elderly people,
older than 85 years for example, for whom the life
expectation is obviously shorter, one may spare a second
operation and leave the implant where it is.
Figure 15, relating to titanium, shows that the risk
of corrosion which has just been mentioned for stainless
steels, does not appear for this metal.

3.4. The diagrams of electrochemical equilibria for high

temperatures.

Before closing the chapter relating to equilibrium

diagrams, I should remind that such diagrams, often
extremely useful and promising, have been set up by many
researchers for aqueous solutions at high temperatures,
for molten salts, and for systems involving a gaseous
phase.

3.4.1. Equilibria in the presence of aqueous solutions

[11,29,30].

Long range work undertaken in the U.S.A. by C.M.

Criss and J.W. Cobble [31] as well as by H.C. Helgeson,
D.H. Kirkham and E.L. Schock [32], concerning the
determination of thermodynamic data for substances
dissolved in water at high temperatures, have allowed many
researchers to set up, without or with use of computers,
potential-pH equilibrium diagrams for high temperatures,
22

3.4.2. Equilibria in the presence of molten salts [33).

Diagrams of electrochemical equilibria relating to

molten salts have been set up for carbonates (R. Buvet,
M.D. Ingram, C.J. Janz, M. Pourbaix), for chlorides (C.
Edeleanu, R. Littlewood), for hydroxydes (B. Tremillon\,
for nitrates (A.J. Arvia, A. Conte, M.D. Ingram, S.L.
Marchiano, N. Ramasubramian, R.A. Rapp), for silicates (R.
Littlewood), for sulfates (R.W. Geist, C.O. Park, A.
Rabrnel, R.A. Rapp).

3.4.3. Equilibria in the presence of gaseous phase

[34-36).

Figure 16 shows an example of an equilibrium diagram

relating to the oxygen-iron system drawn as a funtion of
an electrode potential and temperature, set up in the
frame of a very vast project aimed as the achievement of a
series of Atlases of Chemical and Electrochemical
Equilibria in the Presence of a Gaseous Phase for oxides,
hydrides, carbides, chlorides, fluorides, sulphides and
nitrides of all chemical elements, from 0 to 6000 K.
It shoul be well understood that all diagrams of
electrochemical equiibria which have just been mentioned
should be used only with greatest care, mainly because
they are calculated on the basis of thermodynamic data
which may be incomplete or erroneous, particularly when
they relate to aqueous ions at high temperature or to non
stoichiometric solid substances. It is thus essential that
the exactness and practical validity of these diagrams be
verified and specified by experimental researches which
would, if feasible, include the measurement and the
interpretation of values of electrode potentials.
 23

o 2 :5 8 10 12

+1 ·1

o o o

-1 -1
- -1

• b
o 2. c5 10 12 l'i o 2
pH
o 2 6 10 I~ l'i o 2 o 10 12 1'i
e/~c:!mpolish~ti: 110 pittlA9

+,

-1

c d
o 2 10 12. I~ o .2. 10 12 ,'"I
pH ,of!
Figure 13. BXpierimental areas of general corrosion, is
obviously shorter. one may spare a seconde operation
invnunity, pitting, imperfect and perfect passivity of iron
and leave the implant where it is and. steels in the
presence of solutions 10- 1 molar in Figure 15, relating to
titanium, shows that the risk chloride (3.5 9 Cl- 1- 1 ), at
2S o C; (a) iron and unalloyed of corrosion which has just
been mentioned for stainless carbon steels; (b) AISI 410
chromium steel (draft); (e) steels. does not appear for
this metal. 1\151 304 nickel-chrome steel (draft); (d) AI$1
316 molybdenium-nickel-chromiurn steel (draft).
24

1 (

rn rn ill1J
C1'!rOtrllum sreel
Aisl '30
Chromjl",m. f'tlclf:1 .srfcl ChlOmlu,.,·r1dl:.I·Mol.,./ldtnum srtt 1
AISI :30"1 )'IS! 316

Figure 14. The INCO rubber band test for crevice corrosion
(after F. Lague et al. and J.P. De Greve).

0 "
~ 4 6 - lC ,~

,:;. 1~

EH/V

®.
-- -- -- -- .... ·1
-- -- -- --
-- -- -
-- -- --
-- --
=fro/e dio~ -

o 2 4 6 8 10
pH

Figure 15. Areas of general corrosion and passivation of

titanium at 25°C (draft).
 25

O.
---" ·--:-.L

-"'- ... .:.~ .

······ ... :.5
.......
.... -.. :.~-....

~
"
'"
""
""-2
~ol r..
r." •• -6
~~
-,
h .
l!

.". ~ :l'
"- ~-' -1 -I 0

r
1000 000 000 t·,

Figure 16. Potential-temperature equilibrium diagram for

the oxygen-iron system.

The present text is part of a IllOre extent paper delivered at the Int..rnational Symposi\lll on
"Corrosion Science and Engineering" held in Brussels, March 12-15, 1989 to celebrate the 85th Birthday
of Prof. H. Pourhaix and published in references [37-39}.
The editors acknowledge Corrosion Science (Pergamon Press) for permission for publication.
26

REFERENCES

1. Evans, U.R. (1948), An Introduction to Metallic

Corrosion, pp. VII, XX, Arnold, London.

2. Haber, F. (1898), Uber Stufenweise Reduction des

Nitrobenzols mit begrenztem Kathoden-potential, Z.
Elektrochemie, 1, 506.

3. Haber, F. (1900), Z. Phys. Chemie, 32, 193, 270.

4. Nernst, W. (1889), Z. Phys. Chemie, 1, 129.

5. Christiansen, J.A., Poubaix, M. (1954), Convention

concerning the sign of electromotive forces and electrode
potentials, C.R. 18e Conference Union Internationale de
Chimie Pure et Appliquee, Stockholm, 1953.

6. De Donder, Th. (1936), L'Affinite (redaction nouvelle

par Pierre Van Rysselberghe), Gauthier Villars, Paris,
pp.23 and 97.

7. Wagner, C., Traud, W. (1938), Ueber die Deutung von

Korrosionvorgingen durch Ueberlagerung von
Teilstromspannungs Kurven, Z. Elektrochemie, 44, 391.

8.Pourbaix, M. (1975), Le90ns en Corrosion

Electrochimique, 2e Edition, CEBELCOR.

9. Pourbiax, M. et al. (1963) (1966), (a) Atlas

d'Equilibres Electrochimiques a 25°C, Ed. Gauthier-
Villars et CEBELCOR; (b) Atlas of Electrochemical
Equilibria in Aqueous Solutions (translated from the
French by James Franklin). Pergamon Press & CEBELCOR.
Reimpression by NACE & CEBELCOR.
 27

10. Diagrams of Chemical and Electrochemical Equilibria,

Their Setting-up and Applications, Proc. NATO Advanced
Research Workshop, Brussels (2-5 September 1981), Marcel
Pourbaix and Antoinne Pourbaix (eds.), Rapports Techniques
CEBELCOR, 142, 263 (1982).

11. Equilibrium Diagrams-Localized Corrosion, Proc. Int.

Symp. Electrochemical Society, New Orleans, (7-12 October
1984), Robert D. Frankenthal and J. Kruger (eds.);
Electrochem. Soc. Proc., 84-9 (1984).

12. Porubaix, M. (1938), Thermodynamique et corrosion,

Metaux et Corrosion, li, 189-193.

13. Pourbaix, M. and Pourbaix, A. (1968), Les "Aciers

patinables", (weathering steels) et la Corrosion dans la
construction, Rapports Techniques CEBELCOR, 107, RT.147.

14. Pourbiax, M. (1984), Electrochemical Corrosion of

Metallic Biomaterials, Proc. Conference Biointeractions,
84, London (4-6 January 1984); (a) Biomaterials, 2, 122-
134 (1984); (b) Rapports Techniques CEBELCOR, 147, RT.272
(1984).

15. Delahay, P., Pourbaix, M., Van Rysselberghe, P.

(1951), Diagramme Potential-pH du Plomb. Comportement
Electrochimique du Plomb-Accumulateur aplomb. Loc. cit.,
17.1, 15-28.

16. Pourbaix, M. (1976), Some applications of potential-pH

diagrams to the study of localized corrosion, J.
Electrochem. Soc., 123, 23c-36c.

17. Legrand, J. (1982), Note sur la Protection des

Condenseurs en Al1iage cuivreux par addition de sels
ferreux dans l'eau de mer en circulation, Rapport
Techniques CEBELCOR, 143, RT, 264.4.
28

18. Crouzet, I., Pourbaix, M. (1988), A criterion for

protection of copper-alloys against pitting corrosion in
sea water and brackish water, Rapports Techniques
CEBELCOR, 155, RT, 293.

19. Stubbe, J. and Briquet, J. (1988), Assessment of

FeS04 dosing of a heat exchanger, Rapports Techniques

---.
CEBELCOR, 155, RT, 294.

20. Corrosion and Metal Artifacts-A Dialogue between

Conservators and Archeologists and Corrosion Scientists,
(1977), Gaithersburg, Maryland (17-18 March 1976), NBS
Special Publication 479.

21. Pourbaix, M. (1977), Electrochemical corrosion and

reduction, Loc. cit., 30, 1-16.

22. Garrels, R.M. (1960), Mineral equilibria at low

Temperature and Pressure, Harper, New York.

23. Garrels, R.M. (1982), Applications of equilibrium

diagrams to geology, Loc. cit., 20, section 3.3, 247-
258.

24. Denholm, W.T. (1982), Applications and limitations of

equilibrium diagrams in flotation, Loc. cit., 20, section
3.4, 259-874.

25. Ponnamperuma, F.N. (1977), Physicochemical Properties

of Submerged Soils in relation to Fertility, International
Rice Research Institute - IRRI Research Paper Series no.5
(February).

26. Ponnamperuma, F.N., Tianco, E.M. and Luy, T. (1966),

Redox equilibria in flooded soils: I. The iron hydroxide
systems, Soil Science, 103, 375-381.
 29

27. Ponnamperuma, F.N., Loy, 'l'.A. and Tianco, E.M. (1969),

Redox equilibria in flooded soils: II. The iron hydroxide
systems, Soil science, 108, 48-57.

28. Kurachi, M. (1977), Utility of potential-pH

equilibrium diagrams for the treatment of waste water at
Kamioka mine and refinery, Rapports techniques CEBELCOR,
13 2, RT, 2 4 1 .

29. Van Muylder, J. (1982), High temperature potential-pH

diagrams, Loc. cit., 20, section 1.1, 11-58.

30. Chen, C.M., Aral, K. and Theus, G.J. (1983), Computer

calculated Potential-pH Diagrams to 300°C, Vol. I:
Executive Summary, Vol. II: Handbook of Diagrams, Vol.
III: Users Guide for Computer Program POT-pH-TEMP,
Electric Power Research Institute EPRI.NP 3137.

31. Cobble, J.W. (1982), Measurement and estimation of the

thermodynamic properties of aqueous systems at high
temperature, Loc. cit., 20, section 2.1, 119-136.

32. Schock, E.L. and Hegelson, M.C. (1988), Calculation of

the thermodynamic and transport properties of aqueous
species at high pressures and temperatures. Correlation
algorithms for ionic species and equation of state
predictions to 5 kb and 1000°C, Geochim. Cosmochim. Acta,
52, 2009-2036.

33. Ingram, M.D. (1982), Electrochemical equilibrium

diagrams in the presence of molten salts, Loc. cit., 20,
section 1.3, 91-100.

34. Pourbaix, M., Yang, Xi-Zhen, Zhang, He-Ming and Zhang,

Zhong-Cheng (1986), Some oxidation equilibria: towards the
setting-up of atlases of chemical and electrochemical
equilibria in the presence of a gaseous phase, Corros.
30

Sci., 26, 873-917.

35. Pourbaix, M. (1987), Thermochemical diagrams: some

applications to the study of high temperature corrosion,
Mater. Sci. Eng., 87, 303-317.

36. Pourbaix, M. and Zhang, H. (1988), Equilibrium

diagrams of systems O-S, O-Al, O-Cr, O-Fe, O-S-Al, O-S-Cr
and O-S-Fe. Towards a thermodynamic predetermination of
the circumstances of dry corrosion of metals, Werkst.
Korros., 39, 504-511.

37. Pourbaix, M. (1989), Thermodynamique et

Corrosion,
Proc. International Symposium Corrosion Science and
Engineering, Brussels, March 12-15, Rapp, Robert A. ,
Gokcen, Nev. A., Pourbaix, Antoine (Eds.), Rapports
Techniques Cebelcor, 157, RT. 297, 5-34.

38. Pourbaix, M. (1989), Termodinamico e Corrosao

(translation from the French by Luis Roberto Miranda)
Corrosao e Protec9ao - Publ. ABRACO (Associa9ao Brasi1eira
de Corrosao), l, nQS, 9-18.

39. Pourbaix, fo1. (1990), Thermodynamics and Corrosion,

(translation from the French by Richard DE KRIEK), Corros.
Sci., 30, 963-988.
REVIEW OF MECHANISTIC ANALYSIS BY ELECfROCHEMICAL
TIMPEDANCESPECfROSCOPY

Digby D. Macdonald
SRI International
333 Ravenswood Avenue
Menlo Park, CA 94025

ABSTRACT. The use of electrochemical impedance spectroscopy (EIS) for

mechanistic analysis is briefly reviewed with particular emphasis on the
mathematical techniques that may be used to derive theoretical impedance functions
and to extract kinetic data. The advantages of EIS for kinetic and mechanism
analysis include the fact that it is a steady state technique, that it employs small
signal analysis, and that it is capable of probing relaxations over a very wide
frequency range « 1 mHz to > I MHz) using readily available instrumentation.
EIS also has an enormous advantage over classical transient techniques in that the
validity of the data is readily checked using the Kramers-Kronig transforms. The
principal disadvantage of the method is that many workers are discouraged by the
level of mathematics required to properly analyze impedance data.' This has led
many researchers to analyze their data in terms of simple equivalent electrical
circuits, and hence to ignore the great power of EIS for deriving mechanistic and
kinetic information for processes that occur at electrochemical interfaces.

1. Introduction

Electrochemical impedance spectroscopy (EIS) is now well established [1-5] as a

powerful technique for investigating electrochemical and corrosion systems. The
power of EIS lies in the fact that it is essentially a steady-state technique that is
capable of accessing relaxation phenomena whose relaxation times vary over many
orders of magnitude. The steady-state character permits the use of signal averaging
within a single experiment to gain the desired level of precision, and the wide
frequency band width (l()6 - 104 Hz) that is now available using transfer function
analyzers permits a wide range of interfacial processes to be investigated. These
features generally surpass the equivalent performance characteristics for time
domain experimental techniques [4], so that EIS has rapidly developed as one of the
principal methods for investigating interfacial reaction mechanisms.
In this paper, the foundations of EIS are briefly reviewed with particular
emphasis on mechanistic analysis. Although impedance spectroscopy is not a new
technique, at least in its application in corrosion science and electrochemistry, its
theoretical basis has not been extensively treated in standard texts. Accordingly, we
begin this review by discussing, in very general terms, transfer function analysis
31
M. G. S. Ferreira and C. A. Melendres (eds.).
Electrochemical and Optical Techniques for the Study and Monitoring of Metallic Corrosion. 31--68.
© 1991 Kluwer Academic Publishers.
32

before progressing to the much more specific task of illustrating how EIS is
employed in the study of the kinetics and mechanisms of electrochemical processes.

2. Fundamentals
The response of any conductive system to an electrical perturbation of arbitrary
form may be described by a transfer function

H(s) = V(s)II(s) (1)

where s is the Laplace frequency, and yes) and i(s) are the Laplace transforms of
the time-dependent voltage and current, respectively [6]. In terms of the steady-
state sinusoidal frequency domain, the transfer function becomes

HUro) = F {Vet) 1/F{I(t) I = vUro)IIUro) (2)

where F signifies the Fourier transform, and VUro) and IUro) are the sinusoidal
voltage and current. Provided that the system is linear, causality is obeyed, and that
the interface is stable over the time of sampling [6,7], the transfer function may be
identified as an impedance, ZUro). Because they are vector quantities, HUro) and
ZUro) are complex numbers containing both magnitude and phase information.
From a theoretical viewpoint, the impedance (or, more generally, the transfer
function) is one of the most important quantities that can be measured in
electrochemistry and corrosion science. This is because, if it is sampled over an
infinite bandwidth, the transfer function contains all the information that can be
obtained from the system by purely electrical means.
Since the impedance is a complex number, it is commonly written in the
form

ZUro) = Z' - jZ" (3)

where j =...J-I, and Z' and Z" are frequency-dependent real numbers, which are
related to the magnitude of the impedance and the phase by

IZUro)1 = [Z'2 + Z"2] 1/2 (4)

tan <1> = - Z"(Z' (5)

where <1> is the phase angle. Alternatively, the complex impedance may be
expressed as equation (6)

ZUw) = lZlei<i> (6)

The equivalent mathematical forms of the complex impedance given by equations
(3) and (6) lead directly to the two common methods for displaying impedance data;
as -Z" vs Z' (Nyquist or complex plane) and as log IZI and <1> vs log ro (Bode plane)
plots. Each offers specific advantages; the complex plane frequently is more useful
for mechanistic analysis, because the number of relaxations and their mechanistic
implications (e.g., planar diffusion vs pore diffusion) are often more apparent. On
 33

the other hand, the Bode plane employs frequency directly as the independent
variable, so that more precise comparison between experimental and calculated
impedance spectra can be made. In the past, different authors have tended to
emphasize one or the other forms of presenting impedance data; in fact, both are
useful and should be employed together in any comprehensive impedance analysis.
Over the past decade, considerable advances have been made in developing
ElS as a powerful tool for the analysis of reaction mechanisms in electrochemical
systems (see, for example, Refs. 8-14). The power of EIS for this purpose stems
from the wide frequency range (e.g., < 1 mHz to > 1 MHz) that can be probed
using readily available instrumentation and from the simplified mathematical
structure obtained by linearizing inherently nonlinear electrochemical rate equations.
In our opinion, the precision that can be gained using ElS for mechanistic analyses,
compared with other electrochemical techniques (e.g., potentiodynamic
polarization, "dc" polarization) [4], is the most important advantage of impedance
spectroscopy.
3. Electrodissolution of Aluminum
The application of EIS for mechanistic analysis has been described in detail
elsewhere (e.g., Ref. 1 and references therein). Accordingly, only a brief
introduction to the subject is given here and is illustrated by our recent use of
impedance spectroscopy to explore the electrodissolution of aluminum in alkaline
solutions.
Steady state total, partial anodic, and partial cathodic current/voltage curves
for aluminum in 4M KOH solution at 25°C are shown in Figure 1 [14]. The partial
cathodic curve was determined by oxidizing the evolved hydrogen downstream in a
flowing system using a platinized porous nickel collector electrode and the partial
anodic curve was then calculated as the difference between the total and cathodic
currents, as previously described [15]. The anodic curve demonstrates that
aluminum is a passive metal for E < -1.65 V (vs Hg/HgO, 4M KOH), but at more
positive potentials the metal dissolves in the transpassive mode. Furthermore,
hydrogen evolution is the dominant charge transfer process in the passive zone
whereas metal dissolution dominates at potentials more positive than ca - 1.65
V(Hg/HgO). Finally, the current vs. voltage curve in the transpassive zone is quite
linear at potentials more positive than - - 1.3V, and in this regard aluminum
resembles zinc in its electrochemical behavior in concentrated hydroxide solution
[16]. In this latter system, the linear current/voltage characteristics are attributed to
the existence of a surface Zn(OHh (or ZnO) film whose resistance remains
essentially independent of applied voltage. We have proposed a similar explanation
for aluminum, 17 and we note that surface corrosion product films on aluminum in
hydroxide solutions have been detected experimentally. 18-20
Experimental impedance spectra obtained at 30-80 mV increments over the
potential range of - 1.96V to -1.35V (vs Hg/HgO, 4M KOH) at 25°C are shown in
Figures 2 to 4. In general, the impedance spectra display at least two capacitive
relaxations, with that at low frequencies becoming increasingly dominant as the
potential is made more positive. However, a characteristic feature of these spectra
is the appearance of a loop at intermediate frequencies; this loop is also evident in
the spectra for aluminum in 1 moVdm 3 KOH for a single potential of - 1.5 V (vs
Hg/HgO) as recently published by Brown et al. [18]. The loop apparently is
associated with anodic processes that occur at the interface because it becomes more
evident as the potential traverses the transpassive dissolution region. However,
examination of the partial current/voltage curves shown in Figure 1 reveals that,
over much of the the potential range of interest, both the anodic and cathodic
reactions make significant contributions to the charge transfer properties of the
 34

200

100

E 0

!
f-
Z
w
a:
a:
:::> -100
()

-200

Simulation
Experimental Data
-300

-2.0 -1.5 -1.0 -0.5

POTENTIAL (V vs. Hg/HgO)
JA-7S72-39A

Figure 1. Steady state polarization curves for pure aluminum in a 4 M KOH

at 25°C showing the experimental and calculated total currents
(iT) and both anodic (iA) and cathodic (iH) partial currents.
 35

Experimental Simulated
9KH~
§: .
2.5 KHz

0.9 Htl
o~--------~~~---------- OL---------------~~-----
6.10
Z' (0)
E = -1.95 V

Experimental Simulated,__- _

2 KHz 1 KHz

o~----------~~~------ O~----------------~-L-----
Z' (0) 19.30 20.30
Z' (0)
E = -1.88 V

r:
Experimental

:\5
2 KHz
a 1.6 KHz

N 1 Hz

0
Z'(O)
'".
20.70 Z' (0) 20.90

E=-1.74V
Experimental
1.6 KHz
~f:rsSimulated'2
KHz
c:
N N 2 Hz
I

0 OL-----------~----~-----
20.70
Z' (n) Z' (n)
E=1.70V
JA-7S72-99

Figure 2. Experimental and simulated impedance spectra for aluminum in 4 M KOH at 25°C
as a function of applied potential between -1.95 V and -1.70 V versus Hg/HgO.
Note the difference in scales for the experimental and simulated data.
 §on
36

Experimental
2 KHz

~ 2 Hz
I \
N

OL-----------~~~------ o \
18.8 n Z'In) 20.3 n
z'(n)

P\5
E=-1.68V

Experimental Simulated
2 KHz a 2 KHz
::- 2 Hz
N
I

O~--------~~--~-------- OL-----------.:~---~---
20.8n 20.1 n
Z' (n) z'(n)
E=-1.64V

Experimental Simulated

9. 2 KHz 3 KHz
2 Hz
N ..
OL---------~----~------- OL-------~~------~----
21.3 n 20.1 n
Z' (n) Z' (n)

E= -1.60V

Experimental Simulated
2.5 Hz 2 Hz

oL----~~---~~----
oL-_ _ ~L- _____ ~ _ ___

22.6 n 20.7 n
Z'In) z'(n)

E=-1.50V
JA-7S72-100

Figure 3. Experimental and simulated impedance spectra for aluminum in 4 M KOH at 25°C
as a function of applied potential between -1.68 V and -1.50 V versus Hg/HgO.
Note the difference in scales for the experimental and simulated data.
 37

Experimental Simulated
2 Hz
§
I
2 Hz

o L----=--~O-+-------.;2t-2-.3-n-- I 0
~cr=J 20.6 n
Z'(il) Z' (il)

E = -1.46 V

Experimental

:~
2 Hz

OL--~~----~~-----
20.5il 20.9 il
Z' (il) Z'(il)
E=-1.38V

Experimental

~IO
2 Hz

OL--~~---~F-----
22.7il
Z'(il)
E=-1.35V
JA-7S72-101

Figure 4. Experimental and simulated impedance spectra for aluminum in 4 M KOH at 25°C
as a function of applied potential between -1.46 V and -1.35 V versus Hg/HgO.
Note the difference in scales for the experimental and simulated data.
38

interface. Accordingly, an al2ri.wiseparation of the anodic and cathodic partial

reactions for the purpose of interpreting the experimental impedance data is not
possible.
Any model that successfully accounts for the dissolution of aluminum and
its alloys must explain the various features in the steady-state current/potential
curves and the impedance spectra. During the course of this study, several
hypothetical dissolution and corrosion mechanisms (Table 1) were identified and
analyzed [17] to generate theoretical impedance spectra and steady-state polarization
curves. The simplest mechanism that appears to provide an adequate explanation of
the major features of the experimental data is that given as the first entry in Table 1,
where "ss" represents a bare surface site. The progressive increase in coordination
of the AI(OH)n (n < 3) species is envisaged to result in the gradual removal of an
aluminum atom from the surface plane, such that AI(OHh is probably an
independent molecular species. This species then reacts with a hydroxide ion in a
purely chemical reaction to form an aluminate ion and regenerate a bare surface site.
Additional features of this mechanism are:
(1) The anodic (aluminum corrosion) and cathodic (hydrogen
evolution) partial reactions are "coupled" through their
competition for surface sites (ss).

(2) Langmuir adsorption conditions are assumed.

(3) Charge transfer reactions are assumed to obey the Tafel

equation.

(4) Mass transfer effects are assumed to be negligible.

The impedance for this mechanism is readily derived by expanding the rates of the
individual reactions in terms of dependent and independent variables in the manner
apparently first developed by Frumkin [21] and later used extensively by Lorenz
[22] and Armstrong et al. [23-31]. This method is particularly powerful and easy
to apply in the analysis of reaction mechanisms and its application in the present
case is described in detail in Ref (14).
A successful model must reproduce the general features of the
current/potential curve (Figure 1) including: (1) the active/passive transition at
potentials more negative than -1.65 V, (2) transpassive dissolution, but with ohmic
behavior at potentials more positive than -1.3 V, and (3) hydrogen evolution. The
model must also reproduce the observed impedance spectra, including the changes
that occur as the potential traverses from the cathodic to the anodic regions.
Analysis of the mechanism listed in Table 1 yields a set of kinetic
parameters [14] that reproduce the essential features of the steady-state
current/voltage curves (Figure 1) and the impedance spectra (Figures 2 to 4).
However, the simulated steady-state current/voltage curves are only in semi-
quantitative agreement with the experimental data, with the major discrepancy
arising in the partial cathodic current. This can be corrected for by assuming a
Temkin isotherm, but this simply introduces an additional adjustable parameter
without necessarily enhancing our understanding of the processes that occur at the
interface. Furthermore, introduction of the Temkin isotherm greatly complicates the
mathematical analysis by introducing non-linear terms in potential.
As noted above, the first mechanism listed in Table 1 provides a good
simulation of the impedance characteristics of the interface (Figure 2 to 4). The
calculated data not only reproduce the shapes of the impedance spectra in the
 39

TABLE 1. Theoretical models for the corrosion and electrodissolution of

aluminum alloys in alkaline solutions

Mechanism Reaction Scheme

Irreversible coupled • kl
AI (SS) + 01-1- -+ AI (OH)ads + e-
dissolution/hydrogen
ion-atom evolution k2
AI (OH)ads + 01-1- -+ AI (01-1)2. ads + e-
k3
AI (OH)2. ads + OH- -+ AI (OHb. ads + e-
k4 _
AI (OH)3. ads + OH- -+ AI (OH)4
kS
SS + H20 + e- -+ H + OH-
ks
H + H;P + e- -+ H2 + OH- + SS

II Coupled dissolution! • kl
AI (SS) + 01-1- -+ AI (OH) + e-
H-atom recombination k2
AI (OH) + OH- -+ AI (OH)2 + e-
k3
AI (OH)2 + OH- -+ AI (OH)3 + e-
k4 _
AI (OH)3 + OH- -+ AI (OH)4 + SS
ks
SS + H20 + e- -+ H + OH-
ks
H + H -+ H2 + 2 SS

ill Autocatalytic (I)/hydrogen • kl

AI (SS) + 01-1- -+ AI (OH) + e-
lon-atom evolution
k2
AI (OH) + OH- -+ AI (OH)2 + e-
k3
AI (OH)2 + OH- -+ AI (OHb + e-
k4 _
AI (OH)3 + OH- -+ AI (OH)4 + SS
_ kc
AI OH)4 + AI-+ AI (OH) + AI (01-1)3 + e-
ks
SS + HzO + e- -+ H + OH-
ks
H + HzO + e- -+ Hz + OH- + 2 SS
40

Table 1 (Conduded)

Mechanism Reaction Scheme

• k1
IV Autocatalytic (I)IH-atom AI (SS) + OH- -7 AI (OH) + e-
recombination k2
AI (OH) + OH- -7 AI (OH)2 + e-
k3
AI (OH)2 + OH- -7 AI (OH)3 + e-
k4 _
AI (OHl3 + OH- -7 AI (OH)4 + SS
lie
AI OH)4 + AI-7 AI (OH) + AI (OHb + e-
ks
SS + H20 + e- -7 H + OH-
ko
H + H + -7 H2 + 2 SS

k1
V Autocatalytic (TI)IH-atom AI (SS)· + OH- -7 AI (OH) + e-
recombination k2
AI (OH) + OH- -7 AI (OH)2 + e-
k3
AI (OH)2 + OH- -7 AI (OH)3 + e-

k4
AI (OHb + OH- -7 AI (OH)4 + S
- S

AI OH)4 + AI ~ AI (OH) + AI (OHb + e-

ks
SS + H20 + e- -7 H + OH-
ks
H + H -7 H2 + 2 SS

VI Autocatalytic (TI)/hydrogen • k1
AI (SS) + OH- -7 AI (OH) + e-
lon-atom evolution
k2
AI (OH) + OH- -7 AI (OH)2 + e-
k3
AI (OH)2 + OH- -7 AI (OHl3 + e-
k4 _
AI (OHl3 + OH- -7 AI (OH)4 + SS
lie
AI OH)4 + AI-7 AI (OH) + AI (OH)3 + e-
ks
SS + H20 + e- -7 H + OH-

·SS represents a base surface site, i.e., AI (SS). ks

H + H2O-t. e- -7 H2 + OH- + 2 SS
 41

complex plane, including the development of the loop at intennediate frequencies

with increasing voltage, but also accurately reproduce the low frequency intercept
on the real axis and the frequency distributions, as shown by the frequencies
indicated on the figures. The important feature is that the impedance spectra over a
wide range of voltage (-1.96 Y to -1.35 Y vs Hg/HgO) are accurately simulated by
using a single set of kinetic parameters (kf, bi) and the relatively simple Langmuir
isothenn to describe the behaviors of the aOsorbed intennediates.
While only a single set of kinetic parameters was required to accurately
simulate the impedance characteristics of aluminum in 4M KOH at 25°C, it was
necessary to vary the total number of reaction sites on the surface (rT) with
potential in order to achieve the indicated simulation. A potential-dependent
concentration of active surface sites (e.g., kink sites) has been proposed by Lorenz
et al. [32] and Heusler [33] in the case of metals such as silver and iron, but the
physical viability of this hypothesis is highly controversial. In the case of
aluminum in 4M KOH, as studied in this work, the potential-dependence is
attributed to the existence of a corrosion product film on the metal surface. Indeed,
examination of the data given in Ref. (14) shows that the potential dependence of
the required r T values parallels that of the partial anodic current. This is expected,
assuming that the anodic current is determined by the extent of coverage of the
surface by a porous corrosion product film. The existence of the film may be
included in the model by adding a resistance and capacitance in parallel with the
impedance due to reactions (10) to (15). However, it appears that the contribution
of the film admittance (1;ZfIlm) to the total interfacial impedance is negligible and
hence is not included in the reaction model developed in this work.
A second notable feature of the data required for the simulation is the low
values for the Tafel constants. For an elementary charge transfer process, b(=
anFIRT) is 19.5 V-I for a transfer coefficient (a) of 0.5. Thus, the bi values
obtained in this work [14] indicate a values of the order of 0.1 or less. Similarly
low values were obtained by Keddam et al [34] in their analysis of impedance data
for iron in sulfuric acid. While it is frequently assumed that a = 0.5 for an
elementary charge transfer process, theory [35,36] predicts that this parameter can
in fact have any value between 0 and I, depending upon the relative energies of the
initial, transition, and final states. For strongly interacting species, such as Al and
OH, the reaction coordinate is highly assymetric and activated complex theory [36]
predicts that a should be small (i.e., a «0.5). However, insufficient data are
available for either the configuration or energetics of the initial, transition, or final
states to derive .a. priori meaningful values for a in this work. Nevertheless, the
small values for a, and hence for the Tafel constant b, are not inconsistent with
theory, so that the proposed mechanism is considered to be plausible.
Alternatively, the small apparent values for a may reflect strong repulsive
interactions between adsorbed hydroxyl groups, as embodied in the Temkin
isothenn.3 5 In this case, the rate of reaction appears to be less potential-dependent
than predicted by a simple Butler-Yolmerequation.
The kinetic parameters used in the simulations were selected by "trial and
error", which is an enonnously laborious task for mechanisms that contain multiple
steps. However, one of the great advantages of EIS is that as many simultaneous
equations as are necessary to solve for the kinetic parameters may be generated by
measuring the impedance at the appropriate number of frequencies. Thus, with
reference to the example discussed above, the model contains fourteen kinetic
parameters for any given potential [six (kf, bi) pairs, rT, and Cdil so that fourteen
simultaneous equations are required to solve the problem unequivocally. These
equations can be generated by equating the :malytical expression for the impedance
to the measured values at the fourteen frequencies. However, it does not appear to
be possible to reduce the number of frequencies to seven[(Z', Z") x 7 = 14
42

equations] because Z' and Z" are not independent, as shown by the Kramers-
Kronig transfonns [37-44].
To our knowledge, the rigorous mathematical technique outlined above has
not been completely developed for mechanistic analysis, although we are currently
working on this problem. However, a similar (although simpler) procedure has
been devised by 1. Ross Macdonald [45] for deconvolving equivalent electrical
circuits into their constitutive passive components. In our opinion, development of
the necessary non-linear regression analytical techniques for extracting kinetic
parameters from impedance data will represent a most important advance in
electrochemical impedance spectroscopy.
Finally, it is important to note that the mechanism chosen to represent the
electrodissolution of aluminum in 4M KOH may not be unique, and other
mechanisms may explain the impedance data equally well. Indeed, we have
explored twelve possible mechanisms (Table 1 - six with irreversible reactions and
six with reversible reactions), including some that involve autocatalytic steps of the
type

Al + AI(OHh,ads -+ 2A1(OH)ads (7)

but only the frrst set of mechanisms analyzed above provides a good account of the
observed impedance spectra (Table 2). Nevertheless, an unequivocal choice
between closely related mechanisms can only be made by increasing the accuracy of
the impedance data, thereby increasing the amount of infonnation that can be
extracted from the system. This limitation. which is well known from infonnation
theory, also applies to the assignment of electrical equivalent circuits to charge
transfer processes, and claims that "only one circuit fits the data" merely reflect a
lack of thoroughness on the part of the investigator.

3.1. ACTIVE·TO-PASSIVE TRANSITION

The second example presented in this paper illustrates an alternative method for
mechanistic and kinetic analysis in EIS that was employed extensively by Gerisher
and Mehl [46] and by Epelboin et al. [47-53]. The essence of this technique is that
first order differential equations are established for each of the intennediates
involved in the reaction mechanism and the time dependencies of the intennediate
concentrations are expressed as sinusoidal functions in response to the imposed
sinusoidal voltage or current. Details of this technique are illustrated below by
reference to a primitive model for passivation which yields one of the most
interesting phenomena observed in EIS; negative resistance.
Numerous examples of negative resistance have been reported in the
electrochemical literature (see, for example. Keddam et al . [34]), and in all cases
the condition Re[Z(ro)]oHo < 0 is associated with a passivation event in which the
steady-state current decreases with increasing voltage. Various mechanisms that
predict negative resistance involve the electrochemical adsorption of a blocking
intennediate. It is not our purpose to review these mechanisms here. Instead, we
illustrate the origin of negative resistance by reference to the following reaction
mechanism;

kl
M +A- <=> M-A+e-
(8)
k-l
 43

TABLE 20 Nyquist diagrams calculated with the set of parameters shown in

Table I compared with experimental values for aluminum 4M KOH at 25°C

POTENTIAl (>IS SHE)

Ot5Ct!pllOn a1.SO V 0'.64 V 01.70 V -1.88 01.95

ozol~ or I
ort~ .rln .rW-
1.Ioc/'a,...,,1
200 1:
Ifh 200 1: 200 1: 2001: 60 Z

... - u
.zorn .rtrY\ .rtr "\ orV\ .r~

1301: WO 1: 267601: 61107' 0 1: SSl8 0 z:

I.tedtl.nam ID
.r~ or~ .rVY\ .r~ .rbcfP.
20 z: 3.80 z: SO z: 10.0 z: 20 r

..._ _ mlV ozom .r~ .rb orb

13 0 z: 26760 z: 690740 Z' SSl8 0 z'

M~srn\1 orhO orho orv--p orq or~

S11 0 z: 2.;0 z: 270 z: 2.601: 0<220 z:
E>pe..... n. oZo,cC\
1 orb or h or h orb-
200 z: 200 1: 200 Z 0 200 z: 60 Z
44

k2
M 4 M'1+ + ne-
(9)
In deriving the impedance function for this mechanism, we assume that
Langmuir adsorption conditions apply and that the rates of the elementary charge
transfer reactions can be described by Tafel's law. Designating the fractional
coverage of the surface by M-A as e, and the concentration of adsorbed A at
monolayer coverage as J3 (mol/cm2), we write the conservative equation for
adsorption as

(10)

and the current density (Ncm2 ) in the form

1= F[klC(l-e) - k-le + nk2(1-e)] (11)

where C is the concentration of A- in solution at the interface (mol/cm3) and k}, k-l,
and k2 are potential-dependent rate constants for the elementary reactions. At
steady-state, de/dt = 0 so that Equations (10) and (11) yield

(12)

and

(13)

in which we represent the rate constants as

(14)

(15)

(16)

The small amplitude dynamic response may be obtained by noting that, for a
sinusoidal voltage excitation of frequency ro,

BE = L'l.Eei rot (17)

(18)

and linearization of the rate constants yields

(19)
 45

(20)

k2 - k3. (l + a2E) (21)

By taking the total differential of Equation (10) and substitution of Equations (17)
and (18) we derive the ratio IlS/!:ill as

IlS _ k~alC(l-S) + k~lblS • _1_

!:ill - klC + k_l 1 + jOYt (22)

where the relaxation time (1:) is defined by

(23)

Finall y, by taking the total differential of Equation (11) and by defining the faradaic
admittance as

(24)

we obtain

Yf = A - B/(1 + jOYt) (25)

where

(26)

and

(27)

The total admittance is given by the parallel combination of the faradaic

admittance, Yf, and the admittance for the electrical double layer

Y=Yf+j~ (28)

where Cd is the double layer capacitance (F/cm2). Inversion and rationalization of

Equation (28) therefore yields the impedance as

Zf = Z'(ro) - jZ"(ro) (29)

where
46

Z'(ro) = N(J.2 + 13 2) (30)

Z"(ro) = 13/(;\2 + 13 2) (31)

A= [A - B/(l + W't2)] (32)

13 = [roCd + Ol'tB/(l + W't2)] (33)

Thus, Z'(ro) < 0 when B > A(l + ro2't2). Evidently, this condition will occur as ro
~ 0 provided that B > A, which is true for E > O.
Equation (13) predicts that the steady state current passes through a
maximum as the voltage is increased through zero, as shown in Figure 5.
Likewise, Equation (12) predicts that the surface coverage (9) undergoes a
sigmoidal variation with increasing voltage, with 9 = 0.5 coinciding with the
maximum in current. Clearly, for E < 0, l/(dI/dE) is positive, corresponding to
positive resistance, whereas for E > 0, l/(dI/dE) is negative. Accordingly, negative
values for Z'(ro) can occur only for voltages greater than that for the maximum in
current. That this is so, is clearly shown by the theoretical impedance spectra
(Figure 6). For E «0 (e.g., - 0.4 V), the spectrum appears as a single semi-circle
in the first quadrant. As the voltage is made more positive (e.g., E = 0.1 V), a
second relaxation becomes evident which terminates on the real axis at Z'(ro ~ 0) <
O. Ultimately, for very positive voltages (E = 0.5 V), corresponding to nearly
complete coverage of the surface, almost the entire spectrum lies in the second
quadrant. Changing values for the kinetic constants (e.g., the Tafel constants)
alters the shapes of I(V), 9(V), and the impedance spectra, but does not change the
principal observation that negative resistance occurs only for E > 0 V.
Numerical Kramers-Kronig transforms (49) of the impedance spectra
shown in Figure 6 are displayed in Figures 7 to 9. As expected, only the
impedance spectrum for E < 0 transforms correctly, whereas the transforms of
impedance data calculated for E > 0 fail. As noted in the introduction to this paper,
negative resistance is frequently observed in electrochemical systems and one
example of this behavior is shown in Figure 10 for an aluminum alloy (AI-0.2 Ga-
0.1 In-0.1 TI) in 4M KOH at 25°C [44,45]. In fact, the impedance locus in this
case is found to pass through all four quadrants in the complex plane, ultimately
terminating on the real axis at Re[Z(ro)](J)~ > O. At high frequencies, where the
impedance spectrum is confined to the first quadrant, the transforms are obeyed, as
shown by the data in Figure 11 for ro > 10 rad/s [54]. As the frequency is
decreased and the impedance passes through the left side of the complex plane the
transforms begin to fail. The transforms continue to fail as the locus moves back
into the right side of the complex plane. The failure of the K -K impedance relations
to be obeyed in this case cannot be attributed to the presence of the inductance
because it is well known [41] that impedance loci that are confined to the right half
of the complex plane transform satisfactorily regardless of the sign of the imaginary
component provided that the system is linear, stable, and causal. An example of
this behavior is shown in Figure 12, which presents the impedance locus for the
same alloy/electrolyte system as that in Figure 10, but at a slightly more negative
voltage [54]. Although the impedance locus appears to penetrate the left half plane
at low frequencies, the impedance transforms correctly, (see Figure 13) presumably
because the small negative resistance (if actually present) has negligible influence on
the transfer function.
 47

-3

-4

N -5 1.0
E
!l
~

Cl
.Q -6 0.8

-7 0.6

-8 0.4

0.2

-0.4 -0.2 o 0.2 0.4 0.6

E (voltS)
RA-350522-103

Figure 5. Steady-state current and surface coverage versus voltage curves

for the primitive passivation model.
Parameters: n = 2, [AJ = 10.6 mol/em 3, C dl = 20 ~F/em2, kl = 10.4 em/s,
k. 1 =10. 10 mol/em 2 • s, k2 = 10.8 mollem2 • s, a1 = 18 V'l, b 1 = 18 V'l,
a2 = 18 V'l, ~ = 10-8 mol/em 2.
48

Z"/Q cm 2

42000
E = -0.4 V

28000
190.5

210 14000 28000 42000

E = -0.1 V
140
47.9
70

210 70 140 210

0.6

-210 -140 -70 70 140 210

E = -0.4 V

190.5
2
8 n331.1
.
758.6
27.5 m~z 1738 m~z
I
-42000 -28000 -14000 14000 28000 42000
Z'/il cm 2
RA-3S0S43-38

Figure 6. Calculated impedance spectra as a function of applied

voltage for the primitive passivation model.
(Parameters as given in Figure- 5.)
 49

28.8

Imaginary-to Real
Transform

l'o o . ..
~
N

14.4

Real-to-Imaginary
Transform

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

o 6.68
log (ro, racl/s)
RA-350522-105

Figure 7. Kramers-Kronig fransforms of theoretical impedance spectra

for the primitive passivation model.
E _ -0.4 V. Parameters as given in Figure 5 .• impedance
calculated from the K-K fransform, - impedance calculated from
the passivation model.
50

1600

Imaginary-to Real
800 Transform

400
Real-to-Imaginary
Transform
200

o .....
. ......... "... ...
I

-200

-400
-2.20 o 6.68
log (ro, rad/s)
RA-350522-106

Figure 8. Kramers-Kronig fransforms of theoretical impedance spectra

for the primitive passivation model.
E _ 0.1 V. Parameters as given in Figure 5 .• impedance
calculated from the K-K fransform, - impedance calculated from
the passivation model.
 51

137
--..
•. Imaginary-to Real
...
Transform

E
o
. ._.
9- o . . • ••••• 11 . . . . . . . . . . . . . . . .

N ,,/

67

/\
/ \
Real-to-Imaginary
Transform

Eo / \
...... ..........." /.
...... "-~~ ..... ""....
.........."
::: 0
i-:J

o 6.68
log (ro, radls)
RA·350522-107

Figure 9. Kramers-Kronig fransforms of theoretical impedance spectra

for the primitive passivation model.
E = 0.5 V. Parameters as given in Figure 5 .• impedance
calculated from the K-K fransform, - impedance calculated from
the passivation model.
 52

10

N
E 0
()

~-5
N
-10

-15
5 0 5 10 15 20 25

JA-7572-46A

Figure 10. Nyquist diagram for Alloy No. 21 in 4 M KOH,

at 25°C and at E = -1404 mV HglHgO.
(The applied potential is corrected for IR drop.)
 53

80 ~~----~~----~----~----~-----r----~

(a)

Theoretical Values from KK

40
E
..r:
o
/
N
o

- 40

-1 o 2 3 4 5
LOG w (rad/s)

80
(b)
Theoretical Values from KK

40
/ Experimental Values
E
..r:
.£ 0
/
N
I

- 40

-80

-2 -1 o 2 3 4 5
LOG w (rad/s)
JA-7572-47

Figure 11. Comparison of experimental and KK-transformed impedance data

for Alloy 21 in 4 M KOH at 25°C and at E = -1404 mV versus Hg/HgO.
(a) real component (b) imaginary component.
(The applied potential is corrected for rR drop.)
54

20
10
N
E
u
0

~ -10
N
I -20

-30
-40 -30 -20 0 10 20 30 40

JA-7572-44B

Figure 12. Nyquist diagram for Alloy No. 21,4 M KOH, 25°C,
E = -1430 mV Hg/HgO.
(The applied potential is corrected for IR drop.)
 55

80
(a)

40

E
.t=
E- 0
N

- 40

- 80
-1 0 2 3 4 5
LOG w (rad/s)

80
(b)
Theoretical Values from KK

40 /'xperimental Values

/
E
.t= 0
E-
NI
-40

-80 ~~----~----~~~~~----~----~----~
-1 o 2 3 4 5
LOG w (rad/5)
JA-7572-45A

Figure 13. Comparison of experimental and KK-transformed impedance data

for Alloy 21 in 4 M KOH at 25°C and at E = -1437 mV versus Hg/HgO.
(a) real component (b) imaginary component.
(The applied potential is corrected for IR drop.)
56

In a personal communication, Dr. Takenouti of CNRS, Paris, has

suggested that the apparent failure of the K-K relations to be obeyed by systems
exhibiting negative resistance is illusionary because proper account must be taken of
the form of regulation used when measuring the transfer function. Thus,
impedance measurements strictly involve current regulation (V = Z·I) whereas
regulation of the voltage defines the admittance of the system (I = y. V). Because
EIS is normally carried out by imposing an alternating voltage across the interface,
the transfer function that is actually measured is the admittance rather than the
impedance, so that the K -K relations should be applied to the former and not to the
latter. According to Takenouti, those systems that exhibit negative resistance in the
impedance plane do obey the K-K relations when applied to the admittance plane.

3.2. ANALYSIS OF PASSIVE FILMS

The last example of mechanistic analysis considered in this paper concerns the
properties of passive films on metal surfaces under anodic polarization conditions.
This is an extensively researched subject [55-57] because passivity is the
phenomenon that permits the use of reactive metals, such as Fe, Cr, Ni, AI, Ti, Zn,
and Zr, in oxidizing environments. Despite the enormous amount of work that has
been performed over a period of more than one hundred years, we have only a
rudimentary understanding of the mechanism of passive film growth, principally
because of the difficulty of probing surface films as thin as a few nanometers.
Furthermore, most previous studies of passivity have employed dc, steady-state
techniques, which are not capable of probing directly the relaxations that occur as
charged species move across the film.
The use of EIS to investigate the mechanisms (as opposed to the dielectric
properties) of passive films dates back no more than two decades to the work of
Epelboin, Gabrielli, and Keddam [47-53], Armstrong and coworkers,23-31 and
Macdonald et al. [58-60]. However, much of the early work was concerned with
explaining the form of the impedance locus in terms of a reaction model instead
with extracting kinetic data. More recently, however, great advances have been
made in the application of EIS for extracting kinetic data (rate constants, transfer
coefficients) by using non-linear regression analysis. Since this is probably the
ultimate application of EIS in electrochemistry, we explore one example of this type
of analysis below.
Extensive measurements have been made of the electrochemical impedance
characteristics of passive films on Ni in phosphate and borate buffer systems [61].
These studies, which have been carried out over a range of frequencies from 1 mHz
to 10 kHz, indicate that the impedance of the passive film apparently follows that
predicted by a simple equivalent electrical circuit of a parallel RC network in series
with a Warburg impedance. However, Chao, Lin, and Macdonald [58] developed
the Point Defect Model (PDM) for passive films to provide a more fundamental way
of accounting for the observed impedance in terms of reactions at the metal/film and
film/solution interfaces and the transport of vacancies across the fIlm.
The original PDM, as used to account for the impedance characteristics of
passive metals, was based on the following assumptions [62]:

(1) The metal surface is covered by a continuous film when the

potential is more noble than the Flade potential.

(2) The film contains high concentrations of point defects i.e.,

metal vacancies (V~) and oxygen vacancies (V O)'
 57

Film (MOx/2) I Solution

k3 ,
(2) MM H MX+ (aq) + V ~
k-3

I
(3) m k2
H MM + (X)
2" Vb + Xe' (4) Vb + H20
k4
H 00 + 2W
k-2 k-4

.........- - V X ' _ __
M

---Vb ~

JA-M-320583-2SE

Figure 14. Schematic of physico-chemical processes that occur within a

passive film according to the point defect model.
m = metal atom, MM = metal cation site, 00 z oxygen ion in anion
site V~ = cation vacancy, Vb = anion vacancy.
During film growth, cation vacancies are produced at the film/solution
interface, but are consumed at the metallfilm interface. Likewise,
anion vacancies are formed at the metal/film interface, but are
consumed at the film/solution interface. Consequently, the fluxes of
cation vacancies and anion vacancies are in the directions indicated.
58

(3) The electric field strength across the film is buffered by

electronlhole pair generation and hence is independent of the
applied potential and film thickness.

(4) The reactions occurring at the metaVfilm and film/solution

interfaces are always at equilibrium.

In this present analysis (see Ref. (60)), we modify the PDM by relaxing the fourth
assumption by taking into account the rates of reactions occurring at the metaVfilm
and film/solution interfaces involving the formation and annihilation of vacancies.
These reactions are likely to be rate controlling at higher frequencies. ,
The Point Defect Model [58-66] assumes that metal vacancies (VX), and
oxygen vacancies (V··), which are created at the film/solution and mMlVfilm
interfaces, respective!9, are transported across the film by migration and diffusion
as indicated in Figure 14. These species subsequently are annihilated at other
interfaces by reactions involving either the base metal or the solution. In the
absence of redox species in the solution, the current that passes across the film is
carried by the cation and anion vacancies alone. Accordingly,

Iv = Ivl" + IVa
M (34)

If we now increase the applied voltage by a sufficiently small increment .1V, such
that the response of the system is linear, then the Laplace domain faradaic
impedance is given by

Zf= .1y = .1V

[I(v +6V) - Iv] ~I(V +6V) - Iv] (35)
- -
where I and .1 V are Laplace transforms of I and .1 V, respectively, and s is the
Laplace frequency. In the steady-state frequency domain, the complex faradaic
impedance is obtained by substituting s =joo. Accordingly,

Z(jro) = .1V
jo:{ I(v + 6V)(joo) - I(jOO)] (36)

By solving the Nernst-Planck equations for the movement of cation and

oxygen vacancies across the passive film [60], and by incorporating kinetic effects
in the generation and anhilation of vacancies at the metaVfilm and film/solution
interfaces, it is possible to derive the current response to a small step in voltage and
hence to evaluate the impedance according to Equation (36). Furthermore, by
employing a non-linear fitting algorithm, it is possible to derive a set of parameter
values that provide the best fit of the model to the experimental data. In doing so
we found it necessary to employ a potential-dependent double layer capacitance and
an electric field strength (E) that was indeed independent of the applied voltage and
hence the thickness of the barrier film; the latter being in full concordance with the
view that the electrical properties of thin passive films are analagous to those of
highly-doped junctions (e.g., a "tunnel diode" analog).
We found [60] that the expressions for the impedance could be considerably
simplified if realistic values for the diffusion coefficients for the vacancies were
introduced into the equations. We were able to show that the contribution to the
 59

impedance from reactions at the film/solution interface is negligible at frequencies

above - 10- 10 Hz. Under these circumstances, the impedance arises entirely from
the reactions occurring at the metal/film interface and from the movement of
vacancies through the film.
A typical simulation of the impedance data assuming that oxygen vacancies
alone are mobile in the film is shown in Figure 15, using parameter values that are
given in Table 3 [60]. Figure 15 also shows experimental data for the impedance of
passive films on Ni in O.IN sodium phosphate buffer solution (pH = 11) at 0.1 V
(SCE) for which the simulation was made. At high frequencies the fit is good but
at low frequencies the fit deteriorates, principally because the slope of the
experimental data exceeds 1t/4. A significantly better fit of the model to the
experimental data is obtained if it is assumed that only the cation vacancies are
mobile in the film. In this case, the intercept of the impedance locus with the real
axis is determined by the kinetics of injection of cations into the film at the
metal/film interface. At intermediate frequencies, the slope of the impedance plot in
the Nyquist plane is 45°, however this slope then increases rapidly with decreasing
frequency, to behave in a rather similar fashion to that predicted by a series RC
circuit. Figure 16 compares complex impedance data calculated from the model
assuming that cation vacancies are the charge carriers with the experimental data for
nickel in O.IN phosphate buffer solution (pH = 11) at a potential of 0.1 V(SCE).
It will be noted that both the high and low frequency simulated and experimental
data correspond quite well. Figure 17 illustrates the fit to the experimental data for
0.4 V (SCE). The fit is somewhat poorer, but this may be due in part to the
experimental data that display an unusually large change between the two points of
lowest frequency.
In our previous work [58,59,64] we argued that the impedance
characteristics of thin passive films on nickel could be explained by the transport of
oxygen vacancies, assuming that the metaVfilm and film/solution interfaces are in
electrochemical equilibrium. Indeed, this model predicted that the product Ide.a,
where Ide is the passive current density and a is the Warburg coefficient, should be
independent of the applied voltage and pH, provided that the electric field strength
is not a function of these same variables. This prediction was bourne out by
extensive impedance data for single- and polycrystal nickel in phosphate and borate
buffer solutions at ambient temperature [58,59,61,64]. However, the model was
not able to explain the impedance data at high frequencies, where a partially
resolved semi-circle appears in the complex plane, nor was it able to account for the
systematic deviation of the data at very low frequencies from the purely Warburg
behavior predicted by the theory (slope> 1t/4, Figure 15). Although the low
frequency data are subject to significant error, the deviation was sufficiently strong
to warrant developing the more sophisticated theory outlined in this section.
The high frequency behavior appears to be explained equally well by
assuming that anion vacancies (Figure 15) or cation vacancies (Figure 16) are
mobile in the film. This is so because the high frequency data are dominated by the
kinetics of the reactions that occur at the metal/film interface and only weakly reflect
the transport properties of vacancies in the film. On the other hand, the low
frequency data, which are dominated by transport phenomena, strongly indicate that
the movement of cation vacancies through the film provides a better account of the
experimental impedance data. Accordingly, at least in the case of nickel, the
passive film appears to be predominantly a cation conductor. Furthermore, the
simulations reported here assume that the electric field strength within the film is
independent of applied voltage, thereby confirming our previous conclusion based
on the assumption that the mobile charge carriers are the oxygen vacancies.
Most of the parameter values required for a best fit are well within the range
of expectation, e.g., electric field strength of 1()6 V cm- I , thickness 10 A, diffusion
60

TABLE 3. Parameters used to simulate impedance data for passive films on Ni(III)
at pH 9, 10, 11, at 0.1 and 0.5 V (SCE)

Metal Vacancies Oxygen Vacancies

DM/cm2 s- 1 8.3 x 10- 21 Ddem2 s- 1 1 x 10- 21

kO/em s-I 9.6 x 10- 11 kjmol cm-2 s-I 1.3 x 10- 8

1

kO lmol em- 2 s-I LO x 10-8 k~iem s-I 4.5 x 10- 11

-I

biN 0.1 h2/V 0.1

b_IN 0.1 b-2N 0.1

rJV em-I 1 x 1()6 rJV em-I 1 x 10-6

~ (x =0)/mol.cm- 3 6.02 x 10- 12 C~ (x =0)/mol.cm- 3 6.02 x 10-:

Cp/J..LF cm- 2 24(0. IV) Cp/J..LF cm- 2 24(0.1 V)

50(0.5 V) 50(0.5 V)
 61

100000

90000

80000

70000
o
~ 60000
:r:
~
•
N 50000

40000
o
30000 •
20000

10000

0
0 20000 40000 60000 80000 100000
Z'IOHM
RA-350S22-97

Figure 15. Experimental impedance 0 ,and simulated

impedance _ , data for Ni in a phosphate
buffer at pH 11 and 0.1 V (SeE).
(The simulated impedance data are for oxygen vacancies.)
 62

100000

90000

80000

~
70000

60000
.0
:r:
9N 50000

40000

30000

<J
20000
~
10000 ~

20000 40000 60000 80000 100000

Z'/OHM
RA-3S0S22-98

Figure 16. Experimental impedance 0 , and simulated impedance _ ,

data for Ni in a phosphate buffer at pH 11 and 0.1 V (SeE).
(The simulated impedance data are for metal vacancies.)
 63

100000

90000

80000

70000

~
:::t:
60000 0
•
~
N 50000

40000
•
0
30000

20000
•
0

10000

20000 40000 60000 80000 100000

Z'IOHM
RA-350522-99

Figure 17. Experimental impedance 0 , and simulated

impedance _ , data for Ni in a phosphate buffer
at pH 11 and 0.4 V (SeE).
(The simulated impedance data are for metal vacancies.)
64

coefficients of - 10-20 cm2 s-I, and ~l of - 20 Ilf cm-2. The most disconcerting
values are those for bI, b-I,b2, and b-2 (bi = naiF/RT) - the Tafel constants for the
reactions at the metaVfilm interface (Figure 14). For a transfer coefficient of 0.5,
the b values should be 18 V-I, so that the values of b required to obtain a good fit
imply that a is very small. It should be noted however that the potential
dependence of the experimental impedance is very small, and the low value of b
merely reflects that fact. Other authors have also found it necessary to use small
values of b in order to obtain a good fit of simulated impedance data to experimental
values [14,67-68], possibly reflecting highly assymetric reaction coordinates for the
elementary charge transfer processes that occur at the interfaces.

4. Concluding Remarks

The three examples explored in this paper - the electrodissolution of aluminum, the
active-to-passive transition, and the growth of passive films - were used to illustrate
the enormous power of EIS for mechanistic and kinetic analysis. In particular, we
have argued that quantitative testing of proposed mechanisms is possible provided
that the theoretical impedance functions are derived from the models by using
appropriate mathematical techniques. Indeed, it is the simplicity in mathematical
manipulation that makes EIS so attractive when compared with time domain (or
transient) techniques. Unfortunately, most "mechanistic" studies using EIS
reported in the literature fall far short of this ideal and are concerned with merely
reproducing the general form of the experimental impedance locus or with using
only very limited data (e.g., Z'(o) - Z'(oo» for the subsequent calculation of some
other parameter (e.g., the corrosion rate). The use of EIS in this manner takes
advantage of only a small fraction of the information contained in the impedance
function. Accordingly, greater emphasis should be placed on developing
techniques for optimizing the extent of information transfer from the experimental
data to the mechanistic model. A number of non-linear regression algorithms are
available that will fit an object function (e.g., Z'(ro), Z"(ro), IZ(ro)l, or <I>(ro» to the
experimental data to yield multiple kinetic parameters. We have used one such
model (OPIDES) [69] from the University of Utah to extract kinetic parameters
from impedance data with considerable success.
5. Acknowledgements
The financial support of this work by the US Department of Energy/Basic Energy
Sciences through Contract No. DE-FG-3-84ER45164 is gratefully acknowledged.
Thanks are also due to Drs. Romeo Rocha and Mirna Urquidi-Macdonald for
carrying out some of the calculations reported in this paper.

6. References
1. C. Gabrielli, "Identification of Electrochemical Process by Frequency
Response Analysis," Solartron Instrument Group, SOLARTRON
SCHLUMBERGER, Solartron Electronic Group, Farnborough, Hants, UK
(1980).

2. D. D. Macdonald, and M.C.H. McKubre, "Electrochemical Impedance

Teqhniques in Corrosion Science," Electrochemical Corrosion Testing, STP
272, ASTM, Philadelphia, PA, 1981.
 65

3. M.C.H. McKubre, and D. D. Macdonald, "Electronic Instrumentation for

Electrochemical Studies," in A Comprehensive Treatise of Electrochemistry,
J. O'M Bockris, B. E. Conway, and E. Yeager, Eds., Plenum, NY, 1977.
4. D. D. Macdonald, Transient Techniques in Electrochemistry, Plenum Press,
NY, 1977.

5. M. Sluyters-Rehbach and J. H. Sluyters, Electroanalytical Chemistry, Vol.

4., Ed. A. J. Bard, Marcel Dekker, NY (1970).
6. S. Goldman, Transformation Calculus and Electrical Transients, Prentice
Hall, Inc. NY, 1950.

7. J. R. Macdonald, and M. K. Brachman, Ref. Mod. Phys., 28, 393 (1956).

8. C. Deslouis, I. Epelboin, M. Keddam, and J. C. Lestrade, J. Electroanal.
Chern., 28, 57 (1970); J. M. Coueignoux and D. Schumann, J. Electroanal.
Chern., 17, 245 (1968).
9. 1. Epelboin and R. Wiart, J. Electrochem. Soc., ill, 1577 (L1971).

10. M. Keddam, O. R. Mattos, and H. Takenouti, J. Electrochem. Soc., 128,

257, 266y (1981): M. Keddam, J.-F. Lizee, C. Pallotta and H. Takenouti, J.
Electrochem. Soc., 131, 2016 (1984).
11. 1. Epelboin, C. Gabrielli, M. Keddam, J.-c. Lestrade, and H. Takenouti, J.
Electrochem. Soc., 118, 1632 (1972).

12. I. Epelboin, C. Gabrielli, M. Keddam, and H. Takenouti, Electrochim. Acta,

20, 913 (1975).

13. R. D. Armstrong and K. Edmondson, Electrochim. Acta, 18,937 (1973).

14. D. D. Macdonald, S. Real, S. Smedley, and M. Urquidi-Macdonald, J.

Electrochem. Soc., ill, 2410 (1988).

15. S. Real, M. Urquidi-Macdonald, and D. D. Macdonald, J. Electrochem.

Soc., 135, 1633 (1988).

16. M.C.H. McKubre and D. D. Macdonald, J. Electrochem. Soc., 128, 524

(1981).

17. D. D. Macdonald, S. Real, and M. Urquidi-Macdonald, "Development and

Evaluation of Anode Alloys for Aluminum-Air Batteries," Final Report to
Eltech Systems/DOE Subcontract lOO484-MLM, 1987.
18. O. R. Brown and J. S. Whitley, Electrochim. Acta, 32, 545 (1987).
19. R. Greef and c.F.W. Norman, J. Electrochem. Soc., 132,2362 (1985).
20. H. E. Heusler and W. Allgaier, Werkst. Korros, 22,297 (1971).
66

21. A. N. Frumkin and V.I. Me1ik-Gaik:zyan, Dokl. Akad. Nauk SSSR, 77, 855
(1951).

22. W. Lorenz, Z. Elektrochem, 62, 192 (1958).

23. R. D. Armstrong, R. E. Firman, and H. R. Thirsk, Disc. Faraday Soc., 56,

244 (1973).

24. R. D. Armstrong, 1. Electroanal. Chern., 34,387 (1972).

25. R. D. Armstrong and H. R. Thirsk, Electochim. Acta, 17, 171 (1972).

26. R. D. Armstrong, T. Dickinson, and P. M. Willis, J. Electroanal. Chern.,

48,47 (1973).

27. R. D. Armstrong and K. Edmondson, Electrochim. Acta, 18,937 (1973).

28. R. D. Armstrong, M. F. Bell, and R. E. Firman, J. Electroanal. Chern., 48,

150 (1973).

29. R. D. Armstrong and K. Edmondson, ibid, 53,371 (1974).

30. R. D. Armstrong and M. F. Bell, ibid, 55, 201 (1974).

31. R. D. Armstrong and A. A. Metcalfe, ibid, 71, 5 (1976).

32. H. Schweickert, W. J. Lorenz, and H. Friedburg, J. Electrochem. Soc.,

127, 1693 L(1980).

33. K. E. Heusler, "Recent Findings on Mechanism of Several Types of

Corrosion." DECHEMA-Monogr. 198,93 (1914-1931, Elektrochem. Met.-
Gewinnung, Verab. Korros.), 193-211.

34. M. Keddam, O. R. Mattos, and K. Takenouti, J. Electrochem. Soc., 128,

257,266 (1981).

35. B. E. Conway, Theory and Principles of Electrode Processes, Roland Press,

NY 1965.

36. E. J. Calvo, in Comprehensive Chemical Kinetics, Vol 26, Electrode

Kinetics: Principles and Methodology, Chapter 1, Elsevier, NY (1986).

37. H. Kramers, Phyzik. Z., 30, 521 (1929).

38. R. de L. Kronig, J. Opt. Soc. Am., 12, 547 (1926).

39. V. A. Tyagai, and G. Ya. Kolbasov, Elektrokhmiya, 8, 59 (1972).

40. R. L. Van Meirhaeghe, E. C. Dutoit, F. Cardon, and W. P. Gomes,

Electrochim. Acta., 21, 39 (1976).

41. H. W. Bode, Network Analysis and Feedback Amplifier Design, Nostrand,

NY, Ch. 14, 1945).
 67

42. M. M. Jaksic and J. Newman, J. Electrochem. Soc., 133, 1097 (1986).

43. D. D. Macdonald and M. Urquidi-Macdonald, J. Electrochem. Soc., 132,

2316 (1985).

44. M. Urquidi-Macdonald, S. Real, and D. D. Macdonald, J. Electrochem.

Soc., 133, 2018 (1986).

45. J. Ross Macdonald, Impedance Spectroscopy, John Wiley and Sons, NY

(1987).

46. H. Gerisher and W. Mehl, Z. Electrochem., 59, 1049 (1955).

47. I. Epelboin, M. Keddam, and J. C. Lestrade, Disc. Faraday Soc., 56, 264
(1973).

48. I. Epelboin and M. Keddam, J. Electrochem. Soc., 117, 1052 (1971).

49. I. Epelboin and R. Wiart, ibid, 118, 1577 (1971).

50. I. Epelboin and M. Keddam, Electrochim. Acta, 17,177 (1972).

51. I. Epelboin, M. Ksouri, and R. Wiart, J. Electrochem. Soc., 122, 1206

(1975).
52. I. Epelboin, C. Gabrielli, M. Keddam, and H. Takenouti, Electrochim. Acta,
20, 913 (1975).

53. I. Epelboin, M. Ksouri and R. Wiart, J. Electroanal. Chern., 65, 373 (1975).
54. D. D. Macdonald, S. Real, and M. Urquidi-Macdonald, J. Electrochem.
Soc., submitted for publication (1989).

55. J. Kruger and R. P. Frankenthal (Eds), Passivity in Metals, The

Electrochemical Soc., Princeton, NJ 1978.

56. L. Young, Anodic Oxide Films, Academic Press, NY, 1961.

57. A. T. Frornhold, Jr., Theory of Metal Oxidation, Vol. 1 - Fundamentals,

North Holland, Amsterdam, 1976.
58. C.-Y. Chao, L. F. Lin, and D. D. Macdonald, J. Electrochem. Soc., 129,
1874 (1982).

59. D. D. Macdonald, R. Liang, and B. G. Pound, J. Electrochem. Soc., 134,

2981 (1988).

60. S. Smedley and D. D. Macdonald, "An Electrochemical Impedance Study of

the Passive State on Nickel", in Symp. Proc. Transient Techniques in CO"os.
Sci. and Electrochem., Ed. W. H. Smyrl, D. D. Macdonald, and W. J.
Lorenz, The Electrochemical Soc., Pennington, NJ, 254 (1989).
68

61. D. D. Macdonald, B. G. Pound, M. Urquidi, and B. E. Wilde, Final Report

to Office of Naval Research Contract NOOO14-82-k-0265 (1985).

62. c.- Y. Chao, L. F. Lin, and D. D. Macdonald, J. Electrochem. Soc., 128,

1187 (1981).

63. L. F. Lin, c.-y. Chao, and D. D. Macdonald, J. Electrochem. Soc., 129,

1194 (1981).

64. D. D. Macdonald, M. Urquidi, and B. G. Pound, "Point Defect Model for the
Growth and Breakdown of Passive Films on Metal Surfaces," Proc.
Australassian Corros. Conf., Rotorua, NZ, November (1984).

65. M. Urquidi-Macdonald and D. D. Macdonald, J. Electrochem. Soc., 134,4

(1987).

66. D. D. Macdonald and M. Urquidi-Macdonald, Electrochim. Acta., 31, 1079

(1987).

67. M. Keddam, O. R. Mattos, H. Takenouti, J. Electrochem. Soc., B128 ,257,

266 (1981).

68. D. R. MacFarlane, S. I. Smedley, J. Electrochem. Soc., 133, 2240 (1986).

69. OPTDES, a software system for optimal engineering design, Brigham Young
University, 1986.
 PRINCIPLES AND APPLICATIONS OF PULSE TECHNIQUES TO
CORROSION STUDY

M. M. LOHRENGEL
INSTITUT fur Physikalische Chemie und Elektrochemie
Heinrich-Heine Universitiit Dusseldorf
Universitiitsstr. 1
D-4000 DUsseldorf 1
Federal Republic of Germany

ABSTRACT. The kinetics of the formation and corrosion of passive layers on copper, iron,
aluminium and other valve metals were investigated by single and multiple potentiostatic
pulse techniques in a wide range of time and current. Additionally, photocurrent transients
and mass transients are presented.

1 Introduction

Pulses, applied to electrochemical systems, cause pronounced disturbances in these systems.

The system changes its properties to come closer to equilibrium. To use pulse techniques you
have to overcome three main problems: the synthesis of the pulses, the recording of the
resulting transients and the interpretation.

We have to compare pulse experiments with other techniques. The application of pulses
means, that we are far from equilibrium or steady state conditions and the system changes
rapidly to find a new state closer to equilibrium. We have to extend our measurements to
very short times to record all these processes. Therefore many of the common techniques,
like impedance spectroscopy, optical spectroscopy, potentiodynamic techniques, and any
ex-situ investigations are less applicable because the time to process one sample is usually
in the range of seconds or even more. This is a long period compared with the changes of
the system and these methods produce something like mean values only.

But now back to the pulses: The nature of the disturbance may be any physical parameter
like:

.6.£ potencial
.6.i current density
.6.T temperature
.6.p pressure
.6.c concentration
hu radiation
69
M. C. S. Ferreira and C. A. Melendres (eds.),
Electrochemical and Optical Techniques for the Study and Monitoring of Metallic Corrosion, 69-121.
© 1991 Kluwer Academic Publishers.
70

Usually, pulse techniques is synonymous to potentiostatic or galvanostatic pulse measure-

ments. Nowadays, potentiostatic control is more common in electrochemistry and, hence,
current transients of potentiostatic pulses will become the main topic.

2 Transients of Potentiostatic: Pulse Measurements

References: (1),[2),[3)

Current transients of potentiostatic pulse experiments usually show a fast increase of current
to very high values, a fast decrease followed by an asymptotic decay to very small currents.
There are four main problems:

* the requirement of an efficient potentiostat,

* the resolution of the conventionally registered current signal is poor,
* the electric charge cannot be obtained easily,
* and the knowledge to be taken from the experiment seems to be small.

All these problems limited the application of potentiostatic pulses, but can be solved by
modern equipment.

2.1 Potentiostat

The most important tool to investigate potentiostatic pulse transients is, of course, the
potentiostat. We prefer potentiostats based on operational amplifiers with typical parameters:

Bandwidth 5 MHz
Slew rate 10 V/ JjS
Maximum current 3 A
Feedback compensation of electrolyte resistance

which are sufficient for all applications presented here. The potentiostats are in house
development to enable individually tailored systems.

The potentiostatic system consists of:

Main amplifier
Impedance converter to monitor the reference electrode
Current to voltage converter (eg. feedback compensation)
Summing amplifier for synthesis of the reference voltage
I or 2 autoranging instrumentation amplifiers (see section 2.3)
Minor components, like meters, level indicators, and oscillation indicator

There are numerous different concepts of potentiostats. Mainly designed to avoid potential
deviations caused by the electrolyte resistence, pulsed potentiostats were developed. The
potentiostat is switched off electronically from time to time to determine the potential at
i=O. These concepts produce spikes and noise and, therefore, prevent sensitive measurements
like sine wave modulation techniques. In many cases the problem of electrolyte resistance
is reduced by the use of microelectrodes.
 71

Another interesting concept is a potentiostat of complete digital design. It requires a processor

of 2.0 MIPS or more and a convenient power booster. All these problems can be solved,
but there are no systems as sufficient and versatile as the analog concept.

-1
"-
"-
~2 "-
"" -< Ig i ::: A - Ig
N t
J
C
u
~3 "-
«

'c
........
""'-
'-
.J!l -4 "'-
""'-
-5 <0

-6 1=0 1

-5 -4 -3 -2 -1 0 2
Ig t / s

Fig. 1. Schematic current transient in response to a voltage step.

2.2 Current transients

A typical current transient in response to a potential step is given schematically in a double

logarithmic plot (fig. 1). Initially, the electrode is in steady state condition at EO, almost
free of surface layers. A potential step to €p starts layer formation. The current density
increases up to some Acm- 2 within I or 2 jJS due to a charging of the double layer, then,
Ig i decreases in a first approximation linearly with Ig t down to some jJAcm- 2 after mOre
than 100 s. This is the typical region of stationary corrosion currents of passive electrodes.
Every decade in time requires as similar amount of charge. The use of a double logarithmic
scale illustrates the main problem: the registration. The transient has extreme dynamics in
current and time.
72

2.3 Current Detection

The most common method to monitor currents is to measure the voltage drop of a resistor
by an instrumentation amplifier. Amplifications of more than 10 cannot be used without
loss of performance (slew rate, common mode rejection, and signal to noise ratio). Hence,
the circuit works over 3 decades of current only, but we need 5 to 8 decades.

We tested several concepts. Logarithmic amplifiers, usually realized by operational amplifiers

with thermostatted diodes in the feed back circuit, are too slow, inaccurate and work over
3 to 6 decades only. They cannot handle the high currents which occur during the first /-IS.
Another concept is the common switching of ranges (Fig. 2). The resistor can be switched
from I 0 to some MO, which is a sufficient range. But the commutation cannot be done
by switches or relais, because the response times are too long (minimum more than 100 /-Is).
The system has to switch to a convenient range within I I-'S or less. Switching times like
this can be handled by commercial integrated circuits only up to a few 10 mA, which is
not sufficient. We had to develop special electronic switches based on power-MOSFETs.

[> Ua
1

Fig. 2. Measurement of current by switching of resistors. The rotary switch is replaced by

electronic relais in an advanced concept.

A modified concept which avoids switching of high currents is given in fig. 3. All resistors
are connected in series with diodes in parallel. If the voltage drop at one resistor exceeds
the forward voltage of the diode, the signal line of the amplifier is switched to the next
(lower) resistor by a commercial analog multiplexer. Trigger signals for switching come
from voltage comparators at the output of the amplifier. Advantage of this concept is the
'smooth' switching of the diodes in the power line, which reduces the oscillation tendency
of the potentiostat. Drawback is a very complex frequency dependence of the circuit.
 73

C> Ua
I

Fig. 3. Autoranging of an instrumentation amplifier using the forward voltage of diodes.

underflow overflow
~
o
-1

-2 I

~
:::- -3 I /-------------
~ I /
/
/------------
-4 I /
/

, /i--------------
-5 I / / I
I
I r /
I
-6
o 2 4 6 8 10
Ua/V

Fig. 4. Transfer characteristics of the autoranging instrumentation amplifier.

74

Both concepts are sufficient and in use. Fig. 4 shows the logarithm of the input current
vs. output voltage of the instrumentation amplifier. The linear ranges are curved due to
the logarithmic scale. When the output voltage strikes the borders of overflow or underflow,
the amplifier is switched to a more convenient range.

2.4 Time and Charge

The current has to be integrated to get the charge. This cannot be done by an analog
integrator, because the dynamic in time is to high. The current signal has to be digitized
and stored. This must be done with 106 samples/s to record the fast rise of current in the
beginning. 106 samples/s accumulate 109 samples within 1000 s, which are far too much
data to be stored or processed with typical laboratory equipment. We can see from fig. I
that the dynamics of current are reduced by time. Hence, we can reduce the number of
samples with increasing time retaining a sufficient accuracy. Fig. 5 shows the output of an
especially developed non-linear rate generator vs. time. When triggered, the generator starts
with I MHz, reduces to 100 kHz at 10 /!s, to 10 kHz at 100 /!s and so on. The sample size
N is given by the following formula

,
1 MHz

1 kHz

1 Hz

1 mHz
,
1 f1s 1 ms 1s 1000 s
time

Fig. 5. Decrease of the sample rate with time to reduce the number of samples.

(I) N=1+9(lgt+6)

Combined with a transient recorder, which can be triggered externally, a sweep of 1000 s
requires less than 100 samples. These data can be stored and integrated without problems.
 75

nonlinear
digital rate

f
generator

range laboratory
current
transient
recorder - computer
DEC 11/73
amplifier current

potential
CE
I
G WE
r
/"1.\
R;- potentiostat graphic terminal
plotter

Fig. 6. Flow sheet of signals of the measuring equipment.

Fig. 6 shows the measuring equipment. It consists of an electrochemical cell and a fast rise
potentiostat. The multi-channel transient recorder accepts the trigger signals from the rate
generator and collects the data of potential, current, and current range. To guarantee that
the measurements are not disturbed by switching the range, this circuit is strobed when a
sample is taken. The computer controls the manifold trigger signals, processes and plots
the data and simplifies the use of the equipment.

2.5 Example: Passivation of Aluminum

References: (4)

The techniques described above can be applied to almost every system. In some cases a
complete kinetic analysis can be done by data of potentiostatic pulse transients. The
passivation of aluminum is a suitable system to demonstrate the method.

2.5.1 High Field Model

The growth of electronically insulating oxide layers should obey the high field mechanism
(5),(6). Former measurements confirmed this for aluminum oxide at times longer than a
few seconds, but we will see, that there is a more complex behaviour at short times.

The high field theory assumes a thermally activated, field supported hopping mechanism
of the ions. The current density i depends on the potential drop l::..¢ within the layer and
the oxide thickness d:

(2)
76

io and P are constants, which are specific for the system. Hence, i is a function of the field
strength E = !::.~/d only. The time dependence of the thickness is given by a differential
equation,

(3) dd
-= Ai o exp(f3E)
dt

which cannot be solved exactly. An approximate integration yields the well-known inverse
logarithmic law:

(4) 1
-=A'-Blgt
d

In the following, we will use an exact numeric integration rather than eq. (4) .

.....
I
E
u
> 10 10 6
.......
W
limited
by do

o 5 10 15 20 2S
d / nm

Fig 7. Experimentally accessible values of field strength and Al 20 g layer thickness.

2.5.2 Experiments

The measurements were carried out on electropolished aluminum wires in an acetate buffer
(pH=5.9). The initial oxide thickness do of these specimen was about 4.5 nm.
 77

do = 7.7 nm E = 8.83'10 6 V/cm

0
/ do = 10.9 nm E = 8.07,10 6 V/cm
N

~
I
E
u
-I

-2
/
........
-3
~
-4
do = 14.1 nm E = 7.66'10 6 V/cm
-5 do = 17.3 nm E = 7.40'10 6 V/cm

-6 corrosion current

-5 -4 -3 -2 -I o 2
Ig t / S

Fig. 8. Current transients of aluminum oxide formation (full lines). Curves calculated from
eq. (3) are dashed.

4
do= 7.7 nm E = 8.83'10 6 V/cm - - - -

N
I
do = 10.9 nm
E
u
u 3
E
........ do =14.1 nm E = 7.66'10 6 V/cm
o 2 do = 17.3 nm

-5 -4 -3 -2 -1 o 2
Ig t / 5

Fig. 9. Corresponding charge, calculated from the current.

78

In principle there are numberless combinations of thickness and field strength, but
experimentally there is only a small window accessible to the operator. Fig. 7 shows these
limits in an E vs. d plot. The limits for the thickness are determined by the initial thickness
(low value) and by the maximum output voltage of the potentiostat (high value). The lower
limit for the field strength is given by that point where the oxide formation current comes
in the magnitude of the corrosion current of a few J'Acm- 2 •

Fig. 8 shows some current transients: within the first 20 J's the current of the capacitive
charge is measured, decreasing very fast. After a period of time, which varies from 100 J's
to 1 s, a minimum occurs, then the current increases again. Then, i reaches a maximum
the temporary position of which is depending on potential and thickness, and then decreases
linearly with a slope of about -\. After 100 s the value of the corrosion current of a few
J'Acm- 2 is reached and the measured current becomes constant .

. 13 do = 7.7 nm E = 8.83'10 6 V/em

.12
.11
......
I
.1 do = 10.9 nm E == 8.07.10 6 V /em
E
=
......
.09
...... . 08
I
-c do = 14.1 nm E == 7.66,10 6 V/em
.07
.06
do = 11.3 nm E == 7.40'10 6 V/em
.05
-5 -4 -3 -2 -1 o 1
19 t / s

Fig. 10. Reciprocal layer thickness d- 1 vs. Ig t.

Using the charge data from fig. 9, an analysis due to the inverse logarithmic law was done.
The charges were converted to layer thicknesses, the results are given in fig. 10 in a d- 1
vs. 19 t plot. The data show the expected linear regions for longer periods of times. Up to
ms or s , depending on the potential, the thicknesses are constant, because the time is too
short or the current too low to form additional layer. The slopes of the lines show a linear
dependence on the potential (Fig. II)

The oxide formation coefficient .8",44 nm/V corresponds to the values already published
(71.
 79

300

-, 200
::
"-'Q" )~0::
-;:;
,
u
u
100

o+o----~2-----4~--~6----~8~---1~0~
E (HESS) / V

Fig. 11. Reciprocal slopes d 19 t / d d- 1 , as taken from the preceeding fig., vs. potential E.

Everything seemed to be perfect, but computer simulations of corresponding current

transients from eq. (2) and an exact numeric integration of eq. (3) show strong deviations
(Fig. 8). The calculated current density is constant in the beginning, which is reasonable
as the field strength and the layer thickness do not change in short times. Then the current
decreases due to a decrease of E caused by the growing layer. A minimum and the following
overshoot is not observed.

Our first idea to explain this different behavior was to test the validity of the high field
theory. Fig. 12 shows the logarithm of the peak current Ig imax vs. field strength E for
many transients. A straight line as expected from eq. (2) is yielded for many combinations
of d and E:

(5) 19 i max = .4343' f3E + Jg io

The formation coefficient taken from fig. 12 is ,8=44 nm/V. In the following this value is
used.

Deviations are found for high currents only, when the potentiostatic control becomes
insufficient. Hence, the current depends on the field strength only. This means:

• the high field mechanism is valid,

• the potential drops at the phase boundaries are almost constant,
• the determination of the field strength is correct.
80

We have to return to eq. (2). The constant io

(6)

is the product of the charge carrier concentration p. the frequency factor 11, the jump
distance 6, and an activation energy U·. The charge carrier concentration p is of interest
only and thus we can abbreviate eq. (6) to

(7)
io = P. W

d/nm

l-
0 A 4,5 • ...
0 6,1 0

.
c
-1 '1,1
N
I
. 9,3 •
I"..
E
u • 10,9
c:t -2 • 14,1 0
/
.1
...... /
>< D

.E
I!)
-3 /
Ii• ..
I
.:EF
-4 d • P= 44 nm/V
I-
-5
/
5 6 '1 8 9 10 11
E / 106 V cm- 1
Fig. 12. Peak current Ig imax vs. field strength E,

The factor w represents the mobility of the charge carriers. If we assume that the current
is supported by moving ions at interstitial positions, p may become a function of time t
and location x. The lattice is exhausted of mobile ions at negligible field strength and they
must be injected through a phase boundary to the oxide. The oxide must be filled up with
ions until it can grow on. The potential drops at the interfaces seem to be independent of
 81

the potential, hence, the injection process should be close to equilibrium and not rate
limiting. The movement of the ions through the oxide is limiting. The mean velocity v of
these ions depends on the field strength,

(8) i
v=-=wexp(I3E)
p

The time to pass through a layer of the thickness d is

(9) d
t=----
wexp(I3E)

If this transition time t corresponds to the occurrence of the maximum of the measured
transients t max , a conversion of eq. (9)

(10) I 9 t max - I 9 do = - .4343 . 13 E - I9 w

should yield a straight line. Fig. 13 shows the corresponding plot for many transients. Eq.
(10) is fullfilled. For very short times, deviations can be seen caused by the potentiostat.

An improved calculation of current transients has to consider the transit of the ions. We
can separate the current arbitraryly. The first process is the formation of a space charge
of ions until the layer is filled up (t<t max). The concentration p is a function of x and t:

(11 )

The value of f(x,t) is close to I for t>tmax.

The second process is the growth of the layer according to eq. (2) and (3). The concentration
of mobile ions p is almost constant.

The current of the space charge formation is

(12) fwexp(ilE) )

. d
(
~ [ f(x.t)xdx
1 = -:....----'--------:.-
dt
if
(13) d
t<t = 0
max wexp(I3E)

At t=tmax i becomes equal to the current calculated from eq. (2).

82

d/nm
0
\\ " 4.5
o 6.1
o 1.7
.. 9.3
E
c '1 • 10.9
'-..
0
\, • 14.1
'C \.
-2
-.
\
~ '1. ..
I
III
.t.6
•
'-..
)(
IQ

...
E 0
•
••
~ -4 ... 0

.0
• A

".
-6
5 6 7 8 9 10 11
E / 10 6 V cm- 1

Fig. 13. Plot corresponding to eq. (10). Data taken from measurements of microelektrodes
are marked with an asterix.

It is hard to get more details concerning the space charge formation. The slope of the
experimental curves before the maximum is reached (fig. 8) d Ig ild Ig t is about 3. Simple
simulations assuming a constant field strength show slopes of I. Hence, we must suppose
a variable field strength and a nonlinear movement of the ions through the layer. This is
reasonable, because a charge carrier concentration p > 10 20 C/cm- S , which can be estimated
from eq. (7) and (10), should influence the field strength. Eq. (9) is a rough but sufficient
approximation only.

But how to explain the constant field strength when the ion space charge is built up? The
transport numbers of aluminum and oxygen ions from [8) are close to .5 which means, that
there are comparable numbers of positive and negative ions in the oxide and, hence, there
is no excess charge influencing the field strength.

If this is true, the initial phase of oxide formation looks like this: A13+ and 0 2 - ions are
injected from the phase boundaries to the oxide. These ions reduce the local field strength
and, as the potential drop in the oxide is almost constant, increase the field strength between
the ion clouds. When the clouds meet each other, the field strength decreases again and
then becomes constant.
 83

The information concerning details of this process is poor and we preferred an empirical
function to describe this process

(14) 1 1
P=Po 1 + (I max
)3=PO 1 + ( d 0
)3
1 1w exp (~E)

Hence, the fundamental equations (2) and (3) change to

(15)

and

(16)
dd AwPOexp(~E)
dt
1+ ( do
twexp(j3E)
)3

0
-1
N
I
E -2
u bO
«
....... -3 .
c:
'6'0
ta

~ -4 ..
.::
u
IIJ
»
reorganisation

-5 ..!
high field
::c;:,
IIJ

-6 0
~

-5 -4 -3 -2 -1 o 1 2
Ig t / s

Fig. 14. Separation of current transients into 4 regions: Charging of double layer capacity,
reorganisation processes, formation of ion space charge, and growth corresponding to the
high field model.
84

These equations were used in the following section for computations.

2.5.3 Model

The experimental current transients can be divided into 4 regions. First of all the electronic
charging of the double layer takes place, a process which is finished after 10 JjS. The
following region has a slope d Ig i/dt ~ -I and is independent of the potential. The charge
consumed during this region is always less than 10 JjCcm- 2 and can be neglected. The
current may be caused by charging of the interfaces or by a reorganisation of the oxide
when the field strength increases. The third region starts with the re-increase of the current
and ends in the maximum. Als+ ions are injected from the metal into the oxide and 0 2 -
ions from the double layer and move to the respective opposite side. This process is dominated
by the field strength (eq. (9». When the ions meet the opposite interface the oxide starts
growing. The concentration of ions in interstitial positions remains almost constant and is
independent of field strength and time. The current shows a small overshoot which is caused
by the delayed start of oxide formation. In the last region, the slope d Ig i/dt is equal to
-I and the high field mechanism is valid (eq. (4». This region ends when the decreasing
oxide formation current comes into the magnitude of the corrosion current.

1
'\
0 I \
\ do = 4.5 nm
-1 I
I \,
E = 8.5 106 V/cm do = 10·2 nm
"
E = 8.2 10 V/cm
-2 r,/
N
I do = 14'6 nm
E E = 7.8 10 V/cm
v
c:r -3
/'

--
~ -4
-5
do = 9.3 nm
-6 E = 7.4 106 V/cm

-5 -4 -3 -2 -1 o 1 2
Igl/ s

Fig. 15. Experimental current transients and corresponding calculated curves.

Calculated and measured transients fit well (fig. 15), only for t<IO ms the appearance of
the maximum is delayed. This is caused by the limited capability of the potentiostat.
 85

2.6 Mass Transients

A very recent tool for electrochemical investigations is the Electrochemical Quartz Micro
Balance (EQMB). It enables an in-situ mass detection of electrode surface layers with a
resolution of 10 ng/cm 2 which is in the range of monolayers. Hence, it is possible to record
mass transients.

Oscillator

(ill \
(
\
..... /
J

(a) (b)
Working electrode

Fig. 16. Cylindrical quartz substrate coated with 2 metal electrodes (a). Exposure of the
quartz to the electrolyte (b).

The equipment is not complex. Both sides of a quartz substrate are vacuum coated with
the metal of interest (fig. 16a). In some cases the adhesion is a problem. One interface of
the quartz is exposed to the electrolyte and represents the electrochemical electrode (fig.
16b). Under isotonic and thermostatic conditions the freqency of the quartz oscillator is a
function of the mass fixed to the quartz surface:

(17)

The constant was determined to K = 20 ng cm- 2 Hz-I. A 24 bit frequency counter was
realized as a plug-in board for a laboratory computer. The principles of the equipment are
shown in fig. 17.

The response time of the EQMB is relative long. Typically, a 5 MHz quartz is used and
frequency changes are in the range of ppm. This requires counting periods of .5 s or more.
A fast response could be obtained changing from counting to time interval measurements.
86

potentiostat

24 bi t
binary
counter

Fig. 17. Schematic plot of the equipment used to measure mass transients.

10

~
-10
t>
....,.
-20 :::c
N

-30
100
":
E

mfJ£J
u
<{ 50
::1.

0
0 1.0 2.0 3.0 4.0 5.0
E (HESS) IV

Fig. 18. Current density of potentiodynamic oxide formation and frequency response of
the EQMB in a acetate buffer (pH=5.9), sweep rate 20 mY/so
 87

This means, however, to measure a time interval of 200 ns with a precision of ppm, which
is technical almost impossible. A better idea is the combination of counting and time interval
measurement, in this case response times down to I ms can be imagined.

There is only a small number of investigations concerning passivation. One work 19) indeed
refers to the composition of aluminum oxide and concerns the preceding sections.

o o

M..
-10 6ii
--
-·~7··..--. .... - - _
0.2

-- - -
N fUr AI,O, ..... '. - _
::r: C>
3

-
~
M f ·· / ..
urAI,o,·H,O .
-2
1.0
<l
2V
-20 0.4

11Hz

-30 +----.---.------r---,----.---'-0.6
o 0.2 0.4 0.6 O.B 1.0
!J.Q'me

Fig. 19. Mass changes of the oxide layer vs. charge consumption.

Quartzes were vacuum coated with aluminum and potentiodynamically anodized (fig. 18).
The sweep width of every following sweep exceeds that of its predecessor. The corresponding
current is zero up to the preceeding sweep limit and then increases due to further oxide
formation (fig. 18). In the upper part of the fig. the matching frequency response of the
EQMB is plotted. The mass is calculated from eQ. (17) and plotted vs. anodic charge in fig.
19. The anodic sweeps show straight lines according to an oxide thickness which depends
linearly on the potential. In the cathodic sweep no charge is consumed, but the mass shows
further rise. For comparisons the lines calculated for the compositions AI 20 3 and AI 20 3·H20
are given. The correspondence is nice: during the anodic sweep AI 20 3 is formed obviously,
which incorporates water when the field strength is lowered. The process seems to be
reversible.

3 Multiple Pulse Techniques

Usually the information which is taken from single pulse experiments is not sufficient to
develop a model. Sophisticated combinations of pulses may yield much more information.
88

3.1 Potentiostatic/Potentiodynamic Pulse Measurements

Potentiodynamic polarization curves are very instructive for a qualitative overview of oxide
formation. For quantitative evaluation, however, potential and time must be well defined,
and, thus, potentiostatic formation of passive layers is preferable. Many data of interest
can be taken from the corresponding current transients.

Electrochemical investigations of the reduction kinetics is almost impossible for valve metals.
A partial reduction, if possible, is accompanied by strong hydrogen evolution. Easier is a
measurement of the reduction kinetics of more noble metals, like iron, copper, gold, and
platinum, when the oxide layers can be reduced quantitatively with no or negligible hydrogen
evolution in parallel.

Convenient is a double pulse experiment. The oxide is formed in a first pulse and reduced
in a second one. Side reactions, like corrosion or oxygen evolution can be detected by a
comparison of anodic and cathodic charge. The reduction can be done also by a potentiostatic
pulse, but more information is taken from a potentiodynamic reduction. One prerequisite,
of course, is a fast sweep to avoid any reorganisation of the oxide. Sufficient are sweep
rates of .1 Vis to 10 Vis typically.

The critical point of these measurements is the initial sweep potential. It must not be
identical to the polarization potential (p, because further oxide growth could occur during
the first part of the reduction sweep, especially when tp is much smaller than 1 s. It must
be found experimentally, so that the current in the beginning of the sweep is almost zero.

The technique to record current and charge of the anodic polarization pulse has been
described in section 2. The reduction spectra can be recorded and integrated using the
common techniques.

3.1.1 Formation and Corrosion of Copper Oxides in Alkaline Solutions

References: 110),(11),112),113)

Passive films grow on many metals homogeneously, but in some cases a stepwise change
of stoichiometry is discussed, assuming a multilayer structure (14),115). The growth of the
outer oxide depends strongly on the electronic properties and the ionic conductivity of the
inner layer.

The passive film on copper in alkaline solutions represents such a system which is of
interesst for corrosion protection and photoelectrochemical investigations as one of the few
p-type semiconductors (16). A detailed description of the peak structures is given in (17).

3.1.1.1 Overview

Fig. 20 presents a good overview of the oxide formation and reduction on copper at pH '" 13.
Potentiodynamic measurements were done with a rotating double semi ring disk electrode
(fig. 20). Oxide was formed and reduced at the copper disk and corrosion products were
detected at the semicircular Pt rings. One half ring was potentiostatted to .55 V to reduce
Cu 2 +, the other to 1.3 V to oxidize Cu+ ions.
 89

0.6

N 0.4
1
E
u 0.2
c:(
E 0
•I
.........
~
.!::!
.:P -0.2
-0.4

0.20
N
I 0.16
E .........
u 0.12 3
~ 0.08
):0
n
......... 3I
N 0.04 I
be €ring2 = .55 V (Cu 2+ ...,.. Cu+) ~

.:.r:
C
o ~ ! 002
~ €ringl = 1.3 V (Cu+ -;.. Cu 2+)
~~~~~~~===r~~~~0
Q1 0.3 0.5 0.7 0.9 1.1 1.3 1.5
€ (HESS) / V

P'~' Fig. 20. Rotating copper disk electrode with two

platinum semi rings (left). Potentiodynamic polarization
curve measured at the Cu disk (above). The corrosion
products Cu 2+ and Cu+. respectively. were detected at
the two Pt semi rings simultaneously.

The potentiodynamic sweep starts at 0 V with a layer free electrode. The first anodic peak
at .6 V is the formation of CUzO and is accompanied by a slight increase of Cu+ corrosion.
In a second peak at .9 V CU20 is partially transformed to CuO and additional oxide is
formed. A strong formation of soluble Cu 2+ in parallel is observed. The oxide formation
in this region (.9 V to 1.3 V) is strongly influenced by the speed of rotation, at higher
speeds the fomation of Cu-I1-oxides is totally suppressed. This dependence and the strong
corrosion indicate a participation of soluble components. e.g. a corrosion precipitation
mechanism. A long potentiostatic polarization at 1 V yields thick black oxide layers.

The cathodic sweep typically shows two reduction peaks. In the first one at .5 V the
Cu-I1-oxides are reduced to CUzO. which is reduced in a second peak at .2 V. together
with the Cu-I-oxide initially formed. to Cu.
90

The nature of the oxides is quite well confirmed by many investigations 1181. The inner
layer of CU20 is covered with CuO at higher potential, which is transformed to CU(OH)2
at the electrolyte interface. The thick visible layers consist of Cu(OHh·yH 20. In the following
these three layers are referred to as CU20, CuO, and Cu(OHh, respectively, for reasons of
clearness.

The CU20 and CuO layers behave like dielectric media at lower potentials and grow according
to the high field mechanism. A simple capacitor model should describe the system:

(I8) 1 d CU20 d cuo

-=--+--+-= +---+-
C CCU20 Ccuo CH DD cU20 DDcuo CH

Eq. (I8) is fullfilled at lower potentials and reflects the changes of the layer thicknesses d
correctly, but there are strong deviations at higher potentials.

Potential ~
O~~y'--1.6
~O
V
Capacity
measurement
0.4 V --

----oV

Time

Fig. 21. Potential time diagram to prepare the capacity measurements shown in the next
fig.

To explain this and to avoid problems of a changing layer thickness a modified experiment
was done (fig. 21). Copper oxides were formed by a potentiodynamic sweep to 1.6 V (3.4 nm
total thickness) and reduced to a single CU20 layer by a cathodic sweep to .4 V. In a
subsequent sweep to 1.6 V parts of the oxide are converted to CuO, but further oxide
growth is almost negligible. The capacity values for this sweep are given in fig. 22 (full
line).

The capacity is constant up to .7 V. When CU20 is partially oxidized to CuO the capacity
doubles to 10 /.'F cm- 2 and increases to values up to 50 /.'F cm- 2 at 1.6 V. The first step
 91

25r-------------------------~~

O}

20 0,6

0,5
N
I 15
E
1.1
U. 0,4 mn
:.. ,
........ , ........
U I -c
10 ---'"
/ / n~w equivale~
0,3 "
/

circuit
,
, / 0,2
5 -------- ~J

0,1

0r--.---.--.---,--,--~--~_40
0,2 0,4 0,6 0,8 1,0 1.2 1,4 1,6

€ (HESS) / V or U / V
Fig. 22. Capacity of a copper electrode in dependence on the potential (full line) and
corresponding values of the equivalent circuit (dashed line).

could be explained by an extreme high dielectric constant of CuO, but the second increase
is not explained by simple models. This increase is reversible and must reflect changes of
the semiconducting properties of the oxides.

The vanishing of the oxide capacitors in series (eq. (18» indicates that the voltage drops
in the oxides are limited to constant values at higher potentials and, hence, the capacitors
cannot be charged any more.

Reasons for a limited potential drop in a semiconducting oxide is an accumulation of holes

at the electrolyte interface, when the valence band of the oxide comes in the region of the
Fermi level. The potential drop in the Helmholtz increases as a result, the potential drop
in the oxide remains almost constant.
92

0,441l F

1N4148 1N4148

10 kQ SOOQ 100Q

Fig. 23. Equivalent circuit diagram using diodes to simulate the limited potential drop
within the oxides.

To assure that this model fits to the experiment, we tried to design an equivalent circuit
with similar properties. A simple electronic component equivalent to an oxide reaching the
inversion region is the diode. At low potentials below the threshold voltage the resistance
is high, but at higher voltages the diode becomes conducting and the potential drop is
limited to the threshold voltage of about .7 V. A corresponding equivalent circuit is shown
in fig. 23.

Each oxide layer is represented by a capacitor, a diode and a resistor. The capacitor represents
the oxide film, taken as a dielectric medium without an appreciable space charge, the diode
the electronic effects of the band structure. The resistor simulates the ionic conductivity.

Fig. 22 compares the simulated curve with the measured capacity. The correspondence is
very good, i.e. the model of limited potential drops in the oxide layers is consistent.

This model explains the structure of the copper oxides. At lower potentials CU20 is formed
which accumulates holes at the electrolyte interface with increasing potential. This is
identical to a partial oxidation to Cu 2+. The increasing potential drop in the Helmholtz
layer enables a transfer of these ions and a formation of a CuO nuclei. A further increase
of potential accumulates holes at the CuO/electrolyte interface and the potential drop in
the Helmholtz layer increases again. This enhances the corrosion of Cu 2+ and forms hydrated
layers of Cu(OHh, depending on the pH value.

3.1.1.2 Pulse Measurements

Fig. 24 illustrates the realization of the pulse measurements. The oxides are formed by a
potentiostatic pulse, which defines formation potential Ep and time tp. The anodic current
transient is recorded as described in section 2.3.
 93

€ t
d€ / dt = 10 V / 5

t=O time

o __ ~ r-------l

time
c
,- ... -
C (t) C(€},'
,
I
I

I
-_ ... \ /
'---'"
OL-----~--------------~
time
q

qcorr
O~L---~~------~--~--~
time

Fig. 24. Schematic plot of potentiostatic oxide formation and potentiodynamic reduction:
potential E, current density i, capacity C, and charge q.

The potentiodynamic reduction spectra are separated into the different peaks and integrated
to get the amount of the different oxides formed.

The capacity is recorded during the anodic polarization and the cathodic reduction to
monitor the dielectric properties and thickness changes of the oxides. The technique to
record capacity transients is described in section 4.2.

Each measurement (characterized by Ep and tp ) yields the following data

94

.06,---------------------,
Ep 0.7 V =
.05

.04

-
I
E
u
.03
<C

.~ .02

.01

O~~~~=-~~~-=-~--=-=--=-~--=--~--~
0.600 0.500 0.400 0.300 0.200 0.100 o -0.100
€(HESS) / V

Fig. 25. Potentiodynamic reduction curves after oxide formation at .7 V for various times
at pH = 9.

ip(t) the anodic current transient,

qa the corresponding total anodic charge
Ca(t) the anodic capacity transient
ic(t) the cathodic reduction spectrum
Ql .. qn the charge of the different oxide species
Cc(t) the capacity of the reduction spectrum

Potentiodynamic reduction curves at .7 V are given in fig. 25 for different formation times
and pH = 9. The total reduction charge grows from 80 J.lC cm- 2 (10 J.ls) to 800 J.lC cm- 2
(1000 s). CU20 only should be formed at .7 V, but we see a complex evolution of peak
structures. The formation starts with one peak at .5 V for short times. When the polarization
time exceeds I s, a second peak at .2 V appears. After 1000 s the first peak has vanished
and the well known CU20 peak at .1 V remains only.

Fig. 26 shows reduction spectra for oxide formation at .6 V and 1.2 V in dependence on
time. The variation of charge from .03 to 2.5 mC cm- 2 requires a standardized size of the
spectra to compare shape and peak potentials. A more precise representation of the peak
potential evolution is given in fig. 27.
 95

€(HESS) / V

-04 0 04 0.8 -0.4 0 0.4 O.B

€p = 0.6 V A €p = 1.2 V

1000 S 1000s

) ~ r'-- r-
100s r- 100 s

10 s
10 s
-V r--- ~ 1s
1s
I -V 11\\ 100ms

1- f--/ \
100 ms
10 ms
10ms I - f.-'
V\ 1ms
1 ms
I - f.-' h 100 Ils
100 Ils I - i--"' i'
lOlls

Fig. 26. Standardized potentiodynamic reduction curves after oxide formation at .6 V and
1.2 V. Parameter is the formation time.

O.B €p = 0.6 V €p = 1.0 V 6p=1~ Cu(OHh

0.6

~ ~
CuOx

~
~ 0.4
..
I
I
.lO: I \
I
II I I
Q, I
, CuO

\
W 0.2

\
o
~
-4 -2 0 2 -4 -2 0 2 -4 -2 0
Ig tp I s

Fig. 27. Potentials of the peak maxima fpeak in the reduction spectra vs. Ig tp for a formation
at .6 V, I V and 1.6 V.
96

The conversion of the species formed at short times to CU20 is also seen at .6 V, but at
1.2 V this peak corresponds to the CuO reduction. The charge exceeds a monolayer of
oxide, but the nature of this species remains Questionable, as it combines properties of
CU20 and CuO. We designated this species CuOx , to indicate this uncertainty. It may be
interpreted as a amorphous mixture of Cu+, Cu2+, 0 2-, OH-, and H 20.

CuOx ... CuO + CU20

N
I -1
E
u
ct
....... -2 c
.~
Ii.
..!' -3
~ c

c
-4
c
0 0
-5

0.6 1.0 1.2 14 16

€p(HESS) I V

Fig. 28. Total anodic current density at peak separation vs. €p.

The splitting of CuOx into CuO and CU20 happens with increasing potential at shorter
times and at greater total reduction charges. The critical parameter for splitting may be
the anodic current density. Above a critical value the high field strength prevents a separation
into oxides with different properties. The logarithm of the critical current density in
dependence on the potential is given in fig. 28. i.plit is constant up to 1.2 V and then Ig i.p1it
increases linearly with with a b factor of 160 mY.

An additional peak appears at .8 V for fp ~l.l V (fig. 26 and 27). It shows no interaction
with the other peaks and its charge is always less than 20% of the total charge. This species
is called Cu(OHh

A comparison of the different charges obtained from the pulse experiments is given in fig.
29.

The total anodic charge Qa is the sum of the different partial charges: corrosion Qeorro oxygen
evolution QQ, formation of the different oxide species Ql ... Qn' and side reactions Q.
 97

tp = 1005 (a) €p = 1.5 V (b)

4 4

qa •

N N
I qcorr I
E E

-
u u

-
t
U c.~1..0~) U
E E
i:T i:T

CuO

CuOx t
CU20
J •
0.6 08 1.0 12 1.4 16 -4 -2

E(HESS) / V Ig tp / 5

Fig. 29. Subdivision of the total charge qa into the charges of the different oxides and
corrosion in dependence on €p (t p = 100 s) and on Ig tp (€p = 1.5 V).

(19) qa = q corr + q a + q 1 + ... + qn+ qs

Side reactions are caused by contaminations which can be avoided, oxygen evolution is
impossible below 1.23 V.

Every pulse measurement yields an anodic charge and a set of reduction charges of the
different oxides. The charge of anodic corrosion is calculated from eq. (19). Fig. 29a
illustrates eq. (19) in dependence on the potential, fig. 29b in dependence on the polarization
time. At potentials above .9 V the CuOx layer has separated into CuO and CU20 and the
reduction charge of the latter has become constant due to the limited potential drop in the
oxide layer (fig. 29a). The CuO layer thickness tends also to become constant.

From the slopes of the curves of fig. 29b the corresponding current densities in dependence
on the polarization time can be calculated, as d qn/d t = in. We yield therefore the anodic
current ia(€p,t p)' the formation currents of the oxides in(€p,tp)' and the corrosion current
ico....(€p,tp).
98

This section shows that the kinetics of copper oxides are very complex and, hence, cannot
be discussed in detail. A model which explains most of the effects is presented in the next
section.

N
I .D. _
.D.
Eu Ep = 1.3 V rig
ct >
.. --
c ~
....... .......
80
o 0
0
0 ~
.jJ U 3 3
~
w n n
n n
60 3I 3I
~
0.6

-3
40
05
-4
20
1.4 2.8
-5 0.4

-6 N
I 1.0 2.0
0.3 E
u
u..
~
.......
u
0.2

0.2 0.4
A B c o
-5 -4 -3 -2 -1 0 2 3
Ig t / 5

Fig. 30. Time dependence of capacity C, corrosion current density icom peak potential of
reduction of CuO to CU20 fr' and the reduction charges of CuO and CU20.

3.1.1.3 Summary

A model of copper passivation and corrosion in alkaline solutions is presented. The oxide
layers were formed at polarization potentials of .6 V ~ €p ~ 1.6 V. Measurements were
 99

carried out at pH values of 9, 12, 13, and 14 and led to a model, which separates the
investigated time range from 10 JlS to 1000 s into four regions with different kinetics. These
regions are clearly separated at 1.3 V and pH 14 (fig. 30).

In the first region (A in fig. 30), up to some ms, the high field mechanism is valid. The
oxide formed in this region is amorphous (CuOx ) and behaves like a dielectric material.
The formation current and the corrosion current are almost proportional, i.e. that the field
strength in the oxide and in the Helmholtz layer are similar. The total oxide charge is about
1 mC cm- 2. Cu 2+ ions move through the double layer and form surface nuclei of Cu(OHh
in more alkaline solutions. These nuclei form hydroxide islands growing linearly with Ig t.

When the ionic current through the layer becomes smaller than a critical value of some
rnA cm- 2, a new region starts (B). The charge carrier concentration decreases and the
structure of the oxide becomes more crystalline (CU20), which is reflected in a different
shape of the reduction spectra. The formation of an enrichment layer causes an almost
costant potential drop in the oxide and thus also in the Helmholtz layer. The oxide behaves
like a semiconductor.

The spreading islands of Cu(OHh meet in the time range 100 ms to I s (C) and cover the
electrode completely. The field strength in the oxide increases and it is dehydrated to CuO.
Some parts of CU20 are transformed to CuO by enrichment of holes at the interface
CU20/CUO.

The layer thicknesses of CuO and CU20 become constant in the last region (D), when the
polarization time exceeds lOs. The potential drop in the Helmholtz layer and, hence, the
corrosion current is much smaller now. Corrosion current and fomation current are equal,
the system is quasi-stationary now.

All these four regions are well pronounced at higher potentials and pH> 12. At medium
potentials the corrosion is smaller for short polarization times and there is less precipitation
of Cu(OH)2' At lower potential or pH < 12 the regions Band C are not found. The less
soluble cuprate prevents a corrosion precipitation mechanism, CuO is directly formed by
transferring Cu 2+ ions and is therefore less hydrated.

3.2 Triple Transients: Ageing Effects in Oxide Layers

References: (19J

When metal electrodes are passivated the oxide layer grows and changes its properties. This
effect may be caused by the thickness changes only, but, frequently, there is an additional
effect in time, an ageing process. This ageing process is fast and takes place in less than
1 s. A separation can be done by measurements of triple transients.

The formation of oxide layers on gold electrodes in I n HCIO. is given as an example. Fig.
31 shows details of the technique. An oxide-free gold electrode is covered with gold oxide
by a short potentiostatic pulse (polarization potential €p varies from 1.4 V to 2 V). The
pulse must be short to avoid ageing during formation, hence, we used 1 ms as a typical
value. Current and charge are recorded. The charge varies from 100 to 600 JlCcm- 2,
corresponding to mean layer thicknesses from .05 to.35 nm. This is the region of a monolayer.
100

t
Ep ta 1.5
Ea
en
1,0
--
c:::
0.5

N -200
I
E
u -100 oxide
<X: formation i
E 0
i~·ge;ng
-- 100
200
reduction

200 0
--
"l:
n
n
3I
I>.)

0
0 Z 3
t / ms

Fig. 31. Schematic plot of potentiostatic oxide formation on gold at 1.5 V for I ms, ageing
at 1.4 V for I ms, and potentiodynamic reduction (sweep rate 500 VIs). From top: potential
c, current density i, and charge q vs. time.

Then the layer is aged at a potential Ca for an ageing time ta variing from 100 j1S to 1000 s.
Again, current and charge are recorded. At the beginning of the ageing pulse, a small
current density of double layer charging and oxide reduction or further oxide growth is
observed depending on the value of Ca. Finally, the aged oxide is reduced by a cathodic
potentiodynamic sweep with dc/dt '" 500 VIs. An advantage of a potentiodynamic sweep
is the detailed peak structure of the reduction spectra, which is a sensitive indicator to the
chemical structure of the oxide. The sweep must be fast to avoid further ageing during
reduction.

In most experiments the ageing potential Ca was chosen so that no current was observed in
the ageing period, i.e. the charge was constant. This is the condition for an exclusive ageing
process. Alternatively, the variation of €a makes it possible to investigate the potential
dependence of the ageing process.
 101

o
"'--10005
,-100ms
~lms 50

e = 1.45 V
\

100 ?,-
n
3I
150 N
Ea = 1.36 V

0,3
N
I \--1000s
E

-
f.I
Peakl
<C 0,2
E 1 lOOms
/~\
I ,1ms . I\)
\ / I·
0,1 I \ :,.....--_/ ~
/ "- A .'!
/ _/.

/ /

0
--
1,4 1,2 1,0 0,8 0,6
E/V
Fig. 32. Potentiodynamic reduction spectra after various ageing times ta and corresponding
charges.

Typical reduction spectra are plotted in fig.32 for !p=1.45 Y, !a=1.36 Y, and ageing times
of 1 ms, 100 ms, and 1000 s. The reduction charge is constant, but the reduction curve
changes completely. Three peaks are observed after ta=1 ms, two peaks can be distinguished
after 100 ms, but finally one peak remains after 1000 s, the well known reduction peak
from conventional measurements. This means, that three species produced first are converted
into one species after extended ageing.

Fig. 33 shows the dependence of the peak potentials on Ig tao Peak I shifts to cathodic
potentials before vanishing after ta=100 ms. The potentials of peak 2 and 3 are almost
constant up to I s, but than they overlap more and more and only one peak is observed
after 1000 S.

This indicates two distinct processes, the reduction of a more reversible species (Peak I)
and the formation of a stable one (Overvoltage of reduction of about 500 mY, peaks 2 and
3) rather than a continous conversion, ego a reorganisation process of the crystalline structure.
102

Equivalent experiments were carried out at higher polarization potentials fp from 1.6 to
2 V. Fig. 34 shows typical reduction curves for fp=2 V and f a =1.45 V. The charge is constant
(q=600 J.lCcm- 2 • corresponding to a layer thickness of .35 nm) and the reduction spectra
differ only slightly. After 1 ms only one broad peak is observed which sharpens with
increasing tao

The first peak from fig. 32 at 1.3 V is not observed. but if the amount of that species is
as small as at 1.3 V (about 50 J.lCcm- 2 ). the resolution would not suffice to detect it. Hence.
an investigation of the properties of this more reversible species has to be carried out at
lower potentials. At higher potentials the only process to be observed is a decrease of the
haltwidth of the peak. indicating an increase of homogeneity of the oxide. This process is
proportional to Ig ta and takes place up to more than 104 S.

1,3 (a) (b)

1,2
\~ \
= 1.45 V
E
Ea = 1.36 V

-
>
W
1,0
-.---.
_.--...,
.~kZ

."".~ ,=:=::::::::::::::..-----.
.-.-._
'-._
€IV~NI
._.-.-._.............-==:::-1,6 [\381
' - . _ -is [1.401
"1.9 [1.401
0,9 '--2.0 [1~51

-.-.- ._.-·-Peok3
0,8

Ig ta / S

Fig. 33. Potential of peak maximum € as a function of Ig tao

According to measurements of several authors /201./211 the oxide layers on gold should
consist of AU20s in the monomolecular region, and this should be the composition of the
stable species. The nature of the more reversible species is not so clear. To explain the
pH-dependence of the oxide formation. a reversible adsorption of OH- ions with a coverage
of about 15% was assumed /221. This is similar to the amount of the reversible species.

There is a strong potential dependence of the ageing process. Hence. it must be an elec-
trochemical process. The anodic oxide growth competes with a cathodic reduction of the
reversible species.
 103

During the rapid oxide formation within I ms an appreciable amount of adsorbed OH-
ions is formed (Peak I). At suitable surface sites oxide is formed by place exchange reactions.
The rate of these reaction depend on the local field strength which decreases with increasing
coverage \22\. At other less convenient sites the OH- ions may be desorbed cathodically.

200

-c
400 n
n
3I
E = 2 V Ea = 1.45 V N

500

1.5

N
I
E 1.0

-
u
c:(
E
. - 0,5

0
1,4 1.2 1.0 0,8 0.6
€/V

Fig. 34. Potentiodynamic reduction of oxide formed at €p=2 V for various ageing times.

The less reversible species (Peak 2 and 3) overlap with the stable peak and cannot be
separated with sufficient accuracy. The reduction potentials are close to that of the stable
peak and their conversion to the stable oxide is continuous. Hence, this process may be
reorientation to a less amorphous state. Possibly, there is some influence of the grain
orientation of the polycrystalline gold electrode. It is well known that there is a strong
dependence of oxide formation on the orientation which vanishes during reduction.
104

4 Capacity Measurements by Pulses

References: 1231,(24)

Current transients of potentiostatic pulse measurements can be used to determine the capacity
of the electrode. The method has all the advantages and limits of transient measurements
and is therefore suitable for systems which change their properties in very short times.

4.1 Principles

6(=50 mV

40

':'E 20
u
4: 0
.§
-20

-40

':" 0.1
E
u
~0.05
er-
0

o 5 10 15 20
t/flS
Fig. 35. Principles of capacity measurements by potentiostatic pulses. From top: potential,
currrent and charge vs. time.

The principles are shown in fig. 35. During potentiostatic or potentiodynamic polarisation
short rectangular pulses with a small amplitude t.u are superimposed. Short current pulses
indicate the charging of the electrode capacity. The capacitive charge t.q is obtained by
simultaneous integration of the current i:
 105

(20)

c=-=--
6.Q it idt

6.U 6.U

Many effects can cause errors. First of all, the pulse has to be short to avoid faradayic
processes caused by the pulse itself. Using a pre-pulse of I I'S to accelerate the rise time
of the potentiostat a pulse length of 10 IJS is sufficient, corresponding to frequencies of
about 50 kHz. Keeping the potential dependance of capacity in mind, the pulse amplitude
has also to be low, because the capacity values from this method are not exactly the
differential capacity from eq. (20) but its mean value over a range given by the amplitude.
Pulse amplitudes of 20 to 50 mY are adequate.

Further problems are produced by random faradayic processes during the pulse. The
corresponding currents overlaying the capacitive current must be separated. DC components
and lower frequencies « 10 kHz) are suppressed by an active analog filter. Linear gradients
passing this filter are compensated by a modified sample-and-hold amplifier, random
gradients cannot be separated.

The integration of the filtered current signal is done by operational amplifiers (bandwidth
10 MHz) and digitizing of the output signal with a laboratory computer or by digitizing
and storage of the current signal in a transient recorder followed by numeric integration
in the computer. Both methods give the same results.

The method can be fully automized to yield pseudo-continuous capacity values. Investi-
gations of unknown or complex systems, however, require monitoring of the signal by the
operator to avoid misinterpretations.

DIA
Adder
Po tentios tat DIA

Computer
Current to
Voltage
'--------i Con vert e r
AID

Fig. 36. Block diagram of equipment for capacity measurements.

106

Fig. 36 shows the principles of the equipment. The reference voltage is formed by the
adder, which summs different signal from the computer. The use of digital-to-analog
converters makes fast programmed changes of the pulse shape (amplitude, sign, and length)
possible. The output of the current to voltage converter is filtered, integrated and digitized.

4.2 Continuous Capacity Measurements

A single pulse of 10 /.IS and 20 mY has only negligible influence on polarisation. To minimize
faradayic effects for repetitive pulses reflections on the sign of the pulse should be done.
The pulse should always decrease the overpotential of the electrode, ego an anodic polarization
requires cathodic pulses. Repetitive rates up to 104 S-l can be realized.

Using long polarization times, however, it is advisable to reduce the pulse rate with time
as described in section 2.4.

4.3 Potential Dependence of Capacity

For analysis of passive layer formation and reduction, detailed knowledge of the semi-
conductor properties of these layers is necessary. A suitable method to get data of the band
model is the determination of the potential dependence of the capacity. AMott-Schottky
analysis may give data of charge carrier concentration, flat band potential, and band gap.
During formation and reduction these data change fast, and should be gathered in times
less than 1 ms. This can be done by a slightly modified pulse technique.

Ep ---------------

~E
1,3

-- 1,1
> 1,2
w

1,0
0,9

0 100 200 300 400 500

t/~s
Fig. 37. Pulse complex to measure the potential dependence of capacity.
 107

The measurement is realized by a special stepped pulse (fig. 37). Every step represents a
mesuring potential. The capacity is obtained by superimposing a pulse of inverse sign. The
step height is typical a few 100 mY, the step length may be 80 to 1001's. Hence, the
complete measurement requires 500 to 1000 I's, depending on the number of steps. This
makes experiments possible even at potentials where the layer is unstable, ego a few 100 mY
below the reduction potential. The disturbance of the passive layer is negligible. Using
repetive measurements the rate should be not too high. This is again possible by a repetition
rate decreasing with time (see section 2.4). The running off is given in fig. 38. The first
data are taken at I ms, the next at 10 ms, 100 ms, and so on. This means 7 cycles within
1000 s with a sufficient resolution in time.

0
~
I
<;" -1
E I
u I
<c -2
--- I
I
g'-3
I
-4
I

-5

-8 -7 -6 -5 -4 -3 -2 -1 0 2 3
log tIs
Fig. 38. Schematic anodic transient of passivation. The moments to determine the potential
dependent capacity are marked.

4.4 Examples

The potential dependence of the capacity is a very complex topic, especially in the context
of passivation. Two examples, concerning the passive layers on zirconium [25),[26) and iron
[21),[28) are presented.

Zirconium is a typical valve metal with an oxide growth according to the high field
mechanism. The layer thickness d is a linear function of the polarization potential €p

(21)
108

14

12 tp=1000s a a

N
I
E
l.I
10
'0' ~~ 0 0 :

100~:0:
: : ~\
I.L. 8
::..
-.... 6
u
4 10 ms A \ \0
~.l A
A .. "o~~~
! -
a

0
-2 -1. 5 -\ -.5 o
E/V
Fig. 39. Potential dependent capacity of a zirconium electrode passivated at 2 V for different
waiting times at -.7 V.

14

12
N
I 10
E
l.I
I.L. 8
::..
-.... 6
u
4

2
0
-2 -1. 5 -1 -. 5 o
E/V
Fig. 40. Potential dependent capacity of an zirconium electrode polarized at 2 V for 1000 s
and at -.7 V for 1000 s for different waiting times at 0 V.
 \09

with an oxide formation factor k of 2.6 nm/V. Potentiodynamic measurements showed an

slight increase in capacity in the cathodic sweep at potentials ! < -.7 V. A detailed
investigation was impossible, because the current of hydrogen evolution was too high. The
phenomenon was therefore investigated by the pulse techniques described above.

The zirconium electrodes were passivated at 2 V for 1000 s. An oxide layer of 10 nm was
formed. The layer was stabilised at 0 V for some seconds and then potentiostatted to -.7 V
for 1000 s. A hydrogen evolution of about I mA cm- 2 was observed. The potential dependent
capacity was measured from 0 V to -2 V at different times 10 ms < t < 1000 s simulta-
neously (fig. 39) without affecting the layer. Equivalent DC measurements would be
impossible because of the strong hydrogen evolution.

After 10 ms the capacity is almost constant at 3 J.'F cm- 2 • With increasing time a capacity
maximum at -1.3 V is observed. The maximum increases with time and is shifted anodically.
This behavior is similar to a n-type semiconductor and may reflect a doping during hydrogen
evolution. A partial reduction is possible according to

(22)

The stability of the states produced during hydrogen evolution was tested in a modified
experiment. After the cathodic pulse at -.7 V the electrode was potentiostatted again to
o V and the potential dependent capacity was measured (fig. 40). The capacity curves are
similar, but the time dependence is inverse: the maximum vanishes with increasing time
at 0 V.

A Mott-Schottky analysis was done to get quantitative data from a I/O vs. ! plot (fig. 41a
and b). Straight lines are expected from the Mott-Schottky equation:

(23)

Values of the flatband potential !fbare taken from an extrapolation to l/e 2 = 0, the product
of donor concentration N and dielectric constant D can be calculated from the slopes.

Straight lines are obtained for the measurements at -.7 V and at 0 V. The donor concen-
trations, calculated using D .. 31, are N = 2.3-10 19 cm- 3 at 0 V and N = 1.5-10 19 cm- 3 at
-.7 V. Surprising is the fact, that the lines run parallel for different times. This means that
the donor concentration is independent of time for one potential, but the flatband potential
is changing.

The shift of the flatband potential with time is shown in fig. 42. At -.7 V !fb shifts linearly
with Ig t to anodic values, at 0 V the shift is in cathodic direction up to 10 s, then !fb
remains constant. The anodic shift of the flatband potential may be caused by surface
states, which are formed during hydrogen evolution. They effect a positive charging of the
surface which lowers the conduction band at the surface and, hence, shifts !fb anodically.
The surface states are oxidized at higher potentials (0 V) and vanish with Ig t.
110

.25
(a) !
.2
..-..
U. s
::l.
......... "
N . 15 A

--
I
E
\J t p =10ms
......... .1
N
U
.........
~
.05

0
-2 -1. 5 -1 -.5 0
E/V
.25
(b) 0

-u..
.2
,0

:::t.
......... . 15
N A

--
I
E
\J

.........
.1
N
U
.........
...-oj
. 05
'10 ms

0
-2 -1. 5 -1 -.5 0
E/V

Fig. 41a and b. Mott-Schottky representations of the preceeding figures.

Similar measurements were carried out at electropolished iron electrodes in a borate buffer
of pH = 8.4.
 111

-1
,,
, \
A

\
\

-
-1.2
>
12
W
-1.4

-1.6 /
/

/ C p = OV
/
/ o

-3 -2 o 1 2 3
Ig t / s

Fig. 42. Flatband potential €fb in dependence on Ig t.

It is impossible to spread out all the details of passivation on iron electrodes. We have to
reduce to a short example.

Layer free iron electrodes were passivated at I V, 1.2 V, and 1.5 V for 100 s. The passive
layer thickness are 2 nm, 2.5 nm, and 3 nm, respectively. The potential dependent capacity
is measured in a wide range from .2 V to 1.9 V. The values are presented in aMott-Schottky
plot (fig. 43). In principle three regions are recognized: an almost linear increase of I/O
up to .7 V, a plateau up to 1.5 V, followed by a decrease. These regions must be caused
by different electronic states close to the surface.

In parallel, computer simulations were done to calculate equivalent values from different
models. The best fits were achieved using the model given in fig. 44.

The Fermi level EF is .13 eV below the conduction band (CB), which is formed by Fe4s
states. Completely dissociated donor states, presumably lattice defects, are localized .3 V
below the CB. Their concentration is about 3.1020 cm-li and the capacity is dominated by
these states up to .7 V. Surface states with a half width of .2 eV are located .8 eV below
112

.045.---------------------------_r--------~

.04 .Lr = 1.5 V IIIESS ••

•
. 035

N
I .03
N
I

E .025
u
~
::t
(I' = 1.2 V/HESS
N'" .02
I
U
.015

.01
Eo p = l.O V IHESS

OL-.-------,,------~--------._------_r~
o .5 1.5 2
~ - ( I.R IV

Fig. 43. Mott-Schottky analysis of the potential dependent capacity of iron electrodes,
passivated for 100 s at different potentials. Values of a computer simulation are given in
addition (full line).

the CB. Their surface concentration is about 2.10 13 cm- 2 and they dominate the region up
to 1.4 V. The last region is influenced by deep states, 1.4 eV below the CB. They may be
formed by Fe3d states. Their concentration was assumed to be 5.10 20 cm- 3

The model explains the properties for quasi stationary conditions. The dependence of the
oxide structure on the polarization time tp at 1.5 V is shown in fig. 45. The complex
structure of fig. 43 is developed for tp > 10 ms only. The oxide seems to be highly disturbed
for short times.

5 Photocurrent Transients

Photoelectrochemical experiments are usually carried out with low power densities
P < 10 mWcm- 2 . A considerable enhancement of P yields new results concerning photo
corrosion and photoinduced oxide growth.
 113

Fe FeO x
Fe4s CB E-EcleV
ECB 0
EF
ED

Ess
-1

ET

Fig. 44. Band structure of quasi stationary passive layers on iron.

The enhancement is done by focussing laser beams as shown by Butler [29). We used an
continous wave argon laser which emits green light of >. = 514 nm with a total energy of
2.3 W. In most cases the frequency was doubled to 257 nm (energy 150 mW). The resulting
photon energy of 4.8 eV fits to the band gap of many passive layers. The laser beam has
a typical diameter of 500 I-'m which means power densities of 200 W cm- 2 (green) and
10 W cm- 2 (UV), respectively. Focussing to a diameter of 5 I-'m enhances P to 2 MW cm- 2
and 100 kW cm- 2 • This is a factor of about 106 compared with conventional experiments.

There are more advantages to use focussed laser beams. Common photoelectrochemical
investigations are done assuming the surface to be homogeneous. The photocurrent is an
integral response of the surface and yields average values. Effects of different grains, grain
boundaries, and local defects cannot be distinguished.

With a focussed laser beam and an electrode mounted on a numerically mounted x-y stage
(fig. 46), on the other hand, lateral inhomogeneities down to 5 I-'m can be resolved for
microanalysis [30). Further, laser radiation can be used for microscopic marking of the
surface to correlate the local photocurrent with microscopic images.

A disadvantage, of course, is the missing capability of a continous change of the wavelength.

114

..... .-...... _ ....... .-................ .,.- 10

illS

...... .JI" 1 ms

O~~------r-------.-------~------~
o .5 1.5 2
('IV

Fig. 45. Mott-Schottky plot of an iron electrode for different passivation times at 1.5 V.

Electrode

Fig. 46. Schematic representation of the focussed laser beam and the electrode mounted on
a x-y stage for scanning photocurrent measurements.
 115

The illumination of the electrode surface can be done continuos or by light pulses of defined
length. Corresponding photocurrent transients can be recorded.

The fundamental process induced by radiation is illustrated in fig. 47. The electrode is held
at anodic potentials. Photons are absorbed in the passive layer. Electron hole couples are
formed, which are separated by the field strength. The electrons migrate in the conduction
band (CB) to the metal electrode, the holes accumulate in the valence band (VB) at the
electrolyte interface and react with redox systems in the electrolyte. Typically, oxygen is
evolved or additional oxide is formed.

EI eV

Ti

Fig. 47. Formation of electrons and holes by photons in a passive layer.

5.1 Photocurrent Investigations of Valve Metals at High Power Densities

References: (30),(31),(32)

The problems to treat the electrode as a homogeneous surface shall be demonstrated first.
The experiments were carried out on Titanium electrodes with oxide films prepared by
anodic oxidation in I N H2S04 , Photocurrent transients of laser scans on differently prepared
Ti electrodes are given in fig. 48.

The mechanical polished electrode has a homogeneous surface and a small, site independent
photocurrent. The electropolished one shows local differences, which are well pronounced
on the etched electrode. The changes in photocurrent coincides with the grains seen in the
microscope.
116

50

40

-
<c 30
.:::
_0. 20
~ eleetropolished

meehan.
10 polished

O~----.-----r-----r---~.----.----~
-100 -50 o 50 100 150
x/pm

Fig. 48. Local photocurrent measurements of Ti electrodes after different treatment and
passivation at 5 V. The photocurrent was measured at 1.5 V, power 10 JJW, layer thickness
10 nm.

The macroscopic appearence of the etched electrodes is dark blue. Light blue, dark blue,
and violet grains can be distinguished under the microscope. The colours seem to be formed
by a combination of light absorption and interference. Fig. 49 shows photocurrent scans
of the same area of an etched electrode at different potentials f. Three differently coloured
grains, abbreviated as A, B, and C, were choosen and the potential dependence of the
photocurrent iph was plotted in fig. 50.

The Butler-Gartner model/33) predicts a linear correlation of photocurrent and f1/2, which
is true for macroscopic measurements. The grains show a different behaviour. Grain B
shows a linear increase of i ph with f, grain C reaches a saturation value at 1.8 V. The
photocurrent of grain A only, which behaves like an average of Band C, is proportional
to f1/2.

This different behavior is caused by different oxide thickness, substrate orientation and
different electronic properties.

These measurements show, that the common model of a homogeneous oxide layer must be
substituted by microscopic pictures to avoid misinterpretations.
 117

12 Ti02 prepolarized at 20 V

10

8
<C
c
-_c.
.z:. 6

2
A B c
o~----~--~----.-----.----.----~
-100 -50 o 50 100
x/pm

Fig. 49. Local photocurrents in dependence on the potential, passivation at 20 Y, thickness

40 nm, power 10 jJW.

6
<c
-..
.c
_Q.
4
A-IU
A=1(8+CI
2

o 2 3 4 5
€(SHE) / V

Fig. 50. Photocurrent vs. potential, values taken from the preceeding fig.
118

Photocurrent measurements at high power densities is the next topic. The Ti02 passive
layer is a n-type semiconductor with a flat band potential fb = 0 V. The photons produce
electron hole pairs which are separated in the electric field. At anodic potentials the electrons
flow to the metal and holes react at the electrolyte interface with water (oxygen evolution)

(24)

or Ti0 2 (photo corrosion)

(25)

Dependent on these reactions holes accumulate at the surface. This causes an increase of
the potential drop in the Helmholtz layer AIPH. Further oxide formation:

(26)

or corrosion may be a result:

(27)

These two reactions take place without participation of holes.

Photocurrent transients caused by light pulses are shown in fig. 51. The photocurrent
decreases rapidly to a final value in the first transient. In repeated experiments this constant
current only is observed (3 rd transient in fig. 51). Microscopic pictures show changes of
the oxide colour equivale"t to an distinct oxide growth. The decrease of photocurrent in
time corresponds to this oxide growth, while the time independent part is oxygen evolution.
The charge of the photon induced oxide growth, taken from the integrated current difference
of the first and the following runs fits the values obtained by interference microscopic
data. The local layer thickness increases by a factor up to 10.

A microscopic inspection of the electrode surface showed, that the laser induced oxide
growth extends the illuminated area significantly. This may be explained by two models:
a hole diffusion or a corrosion precipitation mechanism.

The potential drop AIPH at the at the dark parts of the surface is less than at the laser spot.
This causes a lateral potential drop at the surface which is no longer an equipotential plane.
This potential drop forces holes accumulated at the laser spot to migrate laterally up to
some 100 J.lm to the dark surface. Using reasonable values of the hole mobility, a lifetime
of the holes of 10- 5 s is calculated. This is long, but seems not to be impossible in the
exhausted oxide film.

Surprising is the negligible influence of grain boundaries on the lateral migration, as they
should represent hole traps.
 119

300

250
light off light on light off
~oo
~

--
::I.
150
.=
_0.

100

50

0
0 2 4 6 8
t / s

Fig. 51. Photocurrent transients of repeated experiments at 1.6 V. Passivation at 5 V, power

density 40 kW cm- 2 •

The other model presumes an increased photocorrosion at the laser spot forming a sphere
oversaturated of Ti4+. Nucleation processes would yield the same radial symmetry as for
the migration of holes. An oxide, produced by precipitation, should have different properties
than an electrochemically formed one. This seems to be true, the photoinduced oxide shows
a higher corrosion rate and vanishes within 2000 s at 0 V by corrosion.

6 Summary

Pulse techniques are suitable tools to investigate electrochemical processes, which occur far
from equilibrium state, like passivation, corrosion, and the kinetics of inhibiting surface
layers. Most of the corresponding problems, e.g. the potentiostat, recording of the current
transients, or the determination of the charge, are solved by modern electronics.

Several systems are presented as examples. The application of single pulse techniques is
limited to some special systems. Multiple pulse techniques are necessary to investigate the
kinetics of more complex systems, e.g. processes in parallel, like oxide formation and
corrosion.
120

References
1. D. Ebling, M. M. Lohrengel, 1. W. Schultze, Proc. Symp. "Transient Techniques in
Corrosion Science and Engineering" Chicago (1988), in press
2. M. M. Lohrengel, D. Ebling, Meeting Fachguppe Angewandte Elektrochemie, Basel
(1988), DECHEMA Monographie 117 (1989)
3. M. M. Lohrengel, D. Ebling, Meeting 27th Tutzing Symposium of DECHEMA, Tut-
zing (1989)
4. D. Ebling, Thesis, Dusseldorf (in preparation)
5. N. Cabrera, N. F. Mott, Rept. Progr. Phys. 12 (1948) 163
6. E. 1. W. Verwey, Physica 2 (1935) 1059
7. A. Guntherschulze, H. Betz, A. Physik 92 (1934) 367
8. 1. A. Davis, B. Dorney, 1. P. S. Pringle, F. Brown, 1. Electrochem. Soc. 112 (1965) 675
9. D. Ebling, M. M. Lohrengel, A. Meyer, A. Thyssen, Dechema Monographie, Frank-
furt (1989), in press
10. H.-D. Speckmann, M. M. Lohrengel, 1. W. Schultze, H.-H. Streblow, Bel'. Bunsenges.
Phys. Chern. 89 (1985) 392
11. M. M. Lohrengel, 1. W. Schultze, H. D. Speckmann, Electrochim. Acta 31 (1986) 123
12. M. M. Lohrengel, 1. W. Schultze, H. D. Speckmann, H.-H. Streblow, Electrochim.
Acta 32 (1987) 733
13. H.-D. Speckmann, Thesis, Dusseldorf (1987)
14. K. 1. Vetter, Electrochim. Acta 16 (1971) 1923
15. M. M. Lohrengel, 1. W. Schultze, Electrochim. Acta 21 (1976) 489
16. M. Yamashita, K. Omura, D. Hirayama, Surf. Sci. 96 (1980) 443
W. Paatsch, Ber. Bunsenges. Phys. Chern. 81 (1977) 645
U. Bertocci, 1. Electrochem. Soc. 125 (1978) 1598
17. H.-H. Streblow, B. Titze, Electrochim. Acta 25 (1980) 839
18. A. Hickling, D. Taylor, Trans. Faraday Soc. 44 (1944) 262
S. M. Abd el Halem, Badr G. Ateya, 1. Electroanal. Chern. 125 (1981) 309
H.-H. Streblow, B. Titze , Electrochim. Acta 25 (1980) 839
N. S. McIntyre, S. Sunder, D. W. Shoesmith, F. W. Stanchell, 1. Vac. Sci. Technol. 18
(1981) 714
D. W. Shoesmith, S. Sunder, M. G. Bailey, G. 1. Wallace, F. W. Stanchell, 1. Electroa-
nal. Chern. 143 (1983) 153
19. 1. W. Schultze, M. M. Lohrengel, Ber. Bunsenges. Phys. Chern. 80 (1976) 552
20. T. Dickinson, A. F. Povey, P. M. A. Sherwood, 1. Chern. Soc., Faraday Trans. I, 71
(1975) 298
21. 1. W. Schultze, Electrochim. Acta 17 (1972) 451
22.1. W. Schultze, K. 1. Vetter, Ber. Bunsenges. Phys. Chern. 75 (1971) 470
23. U. Konig, M. M. Lohrengel, in Grundlagen von Elektrodenreaktionen (ed. 1. W.
Schultze), Dechema Monographien 102 (1985) 99
24. U. KOnig, M. M. Lohrengel, 1. W. Schultze, Ber. Bunsenges. Phys. Chern. 91 (1987)
426
25. P. Meisterjahn, U. KOnig, 1. W. Schultze, Electrochim. Acta 34 (1989) 551
26. P. Meisterjahn, Thesis, Dusseldorf (1988)
27. U. KOnig, 1. W. Schultze, Werkstoffe und Korrosion 39 (1988) 595
 121

28. U. Konig, Thesis, Dusseldorf (1989)

29. M. A. Butler, J. Electrochern: Soc. 130 (1983) 2358
and J. Electrochern. Soc. 131 (1984) 2185
30. K. Leitner, J. W. Schultze, Ber. Bunsenges. Phys. Chern. 92 (1988) 181
31. J. W. Schultze, J. Thietke, Electrochirn. Acta 34 (1989), in press
32. J. Thietke, Thesis, Dusseldorf (1989)
33. W. W. Glirtner, Phys. Rev. 116 (1959) 84
M. A. Butler, J. Appl. Phys. 48 (1977) 1914
LINEAR ELECTROCHEMICAL TECHNIQUES - ARE THEY ALL THE SAME?

RA. COTIIS and S. TURGOOSE,

Corrosion and Protection Centre,
UM/ST, P.O. Box 88,
Manchester MOO /QD, UK.

ABS1RACI". This paper describes a computer simulation which has been used to explore the relationship
between polarization resistance and ac impedance measurements. This has demonsttated that all
measurement techniques which assume a linear electrode response can be related to one another (or, to
quote MacDonald, are sampling the same impedance response). The process of simulating the system has
required a detailed examination of the physical processes concerned in the very simple electrochemical
system which has been investigated, and has revealed some unexpected results. For example, it has been
shown that the polarization resistance for this system is not directly related to the rate of reaction (when the
latter is defined as the impedance at zero frequency), although conventional polarization resistance
measurement techniques may, somewhat fortuitously, give a result which is approximately equal to the
charge transfer resistance, which does give a measure of rate of reaction.

1. Introduction

The work presented in this paper has been concerned with an examination of the premise that a
range of electrochemical techniques which assume a linear response of a metal-solution interface
are essentially equivalent, and can be interrelated through transformations between the time and
frequency domains. While the link between the time and frequency domains is well known, this
view of linear electrochemical methods has given useful insights into the significance of particular
results.
The investigations presented here originated when considering the relationship between ac
impedance measurements 1, measurements of linear polarization resistance2 and transient
measurements for steel in concrete3. These exhibit somewhat contradictory characteristics, in that
polarization resistance measurements made with a sweep period of around 30 seconds appear to
give good correlation with weight loss, whereas ac impedance measurements need to go down to
very low frequencies, and measurement periods of the order of hours, in order to define the
response. Similarly analysis of the transient response of the same electrochemical system has been
used to determine time constants of the equivalent circuit comparable to those which can be
obtained by ac impedance measurements, but in a somewhat shorter time. As any periodic
waveform can be expressed as a sum of sine waves, through the Fourier transform, it seemed
probable that this contradiction could usefully be explored through the transformation of the
polarization resistance and transient information to the frequency domain, to confirm, as we
supposed, that they were actually measuring an appropriate part of the impedance response.
In order to avoid complications due to non-ideal behaviour of real electrochemical systems,
experiments performed to date have been concerned with a computer model of a 'simple' four

123
M. G. S. Ferreira and C. A. Melendres (eds.).
Electrochemical and Optical Techniques/or the Study and Monitoring 0/ Metallic Corrosion. 123-133.
© 1991 Kluwer Academic Publishers.
124

component electrochemical interface. This has described the potential-time behaviour of the
interface in response to a controlled current wavefonn. This pennits the simulation of linear
polarization resistance measurements by using appropriate current waveforms. The resultant
potential waveform, together with the driving current wavefonn, have been transfonned to the
frequency domain by a standard fast Fourier transform (FFf), and the impedance computed as the
voltage spectrum divided by the current spectrum.
The results have demonstrated that the impedance can be measured from the results of
waveforms such as those used for linear polarization resistance measurement, but they have also
shown significant limitations of this method and requirements for measurement frequency and
sensitivity which might not be obvious at first sight. Probably the most useful outcome of this
work was the improvement in our understanding of the physical system. This insight is commonly
obtained by the development of a detailed simulation, but is difficult to convey in a written paper.

2. Nomenclature

The measurement and analysis procedures use a number of similar variables, particularly with
respect to time intervals, and for clarity these are defined here:
Cycle period, 't, the period of one full cycle of the current wavefonn.
Sample period, M, the time between individual samples used for Fourier transfonn analysis. If
there are n samples per cycle, then M x n = 'to
Subsidiary sample period, the period used between calculations of the model. This may be equal
to the sample period if the latter is small, or it may be a fraction of the sample period.

3. The Model

Double

~
~
Rnl.I.nOll
R.",
Bulk
Solution

DItI'usIon
Process

Figure 1. Model Electrochemical Interface

The model used was a simple four component network (figure I), consisting of a solution
resistance, Rsol, a double-layer capacitance, Cu, a diffusion process between the bulk solution and
the metal-solution interface, and an activation-controlled reaction at the metal-solution interface.
While the charge transfer and diffusion processes are, in total, equivalent to a charge transfer
resistance and a Warburg impedance, it may be noted that the point of separation is slightly
 125

different in this model. The diffusion component is used only to determine the surface
concentration of cathodic reactant, while the charge transfer component calculates the
overpotential as a function of the surface concentration and the faradaic current.
In order to simplify the analysis a single cathodic reaction ha~ been considered, with diffusion
of the oxidised state of the cathodic reactant to the metal surface. A practical example of such a
system would be the reduction of dissolved oxygen at a platinum electrode. In order to fix the
steady-state potential of the interface the current waveform applied consists of an average current,
Ide, which is a fraction of the limiting current (typically 25 or 75%) on which a small fluctuation is
superimposed (5% of the limiting current). While this example may seem rather remote from real
corrosion systems, it should be appreciated that it can be thought of as a system with a fixed (e.g.
mass transport limited) anodic current (equal to Ide) and a cathodic current which is under mixed
control. For a corroding system, such as iron in aerated water the system is analogous, except that
the cathodic current (due to oxygen reduction) may be current limited, while the anodic reaction
will be under mixed control. Thus the model used should give a realistic indication of the
behaviour to be expected from a corroding interface, although some of the details are rather
simpler (for example, for iron the response is complicated by the solution oxidation of ferrous to
ferric, which will modify both the limiting current for the cathodic reaction and the diffusion
component of the anodic reaction). It also avoids a problem inherent in systems giving a
diffusional impedance due to anodic species diffusing away from the interface. In the latter case
the diffusional impedance will only be observed if the rate of the anodic back reaction is
significant, and this would therefore also have to be included in the analysis.
The defining equations for the system described are thus:

3.1. CAPACITANCE

dV leap

dt Cdl

3.2. DIFFUSION PROCESS

de

dt

or, when expressed in terms of the finite difference model used for this work,

D~t
Cn,t+~t = Cn,t + (Cn+l,t + Cn-l.t - 2Cn ,tl

3.3. CHARGE TRANSFER

l· k Cs exp (_ EanF
RT
126

3.4. SOLUTION RESISTANCE

V I Rsol

In order properly to test the applicability of the Fourier transform approach to the analysis of
polarization resistance measurements, it is important that the computer model provides a precise
rcsponse over all frequencies of significance. While analytical solutions for the response of an
electrochemical interface to a current transient are available, they necessarily make assumptions
that are not of general validity. Thus individual solutions are available for the response of the
discharge of the double layer capacitance and the diffusion process, but these do not describe the
behaviour in the intermediate regime where both the capacitance and the diffusion contribute to
the response. In addition it is a fundamental assumption of the Fourier transform that the
waveform being analysed is repetitive. If the period of the waveform is sufficiently long the
system may attain near steady-state conditions during the cycle, but for shorter cycles this will not
occur, and the steady-state starting conditions assumed by any analytical solution will not exist.
Therefore the model is based on a numerical analysis, using a finite difference model of the
diffusion process, and calculating the potential for each time step by iteratively balancing the
potential across the double layer capacitance against the potential due to the faradaic current. In
order to allow the system to attain stable conditions, the analysis is run for several cycles before
recording the voltage response. Considerable care is necessary to avoid the introduction of
spurious information at higher frequencies as a result of instability of the analysis. For example,
when a triangular current waveform was analysed using the procedure which had been 'tuned' for
use with the square current waveform, the resultant voltage-time trace appeared perfectly
satisfactory to the unaided eye. However there were small steps in the results, due to an improper
convergence test, and the resultant impedence spectrum showed marked peaks in amplitude.
The program used to create the model voltage-time response was written in Turbo Pascal,
running on an IBM PC AT compatible. The period of the repetitive waveform and the number of
samples taken over one cycle could be independently varied. For longer intervals between
recorded samples, more frequent subsidiary samples were taken, such that the time between
samples was significantly less than the RpCdJ time constant and the time for the concentration of
the elements in the diffusion process to change significantly. When this 'oversampling' process
was used the sample values recorded were the average of all of the voltages determined during the
period of one sample. In effect this corresponds to filtering the measured voltage signal to limit the
high frequency response to the sampling frequency.

4. Data Analysis

Most of the data analysis has been performed using 1024 samples per cycle. These records have
been analysed using a commercial program (Asystant) which provides very easy processing of
waveform data. For longer time records (up to 65536 points per sample) a special-purpose
program is being developed, running on an Opus PC7, which uses an 80386 processor and can
therefore handle very large data arrays.
The stages of the analysis procedure are indicated in figure 2. A conventional discrete Fourier
transform is used to convert the voltage-time and current-time records to complex frequency
spectra. In both cases the complex component of the input time records are taken to be zero
 127

Run simulation to calculate

potential and current for
1024 points through the cycle

!
Transform current and potential
waveforms to the frequency
domain with FFT

l
Remove zero values from
transformed data

!
Divide potential transform by
current transform to calculate
real and Imaginary components
of Impedance.

l
Display results as Bode or
Nyquist plot

Fugure 2. Data Analysis Procedures

throughout, and no windowing or trend removal is used prior to the Fourier transform. The result
of the Fourier transform is an array of real and imaginary amplitudes at a range of frequencies 0,
Ilt, 2ft, ... , (n- l)/'t, where '[ is the period of one cycle, and n is the number of points in that cycle.
For many regular current waveforms (such as the square and triangle waves used in this work) the
amplitude at certain frequencies is zero, and these are removed prior to further calculation, to
avoid division by zero errors. Then the voltage amplitude at each frequency is divided by the
current amplitude at that frequency (using complex division) to give an impedance. This should
then correspond directly to the impedance that would be measured by a transfer function analyser,
and can be plotted as a Nyquist or Bode plot.
H would not be surprising if a calculated electrochemical response were to give a near-perfect
impedance measurement, especially with all calculations being undertaken with double precision
arithmetic. In real systems, even if the electrochemical system obeys the appropriate equations
exactly, the measurement will introduce errors. These have been simulated in this work in one of
two ways. Errors due to noise in the voltage measurement have been simulated by adding
normally- distributed random noise to each measurement point. Errors due to limited resolution of
128

the measurement have been simulated by quantising the voltage values to a suitable proponion of
the full-scale voltage.

5. Results and discussion

5.1. POWER SPECTRUM OF EXCITATION SIGNAL

.p.o~_~r~SP~8~CI~ra~I~D~8n~8~il~y~(m~A~~~H=Z)~__________- - ,
100.-

80
10

60

0.1

40
0.01

20
1.000E-04

1.000E - 05 L--'--'---'--'..LLllL__-'--'---"--'-J.J.l.J-'------"---"--LLL-'~ 0
1 10 100 1000
Frequency (Hz)

Figure 3. Power spectrum of square wave

An important aspect of the use of polarization resistance types of measurement for the
determination of impedance characteristics is the amount of power present in the input signal as a
function of frequency. In the square wave used for most of this work (figure 3) the maximum
power is present in the fundamental frequency (l/'t), and the high frequency power is low relative
to the fundamental. One might therefore expect to get a better impedance spectrum using signals
such as random or pseudo-random noise, which have a more balanced frequency content4 .
However, this is not the primary purpose of this paper, and has not been pursued here.

5.2. IMPEDANCE SPECTRA - THE EFFECT OF CYCLE PERIOD AND SAMPLING INTERVAL.

There are fundamental limits to the information which can be obtained from a sampled waveform.
It is impossible to obtain information about frequencies below that corresponding to the cycle
period. One way in which this can be expressed is that the bandwidth of the analysis is given by
l/cycle period. This limits the low frequency information which can be obtained, panicularly from
measurements with shorter cycle periods. In order to obtain reasonable resolution of the full
Warburg impedance, the cycle period should be at least twice the time required for the diffusion
process to reach near-steady- state. At high frequencies the Nyquist frequency (n /2't) provides an
upper bound to the measurement. A further problem at high frequencies is that of aliasing, in
which frequency components above the Nyquist frequency are effectively measured as being the
corresponding frequency interval below the Nyquist frequency. The perfect triangle and square
 129

Amplitude (Ohms) Phase (degrees)

10000~------------------------------------~---.

80

1000
60

100
40

10
20

1000 sec
1~"-L~~~~~ll-~UUllill~LLun~LLLLUWL-LJiD~0

0.001 0.01 0.1 1 10 100 1000

Frequency (Hz)

Figure 4. Effect of Cycle Peliod on Calculated Impedance

waves used in this work have infinitely high frequency components present, and this results in the
slight 'flattening' observed at the high frequency end of the power spectrum shown in figure 3.
This has significant implications for the measurement procedures examined in this work. It might
be hoped that the result of a measurement with a relatively low Nyquist frequency, when
expressed as an impedance spectrum, would be a section of the spectrum which is correct over the
valid frequency range of the analysis. The results of a series of experiments performed with a
range of cycle periods are given in figure 4. In general it can be seen that the points in the low
frequency end of the impedance spectrum are correct (if sparse), but serious errors can occur at
high frequencies (typically frequencies above about half the Nyquist frequency). For the
reasonably typical interface characteristics used in this work it should be noted that the
measurement sample period must be of the order of I ms or less if the high frequency part of the
impedance spectrum is to be determined with. reasonable accuracy. These difficulties at high
frequencies are thought to be due primarily to aliasing effects, particularly in respect of the current
waveform, together with a mismatch between the effective filtering applied to the current and
potential waveforms. The former problems can be reduced (though never entirely overcome) by
filtering the current waveform to remove frequencies above the Nyquist frequency, prior to
application of the current waveform to the cell. Alternatively the current waveform can be
calculated as the sum of the appropriate series of sine waves, which should eliminate aliasing
problems. The latter approach is extremely computer-intensive, owing to the need to calculate
many sine terms for each measurement time, and a simpler alternative is for the measurements to
be taken with a range of cycle periods and the higher frequency components of each analysis
discarded (reasonable accuracy appears to be achieved for frequencies up to half the Nyquist
130

frequency). These errors also indicate that the frequency response of measurement systems must
be well above the highest time constant to be measured. This constraint applies to transient
measurements as well as measurements aimed at the determination of impedance spectra by the
approach described here.

5.3. IMPEDANCE SPEC1RA - THE EFFECT OF MEASUREMENT NOISE.

PMM (degJ'M8) Amplitude (Ohms) PhaM (degrees)

10000
1 uV noise 80
10uV noise
80

1000 1000
SO
60
t

~J
100 100
40

10 10
20

1~~~~~~~WL~U=~~~
1 0
0.001 0.01 0.1 1 10 100 1000 0.001 0.01 0.1 1 10 100 1000
Frequency (Hz) Frequency (Hz)

Figure 5. The effect of noise on the impedance spectrum

The effect of measurement noise on the calculated spectrum can be seen in figure 5. With a
standard deviation of measurement of 1 I!V (compared to a maximum overpotential of around 6
mY) there is a significant, but probably tolerable scatter in the calculated impedance spectrum.
With 10 I!V of noise the errors become far more significant, although the general form of the
curve can still be distinguished. It should be noted that the results of figure 5 correspond to white
noise (Le. the power spectral density is the same for all frequencies). Other noise sources, such as
mains frequency interference, may have markedly different effects.

5.4. IMPEDANCE SPEC1RA - THE EFFECT OF VOLTAGE QUANTISATION.

Typical data acquisition systems using moderately-priced analog-to-digital converters (ADCs) will
measure to a resolution of 8 to 16 bits (Le. 1 part in 256 to 1 part in 65536). Figure 6 shows the
effect of quantisation of the voltage measurement corresponding to the use of 8 and 12 bit ADCs
and assuming a full scale range of 10 mY. Providing the range of measurement of the ADC is
closely matched to the range of the signal to be measured, a 12 bit ADC can give adequate
performance. An 8 bit conversion will almost inevitably give a rather noisy spectrum, although the
general features of the spectrum can be detected.

5.5. THE APPLICATION OF NUMERICAL MODELS TO IMPEDANCE STUDIES.

This work has been performed with a view to the validation of a particular interpretation of
linear polarization resistance measurements. If it is required to compute the response of a system
which is assumed to be linear, it is probably more efficient to use the Laplace transform method
for the analysis (as discussed by MacDonald in this volume), or, for repetitive waveforms, to
transform from the known solution in the frequency domain. However, these methods do not
 131

12 bit (2.5 uV) quantiza1ion 80 a bit (40 uV) qu8ntiza~on 80

1000 1000
~ ~

100 100
~ ~

10 20 10
20

1 l=CCUD=-J-LLUm~_=''Cill"~'~,-"-' w"'Cill'~,--'-'~,ll'Cill''''--C~UllJ 0 1~~~lliL~Wlli~~=--LlWill~~lliLML~0

~ ~ ~ 1 W ~ 1~ 0.001 0.01 0.1 1 10 100 1000
Frequency (Hz) Frequency (Hz)

Figure 6. Effect of quantization on measured impedance

readily accommodate non-linear responses, and for more complex systems, such as systems in
which the diffusion wave is modified by reactions in solution, analysis of transients by means of a
generalized numerical model may provide a useful method of deriving the impedance spectrum.
This approach is currently being examined for a number of systems.

5.6. THE INTERPRET AnON OF POLARIZAnON RESISTANCE MEASUREMENTS.

Before considering the way in which polarization resistance measurements can be related to the
impedance spectrum, it is useful to examine the component of the impedance spectrum which
should be taken to determine the corrosion rate. For the simple system examined in this work, the
response will include the charge transfer impedance, which is seen at higher frequencies, and the
Warburg impedance, which dominates the response at low frequencies. The polarization
resistance, Rp , may be defined as the slope of the curve of potential versus current measured with a
sufficiently slow sweep rate that the system remains in essentially steady-state conditions, or, in
impedance terms, the impedance at zero frequency. Thus Rp includes both the charge transfer
impedance and the Warburg impedance, together with any solution resistance. However, for the
model considered in this work the Warburg impedance increases as Ide increases, and Rp is not
linearly related to Ide. On the other hand Rei is related to Ide through the relationship:

2 R T
Ret

n FIde
Thus Ret should, at least for this system, be used to determine Ide (or, effectively, the corrosion
rate). For systems in which the impedance response consists of a charge transfer resistance
together with a Warburg impedance due to the diffusion of an anodic product away from the
surface, the Warburg impedance is related to the rate of reaction (since it is related to the flux of
the product away from the interface), and Rp will be approximately inversely proportional to the
rate of reaction. However, as indicated above, the existence of such a Warburg impedance implies
that thc anodic back reaction is significant, and the calculated rate of reaction will not be an
accurate measure of the corrosion rate.
132

Phase (degrees)
1~~Am~p~li=ruoo~(Oh~m=s~)------------------------~~,
80
1400

1200
60
1000

800
40
600

400 ~-----~~~~---
.' - .'
.. .' . ....... . 20

200
oL-____- L______L -_ _ _ _ ~ _ _ _ _ _ _~_ _ _ _~
0
o 2 3 4 5

Frequency (Hz)

Figure 7. Bode plot using linear scales

While the polarization resistance is nominally measured under steady-state conditions, in

practice the measurement will usually be made at a low enough frequency that the behaviour
appears to have attained steady-state by examination of the potential-time or current-time record.
Whether or not this will include any Warburg impedance will depend on the time required for the
diffusion process to approach steady-state. For the case of steel in concrete, where the diffusion
path is very long, there is no possibility of attaining equilibrium. Thus the measured 'polarization
resistance' will depend on the period selected for the measurement. At first sight this appears to be
a rather arbitrary decision, but careful consideration of the relationship between impedance and
frequency suggests that there is a tendency to select a measurement period which will give a
measured polarization resistance which relates reasonably well to the charge transfer resistance.
Roughly-speaking the impedance measured with a polarization resistance technique will be equal
to the amplitude of the impedance at a frequency corresponding to the cyclic frequency of
measurement. The Bode diagrams of figures 4 to 6 tend to mask the variation in the impedance
that will be measured as a function of time, since the log-log scales give a misleading impression
of the relative time intervals. On linear scales (figure 7) it becomes clearer that there is a large
range of cycle periods where Rp will be determined as approximately Ret (ignoring solution
resistance effects). This can also be seen in the hysteresis diagrams of figure 8, in which the
potential for a triangular current waveform is plotted as a function of current for a range of cycle
periods. While the diagram for a 10 second period shows a continuous curvature, and would
probably not be taken as indicative of having attained steady-state, the curve for a 1 second period
does show a reasonably linear region in the second half of each segment, and could easily be
interpreted as the system having reached steady-state. The slope of the straight region of this curve
gives a 'polarization resistance' of about 500 ohm, compared with a value of 347 ohm for Ret, and
about 1400 ohm for Rp. Careful examination reveals that even the 'true' steady-state curve
obtained with a 1000 second period shows a significant curvature, due to the inherent non-linearity
of the electrochemical system.
 133

25~po=~=n=~~(=mV~)________________________________- ,

1000 sec
20 100 sec

L-----c> 10 sec
15 "L---:::::'::::--~ 1 sec

10

51"--------~--
0.065 0.07 0.075 0.08
Current (mA)

Figure 8. Ell hysteresis diagrams

6. References

1. Wegner, F. and Galland, l. (1989) 'Study of corrosion of steel in concrete by electrochemical

impedance measurements', Materials Science Forum, 44/45, 375-386.
2. Gonzalez, l.A., Molina, A., Escudero, M.L. and Andrade, C. (1985) 'Errors in the
electrochemical evaluation of very small corrosion rates - I. Polarization resistance method
applied to corrosion of steel in concrete', Corrosion Science, 25, 917-930.
3. Newton, C.J. and Sykes, I.M. (1988) 'A galvanostatic pulse technique for investigation of steel
corrosion in concrete' Corrosion Science 28, 1051-1074.
4. Castro, E.B., Real, S.G., Saidman, S.B., Vilche, J.R. and Milocco, R.H. (1989) 'Identification
methods in dynamic studies of metallic corrosion processes', Materials Science Forum
44/45,417-432.
INVESTIGATION OF METALLIC CORROSION BY
ELECTROCHEMICAL NOISE TECHNIQUES

C. GABRIELLI, F. HUET and M. KEDDAM

LP15 du CNRS
Physique des Liquides et Electrochimie
Laboratoire de I'Universite Pierre et Marie Curie
Tour 22, 4 place lussieu
75252 Paris Cedex 05 - FRANCE

A B S T R ACT. Many semi-macroscopic phenomena related to corrosion either

localized or uniform are stochastic by nature. This character leads to an
observed random behaviour : fluctuations of the free potential, fluctuations of
the current when a constant potential is maintained. This random signal is
similar to an electric noise and can be analyzed either in the time domain on a
statistical basis or in the frequency domain by measuring the power spectral
density of the electrochemical noise. Some applications in the field of localized
corrosion (pitting, abrasion, stress corrosion cracking) as well as uniform
corrosion are reported in order to see what kind of parameters are accessible
and how they can be related to the corrosion phenomena.

1. Introduction
2. Random aspects of the corrosion phenomena
3. Mathematical background
3.1. FIRST-ORDER PROBABILITY LAWS OF DIFFERENT POINT PROCESSES
3.1.1. Poisson point process
3.1.2. Past-dependent point process
3.2. POWER SPECTRUM OF STOCHASTIC PULSES SEQUENCES
4. Experimental techniques
4.1. STATISTICAL COUNTING
4.2. TIME-RECORDING
4.3. POWER SPECTRAL DENSITY MEASUREMENT
4.3.1. Principle of the measurement
4.3.2. Experimental arrangement
5. Applications
5.1. PITTING CORROSION
5.1.1. Stable pitting
5.1.2. Extreme value prediction of maximum pits
5.1.3. Pre-pitting stage
5.2. CORROSION BY ABRASION
5.3. STRESS CORROSION CRACKING
5.4. UNIFORM CORROSION

135
M. G. S. Ferreira and C. A. Melendres (eds.),
Electrochemical and Optical Techniquesfor the Study and Monitoring of Metallic Corrosion, 135-190.
© 1991 Kluwer Academic Publishers.
136

1. Introduction

The corrosion of the metals is an electrochemical phenomenon. Since no

electrochemical method directly measures the corrosion rate, the use of the
latter is based upon a certain number of hypotheses (e.g. the estimation of the
corrosion rate from impedance measurements is generally based on a model of
the reaction mechanism). Knowledge about the corrosion mechanism is
therefore often the prime necessity in order to verify the validity of the
hypotheses adopted. The metal dissolves through a charge transfer at the
electrochemical interface which occurs at the end of a succession of more or
less coupled elementary phenomena :
- transport of reactive species in the bulk of the solution or through a layer
often associated with chemical reactions in the bulk phase.
- adsorption of the reactive species on the electrode.
- electrochemical and chemical interfacial reactions.
- transport of dissolved species from the corroded area.
Some techniques able to characterize the state of the surface or the
adsorbed species on the interface necessitate the use of vacuun techniques
(LEED, Auger... ), therefore they cannot be employed for an "in-situ" analysis
of the corrosion phenomena. Other techniques using electromagnetic waves
(optics : ellipsometry or X-rays : EXAFS) are beginning to be used in the study
of corrosion but often they can hardly be applied as an alteration due to the
dissolution of the surface occurs. Hence electrical methods are often the only
possible recourse for "in-situ" investigations.
In this framework the steady state current-voltage plot, the impedance
measurement and the noise analysis form a hierarchy of investigation
techniques which give more and more detailed informations. In spite of its
relative complexity the last technique is now often carried out because on the
one hand the steady state plots give informations only on the rate
determining step and on the other hand the impedance gives only averaged
informations over the space and the time. However as corrosion is controlled
by many random parameters, this is far to be sufficient. In opposition,
lectrochemical noise techniques, resolved in time and/or in frequency, allow
the very random nature in space and time of the controlling parameters to be
investigated.
In addition, to the experimental point of view, many semi-macroscopic
phenomena related to corrosion are random by nature : pitting, abrasion, gas
evolution... This special property of the processes makes the classical
deterministic techniques (d-c techniques for I-E plotting or a-c techniques
for impedance measurements) difficult to perform not to say impossible. Even
when the processes seem to be in a steady state regime (e.g. uniform
corrosion) non-linearity problems often remain. This point constrains to use
a perturbing signal of very low amplitude which is hard to detect and to
measure.
If corrosion is regarded as a stochastic process, the time evolution of the
corrosion current (potential) at a given potential (current) is a random
signal: hence it is commonly called electrochemical noise. Therefore time
and frequency processing techniques specially devised for random signals
can be applied.
In the case where the elementary current transients due to the random
events (pit nucleation and growth, bubble growth and departure ... ) are not too
 137

numerous to be discernable, a statistical counting can be done on the time

series displayed by the current in orde:r to evaluate a mean appearance rate
(e.g. pitting rate, evolution rate ... ). However in the case, closer to the cases of
practical interest, where many events occur at the same time so that their
current transients overlap, only a spectral analysis of the random signal will
allow the characteristic parameters of the process to be attained.
Moreover in the case of corrosion where an exciting signal can
sometimes lead to an alteration of the process studied, a simple "listening" of
the metal-electrolyte interface obtained by analyzing the natural fluctuations
of the corrosion potential or of the corrosion current allow the process to be
investigated or the corrosion rate to be estimated.

2. Random aspect of the corrosion phenomena

In the last two decades corrosion has been recognized as a phenomenon

largely controlled by random processes. Although it seems that it was as early
as 1933 that Evans, Mears and Queneau [1] first pointed out the importance of
probability concepts in corrosion phenomena, it was only in the early
seventies that Nathan and Dulaney [2] and Evans [3] remarked that the
statistical approaches have a great practical importance. At the end of the
seventies the Japanese school have given several papers where these
concepts have been systematically applied in the case of localized corrosion
[4-6].
From the microstructural point of view, metals are generally
polycristalline materials, consisting of grains of different forms and
orientation, on the boundaries of which various chemical elements are
concentrated. In addition cracks, inclusions, voids... occur within metals.
These microstructural features cause metals to be susceptible to random
vanatlOns in physical properties and, consequently, in their corrosion
behaviour. Likewise, local variations in flow conditions, temperature, pH, etc ...
are some of the environmental parameters to be considered. Thus the
inherent randomness of the corrosion system may be viewed from the high
degree of complexity of the metal-protection-environment combination that
can produce significant variations in : (I) the active corrosion driving-forces.
(2) the resistance to corrosion and, consequently, (3) the local corrosion rate
even though the nominal metal-protection-environment parameters appear
uniform.
Corrosion is a material degradation process essentially consisting of a
metal surface exposed to a reactive environment that can be divided into two
main classes : general and local. These classes lead to respectively uniform and
localized corrosions. In the first class when the metal dissolution
simultaneously takes place with hydrogen production (case of iron in acidic
medium) the gas evolution gives rise to a random behaviour related to
nucleation growth and detachment sequence of the hydrogen bubbles. In the
second class different modes may be identified. These are dependent of the
type of metal undergoing corrosion and its environment at the time of attack.
One of the most destructive form is pitting corrosion which is characterized
by the presence of a number of small pits on the exposed metal surface.
Mechanical stresses in metals can also encourage corrosion. Protective films
such as passivating films are susceptible to damage when subjected to tensile
138

TRANSIENT TYPE (·1ET AL / EXP. REF.

I
FE 37

38

STAINLESS STEEL 32

][ CHEMICAL BP-EAKDOWN AL 33)39

MECHANICAL BP-EAKDOWN
I\BRASIoN 43)44
SCRATCHING 50
LASEP- ILLUM. 43)44

TABLE 1
 139

stresses or to impacts by particles in flowing corrodents leading to stress

corrosion cracking or impingement/abrasion attack [7].
Several steps can be separated (1) the initial stage which involves such
phenomena as passive film formation, film rupture under mechanical
stresses, film penetration by aggressive anions... (2) early pit growth,
involving processes leading either to the establishment of a self-perpetuating
corrosion cell or to the repassivation of the occluded cell ; and (3) the
propagation of the cavity involving mechanisms of ion transport in
electrolytic solution.
Localized corrosion is generally quite unpredictable with regard to the
time of initiation and seemingly often the place of attack and this lead to a
random behaviour for the observer.
In the case of localized corrosion this multiplicity of random
parameters gives rise to potential or current transients where the pits
nucleation, growth and death sequence occurs in the pre-pitting stage when
stable pitting is not established yet.
Analyses of the time records published in the literature show that two
types of current transients are reported (Table 1). The type I transient is
distinguished by a slow growth which is followed by a sudden death of the pit.
It seems to be characteristic of iron and iron base alloys. The type II transient
is distinguised by a sudden birth which is followed by a slow death of the pit. It
seems to be characteristic of aluminum and non chemical initiation of the
corrosion (abrasion, scratching, laser illumination ... ). The shape of the type I
transient has been interpreted by a sudden decrease of the corrosion rate in
the pit due to a commutation between a multiplicity of the possible states of
dissolution of iron and iron base alloys [8,9]. This transition between active
and passive states would be possible only during the etching state when the
mass transport limitation is not sufficient to lead to brightening. As Sato [10)
has related the limit of the stability of dissolution of a pit to the etching-
brightening transition (for a pit size of a few ~m), stable pits can grow as soon
as brightening occurs, which would explain the critical times found in the
experiments.
It seems that type II transients are related to a sudden failure of the
protective layer (mechanical breakdown, laser illumination or cracking site
on AI) followed by a repassivation process. However it has been shown in
some cases that the repassivation rate is limited by the time constant ReC of
the electrode (where Re and C are respectively the electrolyte resistance and
the double layer capacity of the electrode) and therefore that this rate is
inaccessible, at least for the experimental conditions investigated.
From the theoretical point of view, MacDonald et al. [11,12] proposed a
pitting corrosion theory based on their point defect model of the passive film
breakdown. Normal and log-normal distributions of the pitting potential and
of the induction time are reasonably accounted for by assuming a distribution
function for the diffusivity of the cation vacancies across the film attributed
to local defects. However this entirely deterministic approach does not model
the time frequency behaviour of the passive current fluctuations generated
by the stochastic process of nucleation, growth and repassivation of the
pitting events. On the contrary, from a statistical mechanical approach of a
passive metal dissolution in solutions containing halide ions Okada [13,14) has
proposed a microscopic model of pitting based on the concept of a transitional
140

halide complex which forms before lattice cation dissolution. The rate of the
passive film dissolution has been derived. using the atomistic parameters
relevant to the system. The electrochemical noise generated during the pre-
pitting stage has been predicted by consideration of the competitive
adsorption between the halide and hydroxide ions around the lattice cation on
the passive film surface. However so far there is no comparison between this
theoretical derivation and experimental results.
As an example pitting susceptibility of metals in aggressive media is
classicaJIy examined by measuring the pitting potential which is obtained by
a potentiostatic method and/or the induction time determined under a
potentiostatic condition. When using these techniques. even when careful
experimental procedures are used. wide scatter in the data is commonly
encountered.
This fact confirms that random variation of the experimental data is
intrinsic to pitting corrosion and must be analyzed from a statistical point of
view. In this paper the main techniques employed : statistical counting and
spectral analysis. will be reviewed in order to see what parameters are
accessible and how they can be related to the corrosion processes. As these
techniques are generally based on stochastic models. the report on the
experimental results already obtained in this area wiJI be preceded by a short
mathematical background in order to outline the main properties of the
stochastic point processes which are of basic importance in this topic.

3. Mathematical background

During corrosion. as in many physicochemical processes. the main aspect of

interest is the repeated occurrence of some specific phenomenon (i.e.
nucleation of a gas bubble. nucleation of a pit ... ) at randomly distributed time
instants which give birth to current or potential transients like those
described in Table I (Fig.la). The random stream of successive occurrence of
such point event at time ti is called a point process [15.16] (Fig.lb). The main
object of both the experiment and the model is to characterize its probability
law which is usually called the nucleation law of the corrosion process under
investigation. To simplify. we avoid all pathological cases there is neither
explosion. nor extinction of the point process. nor point event occurring at
predetermined times.
The simplest means to define the probability law is to associate with
each point process a stochastic process N(t) which counts the number of point
events in the time interval (o.t) : it is the counting process (e.g. the number of
pits occurring between 0 and t) (Fig.lc). For a small time interval dt. the
number of point events occurrences dN(t) is supposed to take only the value I
or 0 ; this means that the probability law of the point process is completely
characterized by the knowledge (for k = 1.2.3 ... ) of the probabilities P[dN(q)= I,
dN(t2)=I ..... dN(tk) :: 1] that one point event falls in each of the small intervals
(tl.lt+dtl). (t2. t2+d t2), .... (tk.tk+dtk).
Most of the works reviewed in the literature about stochastic processes
in corrosion research take into account only the first order probability law of
the point process that means quantities depending on one time t component
such as P[dN(t) = 1]. This will be the case in the first part of this section. For a
few years spectral analysis has been used in corrosion research when the
 141

corrosion current is the superposition of elementary current transients born

at times defining a point process ; the power spectral density of the total
current. calculated from the second-order probability law P[dN(t}) = 1.
dN(t2)=1] of the point process. is given in the second part of this section for
different cases encountered in the corrosion field.

I (t)
a

t, t2 t 3 t4 ts t6
t~ b
0 t, ~t3 t4 t5 t6
Xl x1 x3 Xs d
x2 x4 X6

--
~
6
5
~
Z 3
2
1
0
t1 t2t 3 t..; ts t6
TIME

Figure 1. Time diagrams of a point process.

(a) random events occurring at random instants ; (b) occurrence times tk ;
(c) counting process N(t) ; (d) interval sequence x(t).
142

3.1. FIRST-ORDER PROBABILITY LAWS OF DIFFERENT POINT PROCESSES

The reader who wants to go more deeply in this subject can read Larson &
Schubert Vol.II Chapter 7 [15], which is a clear and recent review on point
processes.
Generally speaking the conditional probability that a point event (e.g.
nucleation of a pit) occurs in the small time interval (t,t+dt), knowing the past
of the process N(t') t' ::;; t, i.e. the number of occurrence of the point process up
to time t, can be written :

P[dN(t) =I I N(t') , l' ::;; t] = A(t) dt (1)

and as there is no more than one point event in a small time interval

P[dN(t) = 0 I N(t'), t' ::;; t] = 1 - A(t) dt (2)

So, the conditional expectation E[dN(t) I N(t'), t' ::;; t], which is the
average of the random increment dN(t) of the counting process (knowing the
past of the process before time t over numerous experiments, is :

I
E [dN(t) I N(t'), l' ::;; t] L. x P[dN(t) = x I N(t'), l' ::;; t] A(t) dt (3)
x=Q

where A (t) is the intensity of the point process, that is the number of point
events per time unit at time t.
For a point process another quantity of interest is the interarrival times
or waiting times sequence (e.g. time ellapsed between the nucleations of two
bubbles or two pits) (Fig.ld)

Xn = tn+l - tn (4)

The probability law of Xn, which is easily defined for some special point
processes, completely characterizes the point process. Unfortunately it is
often difficult to relate the probability law of the time intervals Xn and the one
of the counting process N(t).
In the same way the survival probability (e.g. the probability that the
specimen remains unpitted up to time t) can be defined. If

pet) = P [no occurrence of a point event in o,t] (5)

and - pet) = P [occurrence of a point event in o,t] (6)

Hence

P [occurrence of a point process in t,t+dt I no occurrence of a point process

in o,t]
P [occurrence of a point process in t,t+dt and no occurrence of a point process
in o,t] I P [no occurrence of a point process in o,t]
 143

From eq(1) for the p.h.s. and from eq(6) from the r.h.s., one gets :

d(1 - P(t)
A(t) dt (7)
pet)

~ (8)
pet)

t
i.e. P[N(t) - N(t o) = 0 I N(t'),O < l' < to1 exp (- I A (t) d t) (9)
to

These formula have been used by Shibata [6,19] in a more restricted case
(Markov processes).
Generally the present state of the process, in particular its intensity
A (t), is dependent upon its past evolution. However there is a basic process
where the past behaviour does not influence the future of the process : this is
the Poisson process.

3.1.1. Poisson point process. In this case A (t) is independent of N(t'), t' :S t,
hence the knowledge of the counting process N(t') up to time t brings no
information on its future evolution in the time interval (t, t+dt). In other
words this leads to the characteristic property of a Poisson point process : the
increments of the counting process N(t2)-N(tI) and N(t3)-N(t2) for the time
instants tl < t2 < t3 are independent random variables.
Equations (1 to 3) lead to :

E[dN(t)] P[dN(t) = 1] A(t) dt

P[dN(t) = 0] 1 - A(t) dt (10)

The process intensity A (t) is a deterministic function of time

(inhomogeneous Poisson process). If the point process is stationnary, it is
called an homogeneous Poisson process with intensity A (t) = A.. As it is the
simplest point process, the Poisson process plays a prominent role among
point processes and is often used in corrosion modelling.
If m(t) is the expectation of the counting process N(t), i.e. :

t
met) E[N(t)] I A(u) du (11)
o

The probability law of the counting process is given by the Poisson law :

for tl < t2 and n = 0,1,2 ...

144

One of the chief properties of an homogeneous Poisson process is its

interval property : the sequence of the waiting times Xn is a sequence of
independent random variables with the same distribution function :

P [ Xn < x 1 = 1 - e- h (13 )

where A. is the intensity of the Poisson process. The probability density of the
waiting time Xn is then :

d
UI(X) = dx P[xn<xl (14)

However it must be noticed that for an inhomogeneous Poisson process the

waiting times are no longer independent.
Although an evolution without influence of the past on the future is
often a useful assumption for modelling an impulsive random noise, there are
other random phenomena of an impulsive character where such an
assumption would not be realistic. In these cases past-dependent evolution has
to be introduced into the specification of the point process.

3.1.2. Past-dependent point process. For a past-dependent point process

(terminology used by Larson & Schubert [15], the intensity A (t) depends on
the past N(t'), t' $ t, i.e. depends on the number N(t) of point events occurred
up to time t, and the time instants of occurrence of these point events tl, t2,
teNet»~ (e.g. the nucleation law of a pitting process depends on the number of
pits on the specimen). A few special case of past-dependent point processes of
interest in corrosion are now briefly surveyed.

3.1.2.1. Birth and death processes

Let us consider an electrode surface with n pits ; if the birth of a new pit is
independent of n and the death of one of the n pits is proportional to n, the
probability law to have n pits at time t, P(t,n), i.e. the probabilities of this
process (called "immigration-death" process) is given by the following
differential-difference equation

d P(t,n)
- (A. + I1n) P(t,n) + "'P(t,n-I) + l1(n+ 1) . P(t,n+ 1) (15)
dt

where '" and I1n are the transition probabilities respectively of the birth and
the death of a pit.
The probability generating function, defined by :

+00
G(z,t) L P(t,i) zi (16)
i=O

which allows to calculate the probability P(t,i) through the relation

 145

(17)

obeys the partial differential equation [17] :

at
~
(z,t) == A. (z - 1) G(z,t) + ~(1 - z)
.3G
at (z,t) (18)

The solution of this equation is

A.
G(z,t) == [1 + (z-l) e-J.Lt] no exp [~ (z-l) (1-exp(-~t»] (19)

for no pits present on the electrode surface at time t == O.

3.1.2.2. Doubly stochastic point process

In this case an inhomogeneous Poisson process is assumed to have an
intensity function A (t) which is a continuous time stochastic process, i.e. for
consecutive period of time the intensity function is a random variable. This
process is defined at least up to the second order, by its mean E(A) and its
correlation function or power spectral density '¥ A(t).

3.2. POWER SPECfRUM OF STOCHASTIC PULSES SEQUENCES [

During localized corrosion sequences of elementary current transients are

often observed (Fig.2). These sequences can be modelled by assuming the
occurrence of the elementary current transient fk (t) at time tk which form a
point process (Fig.2a). Therefore the overall observed current I(t) can be
written (Fig.2b) :

N(t)
I(t) L fk (t - tbUk) (20)
k==1

where N(t) is the counting process associated with the point process ; the
elementary transient current can depend on another random parameter Uk
(amplitude, time constant, life time ... ).
The total current I(t), which is supposed to be stationnary, can be
characterized, at least up to the second-order, by its mean value <I> and by the
power spectral density (p.s.d.) '¥ j(f) of the current fluctuations i(t) arount the
current mean value : i(t) == I(t) - <I>. This p.s.d. is the Fourier transform of the
correlation function <i(t) i(tH», i.e. (with j2 == -I) :

+00

f e- 2jnf'i; <i(t) i(t+t» dt (21 )

This p.s.d. is defined for frequencies f between -00 and +00 whereas in
measurements only positive frequencies are considered : the right hand-side
146

of eq(21) has to be multiplied by a factor of 2 in order to define the physical

p.s.d. 'I' i (f) in the frequency range (0.+00).
The stochastic process defined by

N(t)
~(t) L O(t-tk) (22)
k=l
is called the random density function and plays a basic role in the theory of
point processes, because it is only characteristic of the sequence of arrival
times tk of the considered point process ; in this equation oCt) is the Dirac
distribution. ~(t) is schematically represented in Fig.lc.

~(t)~--------------------------------~
a

b
I(t)

~ (t.) c

TIME

Figure 2. Decomposition of the total current I(t) in elementary transients.

(a) elementary transients !k(t) occurring at time tk ; (b) total current I(t) ;
(c) random density function /;(t) of the point process.
 147

In the same manner as before. the random function ~ (t) is defined. at least up
to the second-order. by its mean value <~> (mean number of point events per
time unit) and the p.s.d. 'P ~(f) of the fluctuations of ~(t) around <~>. This p.s.d.
has been calculated for different point processes from the second-order
probability law of the point process P[dN(tl)= 1. dN(t2)= 1].
The power spectral density 'Pi(f) (-00 < f < +00) of the fluctuatio.ns of the
total current is now given for the simplified hypothesis where the elementary
transient f(t) is deterministic.
In this case the current I(t) can be expressed from eq(20) by :

N(t)
let) I. f (t-tk) (23 )
k=1

In addition. the current I(t) can be written under the form

I(t) = f(t) of< ~(t)

where of< is the convolution product and [;(t) is given by eq(22).

The mean of the global current I(t). called a filtered point process. and
the p.s.d. of the current fluctuations around <I>. are given by :

<I> = F(O) <~>

(24 )

where F(f) is the Fourier transform of the elementary current transient ret)

+00

F(f) f f(t) e- 2 j7tft dt

If the mean intensity E(A) of the point process, i.e. the mean number of
point events (i.e. pits) per time units <~> is denoted A. the p.s.d. 'P ~ (f) of the
fluctuations of ~ around <~> can be calculated for different point processes
involved in localized corrosion modelling.

- Poisson process : 'P~(f) = A (25)

In this case it can be shown that when the current transient f(t) shows a
sudden birth or a sudden death the high frequency limit of the p.s.d. 'P i (f)
varies like 1/f2 whatever the shape of the transient. A slow birth and a slow
death give rise to a 1/£4 (or 1/[6 ... ) trend.

- Doubly stochastic process [18] : 'P~(f) = A + 'P A(f) (26)

where 'P A (f) is the p.s.d. of the fluctuations of the process intensity A (t)
around its mean value A.
148

- Renewal process [17J : ..tifL

'I'1;(t) == A [1 + 2Re (1-9(f) J (27)
where Re means real part and 9 (f) is the characteristic function of the
waiting times :

+00
9(t) < e2j1tfx > f e j1tfx
2 UI (x) dx (28)
o

1
The mean value of the waiting times is then : <x>==r
The power spectral density 'I' j(t) of the fluctuations of the total current or of
the voltage can be also calculated in some cases where the elementary
transients fk (t) are random functions.

Examples
Two examples of interest in the field of corrosion are given below in order to
illustrate the derivation of the power spectral density : exponential and linear
transients in the case of a Poisson nucleation law.

a) Exponential transient
The current transient is a decreasing exponential having a time constant 1:

f(t) = A e-t/'t for t 2': 0

(29)
f(t) 0 for t <0

The total current is then

N(t)
let) = L
k==1
f (t-tk) (30)

Then

+00
<I> 1" f
-00
f(t) dt (31)

1"A't (32)

A't is the quantity of electricity involved in one transient. Moreover from

eq(24) :

21" 1 F(t) 12

(33 )
 149

This power spectral density (Fig.3) has a white noise plateau in the low
frequency range :

(34)

and decreases in l/f2 in the high frequency range. above the cut-off
frequency

1
fc = - (35)
21t't

b) Linear transient
The elementary voltage transient can be approximated by a linear function of
time such as :

r(t-s.u) =I if t-s < u

v(t-s.u) k.(t-s) ['(t-s) where (36)
r(t-s.u) = 0 if t-s ~ u

where s is the random birth time of the transient which is assumed to follow a
Poisson law N(t) of intensity A and u its life time which is assumed to be
exponentially distributed with parameter a. This is an example where the
elementary transient is a random function.
The total voltage is in this case equal to :

N(t)
Vet) = L V(t-tk.Uk) (37)
k=1

Hence

and (38)
 150

I I ,-
.' I

-.......

'"
.:;:;:
~10-1
a.:
["',AI ,
.. "'"
Cl

A~
2
~1O- c--

~
..J 0 t
<i TIME
L1Q- 3
ec:: ~
o
z 1;
I I l

10
27C f"'t

Figure 3. Power spectral density calculated from eq(33) for the

exponential transient shown in the inset.
10 :r-r

--.... ,

""'"
~
:-- 10-1
Cl
vi
a.:
LJ ..
r(t/U)
o 10- 2 :---- -
W

~
N 0 U t -
..J TIME
3
<{ 10-
L
ec::
o ,,
'\
z 10- ~ ,
--L .J..

Figure 4. Power spectral density calculated from eq(38) for the linear
transient shown in the inset.
 151

4. Experimental techniques

The time series displayed by the current observed during corrosion can be
analyzed either in the time domain or in the frequency domain following if
the transients are discernible or not i.e. following the nucleation rate of the
events. This alternative gives rise to two main types of experimental
techniques. In the case of growing stable pits a statistical counting can only
be performed as spectral analysis requires stationarity. In the case of random
current transients related to the birth and death of pits nuclei (pre-pitting
stage) a time-recording and a spectral analysis of the random transients can
be carried out if the pits are not too numerous. If there are too many pits only
a spectral analysis can be performed.
The regulating device (ultra low noise potentiostat or galvanostat) is
equipped with two amplifiers for measuring current and potential signals
with d.c. polarization rejection. Then either the random current or the
random potential output signals of these two amplifiers can be connected to
the plotter or to the analyzer.

4.1. STATISTICAL COUNTING

Two types of measurements are typically carried out : either the potential of
the electrode is swept and the pitting potential at which the current flowing
through the electrode exceeds a given threshold, is registred, or the electrode
is potentiostatically polarized and the incubation time necessary for the
growing current to equal the current threshold is recorded. This current
threshold is chosen is order to correspond to one pit growing on the electrode
surface.

In order to make the experiment easier, a multichannel pitting

corrosion testing (M.P.T.) apparatus has been devised (Fig.S) which allows
about a dozen. or more. of samples to be tested in the same batch in a potential
control regime. Each sample circuit is switched off when its current is equal
to the chosen threshold and the timer registers the time.
In the case of a sweeping potential when a pit is nucleated and grows on
one sample the current increases rapidly and when it reaches the threshold
value the circuit is switched off and the total observed current falls back to
the total passive current of the whole remaining unpitted samples (Fig.6).
If the experiment is performed N times, a potential distribution :

El ..... Ej .... EN

can be obtained. Hence for each potential E, the number of samples pitted
before E, such as

(39)

allows the distribution function of Prob[E :5 Ed of the pitting potential to be

evaluated by :
Q(E) = Prob[E:5 Ed (40)
N
152

1---------- ------ --...,

1 I
I
Coun~er elec~rode
I

Timer

Figure 5. Scheme of the multichannel pitting corrosion testing apparatus.

The survival probability. that is the probability for an electrode to

remain unpitted at potential E, IS then calculated by :

N-i
P(E) = 1 - Q(E) (41 )
N

An other approach is to calculate the pitting probability at each time,

using the number j. of non-pitted samples. In this case the survival
probability is

(42)
 153

50 - 0.5

40 x
x /- 0.4
w
U
<! 30 0.3 (fj
:L / ......
..... x >.....
f-
Z 20
/ x
x
x
Xx

x 0.2 -1
w
0::
/ I-
<!
0:: / Z
0.1 W
::> 10
U f-
a
(L

0 . 0
/
/
-0.1
0 5 10 15
TIME I rnn

Figure 6. Evaluation of the pitting potential for 304 stainless steel in 3.5%
NaCl. Change in anodic current with increasing potential at a constant sweep
velocity of 20mVlmn. Preset current threshold is 201lA. The X at the top of
current peaks indicated that the circuit was switched off (from [19]).

In the case of a potentiostatic control of N samples with the same

apparatus the incubation times necessary for one sample to equal the current
threshold give a sequence of instants (Fig.7) :

tl, ... , ti, ... tN

which allows the histogram (Fig.7) and the distribution function Q(t) of the
incubation times to be evaluated
i
Q(t) = Prob[t ~ tj] - N (43 )
154

500
ro.1 V -O.36V
400 ~ " 1[1
<t
:1.. 300 *~ I
~
'OL-

II
I
..... IX
w II
I- OJ
L
:0 '\Il!
Z 200 z
w iI
0::
011
0::: 0

:::> 100
U
x x x
0

0 20
TIME I s

Figure 7. Evaluation of the induction time (same conditions as in Fig.6)

for pit generation at a constant potential of 0.36V. MPT on means
Multichannel Pitting Tester on. The X again indicate that the circuit was
opened. In the inset, induction times obtained for 72 pitted specimens at the
same potential (from [19]).

4.2. TIME-RECORDING

Concerning the time records of the current especially in the pre-plttIllg stage
in addition to a classical electrochemical arrangement (potentiostat), a
resistor can be used in the secondary electrode circuit and the current is
measured by means of a differential amplifier. The current can be logged
from a number of cells simultaneously by using a reed relay multiplexer and
an autoranging digital voltmeter. In this case the digitized data are sent to a
 155

microcomputer which takes care of the ensemble averages and other

numerical calculations performed on the time series.

4.3. POWER SPECTRAL DENSITY MEASUREMENT [20]

The characterization up to the second order of stationary random signals can

be done either in the time domain (correlation function of the signal) or in
the frequency domain (power spectral density or psd of the signal which is
the Fourier transform of the correlation function). In the frequency range
used in electrochemistry (l0-3-105Hz) the latter is much more accurate and
easier to interpret.

4.3.1. Principle of the measurement. As the parasitic noises coming from the
mains, the electromagnetic radiations, the electronics of the amplifiers and
the voltage control (potentiostat) can be, in some cases, as high as the useful
electrochemical noise, a special two channel measurement arrangement has
to be used (Fig. B).
The psd measurements are based on a cros-spectral analysis, carried out
on a Fourier analyzer (e.g. Hewlett Packard 545IC). The signals of interest X
and Y at the outputs of the two measurement channels are sampled by an
analog to digital converter at a sampling frequency Fs (0.2Hz to 100kHz)
during the acquisition time T. The samples are stored in memory blocks of size
M (64-4096). The analog to digital conversion is done in 12 bits (including one
for the sign), hence the input dynamics of the converter is 20 log 211 = 66dB
which is largely sufficient in general.
A wired FFf algorithm computes:

M-I
XT(mM) = 6t L x(n6t) exp (-27tj ~n) (44)
n=O

where j2 = -I, 6t = lIFs is the sampling period and 6f = lIT is the frequency
resolution. The estimation of the cross-spectrum of X and Y, '{IT xy(m6f), where
T is the acquisition time of an elementary spectrum, is derived through the
average of N elementary spectra :

I N
'{ITxy(mM) = N L (45)
i=1

where YT,i(m6f) and XT,i(m6f) are the ith concomitant samples of X and Y at
the frequency m6f and * means complex conjugate. The spectrum is obtained
for M/2 frequencies ranging between 0 and Fmax = Fs/2.
The measurement time, for N averages, is the sum of :
- the total acquisition time of the signals X and Y

NM
ta = NT = 2-F
max
(46)
156

- the computing time

NM
tc = 0.22 1024 (47)

as the calculation of an elementary cross-spectrum of 1024 points needs

220ms.
The errors arising from the elementary spectrum determination
onglOate in general from :
- quantification : this error is usually negligible due to the accuracy of the
analog-digital converter (12 bits) ;
- computing : this error is also negligible as the analyzer computes the
Fourier transfonn with 16-bit words (0.1 % error) and as the computation of
elementary spectra is carried out in double precision (32 bits) ;
- estimation : this last error, due to the finiteness of the measurement time NT,
is often preponderant and is such as

( 48)

where 11 is the signal to noise ratio.

4.3.2. Experimental arrangement. In order to diminish some of the

measurement errors special arrangements are used as depicted in Fig.8.
Firstly, the potentiostat (or the galvanostat) and the differential amplifiers (or
the amplifiers) are battery powered in order to eliminate the influence of the
mains. Secondly, the potentiostat, the electrochemical cell, the amplifiers and
the batteries are protected against radiated parasitic noise by a shielding cage.
Finally, a high-pass filter and a low-pass filter are inserted in each
measurement channel in order to eliminate the drift on the one hand and the
spectrum aliasing error on the other hand.
In practice, even after high-pass filtering, the measurement validity is
ascertained by a systematic monitoring of the signal stationarity during the
measurement by means of an oscilloscope.
On a practical point of view the power spectral density (p.s.d.) is
calculated through a Fast Fourier Transform (F.F. T.) algorithm. If the current
time series is sampled each 8 t the maxim urn frequency which can be analyzed
is :

1
f max = 28t (49)

If the F.F.T. is performed over M samples the lowest frequency which

can be analyzed is :

1
fmin = AUt (50)
 157

AMPLIFIER FILTER!>

]
Figure B. Experimental arrangements for noise spectral analysis.
(a) Potentiostatic mode .. (b) Galvanostatic mode.
nx. ny.' Voltage parasitic noises of the measurement channels X and Y with
gain G.
158

The dual measurement channels allows the parasItIc noises nx and ny

coming from the amplifiers, filters and shunt resistors R to be eliminated by
using a cross-spectrum analysis. As an example in potentiostatic mode
(Fig.8a) :

x = G(Ri + nx) and Y = G(Ri + ny) (51 )

the cross-spectrum is equal to

(52)

and is proportional to the current p.s.d. of interest as nx and ny are

independent from each other and independent from i(t). However this is valid
only for an infinite number of averages of elementary spectrum Xy* but the
experimentalist is able to average over a finite number N of elementary
spectrum and therefore the estimation error is :

< 1-
~
L - ...fN (53 )

for a perfect signal to noise ratio (T]~oo in eq(ll».

To obtain a reasonable accuracy between 10 and 100 averages are
necessary which can lead to very long measurement times as this time is equal
to MN.1.t (e.g. to carry out a measurement down to O.OIHz needs 10000s to have a
10% accuracy). Given the poor steady stateness of the electrochemical systems
the spectral analysis is quite difficult to perform down to very low
frequencies.

S. Applications

Experimental results have been obtained by electrochemical noise techniques

in several cases. We shall successively examine applications dealing with
pitting corrosion, stress corrosion cracking, corrosion by abrasion and at last
uniform corrosion.

5.1. PITTING CORROSION

Two main steps of the pitting corrosion phenomemon will be successively

ex amplified : namely the stable pitting stage where the cavity grows without
end and the pre-pitting stage where nucleation and early pit growth and
death can be hardly distinguished and give rise to a unstable pitting which
can be analyzed by ensemble averaging or spectral analysis.

5.1.1. Stable pitting. As this approach requires a large quantity of samples in

order to obtain statistical data, a multichannel pitting corrosion device which
allows multiples of 12 samples (e.g. often 36 or 72) is generally used.
Shibata and Takeyama [19] were among the pionneers of this technique.
The distributions of the pitting potential for types 304 and 316 stainless steel
were plotted in Fig.9 on a normal probability paper by using the technique
 159

described in Fig.6. The observed straight lines show that the random variation
of the pitting potential about its average value obeys a normal distribution. As
the slope of the line is inversely proportional to the standard deviation some
informations about the experimental conditions can be obtained : e.g. for 304
stainless steel polishing with a fine gra.de emery paper leads to less dispersion
than other surface preparation. The dependence of the pitting potential on
the potential sweep velocity is also shown in Fig.9. It is clearly seen that the
higher the sweep rate, the higher the pitting potential.

99.99.-_,----,_---,_-,-_-,_---,_-.--_-.

99.9

99 480

95
I
~ 90
+ 60
~ 70
-;: 60
,50
>- 40_
t:: 30
~ 20
<:0
~ 10
o
a::
5
0...
240mV/m;n

0.1

O. 01 '---~'=_--'--;:l-::-----'----:'---L-~-...J
0.5
VIS. C.E.

Figure 9. Dependence of the pitting potential of type 304 stainless steel on

the potential sweep velocity. N is the total number of specimens examined and
n is the nth specimen on which pitting occurred (from [19]).
 160

'I
(a) IJI

>-
f- W
-.J f-
05 «
«
CIl
0::

0 Z
g:: 10.1 0

-.J
~
0:::
~ W
Z
:; /..3 W
0::: ~
:::>
CI)
10. 2 10. 2 ~

0 70

>-
f-
(b) w
:::J f-
iii «
«10· 1 60:::
CIl
0 Z
,
--
0::: 0 ~

ll.. f-
-.J
10-' 4«
0:::
...., IJI
I

W 0
~ Z
...-
:; 10.3 . 2 W
0::: ~
:::> f-
CI)
0 oll..
0 0.5 1.0 1.5
3
TIME ,10 5

Figure 10. Time evolution of the survival probability and pit generation
rate for (a) : 304 stainless steel (from [19J) and (b) iron (from [21]).

Plots of the survival probability pet) against the exposure time for
constant potential is given in Fig. lOa. The experimental survival probability is
here obtained from the distribution of the induction time (Fig.7). pet) is the
proportion of non·pitted specimens :

n
pet) = 1· I+N (54)

This figure shows that log pet) decreases linearly with time, the well defined
changes of slope with time define 3 pit generation rates, AI, A,2, A3 as from
eq(8) :
 161

d In pet)
A (t)
dt

The first one begins to operate after a short incubation time.

- The second one gives the highest rate.
And then, the third one finally occurs and seems to last up to infinity
therefore the pit generation would be stationary after some time.
These authors have found that the AiS are potential dependent :

Ao =0 for E < Ecrit

Al = al (E - ECrit>}
1.2 = 120 exp a2E for E > Ecrit (55)
1.3 = 130 exp a3E

If the first process is assumed to control the pit generation, the pitting
potential observed obeys the following relation against the potential sweep
velocity v

E pJ' t = (..L
a 1 )1/2 + EcrIl
. (56)

Linear dependence of the pit generation rate on the potential suggests

that the pitting process is controlled not by an electrochemical reaction but
by an electromechanical breakdown of the passive film. Shibata et al. [19]
think that the pitting failure of the passive metal can be due to the presence
of a precursor, or active state, in the film.
On a very large number of passive iron samples (2430 experiments
Heusler et al. [21] have observed that after some incubation time, the
formation of pits resulting from an increase of the local corrosion rate by
several orders of magnitude follows an increase of the chloride concentration
above a critical level while the electrode potential is kept constant. They
found that the probability density which is the derivative of the survival
probability pet) changes with time t in a nearly exponential manner
(Fig.tOc) :

~
dt = B exp (-At) (57)

where the pit generation rate A(t) is time-dependent. The shape of the
function A(t) is independent of the electrode potential and solution
composition and can be fitted to (Fig.IOd) :

A(t) :: [1 - exp (-t/,t]) ] [ a + b exp (-1/'t2) ] (58)

The four parameters a, b, 'tl, 't2, depend on the electrode potential in the
same way ; signifying the parameters as Ai the potential dependence is given
by :
162

(59)

An elementary pittIng probability per unit area 1t (E.P.P.) has been

introduced by Baroux et aI. [22-24].
If the surface S of the sample is decomposed in k independent small
elementary parts oS, the survival probability of the sample Ps associated with
S is ;

(60)

If the pitting probability is supposed to be proportional to the

elementary surface oS

1 - Ps 1toS
Ps (1 - 1tOS)k
and lim Ps exp (- 1tS) (61 )
oS-tO

Therefore ;

1t = --S1 log P (62)

where P is the survival probability.

On the experimental point of view the E.P.P. can be estimated from :

1
1t - - N log N (63)

where N is the total number of samples and the number of non-pitted

samples at time t.
From the elementary pitting probability, a "pseudo-pitting potential",
El, for which 1t = lcm- 2 , can be considered.
Especially for low NaCI concentration, the authors have found that the
solution must be carefully deaerated, both in the preparation vessels and in
the electrochemical cell, by using a continuous N-H flux. However the
agitation induced by this flux must be kept as small as possible, in order to
avoid an artificial increase of the pitting potential.
They found that 1t is potential dependent (Fig.Il) and for a
concentration c of chloride ions follows the law ;

E - E)
1t = c 3 exp---
q, (64 )

where q, = 17mV
and E) = E} * - 120 log (cv- 1/3 ) (65 )
 163

10
N
I
E
u

>-
I-
-1
OJ
<{
OJ
0
Q:
Q..

<.:J
Z
l-
I-
a.. 0.1
>-
Q:
;:!
Z
w
L
w
-1
w 0.01
0.3 0.4
POTENTIAL, VlS.C.E.

Figure 11. Change of the elementary pitting probability versus potential

for 304 L type stainless steel (from [22]).

Concerning the scanning rate v, they have shown that when v is high
enough :

1 E
It = -; exp $ + cte (66)

In addition, so long as the scanning rate is greater than a critical value,

$ does not depend on the chloride content. From these results, a pit generation
rate per unit area g can be defined :
164

E - El *
g = -d1t
dt
= c3 exp <)l
(67)

By using this approach they have been able to characterize the

susceptibility of the non-metallic inclusions to lead to pitting. For industrial
steels, pits are initiated on non-metallic inclusions, but not all the inclusions
have the same harmfulness. By directly measuring the number of inclusions p
per unit area a susceptibility of the inclusion 'Y can be obtained by 'Y = g/p.
They found that the susceptibility to pitting of two steels can be multiplied up
to a factor of 3x 103 for two types of inclusions.

5.1.2. Extreme value prediction of maximum pits [25-28J. The previous

approach is based on the investigation of stable pitting by the analysis of the
growth of "young" pits on a statistical basis which gives the susceptibility to
pitting for a metal and some pit generation rate. However on the engineering
point of view the service life of a device will often be given by the first leak.
Although structures such as pipelines and oil wells may have a large number
of pits and corroded areas of varying degrees of severity. The first
perforation, whose time and location will be a matter of probability. may stop
all use of the structure. For this reason the probability of such a failure must
be known as well as possible. It has been shown that the prediction of the pit
depths can be made from applying extreme value statistical analysis to actual
examinations of pitting at randomly selected sites. This procedure can
substantially reduce the time and expense of physically examining pipelines
or making extensive potential surveys.
It was shown that the distribution of pits on a group of corroding
samples is such that the sample maximum pit depths are distributed according
to the equation :

Pw = exp [ _e-a(w-W) 1 (68)

where Pw is the cumulative probability of occurrence of a pit of depth w, and

W is the characteristic deepest pit depth. If W increases according to the
equation :

W = b log t + c (69)

where t is the exposure time, then the times-to-first performation for a group
of samples are distributed according to the equation :

t .
Ps = exp [- ( U)K 1 (70)

where Ps is the survival function or the probability that the first sample
perforation will not have occurred at exposure time t and u is a characteristic
age (Fig. 12).
 165

REDUCED VARIATE-Y
o 2
200

~
If)
100
I
I-
z 70
o
:2: 50
1
~
<l:
~ 30
Q
z 20
If)
<l:
U
o
I- 10
w
:2:
i= 7

Figure 12. Survival function for a 7-in well casing (from [25J the
straight line fits eq(23) with k = 1.7 and v = 52 months.

5.1.3. Pre-pitting stage. In the early stage of pitting and in some experimental
conditions nucleation-growth and death of a pit sequence can be repeatedly
observed. As this process is stationary either time record analysis or spectral
analysis can be used for investigation.

5.1.3.1. Time record analysis [29-32]

A global stochastic model of pitting which involves a pre-plttmg stage and a
stable pitting stage has been considered by Williams et al [29]. This model has
the features that :
(i) events are nucleated with frequency A. (s-l. cm -2),
(i i) events have a frequency Jl (s-l) of dying,
(iii) events which survive beyond a critical age 'Cc (s) do not die.
166

For certain aspects of the modelling it can be assumed that all pits
evolve according to a deterministic law :

i=cu (71 )

where u is the age of the pit. In this framework particular attention can be
paid to the probability of survival of unpitted specimens (i.e. specimen still in
the pre-pitting stage) and to the ensemble statistics of potentiostatic current-
time transients (Fig.13).

50~ ________________ ~ ________________ ~

«:L
I-
Z
W 25
0::
0::
~
U

O~ ________________- L________________ ~

o 1200 2400
TIME I 5

Figure 13. Example of an ensemble of current time transients

18Cr 13Ni 1 Nb steel. 5cm2 • 0.028M NaCl. 50mVIS.C.E. (from [30]).

The pre-pitting stage is modelled by a birth and death process (eq(l9».

The mean <I> and the variance O'(I) of the current due to the unstable pits can
 167

be obtained for the limiting cases of short and long times. At short times, the
system is dominated by the birth of unstable centers so :

.5!ill.. 1
(72)
<I> ..JSt:t
where S is the area of the specimen.
Whereas for long times the system reaches an equilibrium distribution
of the number n and age of unstable pits, with :

.5!ill.. __ r:K. __1_

(73 )
<I> 'IiS-...r<;>

By comparing these ratios to the same ratios obtained experimentally

the authors have shown that in several cases the agreement between their
model and experiment is good especially for high polarization potentials
(Fig.14).
Concerning the birth of stable pits the expected number of events in
time t is ASt and the probability of survival of a single event for time 't c IS
exp( -1.1. 'tc). Thus the birth rate of stable pits A (s-l) is given by :

A = AS exp (-I.I.'tc) (74 )

168

«
::L2 00
.....
r
z
w
0:::
0:::
100
/
::)
/
u /
z /
«w ,..,. /1
2: 0

1.2

1\
H
V
........
........ 0.6
H
........
b

o.o~ __~__~____~__~__~
o 1000 2000
TIME I 5

Figure 14. Ensemble mean current (a) and standard deviation/mean (b)
in the same experimental conditions as in Fig.J3. Dashed lines are theoretical
results calculated using nucleation rates for stable pitting (from [30 j).
Curve I : 50mV/S.C.E. ; Curve 2 : 200mV/S.C.E.
 169

The authors have shown that the probability P(t,n) = P n of generating n

stable pits at time t is governed by the set of simple differential equations :

dP n
dt = A. Pn-l - A. Pn (75)

with P(1) = 1 at t = 'tc, which generates a Poisson process of intensity A. with the
properties
(i) the expected number of stable pits at any time t > 'tc is :

<n> A. (t - 'tc) « n > = 0 if t < 'tc)

AS (t - 'tc) exp (-Il'tc) (76)

(i i) the probability that no stable pit is formed (unpitted specimen) within a

time t > 'tc is P(O), given by :

In P(O) -AS (t - 'tc) exp (-11 'tc) (77)

P(O) = if t < 'tc

An expected value for the induction time before the onset of stable
pitting <'t> can be obtained by substituting <n> = 1 in eq(76)

(78)

The experimental data show that both the mean and standard deviation
of the number of pits tend to a constant value when the time increases so that,
given A (determined by peak counting) 11 can be calculated from the mean
number of peaks.
The probability of survival of an unpitted specimen has been related to
the nucleation rate for stable pitting A. but the authors have noted that there
is an ambiguity in the definition of the induction time 't c as the time of the
first passage through some arbitrary current level. They found a deviation at
long times from the behaviour predicted by eq(77).
From their analysis they concluded that the initiation of pitting
corrosion requires the generation and maintenance of gradients of acidity
and electrode potential on the scale of the surface roughness of the specimen.
Fluctuations in these gradients, leading to the birth and death of events can
arise because of fluctuations in the boundary layer thickness in the liquid at
the metal surface ; a pit becomes stable when its depth including the surface
roughness significantly exceeds the thickness of the solution boundary layer.
An other example is given by Bohni et al. [33] who have studied the
unstable pitting on aluminum which was decomposed in three steps : pit
170

initiation, pit growth and pit repassivation. The pit initiation rate is fitted to
the function :

A(t) = A exp (-Bt) + C (79)

The values of the parameters A,B,C, are both potential and rotation speed
dependent. They have studied the histogram of the size of the transients and
have shown that the value of the height and length of the transients have a
log-normal distribution. They interpret the exponential decay of the pit
generation rate by assuming that the newly formed passive film on an
attacked and repassivated site is better than before, and therefore that the
weak spots are no more weak. In some experiments they have also noted an
increase of the pit initiation rate during the first 20 to 40mn.
Concerning the probability law of the nucleation rate, Bertocci et aI.
[34] have paid a special attention to the distribution of the nucleation instants
in order to test the validity of the Poisson hypothesis. From the recording of
the fluctuations in the passive current of Fe-Cr alloys in chloride containing
solutions they have analyzed the statistical properties of the sequences of
intervals between current spikes. They have shown that these intervals are
not independent and therefore their statistics cannot follow a renewal process
law (and especially a Poisson law). In addition they have remarked that the
occurrence of the spikes have a tendency to cluster in time.
Some remarks : the main problems and the kinds of results which can
be gained by this approach can be now summarized. Concerning statistical
counting and plots of the survival probability on the experimental point of
view, the main problem is the choice of the threshold level of the current
above which the pit is considered as stable and its current circuit is switched
off : its choice conditions the estimation of the pit rate. However it seems that
the time evolution of the survival probability is roughly similar from one
metal to another. This shape leads to a pit generation rate A which first
increases with time and then decreases down to a constant value after some
time. This change has been described by discrete A values or, more
realistically, by a continuous change in time [21]. In all cases, after some
transient behaviour the pitting seems to become stationary, but this transient
state can be rather long. This point could explain Bertocci's results
concerning the non-adequacy of the Poisson hypothesis. As he found
dependent time intervals his process is not an homogeneous Poisson process
but could be a non-homogeneous one if he did not performed a sufficiently
long observation. However the interpretation of this evolution is still pending
as it can be correlated either to changes in the pitting resistance of the film
itself or to an exhaustion of the discrete sites initially available for pit
nucleation. In addition the conclusions considering that the pit generation
rate is determined only by the instantaneous value of the potential regardless
of the sweeping rate (i.e. of the degree of non-stationarity of the film) are
questionable [35].
The interpretation of the survival probability curve in terms of birth
and death processes [36] seems problematic if only the statistical law of the
stable pits are taken into account. The master equation used by Williams et al.
[29] (eq(l9» :
 171

dP(t,n)
dt = - (A + n~) P(t,n) + (n + 1) . ~ . P(t,n+l) + A . P(t,n-l) (80)

gives for n = 0 (as it is supposed that P(t,-I) "" 0) :

dP(t,o)
dt = - AP(t,O) + ~ P(t,l) (81 )

For relatively low pitting activity, it can be supposed that at most one pit
can exist on the surface, hence P(t,l) = 1 - P(t,o) and the equation used by
Shibata et al. in [19] is obtained. However this simplified model leads to a
survival probability which tends towards a limit ~/(A + ~). Instead the full
sequence of the master equation predicts a survival probability which tends
towards zero for long times. This conclusion is more realistic. In any case the
death rate cannot be obtained by only taking into account the counting of the
stable pits without analyzing the pre-pitting stage [35]. The investigation of
this stage is particularly attractive because its investigation could lead to a
non-destructive technique for evaluating the resistance of metals to localized
attack in a given medium whereas the statistical counting method requires the
start of a stable pit to be efficient.

5.1.3.2. Spectral analysis

Spectral analysis has been used in order to study pre-pitting stage for vanous
metals :
a) Iron : The chemical breakdown of the passive layer on passive iron
depassivated by adding chloride ions has been studied by Keddam et al. [37] by
this technique. The current time recording is given in Fig.15. Microscopic
study of the electrode surface demonstrates that the large peaks correspond to
large pits of spherical shape (radius > 1O~ m). Inside, secondary pits of
polygonal shape can be observed. The current fluctuations arising from the
secondary pits have been studied by spectral analysis (Fig.16). It can be
noticed that for low pitting activity (and especially for passivity) the increase
of the high frequency noise is due to the parasitic instrumentation noise. The
current noise measured during the current growth sometimes reveals a lot of
pitting activity even if the d-c current level is not too different from its value
in passivity.
It is when there is sufficient pitting activity that a spectrum with a
white noise low frequency limit 'I' i (f=O) is observable. When the pitting
activity is too low this part of the spectrum is buried in the background noise.
If the pit generation rate is supposed to follow a Poisson law of intensity A,
eqs(3l) and (32) give the mean of the current fluctuations due to pitting :

<I>=AQ (82)
and the white noise low frequency limit of the p.s.d.

(83)

where Q = F(O) is the charge involved in one current transient. When < I > and
'l'i(O) are known, A and Q can be determined.
172

2.5
0.15
<!
:L <!
...... E
r2.0
....
Z 0.1 t-
W Z
W
0::: 0:::
0::: 0:::
:::> 1.5
U 0.05 ::::>
U

I:.-.J..I..L..J..!..J.1-I.....J.'-I...1...1.'--1..'...1.1--,-,-lit I , , I , I 1.1
o 6 12 18 1/ 43 49 55' 64 70
TIME I mn

Figure 15. Current-time dependence after chloride addition (3.10- 3 M).

Fe in O.5M H2S04 at O.2V/S.S.E. (from [37]).
 173

lil A SQUAR~/HZ
-14

-15
w
....
Ul -16
a
z
~ -17
z
w
a::
a:: -18
:::l
u
lI..
a -jg

c:i -21'1
u)
a: -21

-22
HI
FREOUENCY

Figure 16. Current noise spectrum at different pit activities.

A) Fe in O.5M H2S04 at O.2VIS.S.E. after an hour of passivation, J = 5.5mA.cm-2 ;
B) after chloride addition (3.10- 3 M), J == 150mA.cm- 2 ; C) after chloride addition,
J = 30mA.cm- 2 Cut-off frequency at 3dB attenuation is 60Hz (from [8]).
174

b) Fe-Cr : The current fluctuations during pit llllUaUon in Fe-Cr alloys have
been studied by Bertocci et al. [37,38]. Frequency spectra of these fluctuations
have been recorded before and during pit initiation. No detectable
fluctuations have been observed in the absence of chloride, where pitting
does not occur. In the presence of chloride, current transients can be detected
during the induction period. They have found that once pitting starts the
white noise low frequency limit is often hardly observed (Fig. I?) ; they think
that this is not due to the limited frequency range investigated, but it is rather
because the measurement conditions are non-stationary. They concluded that
the effect of CI- is to increase the chance of local breakdown rather than to
affect the repassivation time.

~
-60
I
,..
t:
u -80 0
IN

=-100
<t

W-140
g
c.. -160
w
v,o -180
z
10-' 1 10
FREQUENCY, Hz

Figure 17. P.s.d. of the random current fluctuations of Fe-Cr alloys

electrode in borate buffer plus 0.1 M NaCi during pitting.
(a) Fe-20%Cr : 300mV/S.C.E., 200mV/S.C.E. (from [37]).
 175

c) Aluminum : Pitting corrosion and activation of aluminum and aluminum

alloys in chloride media have been investigated under galvanostatic
conditions. The potential noise generated by a rotating disc electrode
(lOOOrpm) of an aluminum-3% magnesium alloy in NaCl 0.6M in borate buffer
(Fig.18) have been studied by Bataillon et at. [39]. In the low frequency range
the p.s.d. varies like f-a. When the current density increases the morphology
of the localized corrosion changes as the pitting on the surface becomes more
generalized. As the coefficient a. decreases the authors suggest a possible
relationship between this coefficient and the distribution of the pits on the
surface.
On pure aluminum in 3% sodium chloride solution a white noise low
frequency limit is obtained as shown in Fig.19 where the p.s.d. of the potential
fluctuations have been plotted before chloride ion addition and after chloride
addition. Dawson et al. [40] observed the occurrence of sudden potential drops
accompanied by rough exponential recoveries. They suggest that a breakdown
repair process is occuring on the surface at defect sites of the pure aluminum
oxide film, even in the absence of aggressive ions. They concluded that the
analysis of the p.s.d. of the potential fluctuations provides a means by which
to characterize the corrosion attack present : shallow slopes of -ZOdB/dec. or
less are indicative of pitting corrosion whereas a rolloff of more than -
ZOdB/dec. represents generalized corrosion or a passive state. They have
noticed that hydrogen evolution could be a contributing factor to the high
frequency noise levels observed.

On a general viewpoint concerning the spectral analysis method, some

problems of measurement are inherent to the technique. If the raw p.s.d. data
are plotted, the statistical errors due to the finite number of random samples
taken into account appears as the thickness of the line (e.g. see Figs.16,17).
The p.s.d. data can be processed in a computer by some smoothing techniques
but if great care is not taken the results could be misinterpreted (e.g. see
Fig.19). The drawing is neat because the processed line is very thin but with
only one 1000s sample the lowest frequency theoretically reachable is 1O-3H z
(eq(SO)) with a 100% (eq(SZ)) estimation error. In the case of Fig.19, the
curves are extrapolated down to 10- 5 Hz (for curves 1, 3, 4, it is only an
horizontal extrapolation but for curve 2 a cut-off frequency appears lower
than 1O-3Hz which is very questionable).
It is to be noticed that in all the above cases of iron base alloys, the
current p.s.d. often increases when frequency decreases and does not point to
a white noise low frequency limit which is easier to interpret. When the
elementary current transients are clearly related to a finite electric charge,
this 1/fU behaviour has to be ascribed to the point process itself : either to a
non-stationarity of the pit generation rate or to some special distribution. In
the latter case similarly to any 11f noise in other physical or chemical domain
the interpretation is not clear at all and this special behaviour can lead to
problems even in the counting procedure [41].
 176

10"
I r-------------~------------._------------~------------~
N
I -4~~~------~-------------+-------------r----------~
J

>.....
o -6~----------_+----~~----+_------_+--_+----------~
{(j
0...: -8
~------------+-------------~----~~~~~-+.----------~
~
«
~ -10r-------------+_------------~------------_r~~._~,,~~
Z
w
f-
o -12L-~--------~----------~----------~----------~
0... 10-' 1 10 10 2 10 3
FREQUENCY I Hz

Figure 18. P.s.d. of the random potential fluctuations of AI-3%Mg alloy

electrode borate buffer plus 0.6M NaCI during pitting (from [39J).
0
N
~
I
N
-20 I-
2 -
3
I
Q) -40 I 4 I -...., 1 ___

-0
~
1 ~
c:i -60 -
(/)
n..: -80
~~ -

1\
..J
~ -100
I- I-
Z
W
0 -120 I
c... 10- 4 10- 3 10- 2
FREQUENCY/Hz

Figure 19. P.s.d. of the random potential fluctuations of pure Al electrode

in the absence of Ct :
Curve 1 : at immersion ; Curve 2 : 20mn later ; Curve 3 : just after chloride ions
addition; Curve 4 : 20, 40 and 60mn later (from [40)).
 177

5.2. CORROSION BY ABRASION

The mechanical breakdown of the passive layer have been studied by spectral
analysis by Oltra et al. [8,42,43]. A continuous abrasion of the passive surface
of iron is obtained by an impinging jet of SiC particles in suspension in the
electrolyte (Fig.20). The average size of the mechanical damage is on the order
of IJl m 2. In this case the current transient due to a single impact is hardly
distinguished from the noisy background current. except when only a few SiC
particles impinge the working electrode at the same time (Fig.21).
Fig.22 illustrates the current-time response to a square wave
perturbation by a flow of numerous abrasive particles. The current tends
towards a steady-state (mean dissolution current) through a transient regime.
If the abrasive particles are suddently removed from the flowing solution
(stop-flow) the current returns to its initial value following the same
transients.
The p.s.d. of these fluctuations are plotted in Fig.23A. Two plateaus
followed by an approx. l/f2 decreasing part can be observed. so two cut-off
frequencies can be defined :
(i) one in the low frequency range (about 1Hz).
(ii) the other one in the high frequency range (about 1kHz).
In order to explain the nature of these cut-off frequencies the p.s.d. of
the phototransistor current fluctuations due to the particles concentration
fluctuations in the nozzle has been measured (Fig.24). By correcting the raw
p.s.d. of the genuine depassivation-repassivation events has been obtained
(Fig.23B) and shows only one cut-off in the high frequency range. So the
current fluctuations can be assumed to follow a doubly stochastic Poisson
process. It means that the number of particles per time unit is a random
quantity which has a mean value ~ and a p.s.d. 'P A(f). The p.s.d. of the current
fluctuations can be then written (eqs(31) and (33» :

[}" + 'PA(f) 1 1 I(f) 12 (84 )

where I(f) is the Fourier transform of the elementary current transient i(t)
due to a simple impact which is supposed deterministic in this study. I(f) is
given by eq(33) :

(85)

From the p.s.d. of the current fluctuations. some informations can be

obtained. The remaining high frequency cut-off frequency gives the average
time constant of the current transient (here 80Jls). However this quantity is
not related to the repassivation rate which should be shorter but to the Re C
time constant (Re : electrolyte resistance. C : double layer capacity) of the cell.
 178

Expansio.n
Vessel

Phototransistor

Pump

Figure 20. Abrasion experimental arrangement. The expansion vessel has

a 5 I volume ; the flow rate of the SiC particles varies from 1 to 2m.s- l . The
working electrode (I/J 5mm) is potentiostatically controlled and the pneumatic
pump yields 20 to 2001.h- 1 (from [42]).

30
<t:
:L

I-
Z 20
W
0:::
0:::
::>
U 10

Figure 21. Current transients observed during abrasion in the case when
a few particles impinge the working electrode.
 179

!SpA

<t:
::L
lOpAI--
Z
W
0::
0::
~
U
5pA

100s
TIME I 5

Figure 22. Current-time response to a switch-on and switch-off of a flow

numerous abrasing particles.

IN
I -'3
N';
a'
vi -,4
0..

W
C!1
<5 -,s
Z
f-
Z
W
a::
a:: -'6
:::>
u

-17
,.
FREQUENCY I Hz

Figure 23. Measured (Curve A) and corrected (Curve B) p.s.d. of the

current fluctuations (E = 300mV/S.S.E., SiC.ASTM 150, 9g.l- i ), mean dc current
53J.LA) (from [42]).
 180

-,.

II
IN

N
<::[
:::l~ -:B
I

~
ei
c;:
!
i
w -:9
vi - - - - 1- - --1----I

0
z
-ze
1- !
I I
f- I

I !
Z
w
c:::
c::: -;:: 1 1 !
::J

I
U

-;:;2
Hi'; ;~ ~ L! .:. :0 3 .e ~
FREQUENCY I Hz

Figure 24, P,s,d, of the phototransistor signal fluctuations (same

experimental conditions as in Fig,]8) (from [42]).

This point can be checked for by modifying the electrolyte

conductance, e.g. by using a more dilute solution. But the relative contribution
of Re cannot be separated. In contrast when Re is artificially increased by
adding an extra resistance in series with the working electrode, the cut-off
frequency is decreased, in agreement with the ReC value.
In order to explain the whole electrode control of the electrochemical
response an electrical coupling model has been proposed. In this model the
film repair is assumed to consist of two steps ;
- Firstly, just after the mechanical impact the unaffected passive area supplies
the depassivated patch with the required electrical charge L1 Q for the
repassivation,
 181

- Secondly, the potentiostatic device provides the same charge Ll Q which has
been invested in the last reaction.
Under potentiostatic control, the transient current response
corresponds to this second step.
The low frequency limit of the p . s.d. [ 'I' j(O) ] gives the mean value of
the elementary charge DQ involved in each transient through (eq(74» :

(86)

The abrasion technique coupled with spectral analysis method reveals an

unexpected result of a contribution of the unaffected surface to the repair of
the damaged area. The actual repassivation time constant measured by the
high cut-off frequency of the p.s.d. of current fluctuations is short, in the
case of a restricted area.
This technique gives rise to a lower limit for the measurement of the
repassivation time of a patch of fresh metal surface promoted by passive film
breakdown. This limit is function of the ratio between the surface of the
whole electrode and the damaged one and is equal to the product of the
electrolyte resistance Re by the high frequency capacitance of the whole
surface. Thus the main result of these abrasion experiments is the unexpected
founding on the electrical interaction between the localized depassivated area
with the surrounding passive surface.

5.3. STRAINING TESTS

According to the phenomenological model of the repetitive localized

depassivation at film-covered electrodes it could be expected that noise
analysis can provide characteristic noise spectra during strammg and
cracking tests. However, in spite of the promissing expectations, little work
has been published on this topic.
The transgranular stress corrosion cracking (s.c.c.) of alpha-brass and
copper in nitrite solutions has been studied by Sieradski et al. [44] by
comparing the electrochemical noise to the acoustic emission. Electrochemical
and acoustic transients (related to mechanical rupture) occurred
simultaneously during the crack advance (Fig.25). By complementing this
electrochemical approach by fractographic examination they concluded that
cracking occurred by discontinuous cleavage.
Spectral analysis has been tentatively used to investigate the current
fluctuations during straining of a passive iron electrode in order to validate
the model of repetitive localized depassivation at the emergence of slip steps
by Gabrielli et al. [45]. According to noise spectra shown in Fig.26 straining
increases substantially the noise level in the low frequency half part of the
frequency domain (below 500Hz). In contrast with the high frequency region
this enhanced noise cannot be explained by the residual voltage fluctuations
across the reduced impedance during straining. However it is not at all clear
whether this noise is related to specific contributions of discrete depassivation
events or to a general lIf type noise.
182

Figure 25. Correlation of the peak amplitude of acoustic emiSSIOn (AE)

with the current fluctuations for a growing crack in an alpha-brass specimen
tested in a nitrite solution (from (44)).
~ 1~2~~_ _ _ _ _ _ _ _ _ _~_ _ _ _ _ _ _ _~_ _~

I-
Z
UJ
cr::
cr::
:l
U

FREQUENCY I Hz
Figure 26. P.s.d. of current fluctuations during straining of a passive iron
electrode in H2S04 1M - strain rate = 10-3 .s- l (from (45J).
1001~~ _ _ ,_ _

!II
~ OOt_

FREQUENCY,lb

Figure 27. Change in the spectra of the noise level as function of

frequency for two mechanical states of an alpha-brass electrode in Mattson
solution. (---) stressed; (___ ) unstressed and for 1) t = a ; 2) t = 180mn (from
[46]).
 183

Loto et al. [46] have published noise analysis related to S.C.c. of alpha-
brass in Mattson solution. From the experimental spectra it seems very
difficult to separate surface modifications, such as film formation or pitting,
from crack propagation events. This could be related to the characteristic of
low frequency noise which shows a typical llfu behaviour whatever the
mechanical state of the tested electrode (stressed or unstressed). The
occurrence of cracking seems only to promote an increase of the noise level
on the low frequency plateau but cannot be clearly used as a criterion to
detect and follow the crack propagation as shown in Fig.27.

5.4. STUDY OF THE UNlFORM CORROSION THROUGH ELECTROCHEMICAL NOISE

[ 48,49]

The corrosin rate of a metal immersed in an aggressive medium has been

investigated in the case of the corrosion of iron in acidic medium [47].
At low anodic current density the dissolution of iron in sulfuric acid
medium takes place together with hydrogen evolution. The measured total
current is the sum of the iron dissolution current and the hydrogen evolution
current which are undistinguishable. The hydrogen evolves through bubbles
which give rise to fluctuations in the measured current or voltage. The time
recording of the electrode potential under galvanostatic control is given in
Fig.28 for an electrode facing upwards. An optical observation shows that the
slow increase is due to the growth of an hydrogen bubble although the steep
jump is due to the bubble detachment [48].
The elementary voltage transient can be approximated by a series of
linear function of time, the mean and the p.s.d. of which are from eqs(37) and
(38)

Ak
<Vet»~ =2" (37)
a

2Ak2(3a 2 + 41[2f2)
'I'V(f) (38)
a 2 (a 2 + 41[2f2)2

In Fig.29, example of the p.s.d. of the random fluctuations of the

corrosion potential at a free electrode 'I' V (f) is given. The measurements are
made between 10- 2 to 10 2 Hz. The lower frequency is imposed by the natural
drift of the system which is of the order of 1011 V . m n -1. Above 10 2 Hz the noise
of interest is buried in the parasitic noise coming from the regulation. A
careful inspection of the p.s.d. of the voltage fluctuations shows that below a
frequency of the order of 10- 1Hz the noise can be considered as white, i.e. the
p.s.d. seems independent of the frequency and equal to 'I'v (0) [49].
As the experimental p.s.d. of the voltage fluctuations is in rather good
agreement with the p.s.d. predicted by the theoretical model the characteristic
parameters A, kv, and <V> can be extracted. They are given in Table 2 for two
types of iron immersed in the sulfuric acid solution at pH=O together with the
charge transfer resistance obtained from impedance measurements. It can be
noticed that when the corrosion rate of iron increases (i.e. the charge
transfer resistance decreases), the parameter <V> increases.
184

...
5.'

a.
t\NOOIC
2••

>
1..

..
E

"
"'-<
Ll
....J -1.111
0
>
-2."
CATHOOIC
-3.•

-4.1.1

-5. "
" 2' 3. 41iS
TiME, •
5111 61iJ 70
•• 00 I.'

Figure 28. Time recording of the electrode potential under

galvanostatic control at I = 3.7mA.cm- 2 in the case of iron dissolution in 2N
sulphuric acid medium.

,."-s V·qua~·/H:z

"'1/1 -B

g
-7

"'-<
II
I-
J -8
0
>
•
f. -0

~
"
0
-I'
ei
iii
rL -11

-12
lIi~ 10"':2 lIif]
"
FREQUENCY / HZ

Figure 29. Power spectral density of the fluctuations observed during

iron dissolution (same conditions as in Fig.28).
 185

Table 2

Metal

Iron I 0.715 37 52 50

Iron II 0.357 55 154 11.5

The voltage fluctuations related to bubble evolution are due to a

screening effect of the hydrogen bubble which grown on the electrode
surface. Therefore <v> can be considered as a measure of the volume of
hydrogen evolves on the iron surface. As at the corrosion potential the
hydrogen current is equal to the dissolution current of iron, an increase of
<v> means an increase of the corrosion current of the metal. This analysis of
the fluctuations of the corrosion potential can lead to an effective way to
estimate the corrosion rate without perturbing the electrochemical system.
186

REFERENCES

[1] EVANS U.R., MEARS R.B. and QUENEAU P.E., (1933),

"Corrosion-velocity and corrosion-probability", Engineering 136, 689.

[2] NATHAN C.C. and DULANEY C.L., (1974),

"Localized corrosion as a statistical phenomenon" in R.W. Staehle (ed.),
Localized Corrosion-3, NACE, Houston, pp.184-l89.

[3] EVANS U.R., (1974), "Notes on the localized corrosion meeting and the
corrosion studies at Cambridge" in R.W. Staehle (ed.),
Localized Corrosion-3, NACE, Houston, pp.i44-i46.

[4] SATO N., (1976), "Stochastic process of chloride pit generation in passive
stainless steel", J. Electrochem. Soc. 123, pp.1197-l199.

[5] OKAMOTO G., TACHIBANA K., NISHIYAMA S. and SUGITA T., (1976),
"The analysis of passive current noise of stainless steels under
potentiostatic conditions with and without chloride ions" in R.W. Staehle
and H. Okada (eds.), Passivity and its breakdown on iron and iron base
alloys, NACE, Houston, pp.106-109.

[6] SHIBATA T. and TAKEYAMA T., (1976), "Pitting corrosion as a stochastic

process", Nature 260, 315-316.

[7] SHARLAND S.M., (1987), "A review of the theoretical modelling of

crevice and pitting corrosion", Corrosion Sci. 27, 289-323.

[8] GABRIELLI C., HUET F., KEDDAM M., OLTRA R. and PALLOTTA C, (1983),
"Stochastic aspects of mechanical and chemical breakdown of
passivity", in M. Froment (ed.), Passivity of metals and semiconductors,
Elsevier Sci. Pub., Amsterdam, pp.293-298.

[9] OSTERWALD J., (1981), "Contribution to the theory of anodic pitting"

in R.P. Frankenthal and F. Mansfeld (eds.), Corrosion and corrosion
protection, The Electrochemical Society, Pennington, pp.112-119.

[10] SATO N., (1981), "The stability of pitting dissolution of metals in aqueous
solution", in R.P. Frankenthal and F. Mansfeld (eds.), Corrosion and
corrosion protection, The Electrochemical Society, Pennington,
pp.l01-111.

[11] MACDONALD D.D. and URQUIDI-MACDONALD M., (1986),

"Distribution functions for the breakdown of passive films",
Electrochim. Acta 31, pp.1079-1086.

[12] MACDONALD D.D. and URQUIDI-MACDONALD M., (1987),

"Theoretical distribution functions for the breakdown of passive films",
J. Electrochem. Soc. 134, pp.41-
 187

[13] OKADA T., (1985), "A theory of perturbation-initiated pitting",

J. Electrochem. Soc. 132, pp.537-544.

[14] OKADA T., (1986), "The rate of passive metal dissolution in relation to the
presence of traditional halide complexe in halide solutions",
Corrosion Sci. 26, pp.839-844.

[15] LARSON H.J. and SHUBERT B.O., (1979), "Probabilistic models in

engineering sciences 2", John Wiley, New York.

[16] SNYDER D.L., (1975), "Random point processes, John Wiley, New York.

[17] COX D.R. and MILLER H.D., (1965), "The theory of stochastic processes,
Methuen, London.

[18] STRATONOVICH R.L .. (1963), "Topics in the theory of random noise 1".
Gordon and Breach. New York.

[19] SHIBATA T. and TAKEYAMA T., (1977). "Stochastic theory of pitting

corrosion". Corrosion 33, 243-251.

[20] GABRIELLI C .• HUET F. and KEDDAM M., (1986), "Investigation of

electrochemical processes by an electrochemical noise analysis.
Theoretical and experimental aspects in potentiostatic regime".
Electrochim. Acta 31, 1025-1039.

[21] DOELLING R. and HEUSLER K.E., (1984), "Incubation of pitting corrosion

at passive iron as a stochastic process", Z. Physik. Chern. N.F. 139. 39-58.

[22] BAROUX B .• SALA B .• JOSSIC T. and PINARD J .• (1985). "Methode statistique

de determination du potentiel de piqiire des aciers inoxydables".
Materiaux et Techniques 5. 3-8.

[23] BAROUX B., (1986). "Kinetics of pit generation for stainless steels".
in M. Duprat (ed.), Electrochemical methods in corrosion research.
Materials Science Forum 8, Trans. Techn. Pub .• Suisse, pp.91-103.

[24] BAROUX B., (1986). "The kinetics of pit generation on stainless steels".
Corrosion Sci. 28, 969-986.

[25] ELDREDGE G.G .• (1957). "Analysis of corrosion pitting by extreme-value

statistics and its application to oil well tubing caliper surveys",
Corrosion 13. 51-60.

[26] AZIZ P.M., (1956), "Application of the statistical theory of extreme values
to the analysis of maximum pit depth data for aluminium". Corrosion 12.
495-506.
188

[27] FINLEY H.F., (1967), "An extreme-value statistical analysis of maximum

pit depths and time to first perforation", Corrosion 23, 83-87.

[28] PROVAN J.W. and RODRIGUEZ E.S., (1989), "Development of a Markov

description of pitting corrosion", Corrosion 45, 178-206.

[29] WILLIAMS D.E., WESCOTT C. and FLEISCHMANN M., (1983), "A statistical
approach to the study of localized corrosion", in M. Froment (ed.),
Passivity of metals and semiconductors, Elsevier Sci. Pub., Amsterdam,
pp.217-228.

[30] WILLIAMS D.E., WESCOTT C. and FLEISCHMANN M., (1984), "Studies of the
initiation of pitting corrosion on stainless steels", 1. Electroanal. Chern.
180, 549-564.

[31] WILLIAMS D.E., WESCOTT C. and FLEISCHMANN M., (1985), "Stochastic

models of pitting corrosion of stainless steels", J. Electrochem. Soc. 132,
1796-1811.

[32] WILLIAMS D.E., FLEISCHMANN M., STEWART J. and BROOKS T., (1986),
"Some characteristics of the initiation phase of pitting corrosion of
stainless steel", in M. Duprat (ed.), Electrochemical methods in corrosion
research, Materials Science Forum 8, Trans. Techn. Pub., Suisse,
pp.151-166.

[33] HOLLIGER R. and BOHNI H., (1985), "Evaluation of current transients to

study pit initiation", in M.W. Kendig, U. Bertocci and J.E. Strutt (eds.),
Computer aided acquisition and analysis of corrosion data,
Proc. VoI.85-3, The Electrochemical Soc., Pennington, N.J., pp.200-209.

[34] BERTOCCI U., KOIKE M., LEIGH S., QIU F. and YANG G., (1986), "A statistical
analysis of the fluctuations of the passive current", 1. Electrochem. Soc.
133, 1782-1786.

[35] KEDDAM M., KRARTI M. and PALLOTTA C., (1987), "Some aspects of the
fluctuations of the passive current on stainless steel in presence of
chloride. Their relation to the probabilistic approach of pitting
corrosion", Corrosion 43, 454-458.

[36] SHIBATA T. and TAKEYAMA T., (1981), "Death and birth stochastic
process in pitting corrosion of 17Cr ferritic stainless steels",
in Proceedings of the 8th Int. Congr. Met. Corr. Dechema,
Frankfurt on Main, pp.146-151.

[37] BERTOCCI U. and YE Y.X., (1984), "An examination of current fluctuations

during pit initiation in Fe-Cr alloys", J. E1ectrochem. Soc., 131,
1011-1017.
 189

[38] BERTOCCI U. MULLEN J.L. and YE Y.X., (1983), "Electrochemical noise

measurements for the study of localized corrosion and passivity
breakdown", in M. Froment (ed.), Passivity of metals and
semiconductors, Elsevier Sci. Pub., Amsterdam, pp.229-

[39] BATAILLON C. and FIAUD C., (1986), "Spectral analysis of noise potential
during pitting", in M. Duprat (ed.), Electrochemical methods in
corrosion research, Materials Science Forum 8, Trans. Techn. Pub.,
Suisse, pp.141-150.

[40) URUCHURTU J.e. and DAWSON J.L., (1987), "Noise analysis of pure
aluminium under different pitting conditions", Corrosion 43, 19-26.

[41) VANVLIET C.M. and HANDEL P.H., (1982), "A new transform theorem for
stochastic processes with special application to counting statistics",
Physica 113A, 261-276.

[42] OLTRA R, GABRIELLI e., HUET F. and KEDDAM M., (1986),

"Electrochemical investigation of locally depassivated iron.
A comparison of various techniques", Electrochim. Acta 31, 1501-1511.

[43) GABRIELLI C., HUET F., KEDDAM M., OLTRA R and COLSON J.C., (1986),
"A new insight in the electrochemical response of a depassivated
surface", in M. Duprat (ed.), Electrochemical methods in corrosion
research, Materials Science Forum 8, Trans. Techn. Pub., Suisse,
pp.491-500.

[44] SIERADSKI K. and NEWMAN Re., (1985), "Brittle behaviour of ducitile

metals during stress-corrosion cracking", Philosophical Mag. A51,
95-132.

[45) GABRIELLI e., HUET F., KEDDAM M., COLSON J.C. and OLTRA R, (1984),
"Depassivation of iron by straining and abrasion, a-c impedance and
electrochemical noise measurements", in Proc. of the 9th Intern. Meet.
of MetalIic Corrosion, Toronto 3, pp.427-432.

[46] LOTO e.A. and COTTIS R.A., (1987), "Electrochemical noise generation
during stress corrosion cracking of alpha-brass", Corrosion 43, 499-504.

[47] EPELBOIN I., KEDDAM M. and TAKENOUTI H., (1972), "Use of impedance
measurements for the determination of the instant rate of the metal
corrosion", J. Applied Electrochem. 2, 71-79.

[48] GABRIELLI C., HUET F.and KED DAM M., (1985), "Characterization of
electrolytic bubble evolution by spectral analysis. Application to a
corroding electrode", J. Appl. Electrochem. 15, 503-508.
190

[49] GABRIELLI C.• HUET F .. KEDDAM M. and MACIAS A.. (1986).

"Electrochemical noise generated by a corroding electrode. Effect of
metallurgy and inhibitors", in E. Mc Cafferty and R.J. Brodd (eds.).
Surfaces. inhibition and passivation. VoI.86-1,
The Electrochemical Society. Pennington. N.J.. pp.501-523.

[50] PATTINSON J. and BURSTEIN G.T .• (1985).

"Computer aided collection and analysis of electrochemical current
transients". in M.W. Kendig. U. Bertocci and J.E. Strott (eds.). Computer
aided acquisition and analysis of corrosion data. The Electrochem. Soc ..
Pennington. VoI.85-3. pp.108-129.
PHOTOELECIROCHEMISlRY : THEORETICAL BASIS

M. DA CUNHA BELO
CNRS- CECM,
15, rue Georges Urbain,
94400 Vitry-sur-Seine, France

ABSTRACT. An attempt is made to present the essential features specific to photoelectrochemical

processes on semiconductors. Thus, some basic notions on solid state physic (band structure and electron
distribution) are first introduced. Then the kinetics of electrochemical reactions is described taking into
account the contribution of the two types of charge carriers and the influence of a space charge near the
semiconductor-electrolyte interface. Finally, two approaches for describing electrochemical processes based
on photoexcitation of the semiconductor are discussed: a kinetic approach, and a thermodynamic approach
which introduces the concept of the quasi-Fermi levels. The influence of the mutual position of the
different energy levels on the conditions of stability and corrosion of a semiconductor is examined.

IN1RODUCTION.

The protective character of passivity can be explained by the development of passive layers with
semiconducting properties. This is the reason why some aspects of the band structure and
thermodynamic properties of the semiconductor/electrolyte interface will be briefly presented.
Further, to describe the advances in the theory and practice of photoelectrochemistry, it is
necessary to make acquaintance with the specific features of electrode processes on
semiconductors. It is well known that the latter, as compared with those on metallic electrodes, are
connected, firstly with their electronic structure and in particular with the band gap and the two
types of charge carriers (electrons in the conduction band and holes in the valence band) and,
secondly, with the great influence of the applied electric field on the concentration of these carriers
near the surface. The mutual disposition of the different energy levels for the electrode and the
solution has a decisive influence over the magnitude of the electron transfer currents. It is precisely
on these points that we shall concentrate our attention. Some aspects of the kinetics of
electrochemical reactions are considerated in terms of transport equations, taking in to account the.
two types of charge carriers and the presence of a space charge near the interface.

Photoelectrochemical processes can occur under very diverse regimes depending on the
radiation energy, the depth of light penetration and the thickness of the space charge region. The
length of diffusion of the charge carriers and the recombination time after photogene ration of the
electron-hole pair are also essential factors. Two approaches for describing electrochemical
processes based on photoexcitation of the semiconductot will be introduced: a kinetic approach by
solving the equation describing the kinetic of carrier transfer across the interface, and a
quasi-thermodynamic approach based on the concept of the quasi-Fermi levels.

As a consequence of the electronic structure, the electrochemical decomposition of a

semiconductor material may be either anodic or cathodic. Different conditions of thermodynamic
stability with respect to decomposition processes will be examined.

191
M. G. S. Ferreira and C. A. Melendres (eds.),
Electrochemical and Optical Techniques for the Study and Monitoring of Metallic Corrosion, 191-217.
© 1991 Kluwer Academic Publishers.
192

SOME BASIC NOTIONS IN SOLID STATE PHYSICS

1. The band structure

The energy spectrum of electrons in an ideal crystal consists of intervals of energy filled with
energy levels (allowed bands) separated by intervals in which there are no electronic energy levels
(forbidden bands or band gaps) (1-4). According to the principles of quantum chemistry, the
allowed bands energies are a result of the splitting of the electronic orbitals of individual atoms
(ions) due to the formation of the crystal. This is why the corresponding allowed bands, like the
initial orbitals, are sometimes called s-, p-, d-, etc... bands. Within each band the different states
of the electron are characterized both by their energy E and their quasi-momentum p.

At absolute zero the electrons fill the lowest energy levels. Depending on the number of
electrons, they fill several of the lowest allowed bands leaving the higher bands empty. A crystal in
which the lower bands are completely filled at T = 0, whilst the higher bands are completely
empty, behaves as a dielectric. The metallic electronic structure occurs when at least one allowed
band is only partially filled with electrons at T = O. The upper of the filled bands is called a valence
band, while the lower of the unfilled bands is called a conduction band. It is the distribution of
electrons in these two bands which exerts a decisive influence on the electronic properties of the
crystal. Differences in conductivity between different substances must be a matter of vacant or
partially filled bands. Electrons cannot go into a filled state (pauli principle) and only move
throught partially vacant energy bands. Figure I a, shows the influence of the interatomic spacing
on band picture of a certain solid. If the equilibrium interatomic distance is dM, the crystal will
show metallic conduction. The overlap of the conduction and valence bands permit electron
mobility throught unfilled energy states in the valence band or unoccupied energy states in the
conduction band. The situation relative to interatomic distance, dI, corresponds to a material which
behaves like an insulator. The valence band becomes completely filled and the conduction band
completely empty the energy of the band gap is large when compared with the thermal energy KT
and therefore the electrons cannot be significantly thermally excited into the conduction band.
Finally the interatomic distance, dsc ' represents a situation where the electrons in the valence band
can be excited into the condution band by thermal effects. This is the case of an intrinsic
semiconductor.

A semiconductor may be either intrinsic or extrinsic. In intrinsic semiconductors at T > 0,

the generation of current carriers occurs as a result of the thermal excitation of some of the
electrons from the valence band to the conduction band with the corresponding thermal rupture of
some chemical bonds. Simultaneously an equal number of positively charged holes are created in
the valence band. These holes, like the electrons in the conduction band, may participate in the
flow of current. In the electric field the holes behave like positive particles, possessing a positive
charge equal in absolute value to the charge of the electron. They are also characterized by the
definite values of their quasi-momentum, and thus like electrons in a solid, they may be regarded
as charged quasi-particles.

Extrinsic semi conductivity is due to some kind of lattice defect, i.e., lattice imperfections
which may be caused by atoms leaving their lattice (interstitial sites) giving a Frenkel defect.
Lattice imperfections may also be caused by an excess' or a defect of constituent atoms from the
stoichiometric formula (Schottky defects). Extrinsic conductivity is also obtained when some of
the lattice atoms have been replaced by atoms of a different element. Because the energy required
to form the different types of defects is usually slightly less than the width of the forbidden band,
their respective energy levels are located within the band gap itself. Impurities and defects are
subdivided into donors and acceptors. Donors give up excess electrons to the conduction band of a
semiconductor, thereby creating electronic conductivity (n - type semiconductors). Acceptors
capture valence electrons from atoms of the basic substance (into which they have been
 193

conduction bond

gap »KT

Interatomic
distance

bj E
e

interbond
3

Fig. 1 - a) Schematic representation of the dependence of electron energy upon interatomic distance
(splitting phenomenon). b) Types of transition in a semiconductor: transitions 1 and 2 occur as a
result of thermal excitation. 3, 4 and 5 denote those which occur under the influence of light.
194

introduced) producing hole conductivity (p - type semiconductors) (fig. Ib).

2. The electron distribution

In the absence of external effects (electric field, illumination, etc...) the thennodynamic equilibrium
concentrations of electrons and holes in semiconductors are detennined by the temperature, the
width of the bandgap, the concentration of defects and impurities and finally the dispersion laws,
E(p), for the conduction and the valence bands. Let us note that in a semiconductor the electrons
are concentrated near the lower edge Ec of the conduction band and the holes near the upper edge
Ev of the valence band, within narrow energy ranges of the order of KT.

In any electronic system in thennal equilibrium the number of electrons per unit volume
occupying states in the energy range between E and E + dE is :

nCE)dE = pCE) fCE)dE (1)

where pCE) represents the number of possible states per unit volume (including the spin) and fCE)
is the Fenni-Dirac distribution i.e., the probability of the state energy E to be occupied

E-EF)Jl
f(E) = [ 1 + exp ( ~ U (2)

EF is called Fenni energy and corresponds to a level which has a probability of 1/2 of being
occupied (fig. 2). In the case of insulators and semiconductors, EF is usually located in the band
gap. Therefore, EF is not a level which can actually be occupied by an electron.

Note that for non-degenerate semiconductors the Fenni level EF lies within the band gap
and is several KT from the band edges Ec and Ey. In this case, the Fenni-Dirac distribution may
be approximated by the Boltzmann distribution

f(E)=exp ( EF-E)
~ (3)

At a temperature other than zero, the thennodynamic equilibrium concentration, ilc of electrons in
the conduction band is :

Ec, Top
I1c = J peE) f(E,T) dE (4)
Ec

Near the bottom of the conduction band the density of states is:

(5)

where mc * is the effective mass of an electron near Ec

 195

The number of electrons in the conduction band is :

(6)

Similarly the expression for the thennodynamic eqUilibrium concentration of positive holes
in the valence band, Pv, may be readily obtained. The probability for a state of energy E to be
unoccupied is given by :

1 - fl:E) = [ 1 + exp (Ir.i'

EF-E)~-1
U (7)

the number of holes per unit volume may thus be written:

Ev
nv = f Py(E) [1 - fl:E,T)] dE (8)
Ev, bottom

where the density of states Pv(E) near Ey is :

(9)

where mp* represents the effective mass of a hole near the top of the valence band. If EF lies more
than about 4 KT above Ey, the number of holes in the valence band is :

(10)

3. Light absorption by semiconductors

In examining the optical properties of semiconductors it is possible to single out schematically

several different regions : a region of small absorption for infrared for the lowest frequencies
through interaction of light with the oscillations of the lattice (phonons) ; a region characterized by
electronic transitions [visible and the adj acent parts of the spectrum on both sides J. In this region
an important phenomenon is observed. This is a sharp growth of absorption caused by the
interband optical transitions in which the energy of the photon is used to create electron hole pairs.
This is called fundamental absorption.

Interband transitions are subdivided into direct and indirect. There are semiconductors in
which the bottom of the conduction band and the top of the valence band are located at the same or
almost the same point of the momentum space and there are semiconductors in which the maxima
and the minima of the bands do not coincide (fig. 3a and b).

In sufficient pure crystals the indirect transitions are usually less probable than direct ones.
In contrast, in relatively heavily doped semiconductors the probability of indirect transitions is
relatively larger due to the interactions between electrons and impurities (fig. 3c).

4. The semiconductor.liquide interface

To describe the electrochemical and photoelectrochemical behaviours of semiconductor electrodes

in tenns of energy levels, electric potential (inner or Galvanic potential), applied potential
196

NQf"maty empty
bond
,-----""""l--, E C. 109

Lr-----'
__ _______ Ec
EF

lm~iY· -t2;;i /-

r"--- - - - , E,
,,
, L _ _ _ _ _-..,.-' E , bottom
Normoly fitt~
v
I

,I,
b,,,'
If (E)
H

Fig. 2 _Schematic representation of lhe Fenni-Dirac distribUtion function PCE), band structure and
density of states function p(E) for an n-type semiconductor.

E E
oj tondudion bond bj oj

E,
r'- ,

vnLtrK:t bond
L , - -K

Fig. 3 - The scheme for direct ( 1) and indirect (2) optical transitions from the valence band 10 the
conduction band: 3.) maxima and minima of the bands coincident; b) maxima and minima of Ihe
bands non-coincident ; c) opticallnJnsitions when some or the electron levels in the conduction
band are occupied (n-type semiconductor).
 197

(polarization potential) we have to take into account that the surface physicist's convention for the
sign of potentials is not the same as that of the electro chemist. It is an established tradition in
electrochemistry that the positive direction of electrode potential is chosen to correspond to
increasing positive charge on the electrode. The potential of the reference electrode, is related to the
potential1jf(el) = 1jf(_00 ) in the bulk of the solution. In semiconductor surface physics, potentials
are read from the bulk of the semiconductor, I\>sc = 1\>(00). Then an enrichment of the space charge
layer with electrons corresponds to a positive potential at the surface in accordance with the
Boltzmann distribution. As a consequence the directions of the accepted potentials axes of
electrochemistry and semiconductor physics are directly opposed.

When a semiconductor is placed in contact with a solution we have to consider the formation
of surface charge as in the case of the metal. However, metal-semiconductor or
semiconductor-electrolyte junctions differ from one another in that in the latter, the charge in the
semiconductor side of the interface extends over a much larger distance (due to the lower
concentration of charges carriers) than in the former. This fact, which is true either for intrinsic or
extrinsic semiconductors of both types, gives rise to the so called space charge layer; also named
depletion or exhaustion region because the subsurface region and the surface itself are depleted in
free carriers (ionized donors in n-type and ionized acceptors in p-type semiconductor). The
building of the space charge promotes a drop of potential close to the semiconductor surface and
leads to a change in the electrons potential energy with the distance from the surface to the interior
of the semiconductor. Since the charge on the electron is negative, the bands are bent downwards
if I\>sc > 0, and upwards if I\>sc < 0, with a total magnitude of band bending equal to Ie~l\>scl

°
In the special case when I\>sc = the bands remain unbent ("flat") right up to the surface. At
the flat band potential, by definition, the potential on the semiconductor surface coincides with the
value of the potential in the semiconductor bulk.

The different forms of space charge layer that can be developed on a semiconductor are
illustrated in figure 4. The depletion layer (figure 4a) appears when the majority carriers are
extracted in moderate amounts. The surface region is depleted of majority carriers and minority
carriers are not present, so the surface region is depleted of both forms of surface carriers. The
surface region is very insulating except when the drop potential in this region is very small. The
space charge to insert into Poisson's equation is due only to ions or trapped charges in the depleted
region. Figure 4b corresponds to with a situation of accumulation layer which is obtained when
majority carriers are injected into the semiconductor from the surface, and these excess majority
carriers form the space charge. The inversion layer in figure 4c, results when the number of
majority carriers extracted is very high. The band bending cannot supply all the majority carriers
required and minority carriers have to come from the minority carrier band. In the case of an n-type
semiconductor (figure 4c), some electrons are extracted from the valence band. Thus holes are
injected in this band. The term inversion describes in fact the situation of an n-type semiconductor
which is inverted to p-type at the surface.

The situations represented in figure 4 can be produced, in principle, by varying the electrode
potential. Under cathodic polarization the bands bend down at the surface : there is an
accumulation layer with an n-type semiconductor and' a depletion layer in the case of a p-type
semiconductor. Under anodic polarization the bands bend up and a situation of depletion and
accumulation at the surface represents respectively the n- and p-type semiconductor.

If under the influence of polarization alllhe energy levels at the surface remain constant with
regard to the positions of the energy levels in lhe electrolyte. the energy band edges are considered
to be pinned at the surface (figure 5a and b). This is the case when the potential drop in the
Helmholtz double layer remains pratically unchanged. Conversely if we assume that when the
semiconductor is polarized. it is the potentiel drop in the Helmholtz double layer that changes, then
198

oj b) c)I--_ _ _-,

___..:.......;~ Ec
I------'-"'-"'-'X- EI'

Ev Ev

Fig. 4 - Types of space charge layers on n-type semiconductor (a) depletion layer;
(b) accwnuiation layer; (c) inversion layer.

aJ E b) E c) E

e6. H

Ec
f----EF

EV

EV ----EV

Fig. 5 - The energy scheme for a semiconductor-electrolyte junction to which an external voltage is
applied. Transition from a to b, illustrates the pinning of the band edges and transition from a to c,
the pinning of the Fermi level at the surface of a semiconductor electrode.
 199

the energy levels at the surface are shifted with regard to the energy levels in the solution by a
value of 8$H (figure 5c).

THERMODYNAMIC ASPECfS

For a metal the thermodynamic work function (Wt) is interpreted as the work necessary to
withdraw an electron from the Fermi level EF to the vacuum. In the case of a semiconductor Wt
does not correspond to EF because this energy level is located in the band gap. However. the
thermoemission current which measures the magnitude of Wt • is dependent not only on the energy
of the electrons but also on their composition. For semiconductors. the number of electrons is
proportional to exp (EF - Ec)!KT. Thus the thermoemission depends firstly on the number of

electrons in the conduction band. and secondly, on a factor exp [- irJ which gives the

probability of thermoemission of electrons from the bottom of the conduction band and where X is
the electron affinity.

The work function in semiconductors is highly dependent on surface an space charges and
varies with the applied potential. On the contrary X remains constant. When the bands are bent at
the surface (fig. 6), thermodynamic equilibrium in the electrode/solution system is achieved by
electron-ion exchange at the interface.

Galvani potential results from the difference in electrostatic potentials between the two
contacting phases and compensates for the difference in their chemical potentials. This electrode
potential is made up of the potential drops at the various interfaces (metal/semiconductor.
semiconductor/electrolyte contacts and also of ohmic voltage drops in the bulk phases when the
electric current is flowing). It should be pointed out that the electrode potential. with respect to a
given reference electrode, is. unlike the Galvani potential at semiconductor electrolyte interface.
only determined by the chemical potentials of the redox couple in solution i.e .• the semiconductor
behaves as an inert electrode.

The change in conduction mechanism at the interface implies the occurrence of an

electrochemical reaction. to which the semiconductor itself either mayor may not participate. We
will first treat the latter case, in which hence the electrochemical reaction is essentially an electron
transfer process. This transfer is assumed to take place by electron tunneling from occupied levels
at one side to empty levels at the other side of the interface. Thus to describe the electrochemical
behaviour of the semiconductor we have to take into account the energy distribution curves of
occupied and empty levels in the redox system.

Let us consider a cycle represented in figure 7a. consisting of a first process in which an
electron is removed from the reducing agent and taken out of the sphere of influence of the
electrolyte without changing the solvation structure (Franck-Condon principle) ; this process
requires an energy 1. In a second consecutive process the oxidized species relax towards its most
stable solvation structure and an energy ER is released. ,The following reaction step (3) represents
the capture of an electron by the oxidizing agent without change in solvation structure.
accompanied by the release of an amount of energy A. The last step (4) describes the relaxation of
the reducing agent formed towards its most stable solvation structure; for simplicity. it is assumed
that the amount of energy released in the second step is the same as in the last step. This amount of
energy is known as the reorganization energy ER of the redox system.

As the total energy change in the cycle is zero. it follows :

 200

U vacuum
aJ E

w :x w
Ec
...y._ _ _
- - - - - EF t:

-'---l---E v
- - - - Ev

Fig. 6 - Energy diagram, X the electronic affinity, W the work function and Sthe ionisation energy
a) at flat band b) under anodic polarization.

E
b)

vocum level
oj !
-A

iR
I
EOox

E~ed
W(E)
O(E)

Fig. 7 - (a) the most stable occupied and empty electrons level in a redox electrolyte. (b) energy
scheme of a redox electrolyte.
 201

o 0
I-A=2ER=E ox -E re d (11)

where EO and EO d are the most stable respectively empty and occupied levels in the redox
ox re
electrolyte.

Thermal fluctuations in the solvation and inner sphere structures lead to distribution of
energy levels W(E) around the most stable configurations.

When nred and nox ions of reducing agent and oxidizing agent respectively are present per
unit volume of electrolyte solution (these quantities being proportional to the molar concentrations
cred and cox), the numbers of ions of reducing and oxidizing agent per unit volume and unit
energy, Dred(E) and Dox(E), i.e. the numbers of filled and of empty electron levels in the redox
electrolyte per unit volume and unit energy, are given by the following expressions respectively:

(12)

Dox(E) =Wox(E).nox (13)

In figure 7b, the functions W(E) and D(E) are schematically represented in the case where
cred/cox = 4. In view of the definition of the Fermi level, it is clear that the intersection of the
Dred(E) and Dox(E) curves (doned lines in the figure) constitutes the Fermi level of the redox
electrolyte at the given composition.

ELECfRODE KINEfICS

L Quantum electrode kinetics

Electrode kinetics basically deals with charge transfer processes such as electron and proton
transfer at the solution-electrolyte interface. Since the electrons and protons are quantum particles,
the theory of quantum mechanics can be applied. Electrode kinetics concerns the rate of transfer of
an electron between two electronic levels of equal energy and can be described by time-dependent
perturbation theory. One electronic level is in the electrode and the other is in the acceptor ion at the
interface. This process of transfer of an electron between equal energy levels is termed radiationless
transfer. If we consider the interfacial electron transfer step as rate determining, the most general
expression for the electrochemical current density is given by :

i C =e JJ C(x) G(E, x) W(E,x) peE) ftE) dE dx (14)

where e is the electronic charge, C(x) is the concentration of acceptor or donor atoms or ions at any
distance x from the metal electrode, G(E,x) is the probability of having the acceptor (for the
cathodic case) or donor (for the anodic case) at any energy level E and at a distance x from the
electrode, W(E,x) is the transition probability of an electron from the electron level in a metal to an
acceptor ion in the solution (for the cathodic case) or vice versa.

Since most of the acceptor and donor ions are situated in the outer Helmholtz plane, one
may remove the integral over distance and write in the simplified form as :

i C = eC f G(E) WeE) peE) ftE) dE (15)

From these different terms, the most widely treated by quantum mechanics is WeE), the
202

transition probability of electrons at the interface. The probability GCE), has been done by all
workers in terms of classical distribution function, and has been included in the term WeE). The
Fermi distribution fCE) and the density of states p(x) are obtained using respectively the quantum
statistics and the quantum mechanical treatment of the free electron theory.

2. Kinetics of semiconductor reactions

A theoretical treatment of semiconductor reactions was first put forward by Gerischer (5) (6).
Other treatments were made by Levich (7), and Dogonadze (8). The electrode reactions have been
reviewed by different authors (9) (10). There are some important differences between the electron
transfer processes at a metal-solution interface and that at a semiconductor-solution interface. In the
case of semiconductors, electrons both from the conduction and holes from the valence band
participate to the transfer process. Further the electric field penetrates into the semiconductor (in the
space charge region). As a consequence, the interfacial drop is partially inside the semiconductor,
partially in the interface i.e., in the Helmholtz double layer.

The following two reduction processes have to be considered:

(16)

Ox +± Red + p+ (17)

where the symbol p+ signifies a positive charge (hole) in the valence band. In the case of a metallic
electrode only the first of these two reactions can take place. The transfer of charges consisting in
removal of electrons from the semiconductor or injection of holes contribute to cathodic current,
and those consisting of the injection of electrons or removal of holes contribute to anodic current
(figure 8).

Due to the above-mentioned differences i.e., the participation of two types of charge carriers
and also as a consequence of the influence of the applied electric field on the concentration of these
carriers near the surface, the rate of electron transfer to the redox system in a cathodic process
occurring over all energy ranges is proportional to the number of occupied states in the
semiconductor and the number of empty states in the redox system, and this can be represented by
the expression:

+00

iC = e f WeE) D (E) D occup (E) dE

ox sc (18)

The corresponding expression for the anodic current density due to transfer of electrons from
the reduced species in the electrolyte to the empty electron levels and holes in the semiconductor is
given as:

i a = e f + 00 WeE) Dred(E) D vacant (E) dE (19)

_ 00 sc

where Dox and Dred are functions of the density of the vacant and occupied states in solution
vacant occupied
while D and D are the corresponding terms for the semiconductor electrOde with
sc sc
vacant occup . .
D = peE) feE) and D = peE) [1 - feE)]. Taking mto account Eq. (4) and (8) we obtain
oc sc
 203

for the partial cathodic currents, i~ and i~, passing through the conduction and valence band the
following expressions :

.c .Bc, Top
1 =eJ W(E) DoX<E) D OCCup (E) dE (20)
n Ec sc

i C =e
p
rEv, bottom
W(E)DoX<E)D OCCup (E) dE
sc
(21)

a a
and for the partial anodic currents, in and ip'

.a .Bc, Top
1 =eJ W(E) D (E) D vacant (E) dE (22)
n Ec red sc

ia = e
p
r
Ev, bottom
W(E) D
red
(E) D vacant (E) dE
sc
(23)

the distribution of oxidized, Dox' and reduced, Dred, species in solution have been expressed as

Cox [
(E_EO-y.~
ox
Dox= exp - 4KTE (24)
...; 41t KT ER,ox R, 0

(25)

o 0
Cox and ered are the concentrations of the oxidized and reduced species, E
ox and Ered correspond
to the energy levels indicated in figure 8. ER,red, and ER,ox are the reorganization energies.

Figure 9 shows schematically the distribution of electron energy states at the interface of a
solution containing a redox couple in eqUilibrium with n- and p-type semiconductor. At
equilibrium the electrochemical potentiels of the solide and the solution coincide. Thus the
distributions of the vacant and occupied electron states in the solution, Dox and Dred' are shifted
with respect to the positions of V redox by ER,ox and ER,red. They may overlap or not overlap
. ... vacant occup . . ..
With the distrIbutions of the electron states, Dsc and Dsc ' ill the solId. A contnbutlOn to the
electron current is only made by these energy intervals for which there is overlap of the density of
states.

The electrode polarization leads, firstly to a change in the absolute magnitudes of the
densities of states, D sc ' Dox and Dred and, secondly, to a change of their relative positions.
204

aj E E

Fig. 8 - Partial electron-transfer anodic and cathodic currents in the conduction band and in the
valence band. Semiconductor placed in contact with a redox system.

E E
0) b)

EC ER EC
EF - - -
ER
Do<

o red
ER
EF--- -
EV EV ER

Fig. 9 - Distribution of energy states at the junction between (a) an n-type semiconductor, (b) a
p-type semiconductor and an electrolyte which contains a redox couple.
 205

Consequently, the energy intervals for their overlap also change.

Assuming different boundary conditions and particularly that charge transfer occurs only
from energy regions close to the bandgap, the partial cathodic and anodic currents can be written
(11).

iC =ecox s ( lIT )1/2 k(E c s)ns exp [_ (Ec,s - ~~X - EW2 ] (26)
n , 47tER' R

.c ( lIT )112 [ (Ev,s - Vredox - ER)2] (27)

1 = ecOX s - - k(Ev S)Nv exp - 4KTE
p '47tER' R

(28)

.a ( lIT )112 [ (EV,S - Vredox + ER)2]

1 = eCred s -- k(EV S)pS exp - 4KTE (29)
p '47tER' R

where the index s means surface. Vredox is the Fermi energy of the redox electrolyte and arises
from the analogy of the Fermi level, EF ' on the semiconductor
o 0 . .
Vredox = E ox - ER,ox = Ered + ER,red (see fig. 8). EquatIons (26) - (29) show that the cathodIC
c
current through the conduction band in is proportional to the surface electron concentration, n s ' in
the surface while the cathodic current through the valence band is proportional to the density of
a a
states Nv . Similarly the components of the anodic current, i and i ,are proportional to the density
n p
of states, N c ' and the hole concentration, Ps' in the surface respectively. Note that there is a
exponential relationship between the values of ns and Ps and the applied potential while Nc and N v
are approximately potential independent.

If we consider that the entire additional potential drop is concentrated in the layer of the
semiconductor near the surface (Le., when the band edges are pinned), then we get the following
relations for the partial currents :

(30)

(31)

where i 0 and iO are the equilibrium currents without polarization. The dependence of the currents in
n p
and ip on applied potential is shown in figure 10. One observes that the resultant electron current in
o
increases exponentially under cathodic polarization whilst remaining limited by the value of in
206

under anodic polarization. In contrast, the current ip increases under anodic polarization
o
(exponential dependence) and is limited by the value ofip under cathodic polarization.

STABILITY AND DECOMPOSmON OF SEMICONDUCTORS

A semiconductor placed in contact with an electrolyte can decompose into electrochemical reactions
either by oxidation or reduction. The oxidation process is caused by reactions with holes (12) (13)
while the reduction process results from reactions involving electrons. Thus, in the case of a
binary semiconductor MX the equations for the partial electrochemical reaction of cathodic
decomposition can be written as :

(32)

where M and X are respectively the electropositive and electronegative components of the
compound.

On the other hand the reaction of anodic decomposition is :

(33)

Combining reactions (32) or (33) with the electrode reaction of a reference electrode (for instance
the standard hydrogen electrode) we obtain for the cathodic decomposition reaction

(34)

and for the anodic decomposition

(35)

The calculation of the corresponding change in free energy for these reactions from the
thermodynamic properties of the reagents makes it possible to determine the electrochemical
potentials (standard equilibrium potentials) of cathodic decomposition, Vdec,n' and anodic
decomposition, Vdec p' Knowing these potentials one can predict the stability of the semiconductor
from the thermodynamic criteria for the occurence of an electrode reaction. according to which a
corrosion reaction is possible if :

V> Vdec,n (or E < Edec,n) (36)

for the cathodic reaction, and

V <Vdec,p (orE> Edec,p> (37)

for the anodic reaction.

An energy level model of electrode stability has been presented by different workers (14)
(15). In figure 11 the relative positions of V dec,n and V dec,p are compared with the positions of
the semiconductor band edges. The semiconductor is absolutly stable. if the levels for cathodic and
anodic decomposition are beyond the limits of the band gap (figure IIa). However such condition
 207

Log i
p-type
n-type
/ ·0
IP

·c
IP
.(]
In

n cathodic- - n anodic

Fig. 10 - Typical current voltage curves for the partial electron-transfer currents in the conduction
band and in the valence band.

oj E b) E c) E d) E
Vdec, n Vdec,n

Ec Ec Vdecn Ec Vdec,n Ec
Vdec P V
Ev Ev Ev Ev
Vd~S

Fig. 11 - Scheme illustrating the thermodynamic stability of a semiconductor with respect to

decomposition: a) the semiconductor is absolutely stable; b) stable with respect to cathodic
decomposition; c) stable with respect to anodic decomposition; d) unstable.
208

does not occur with usual materials and more often it is found that one of the decomposition
potentials lies within the band gap, making the correspondent decomposition processes possible
(figure lIb, and c). Finally, the semiconductor is susceptible to both types of decomposition ifthe
two potentials Vdec,n and Vdec,p are within the band gap (figure lId).

Note that in a more realistic situation, we also have to consider redox reactions with species
in the electrolyte. Then the conditions from corrosion of the semiconductor can be written as:

VRedox > Vdec,n (38)

for the cathodic decomposition, and

VRedox < Vdec,p (39)

for the anodic decomposition.

These thermodynamic conditions for decomposition, however, do not imply that the
decomposition really occurs. Kinetic aspects must be taken into account. For instance, if the redox
level, V redox, supplying an electron to a valence band hole is substantially above EV, the
transition may be slow, allowing corrosion to proceed. Further it is not only the concentration of
holes which controls anodic decomposition (or electrons in cathodic decomposition). The position
of the band edges at the surface which depends on the voltage drop ~PH in the Helmholtz double
layer are also important parameters.

THEORETICAL ASPECTS OF THE PHOTOELECTROCHEMISTRY

If the semiconductor electrode is illuminated by light of a suitable energy hv, electrons can be
excited from occupied to unoccupied states. A phenomenological presentation of the processes is
represented in figure 12 assuming that the electronic structure near the interface can be described
by a Mott-Schottky barrier model (i.e., with a parabolic band bending). The electron pair
formation is the basic processes for photoelectrochemical effects. Further, if a depletion layer
exists at the semiconductor/electrolyte interface, the electron-hole pair is separated under the
electric field formed in the space charge region. This is shown in figure 13 for n-type and p-type
semiconductors. According to the sign of the field, anodic or cathodic photo currents are obtained.

It is important to note that the electric field, which is present in the space charge region at
equilibrium conditions is disturbed under the influence of light as a consequence of charge
separation. The counter field which results from this phenomenon shifts the Fermi level in the bulk
relative to the energy levels in solution. Figure 14 shows the polarization effect of the
semiconductor electrode under illumination (photopotential change of -e~<1>sc).

If we consider the semiconductor-electrolyte interface as a simple Schottky barrier, we can

describe the behavior of the photocurrent upon external parameters such as wavelength, applied
potential, conductivity, etc ...

The general equations used by several authors relative to hole concentration are the
following (14)-(16)
 209

H.L

Fig. 12 - Schematic representation of a semiconductor interface, and typical phototransitions

involved in photoelectrochemical processes (b, sc and s means bulk, space charge and surface
respectively).

oj E bJ E

EC
hv

donor
EF-------
EV hv

EV

Fig. 13 - Effect of illumination in depletion layer and electron transfer reactions at the
semiconductor-electrolyte interface of an n-type (a) and a p-type (b) semiconductor.
210

Dp + *
d2 (x) d
+ IIp dx (p*(x) ~(x) + Up(x) - IXA. llq exp (- IXA.(X» (40)

and

d*(x)
i*(p) = e Dp T + e IIp p2(x) ~(x) (41)

which enables the recovery of the photocurrent i*. In these equations dp is the diffusion coefficient
of holes, p*(x) is the concentration of these holes at distance x, Jlp is the hole mobility, Sex) is the
value of the electric field at distance x, Up(x) is the recombination current at distance x, IXA. is the
optical absortion and llq is the quantum efficiency.

The theoretical calculation of the real charge distiribution of the mobile charge carriers in the
stationary state of illumination is extremely complicated. Some solutions have been outlined based
on simplifying assumptions.

For the ideal Schottky barrier, the model which describes the photocurrent behaviour has
been proposed first by Ganner (17) for the metal-semiconductor junction and then extended to
semiconductor electrolyte junction by Butler (18).

The Butler model, which therefore considers that the semiconductor-electrolyte junction
closely looks like a Schottky barrier, is very useful for determining the band structure parameters.
It assumes mainly that all the carriers generated in the depletion region contribute to the
photocurrent. These carriers consist of two basic components: the carriers generated within the
depletion region and the other relative to carriers produced in the bulk.

Following Gartner (17), for photons incident upon the interface, the generation of
electron-holes pairs is given by :

g(x) = J oCt. exp ( - ax) (42)

where 10 is the photon flux, IX the optical absorption and x the distance into the semiconductor.
The photocurrent flowing through the semiconductor consists of charge carriers generated by the
light in the space charge region and charge carriers generated in the bulk that diffuse to the
depletion layer. If it is assumed that all carriers generated in the depletion region contribute to the
photocurrent, then this component is given by :

* w
id =-e! g(x)dx=eJo (e- IXW -1) (43)
ep 0

where w is the depletion-layer width and is given by

(44)
 211

to and t are the dielectric constants of the vacuum and semiconductor. ND is the dopping
concentration. K and T have the usual meaning. and Vfb is the flat band potential. On the other
hand. the diffusion current of the charge is given by

(45)

Combining (43) and (44) we obtain for the absolute value of the photocurrent

(46)

This equation shows how the photocurrent depends on the radiation characteristics (through 0:(00)
and the light flux Jo ). the kinetic characteristics of the minority carriers (through Lp), the
concentration of the majority carriers and the electrode potential (through w).

Materials with wide band gap have small hole mobilities. I1p. and thus the diffusion lengths
may be dominated by different mechanisms. Diffusion-controled bulk recombination often
determines carrier lifetimes rather than trapping factors. The diffusion length is defined as

(47)

where 1> is the hole lifetime. The Einstein relationship

(48)

relates the diffusion constant to the hole mobility. The hole recombination rate is described by

I 41te
-=-(11 +1te)Nn (49)
1> t P
Combining Eqs. (47)-(49) we obtain the following expression for the hole diffusion length :

(50)

1. Quasi Fermi levels

The kinetics ofphotoelectrochemical reactions considered in terms of the transport equations. gives
an accurate solution. However, in some cases-for example. with relatively complicated boundary
conditions a quasi-thermodynamic approach based on the concepts of quasi-Fermi levels proves to
be more effective than the proceeding kinetic approach.

In the region of a semiconductor, where light is absorbed and electron-hole pairs are
212

ellEI~~
!II

-------- _/ '"
hV

EC _ _ _ _ _ _ _-l
EF

EV

Fig. 14 - Influence of light on band bending of an n-type semiconductor in depletion layer region.

E E
aJ b)

Ec
hv E; / - - Freacflon
/
I hv
I
I •
F reaction ,/ IPEF
EF~--~- __
EV

Fig. 15 - Quasi Fermi levels in depletion layers of boIh n-type (a) an p-type (b) semiconductor:
steady state wiIh electrochemical consumption of minority carriers.
 213

generated. equilibrium between electrons and holes is no longer preseIVed. Both electrons an holes
are in an excess and in the steady state. recombination and transport locally compensate for the
generation rate. However. the distribution of electrons and holes over the quantum states of their
respective bands remains in thermal eqUilibrium. As a consequence, the increase in free energy can
be described by the excess concentrations. and the free energy of the charge carriers can be locally
* and pEp-* Unlike the case of complete
represented by their respective quasi-Fermi levels (12). nEF
* * * *
thermodynamic equilibrium where nEF =pEF =EF, the values of nE F and pE F are different from
each other (See figure 15) and can be determined by the following relations

(51)

(52)

If we denot the excess concentrations by t.n* and t.p* we can write

n* = n + t. * p* = p + t. * (53)
n P
where n and p are the electron and hole concentrations in the dark.

If the intensity of the illumination is not too high, the magnitudes of t. n* and t. p* are
proportional to the logarithm of the light intensity. The further shift in the quasi-Fermi levels
become less pronounced due to the increase in recombination.

When the semiconductor is illuminated. the transfer of electrons and holes to the solution
increases as a result of the increase of charge carriers. In these conditions the potentials V p and
V n. at which the electron and hole concentration respectively become zero is no longer coincident
with each other or with the equilibrium potential V redox. In other words, under the influence of
light. it is as if the semiconductor electrode has two values of electrode potential simultaneously.
one which determines the rate of reaction with the participation of holes from the valence band.
while the other determines the rate of reaction with the participation of electrons from the
conduction band.

2. Photocorrosion

Let us now examine the influence of light on the corrosion processes assuming that the role of light
is restricted to the excitation of the electron-hole system in the semiconductor. Since illumination
essentially affects the minority carrier concentration, corrosion changes can be related to an
effective increase in the current for the reactions occurring through the minority carrier band. That
is to say, with the increase of the anodic current of decomposition, i~. in the case of an n-type

semiconductor and with the increase of the cathodic current i c relative to the redox system in the
n
case of a p-type semiconductor. Figure 16 shows the influence of the n- and p-type conductivity
on the increasing rate of the corresponding partial reactions. The direction of the shift of the
214

_ _ _ Ip . o·
'0
~...---Ip

Vdec, p Vccirr Vcorr Vredox

p_type '0
In

-" \
\
\
\
\
\
\

Vdec, p V~orr Vcbrr Vredox

Fig, 16 - Influence of light on the electrochemical and corrosion behaviour of an n-type and a
p-type semiconductor.
 215

corrosion potential is also indicated.

If we consider the problem of the photocorrosion on the basis of the quasi thermodynamic
approch developped by Gerischer (12) and Bard and Wrighton (13), we have to take into account
the "splitting" of the Fermi energy level, EF, and the formation of both nE;
and pE; quasi-Fermi
energy levels. Then, the conditions for cathodic photodecomposition with the participation of the
conduction band should be

(54)

where Vn is the electrochemical potential which corresponds to nE; in the semiconductor surface.

On the other hand the anodic photodecomposition reaction with the participation of holes
occurs when:

Vp > Vdec,p (55)

where VP is the electrochemical potential which corresponds to pE; in the semiconductor surface.
Eqs. (54) and (55) establish the thermodynamic condition favorable to cathodic or anodic
decomposition. However, as for situations defined by Eqs. (36) to (39), the kinetic aspects often
play the most determinant role.

When examining photocorrosion, one should bear in mind the possibility of competition
between photodecomposition and other electrode reactions. Indeed, the photocorrosion of a
semiconductor may be suppressed by the addition of suitable agents with sufficient oxidizing or
reducing influence. For instance, anodic photocorrosion is avoided if a redox system with a redox
potential more negative than the potential for semiconductor decomposition:

Vredox> Vdec,p (i.e., Eredox < Edec,p) (56)

3. Role of surface states

Surface states playa very important role in the electrochemical behaviour of semiconductors. They
can easily exchange electrons with redox couples in solution and increase the rate of recombination
processes under illumination. Electron exchanges with the bulk of the semiconductor will also be
fast if the surface states have energy levels close to one of the band edges. Surface states may be
described as having a donor or acceptor character. Donor states too far from the band edges can
only remove holes from the valence band but cannot inject electrons into the conduction band. This
results from the fact that surface states only pick up charge in the downhill direction with respect to
energy. Only in the case of very highly doped semiconductors are the reverse processes (injection
of holes in the valence band or injection of electrons in the conduction band) possible. Another
considerable effect of surface states, when present in considerable concentration, is that of their
contribution to the charge of the Helmholtz double layer. Further, surface states can pin the
quasi-Fermi level by fast recombination processes or rapid electron transfer to a redox system in
solution.

Figure 17 shows an n-type semiconductor with donor surface states in contact with an
electrolyte. In this case surface states are neutral in the occupied state and positively charged in the
vacant state. At equilibrium (figure 17a) this positive charge shifts the band edges downwards
compared with their position at flat-band potential i.e., when all surface states are occupied. Under
216

the influence of light there is a possible increase in the positive charge by hole capture from the
valence band (cathodic polarization) and a further downwards shift of the band edges. On the other
hand, in the bulk the band edges shift upwards as a consequence of the decreased space charge
(figure 17b).

Finally the electrode stability can be related with the light intensity through its influence on
the energy of the quasi-Fermi levels and with the concentration of the surface states which affect
band bending and electrode kinetics. It is the influence of surface states on the electron and hole
capture rates that deteImines pinning and establishes the conditions of competition between anodic
decomposition of the semiconductor and partial reaction of the redox system.

Figure 18 shows an n-type semiconductor with surface states of acceptor character which
are negatively charged if occupied and neutral if vacant. In this case the accumulation of charge
promotes an uplifting of the band edges at the interface (cathodic polarization) which improves
stability against anodic decomposition but may introduce cathodic decomposition
*
if nEF,S<Edec,n'

REFERENCES

1. Kittel, C. (1983) Introduction ala Physique de !'Etat Solide 5e edition, Dunod Paris.

2. Mathieu, H. (1987) Physique des semiconducteurs et des compos ants electroniques,

Masson Paris.

3. Sze, S.M. (1981) Physics of semiconductor Devices 2nd edition, John Wilew& Sons

4. Jarzebski, Z.M. (1973) Oxide Semiconductors, International Series of Monographs in the

Science of the Solid state Vol. 4, Pergamon Press.

5. Gerischer, H. (1961) Advances in Electrochemistry and Electrochemical Engineering,

P. Delahay, ed., Vol. 1, Wiley (Intersience), New-York.

6. Gerischer, H., (1970) Physical Chemistry An Advanced Treatise, H. Eyring, ed., Vol. 9A,
Academic Press, New-York.

7. Levich, V.G., (1970) Physical Chemistry An Advanced Treatise, H. Eyring,

D. Henderson, and Y. Jost, eds., Vol. 9B, Chapter 12, Academic Press, New-York.

8. Dogonadze, RR (1971) Reactions of Molecules at Electrodes, N.S. Hugh, ed., Wiley

(Interscience), New-York.

9. Myamlin, V.A. and Pleskov, Y.V. (1967) Electrochemistry of Semiconductors, Plenum

Press, New-York.

10. Vijh, AK. (1973) Electrochemistry of Metals and Semiconductors, Marcel Dekker,
New-York.

11. Pleskov, Yo.V. and Gurevich, Yo, Ya (1986) Semiconductor Photoelectrochemistry,

Consultants Bureau New-York.
 217

12. Gerischer, H. (1977) Semiconductor Liquid-Junction Solar Cells, Proceedings of a

conference on the Electrochemistry and Physics of Semiconductor liquid interface under
illumination, Edited by Adam Heiler, Vol. 77.3 (the E1ectrochem. Society, Princeton).

13. Bard, A.J. and Wrighton, M.S. (1977) ThernlOdynamic potential for the anodic dissolution
of n-type semiconductors, J. E1ectrochem. Soc., 124, 1706-1710.

14. Wilson, R.H. (1977) A model for the current voltage curve of photoexited semiconductor
electrodes, 1. Appl. Phys. 48, 4292-4297.

15. Reiss, H. (1978) Photocharacteristics for electrolyte-semiconductor junctions,

J. Electrochem. Soc. 125, 937-948.

16. Jarret, H.S. (1981) Photocurrent conversion efficiency in a Schottky barrier, J. Appl. Phys.
52,4681-4689.

17. Glirtner, W.W. (1959) Depletion-layer photoeffects in semiconductors, Phys. Rev. 116,
84-87.

18. Butler, M.A. (1977) Photoelectrolysis and physical properties of the semiconducting
electrode W03, J. Appl. Phys. 48 nOS, 1914-1920.
PHOTOELECfROCHEMISTRY: APPLICATION TO THE STUDY OF PASSIVE FILMS

M. DA CUNHA BELO
CNRS - CECM,
15, rue Georges Urbain,
94400 Vitry-sur-Seine, France

ABSTRACT. The application of photoelectrochemistry as a very important "in situ" technique for
the study of the passive fIlms is examined. Experimental techniques, specific features of the
electronic structure of the passive film and capacity, approach of the interface film-electrolyte
(Mott-Schottky plots) are firstly presented. Then, photoelectrochemical measurements and their
application to the study of passive fIlms formed on iron, stainless steel and titanium are discussed.
The consequences of irradiation and implantation treatments on the electronic structure and
amorphization of the passive film are also examined. It is generally accepted that passive films
might be largely non stoichiometric, highly disordered and possibly amorphous.
Photo electrochemistry appears as an interesting in situ technique for the study of structural
properties. Results illustrating the possibilities of this method are completed by a discussion of the
degree of order to be expected in vitreous and monocrystalline materials.

INTRODUCTION

Passive films have been intensively investigated in order to improve the corrosion resistance of
metallic materials in innumerable practical applications. Obviously, published data marked an
important contribution from a phenomenological point of view and are of great technological
interest. However, the understanding of such films in terms of electronic structure is still in need
of further research.

Recent researches (1-5) present photoelectrochemistry as a very interesting "in situ"

technique for the study of passivity. Indeed, most of the metallic electrodes show photoeffects
occurring under the influence of light. The use of photoelectrochemistry for the study of passivity
is largely based on the fact that the passive fIlm formed on different metals and alloys exhibits
semiconductive properties. Photoelectrochemical measurements can be carried out in situ and
directly related to electrochemical parameters such as potential and current. If we assume that the
concepts of solid state physics may be introduced to describe these phenomena, information can be
obtained on the electronic structure of the passive [tlm.

Experimental techniques and particular chemical and structural aspects of the passive film
will be firstly presented. Then, photoelectrochemical measurements will be discussed. Emphasis
will be put on passive films formed most particularly on iron, stainless steels and titanium base
alloys. The influence of the applied potential and pH of the corrosive medium on the electronic
structure of the passive film will be examined. The consequences of irradiation and implantation
treatments on the electronic structure and amorphization of the passive film are also discussed.

219
M. G. S. Ferreira and C. A. Melendres (eds.),
Electrochemical and Optical Techniquesfor the Study and Monitoring of Metallic Corrosion, 219-239.
© 1991 Kluwer Academic Publishers.
220

Several workers consider that passive films might be largely non stoichiometric, highly
disordered and possibly amorphous and recent work tries to point out the impact of the amorphous
character of the passive film on photoelectrochemical results. Published data demonstrate that
photoelectrochemistry is a suitable "in situ" technique for determining structural properties.
Experimental data illustrating the sensitivity of the method is completed by a discussion of the
degree of order to be expected in vitreous and noncrystalline materials.

1. Experimental techniques

The experimental techniques used in photoelectrochemical studies of passive fIlms associate the
classical equipment used in electrochemical studies of the passivation to the equipment employed in
photoelectrochemical research with semiconductors. In Fig. I the diagram of a typical experimental
set-up is outlined. Basically, the instrumentation for photoelectrochemical measurements consists
of a conventional three electrode cell, placed under potentiostatic control, with the working
electrode submitted to the light beam influence.

Photoeffects, in the case of passive fIlms, are very small and modulation techniques have to
be employed. The light beam is chopped and the modulated signal is amplified with reference to
the chopper frequency by a lock-in amplifier. The latter is essentially a narrow band filter. The
modulated signal is fed through the filter and the rectified output is smoothed by a reference
channel integrator with a variable time constant. The overall fIlter bandwidth as determined by the
time constant can be made much narrower than the bandwidth of conventional filters. As a
consequence this bandwidth allows measurements to be made even in the presence of noise which
is many orders of magnitude larger than the coherent signal. Under stable electrochemical
conditions, photo currents obtained from the metal-passive film-electrolyte system are typically
represented by several nA.cm- 2 . For this reason it is necessary to dispose of a monochromator
with high luminosity in a wide wavelength range. The choice of the light source is detennined
firstly by the value of the bandgap energy of the semiconductor and, second, by the flat
characteristic for the intensity change with the wavelength. Fig.2 shows the typical output spectra
for different sources. The Xenon arc lamp source (150 to 450 W) is usually employed on the study
of passive films because experiments are carried out in the wavelength range from 200 to 500
nanometers. The light beam is chopped and the modulated signal is amplified with reference to the
chopper frequency by the lock-in amplifier. The mechanical light chopper is usually employed on
the frequency range from a few Hz to several KHz. A very good jiter performance is necessary.
The rate of light modulation is followed by a photodiode. Calibration curves of the light source are
determined by a detector of known spectral sensitivity. The knowledge of the number of incident
or, if possible, adsorbed photons allows for a calculation of the quantum yield.

2. The metal-passive film-electrolyte system

2.1. CHEMICAL ASPECfS

In order to correlate photoelectrochemical results with solid state properties, it is necessary to have
an adequate description of the chemical composition of the passive film. This is possible with
Auger electron spectroscopy and sputter-etching of solid surface. The composition of a surface
layer in depth is established using high energy LMV transitions in order to obtain concentration
gradients relative to the different elements in the oxidized state. The analytical approach introduces
the sequential layer sputtering (S L S) model developed by S. Hofmann (6) which is expected to
be suitable especially for depth analysis of very thin films.
 221

Mono
~-I--...l:-'\'----_!'--------I-~ chromotor

Lock in
amplifier

x-v

Potentiostot

Figure 1. Schematic representation of an experimental set-up for photoelectrochemical studies on

passive films.

CIJ
g 1000 watt mercury lamp
il
'0
El00
u
CIJ~
/
g1J
o
'OE
._~gg 10
ON'
,!:;E
~.!2..
a.1J1
U):t:: hologen(filomentj lamp
~ 1.0
e'
u
~ 20LO~--3~0~O----4~OO~--~50~O----6~OO----~7~0~O----80~0--~900
Vovelength (no nometers )

Figure 2. Typical output spectra of different light sources.

222

Fig.3 shows the depth profiling analysis of a passive film formed on the surface of an AISI
304-type austenitic stainless steel in a borate-boric solution (PH 9.2) at room temperature after 2h
of film growth at O.8Y/SCE (7). It appears that according to its composition the passive film can
be divided in two layers: a Cr-rich inner layer and a Fe-rich outer layer. Auger analysis reveals
that the film formation potential and the temperature of the corrosive medium have little effect on
thickness and chemical composition of the inner layer. On the contrary the change of experimental
conditions affects both composition and thickness of the outer layer.

2.2. ELECTRONIC STRUcrURE

The difficulties in the characterization of the passive films from the electronic (structure) point of
view arise from the high doping levels, the deep trap and the presence of surface states. In
addition, the chemical composition of these films is usually not uniform. Furthermore, surface
charge regions can exist, not only at the film-electrolyte interface, but also at the contact between
the metal and the passive film. Indeed, for a common case of a semiconductor oxide growing on a
metal, there is no sudden discontinuity in potential across the metal-oxide interface. This is due to
the relatively low conductivity of the semiconductor oxide relative to the metal. Growth of the
passive film may be accompanied by the development of a space charge in the region near the metal
interface. The extent of the charge distribution in the oxide is given approximately by the Debye
Length in the semiconducting oxide :

ctsc = (E£oKT/e2ND) 1/ 2 (1)

where E is the dielectric constant, Eo the permittivity of the vacuum and ND is the donor
concentration for an n-type material.

The electron properties therefore vary with thickness and distance from the metal.

The complexity of the system metal-passive film-electrolyte makes the analysis of the
photoelectrochemical measurements very difficult. In a recent paper Gerischer (8) discusses, on
the basis of very simplified models, the influence of the applied potential with regard to the
photocurrents for different types of passive fIlms. It is supposed that the contact between a metal
and a passive film can be either blocking or ohmic, depending on the difference between the work
function of the metal and electron affinity of the conduction band or the ionization energy of
electrons in the valence band. The situation in the case of insulating, n-type and p-type oxide films
is successively examined.

Fig.4 gives a schematic representation of the n-type oxide film with homogeneous doping
concentration. In the case of a blocking metal-oxide contact, and, assuming that a depletion layer
exists at the same interface, the electronic structure of the oxide under illumination corresponds to
the schema of Fig.4a. The photocurrent flow will be either cathodic or anodic, depending on the
applied voltage. Fig.4b shows the case of an n-type oxide layer with an ohmic contact at metal-
oxide contact. The depletion is situated at the contact to the electrolyte. Under illumination the
photoelectrochemical behaviour of this kind of passive film will correspond to a situation of an n-
type semiconductor with an infinite extension if the oxide is thicker than the space charge region
plus the mean diffusion length of the minority carriers.

The electronic structure of an oxide film with variable doping concentration is obviously
more representative of the electronic structure of the most common passive films. Fig.5 shows the
situation of an oxide film with an n-type oxide at metal contact but nearly insulating in the bulk
 223

;::- OJ OXIDE....---}- METAL

~ / L(Fe,cr,Ni,O,C) ox

c
z

Fe

Figure 3. Depth concentration profiles for different elements in AISI 304-type passivated stainless
steeL

a)
b)
E E
F ref

Ey
Metal Oxide Electrolyte

Figure 4. n-type oxide film with blocking (a) and ohmic contact (b) under anodic polarization.
Figure 5. Oxide film with variable doping concentration: n-type at metal contact but nearly
insulating in the bulk.

f---,--V redox
Ec ell I{> ox
EF+-----·--=--+--~--

Ev
METAL OX) DE - Red9x system

Figure 6. Electron transfer processes in a thin passive film exposed to a redox system under anodic
polarization. (1) tunnel process into the conduction band. (2) resonance tunnel process via a low
energy donor tenn. (3) direct tunnel process at the Fenni level. (4) tunneling from the valence band
concerning hole generation.
 225

(n-i junction). Since an electric field exists inside the oxide, illumination will create a negative or
positive photovoltage.

These simplified models do not include surface states and deep doping levels which
fundamentally affect distribution and transport of charges. Further it is not sure if tunnel processes
must be taken into account. Tunneling is introduced by several workers in the mechanism of
electron transport. The influence of the applied potential on the anodic current when the electron
transfer proceeds as a tunnel process has been examined in detail (9). Fig.6 is an illustration of the
various electron tunnel transfer mechanisms.

2.3. CAPACITY MEASUREMENTS

Very important information on electronic properties of the passive film can be obtained by capacity
measurements which reflect the charge distribution in the fllm.

In the approximation of the ideal Schottky barrier model no surface states exist. So, the total
charge of one sign in the space charge region equals the charge of opposite sign in the Helmotz
layer. The junction consequently behaves like a parallel plane condensor. One can then derive the
space charge capacitance. As seen earlier this capitance is dependant on the bias and hence is called
the differential capitance. Therefore the depletion layer capitance per unit of surface is dermed as :

(2)

where dQsc is the incremental increase in charge per unit area upon an incremental charge of the
applied voltage dY.

The derivation of C is based upon the well known Poisson's equation which when applied for one
dimension is :
d2¢(x) -p(x)
ctxr=-;;- (3)

with p(x) the charge density at the position x and cIl(x) the potential at the same position.

Integration of this equation, taking account of the boundary conditions, leads to an important
relationship between the differential capitance and the applied potential known as the C-V Mott-
Schottky relation:

l/C2 = (2/€oEeND) (V-Vbf-kT/e) for an n-type SC (4)

or l/C2 = (2/€oEeNA) (-V+Vbf-kT/e) for ap-type SC (5)

where ND and NA are the donors and acceptors concentrations respectively, Vbf the flat-band
potential.

It is clear that by plotting l/C2 versus V a straight line should result for the junction, the
slope of which gives the impurity concentration of the substrate and the intersection (at l/C2 = 0)
the flat-band potential Vfb (more precisely Vfb+kT/e for the n-type or Vfb-kT/e for the p-type
semiconductor).
226

Integration of the one dimensional form of the Poisson's equation also permits the
determination of several other parameters such as the potential drop in the Helmotz layer (<I>H), the
space charge width (Wsc), the electric field E etc ... Thus the Helmotz potential drop is given as :

<l>H= (2eaEeND)I/2 dH (<I>sc-kT/e)l/2 (6)

fJ?-
with dH the width of the Helmotz layer and <I>sc the space charge layer potential drop. It is easy to
check that the potential drop Se/EL occurs practically exclusively over the space charge region of
the semiconductor when the doping level is relatively low.

The space charge width wsc and the electric field E are given by :

rosc = (2e"E/eND) 1/2 01- Yfb-kT/e) for an n-type semiconductor (7)

eND
E=- (8)
fJ?-
The great interest of the capitance measurement is undoubtly the possible contribution in the
determination of the electronic structure of the semiconductor which comes from the knowledge of
the doping level, the flat-band potential and also the band gap'" Eg and the effective density of
states Nc. The relations are successively:

EF.fb = -eYfb (9)

this relation gives the Fermi level of the semiconductor at flat band situation (EF. fb)
ND
Ee,fb =EF,fb-kTLnNC (10)

Ee,fb is also the position of the conduction band edge at the surface Ee,s at all values of Y
(applied potential) under conditions of depletion.

So, with the knowledge of the band gap /',.Eg, the position of the valence band edge at the
surface Ey,s is given by :

Ey,s = Ec,s-/',.Eg (11)

One can therefore characterize the semiconductor electrolyte junction from a structural point of
view.

3. Photoelectrochemical measurements

An important factor in photoelectrochemical measurements (when classical semiconductor material

are considered) is the depth of light penetration. Indeed, as a consequence of hole-electron
generation, this depth is relatively large for light of comparatively long wavelength but becomes
less than the thickness of the space charge region for short wavelengths.

In the case of passive films, both the magnitude of space charge region, W, and the
thickness of the film are small as compared with the depth of light penetration. Further, as a
consequence of the lack of information about traps and potential distributions, we do not know
how light intensity decreases with increasing distance from the surface.
 227

Theoretically when very thin films are illuminated the possible absorption or reflection at the
metal-passive film interface has also to be considered. For a passive film of thickness d, the
photocurrent can be expressed by (5) :

iph = eTl1o[l-exp(-aA,d)] [1 +R exp(-aA,d)] (12)

where '11 is the quantum efficiency, Jo the photon flux, R the reflectivity of the metal, aA, the light
absorption and d the thickness of the passive film.

Obviously quantum efficiency is not the essential problem in the study of the passive films.
Photoelectrochemical effects can not be thought in terms of direct energy conversion and we can
develop a phenomenological approach which neglects quantitative aspects when current densities
are considered.

Usually, in photoelectrochemistry, the conditions of illumination are such that the system is
in a steady state. The solution of the general expression for hole concentration and the related
current (see Eqs(40) and (41) in : Photo electrochemistry Theoretical Basis) yields a simple
expression which corresponds to that found by Gartner for a metal-semiconductor function (10).

iph = eJo[ l-exp(-aA,W)/l +a.).l.pJ (13)

where W is the thickness of the space charge region and Lp the hole diffusion length.

The restrictive conditions in this expression are :

-the recombination rate in both surface and space charge region is zero
-the recombinations in the bulk are only due to diffusion
-there is no change in the hole concentration through the space charge region
-the electron transfer is not a limiting step
-the band bending is greater than leV.

In the latter expression we describe only the semiconductor part of the junction : the
electrolyte is considered equivalent to a metal. Further it is also supposed that the excess electron
current due to majority carriers is negligible. This is not always true when the photoelectrochemical
behaviour of passive films is examined.

If aA, is small enough so that aA,Lp « aA,W, then Eq. (13) becomes:

iph = aA,eJoW = Cte(V-Vfb) 1/2 (14)

where W = Wo(V-Vfb)l/2 (15)

and Wo = (2ec,eleND)I!2 (16)

where Eo and E are the dielectric constants of the vacuum and semiconductor and ND the doping
concentration.

In this case the linear relationship between i2p h and (V -Vbf) reflects the characteristic
dependence of the depletion layer, W, on the potential. The straight lines obtained intersect the
potential axis at one point which practically coincides with the flat-band potential. Note that this
228

photoeIectrochemical method is analogous to the Mott-Schottky dependence of CSC- 2 on <I>SC

(Eqs 4 and 5).

Here the same comments should be made about the possible influence of the Helmholtz
capacity and other factors frequently discussed when the differential capacity method is used.

At constant flux and if we suppose aW « 1 and aLp « 1, Eq.(13) reduces to :

(17)

the straight line obtained with the iph vs W plot intersect the W axis at one point which
corresponds to the value ofLp.

Optical transitions for semiconductors can be obtained from the variation of the absorption
coefficient aft. near the band edge. If we consider K = 0 allowed transitions, then it has been
shown (11) that the photon absorption is expressed by :

a", = AChv-Eg)Yll2/hv (18)

where A is a constant.

The optical band gap values, E g , of the passive films can be obtained from the photo current
spectra at the constant potential by means of the expression:

iph.hv = Cte ChV-EF)n/2 (19)

The value of n depends on wether the transition between the states of the valence band and those of
the conduction band are direct (n = 1) or indirect (n = 4). In the use ofEq.19 it is supposed that the
photocurrent is proportional to aft., then the minimum enerfY required for direct or indirect
transitions can be evaluated from (iph.hv)l/2 vs hv and (iph.hv) rl. vs hv plots respectively.

In the case of highly disordered fIlms, the potential dependence of the photocurrent may be
altered due to the presence of a high number of localized states in the band gap. If a high field is
present, the removal of the carriers from their respective traps is lowered by the amount
~~ 1/2.Consequently, the photocurrent is given by:

iph = exp(~ 1/2/KT) (20)

when ~ = (e 3/f£o)1/2 and ~ the field.

The field de~ndence on the escape probability is called Poole-Frenkel effect (12). Straight lines in
In(iph) vs ~ f/2 plots indicate a Poole-Frenkel behaviour.

The Poole-Frenkel effect is commonly observed for field ~> 104 V.cm- l . However for very
high fields it is admitted that the barrier becomes thin enough and tunneling processes are possible.

Another phenomenon frequently observed with amorphous materials (but rarely found in
crystalline solids) is the so-called Urbach tail (12) described by the following relationship:
 229

<X =a.,exp ['Y (hv-Eg) IKTJ (21)

where <Xo is a constant and 'Y, the slope of relation (21), corresponds to the inverse of an energy.
Relation (21) can be related with the number of defects (disorder, impurities) in the semiconductor.
The more 'Yis large the more the crystal is perfect. Note that equation (21) allows the determination
of the bandgap which is usually assumed to be at the point where In <X ceases to be linear with hv.

Urbach tails and Poole-Frenkel effects have been found for passive films formed on
different metallic materials (iron, titanium). For many workers, the analysis of the photocurrent
reveals the influence of a large number of localized states and the photoelectrochemical behaviour
of the passive film is closely associated with that of an amorphous semiconductor. It is a matter of
fact that knowledge of the exact nature of these localized states is still in need of further
investigation.

3.1. IRON AND STAINLESS STEEL PASSIVE FILMS

Several investigations of passive iron have been performed in the last few years (2) (13). Band gap
energies between 1.9 and 2.5 eV are reported and spectral results, specially the photocurrent
behaviour according to Urbach tail, suggest a noncrystalline passive film. This is supported by the
fact that the Poole-Frenkel potential dependence is observed. It is concluded that the photoresponse
involves localized states and that the rate-determining step in the photoexcitation processes is
probably the removal of the photoexcited carrier from its respective trap .

Fig.7 shows the influence of the pH of the corrosive medium on electrochemical and
photoelectrochemical behaviour of iron passive films (14). Both monochromatic (). = 37Onm) and
polychromatic illumination has been used. It is interesting to observe that the photocurrent intensity
depends on solution pH applied potential. However data relative to the thermodynamic equilibrium
pHcpotential diagrams of aqueous solutions appears as the more determinant factor. Indeed, if we
introduce photoelectrochemical results in Pourbaix diagrams (Fig.8), we observe that the
photocurrent event can be associated with the access to a potential region where yFe203 is the
oxide thermodynamically stable. Further in solutions of lower pH the oxygen evolution is
necessary to obtain measurable photoeffects.

Figure 9 shows the i2ph vs potential plots (according to Eq. 14) obtained with passive film
formed at O.SV/SCE in a borate solution (pH 9.2). The linear relationship observed is
representative of the photoelectrochemical behaviour of an n-type semiconductor-electrolyte
junction under illumination and reflects the potential dependence of the depletion layer thickness.
The flat-band potential, Vfb, obtained by extrapolation of the i2ph vs V plot to iph = 0 is higher by
several hundred millivolts than Vfb values obtained from capacity measurements. The presence of
surface states and recombination processes may be responsible of such a deviation.

Figure 10 shows the (iph.hv) 1/2 vs hv plots (according with Eq 19 for n = 4) for iron
passive films formed in solutions of different pH. It can be observed that the photocurrent intensity
is strongly dependent of both the film formation potential and pH of the electrolyte. It also appears
that the band gap of the various passive films is a constant with regard to the pH of the corrosive
medium. It was also observed that the band gap does not depend of the value of the formation
potential. In all different experimental conditions the band gap energy is of 2.67 eV (14). The
question arises of which optical transition generates electron-hole pair charge carriers. Cahan et al
(15) using in situ automatic ellipsometry obtain a minor peak at 2.2 eV and a more intense peak at
above 3.6 eV which are associated to a d-d transition and charge transfer transition (~p ~ Fe3d)
230

respectively. In ferrites with spinel structure, we have to consider crystal-field transitions which
depend strongly on the crystal-field strength and charge transfer transitions. The latter are indicated
in schematical band-structure diagram shown in figure 11. It can be seen that holes can be
generated in both Fe o3+ and oxygen 2p band. What is the matter with iron passive film where
transitions assigned with Fe3+ in tetrahedral and octahedral sites also have to be considered ?

For passive films on iron electrodes, Stimming found different exponents n depending on
the electrolyte and the electrode potential (2). It is supposed that since the passive film on iron is a
few atomic layers thick, no well defined periodicity can be expected normal to the surface. Only a
short range order exists due to the high number of nonequilibrium positions of the ions. Thus, to
interpret the photochemical results, it is necessary to take into account the large number of bulk
localized states. Further, the analysis of photocurrents demonstrates the influence of a large
number of surface states.

Remember that the phenomenon of passivation of iron has been explained by the formation
of an oxide film having a structural and a chemical nature that place it between magnetite (Fe304)
and hematite (Fe203) (17). Several authors assume that the iron oxide with composition and
structure close to yFe203 plays the most important role (18). These oxides have an inverse spinel
structure. As it is known, in the inverse spinel, the unit cell contains 32 octahedral sites and 64
tetrahedral sites. In this structure 32 Oxygen atoms occupy a nearly close packed f.c.c. sublattice
and the metal ions are distributed over the tetrahedral and octahedral sites. The localization of Fe2+
ions in octahedral or tetrahedral positions is a determinant factor to both the reactivity of the oxide
and its conductivity.

For iron passive film Stimming et al (9) associate the shallow and deep donor levels
respectively to the ions in the tetrahedral and octahedral positions in the unit cell of a spinel
structure. Simoes et al (19) also detected (by Mott-Schottky plots) the existence of shallow and
deep levels for passive films formed on stainless steels. It is supposed in this work that the
particular resistance of passive ftlms formed on stainless steel can be related with the substitution
ofFe3+ by Cr3+ in octahedral sites.

Figure 12a shows the influence of the film formation potential on the evolution of these
structural factors. It appears that for potentials below OV/SCE the capacity measurements reveals
the presence of both shallow and deep donor levels. Above OV the shallow donor level is detected.
It is interesting to realize that the potential of OV is precisely the potential at which the photocurrent
is detected when the electrode potential increases (Fig.12b). This potential (OV) corresponds to the
minimum of the current intensity in the classical polarization curve represented in Fig.12c. Note
that an equivalent correlation between electrochemical and photoelectrochemical behaviour can be
observed in the case of the iron passive film (Fig.9b and c). Besides, the potential of OV also
corresponds to a change in the electrochemical behaviour of iron alloys (20). According to
equilibrium thermodynamic pH-potential diagrams the potential region below OV is dominated by
the stability of Fe304, whereas for higher potentials, the stable iron oxide is yFe203.

3.2. TITANIUM

Titanium and related alloys form a class of corrosion resistant materials not only of scientific
interest, but also of considerable practical importance. The oxide film is known to be Ti02, a
semiconductor oxide which has been extensively studied in its bulk form. The fact that quantum
efficiencies are high in Ti02 explains the high number of researches involving
photoelectrochemistry.The thickness of films formed by anodic oxidation is often expressed in
 231

300 0)
pH 6.9
Fe 3O,
15
200 ,,'.."\ 4
~,'. \
\
I \ 10
I \

,
I I
I I
\ 2
100 I I
i
I
I \ ?'-t--_i 5
I \
I
I ,
" .
a
-
f
-0.5 0 O.S
b) ..-
1,5 15 '"I
pH 9.2 '"I
'", 10 E E
(,) (,)
H2 O/0 2
E <{
10
<t:
~
en
<t: 5 Fe) 0, U)
I I
U)
0,5 0 0
I
5
-
~ T'""
0
£
..c: 0.
~'~"~~------ 0.
a c) -0,5 a 0,5
0

pH 11.6 2,0
300

1.5 15
Fe30,
200 IFe OOH
10

100 "' .... - ....

--:;' \ 0,5 5
,''' \
I
a -0,5

Figure 7. Influence of the solution pH on electrochemical and photoelectrochemical behaviour of

iron passive films.
232

W
I
Z
:> F.··
1.2 0-~

-o,S
F.
-1,2

-1,6

-2 -1 0 10 12 14 pH

Figure 8, Equilibrium pH-potential diagram, Ephl, Eph2, Eph3 and Eph4 is the lower applied
potential at which the photo current can be detected,

--:z...........
Eu E,
«
~
".
.s
.c ~
l.L

,-"- ~
2J0 3 40

20

Vfb 05 V(SCE)

Figure 9. Plot of l/e2 and i2ph vs potential obtained with iron passive film fonned at O.5V/ SCE
 233

~ \

if'0' "
~""' /~,\-9,2~PH 0.2
;>
..tl ~11':,~
~~
~

\ /'
~. " y
03

\ .. 6

\~
~\~\
02 0.1

.. 6 \\
\ \\\
A6 \\

0.1 ~'~
\~
"i
'\.
\\\

"
3,5 3 2.5 E (eV)

Figure 10. Plots (iph.hv) 1/2 vs hv for iron passive films formed in solutions of different pH.

E Fe:4s

Fe~+: 3d 6

J....-----t--- Fe~+ : 3d 6

B A
Fe~+: 3d S
Fe~+: 3d S

~@,O:2p
- __ N(E)

Figure 11. Energy diagram for Fe203 (After 16).

234

terms of the apparent growth rate, i.e. Angstrams (or nm) per volt of applied voltage.
Schematically, anodic films formed at very high potentials are crystalline, while those fonned at
lower potentials are amorphous.

Recent investigations demonstrate the influence of the structural disorder on

photoelectrochemical spectra. Stimming (I) shows that for films fonned at potentials higher than
SOV the photoelectrochemical results correspond to that of crystallized oxide and can be compared
to those obtained with bulk Ti02 electrodes. Films of intermediate thickness (formation potential
between 30 and 60V) show also clear features of crystalline semiconductors, from the spectra as
well as from the potential dependence for band to band excitations. However, their present low
quantum efficiencies at high photon energy which suggest a high number of recombination centers
in the band gap. In this case band gaps for both direct and indirect transitions can be obtained.
Conversely for very thin films no direct transitions are observed. This and the fact that the
Poole-Frenkel effect corresponds to photon energies above the band gap energies of crystalline
Ti02 indicates that the thin films are amorphous.

3.3. lRRADIATION EFFECT ON PROPERTIES OF PASSIVE FILM FORMED ON A AISI 304

TYPE STAINLESS STEEL

Considerable work has been carried out on the effects of irradiation on the corrosion of various
materials. This research was motivated by the development of nuclear waste storage technology.
Previous studies have mainly been concerned with the chemistry of radiolysis products and with
their influence on metallic corrosion (21).

Gorse et al (22) show that the direct influence of irradiation on the passivation process of an
austenitic stainless steel type 304 can be obtained by photoelectrochemical measurements. The
experiments ~ pe~~d in a buffered borate solution at pH 9.2 under the influence of a particles
( flux: 3.10+ a cm s ). To minimize the irradiation of the electrolyte, the stainless steel sample
(a thin foil at most lO!lm thick, otherwise only a implantation in the bulk of the sample would
occur) is inserted between the source of irradiation and the electrolyte. With this geometry, the
particles bombard the alloy and the passive film, but only very few penetrate the electrolyte, having
lost more than 80% of their initial energy in the alloy for the more energetic ones. One shows that
for a steel foil thicker than 8.3 JlID, the energy deposited in the electrolyte per incident particle is
very weak.

Fig.13 shows the irradiation effect on photocurrent obtained with a stainless steel electrode
passivated at 0.57SV/SCE. Irradiation increases the photocurrent by a factor 2 or 3. Conversely, a
irradiation has no effect on the band gap and the flat band potential. These results are compatible
with some modification of the atomic and electronic structure of the passive film under irradiation.
Indeed, even when the energy of the impinging a particles (S4.8 MeV) is too low to produce
direct atomic displacements in the passive film, such double charged projectiles lead to numerous
ionizations processes.

From the point of view of the electronic structure, it is assumed that the passive film can be
compared to a highly doped n-type semiconductor. One shows that a irradiation decreases the
donor concentration (obtained by capacity measurements), increases the space charge region and
the diffusion length for minority carriers. The latter point is consistent with a decay of the number
of trapping centers for holes.

3.4. OXIDE FILMS MODIFIED BY ION IMPLANTATION

 235

t')
10' I'A/cm2
'E
u C/
;i
Q
Z
22 o 0.4 VI SCf'
10

21
10

:Jb/ 0
10
20
Qot / ,
o 0'.4 o'a V/SCE

-0.4 o 0.4 0,8

Potential V/SCE

Figure 12. Influence of film fonnation potential on: a) shallow (ND1), deep (NoV and total
donor levels b) photocurrent - potential curves c) polarization current-potential curves

30

20

10

Figure 13. Irradiation effect on photocurrent only the passive film fonned on steel foil with 8.3 IJlll
is irradiated.
236

Oxide films can be modified by ion implantation. This surface treatment induces doping and
stoichiometric changes as well as radiation damage. Schultze et al (23) examined the influence of
the implantation of Ti+ and 0+ with various concentrations on Ti02 films. They show that
photoelectrochemical measurements give information on structural changes produced by
implantation treatment. It appears that the indirect band gap energy obtained with n = 4 in Eq. 19
shifts from 3.43 eV to 3.23 eV when the non-implanted oxide is implanted with 3% Ti+. Then the
band gap energy increases to 3.36 eV with further Ti+ implantation (10% Ti+). The effect of 0+
implantation is comparatively small. The photocurrent spectra relative to the implantation treatment
also reveals an additional peak around 2.6 e V which can be related with induced defect states.

The different implanted oxide fIlms also present a smaller slope of the Urbach tail than the
non-implanted oxide. Thus the factor of order 'Y (see Eq. 21) decreases this means a higher
disorder energy which can be correlated with an amorphization of the oxide film.

4. Structure of the passive films : the crystalline-amorphous controversy

Oxide perfection is considered a major factor in controlling the rate of oxidation and corrosion of
metallic materials. Several researches have investigated the possible importance of the
monocrystalline structure of the passive films formed during corrosion processes (24) (25). It has
pointed out that the development of mixed oxides involving polyanions increases corrosion
resistance. Studies by electron diffraction showed that when Cr is added to Fe in an alloy, the
structure of the passive film formed changes from crystalline to noncrystalline. Okamoto (26) has
emphasized the importance of the amorphous nature of passivating films on stainless steels and
studies conducted with different amorphous metallic materials also support the idea that a
noncrystalline structure improves corrosion resistance (27).

In general, the passive films formed at room temperature are highly disordered and look
more like amorphous materials more than like crystalline one. Unfortunately, published data give
li ttle information with regard to the structural characteristics of noncrystalline films.

What is the exact structure of the passive film formed in aqueous solutions? What degree
of order and what kind of defects exist in these very thin protective oxides? These questions are
also pertinent to an area closely related to oxidation, the formation and properties of anodic oxide
films, many of which are noncrystalline. It is a matter of fact that the structural information
obtained from X-Ray or electron diffraction, even when high efficiency electron diffraction
techniques are employed, is insufficient to characterize these films on a submicroscopic scale.

The nature of defects in crystalline materials and their influence on the course of oxidation is
well established. Although the same is not yet true for noncrystalline materials, some attempts to
define these conceptually different defects have been initiated (28)-(32).

The vitreous state refers to a class of noncrystalline solids with a high degree of short-range
(chemical and physical) and long-range chemical order. In contrast, the term amorphous refers to
the class of noncrystalline solids which present only partial short-range physical and chemical
order and total long-range physical and chemical disorder.

Anodic oxidation of a crystalline metal can produce a vitreous material which is a metastable
state with respect to crystalline oxide, so that kinetics dominates formation for the various levels of
noncrystallinity. It is admitted that a mechanism of self-diffusion itself is responsible for
stabilization of the vitreous structure. Other factors have to be considered. For instance, water can
 237

promote vitreous oxide forming multiple structures of equivalent energy or destroy vitreous oxide
by weakening network bonding, allowing crystallization.

The slow logarithmic kinetics of passive film growth can also indicate the presence of
vitreous oxide, but additional proof is needed to confirm its existence. What is not clear is wether
the oxide resulting from low temperature oxidation is vitreous or microcrystalline. Single crystal
oxide on a metal requires an epitaxial relationship between the two crystal lattices. This is seldom,
if ever found, even when the metal is a single crystal. On the contrary, the vitreous state can
accommodate varying crystal lattices or even a polycrystalline surface. Apparently, quasi perfect
solid-solid interfaces can only be obtained when either a quasi perfect epitaxial relationship exists
between two single crystals or at least one side of the interface is noncrystalline.

The existence of a hybrid character due to partially ionic and partially covalent bonds, a
small difference between the standard free energies of various oxides of the same metal and ligand
effects may be also factors of formation and stability of noncrystalline structures. In addition
transition metal oxides such as Cr203 may exhibit a departure from the regular structure due to
ligand effects. The structural complexity of the vitreous structure is manifested in the very large
unit cell that is divided into subcells. These structures have been named infinitely adaptative
because they offer the possibility of varying the composition ( stoichiometry) by changing the
relative numbers of subunits. This is the case of the oxide formed by anodic oxidation of tantalum
which results in a noncrystalline Ta205 film whose dielectric and optical properties indicate that
there is a high degree of short-range order in the structure.

What is the concept of band structure which describes more exactly the electronic structure
of a noncrystalline oxide? In accordance with Ioffe's finding (33), it would appear that the band
structure does not depend on long-range order but is instead determined by the chemical bonding,
the type of coordination and the type of short-range order. Long-range order, i.e. long-range
periodicity, may be a sufficient condition for the development of energy bands, but is not
necessary one. Even the relatively short range order, represented in a liquid by its radial
distribution function and appearing only in spherical coordinates,is a state of order.

Short range order is also sufficient to give rise to carrier transport. In fact the parameters on
which semiconductivity, 0, usually depends are the energy interval E and the mobility J.!. The
overlap of electronic wave functions on neighbouring atoms is a very important factor. The bigger
the overlap, the greater the band width and the larger the mobility. The overlap in tum depends on
the particular orbitals involved and on their separation and relative orientation, both of which are
determined by the position occupied by the atomic nuclei in the crystal.

REFERENCES

l. Leitner, K., Schultze, J.W. and Stimming, U. (1986) Photoelectrochemistry Investigations

of Passive Films on Titanium Electrodes, J. Electrochem. Soc., Vol. 133, 1561-1568.

2. Searson, P.C., Latanision, R.M. and Stimming, U. (1988) Analysis of the

Photoelectrochemical Response of the Passive Film on Iron in Neutral Solutions.].
Electrochem. Soc., Vo1.135, 1358-1363.

3. Di Paola, A., Di Quarto, F. and Sunseri, C. (1986) A Photoelectrochemical Characterization

of Passive Films on Stainless Steels, Corrosion Science, Vo1.26, 935-948.
238

4. Stimming, U. (1986) Photoelectrochemical Studies of Passive Films. (Review article),

Electrochimica Acta, Vol.31, 415-429.

5. Simoes, A.M.P., Da Cunha Belo, M., Ferreira, M.G.S. and Rondot, B. (1987)
Characterization of Passive Films on Stainless Steel by Photoelectrochemical and Impedance
Measurements, Portugalire Electrochimica Acta, Vol.5, 59-69.

6. Sanz, J.M. and Hofinann, S. (1986) Qualitative AES Depth Profiling of Very Thin Layers,
Surface and Interface Analysis, Vol.8, 147-157.

7. Lorang, G., Basile, F., Da Cunha Belo, M. and Langeron, J.P. (1988) Quantitative
Analysis of Passive Films Formed on Stainless Steels, Surface and Interface Analysis,
Vol.12, 424-428.

8. Gerischer, H. (1989) Models for the Discussion of the Photoelectrochemical Response of

Oxide Layers on Metals, Corrosion Science, Vol.29, 57-266.

9. Stimming, U. and Schultze, J.W. (1979) A Semiconductor Model of the Passive Layer on
Iron Electrodes and its Application to the Electrochemical Reactions, Electrochimica Acta,
Vol.24, 859-869.

10. Gartner, W.W. (1959) Depletion Layer Photoeffects in Semiconductors, Phys. Rev.,
Vol.1l6, 84-87.

11. Johnson, E.I. (1967) in Semiconductor and Semimetals, Vol.3, chap.6, edited by
R.K.Willardson and A.C.Beer (Academic Press, New York).

12. Mott, N.F. and Davis, E.A. (1979) Electronic Processes in Noncrystalline Solids,
Oarendon Press, Oxford.

13. Aazumi, K., Ohtsuka, T. and Sato, N. (1986) Spectroscopic Photoresponse of the Passive
Film Formed on Iron, J. Electrochem. Soc., Vo1.l33, 1326-1328.

14. Faty, S. (1986) Etude de la passivation du Fer par mesures d'impectance et de

photoelectrochimie, These Universite Pierre et Marie Curie (paris).

15. Chen, C.T. and Cahan, B.D. (1982) Nature of the Passive Film on Iron, I. Electrochem.
Soc., Vol. 129, 17-26.

16. Haart, L.G.I. and Blasse, G. (1985) Photoelectrochemical Properties of Ferfites with the
Spinel Structure, J. Electrochem. Soc., Vol. 132,2933-2938.

17. Nagayama, M. and Cohen, M. (1962) The Anodic Oxidation of Iron in a Neutral Solutions,
1. Electrochem. Soc., Vol. 109,781-790.

18. Bloom, M.C. and Goldenberg, L. (1965) y Fe203 and Passivity of Iron, Corrosion
Science, Vol. 5, 623-630.

19. Simoes, A.M.P., Ferreira, M.G.S., Rondot, T B. and Da Cunha Belo, M., to be published
in I. Electrochem. Soc.
 239

20. Da Cunha Belo, M., Rondot, B., Cadet, P., Takenouti, Beranger, G. and Ben Salah, N
(1984) Comportement electrochimique et resistance ala corrosion des alliages amorphes a
base de fer, Metaux, Corrosion, Industrie, N°7l0, 321-335.

21. Marsh, G.P., Taylor, K.J., Bryan, G. and Worthington, E. (1986) The Influence of
Radiation on the Corrosion of Stainless Steel, Corrosion Science, Vol. 26, 971-982.

23. Gorse, D., Rondot, B. and Da Cunha Belo, M. (1989) Irradiation Effect on Properties of
Passive Film Formed on a AISI 304 type Stainless Steel, C. R Acad. Sci., Paris, t.308,
serie II, 369-375.

23 Schultze, J.W., Elfenthal, L., Leitner, K. and Mayer, O. (1988) Modification of Oxide
Films by Ion Implantation: TiOz-films modified by Ti+ and 0+ as exemple, Electrochimica
Acta, Vol. 33,911-925.

24. Revesz, G. and Kruger, J. (1978) The role of Noncrystalline Films in Passivation and
Breakdown of Passivation. Passivity of metals, Edited by RP. Frankenthal and J. Kruger
the Electrochem. Society, INC, Princeton.

25. Revesz, G. and Fehlner, F.P. (1981) The role of Noncrystalline Films in the Oxidation and
Corrosion of Metals, Oxidation of metals, Vol. 15,297-321.

26. Okamoto, G. (1972) Proceedings of the fifth International Congress on Metallic Corrosion,
Tokyo, Japan (Published by NACE).

27 Hashimoto, K., Naka, M., Noguchi, J., Asami, K. and Masumoto, T. (1978) Passivity of
Metals Edited by P. Frankenthal and J. Kruger (the Electrochemical Society, Inc Princeton).

28. Takayama, S. (1976) Amorphous Structures and Their Formation and Stability (Review), I.
of Materials Science, VoLll, 164-185.

29. Waseda, Y. (1980) The Structure of Noncrystalline Materials, Mc Graw-Hill International

Book Company.

30. Turnbull, D. Cohen, M.H. (1960) Modem Aspects of the Vitreous State, VoU, Mackensie
J.D., ed. (Butterworths, London).

31. Griscom, D.L. (1976) Defects and Their Structure in Nonmetallic Solids, Henderson B. and
Hughes A.E., eds. (plenum Publishing Co., New York).

32. Kingery, W.D. Bowen, H.K. Uhlmann, D.R (1976) Introduction to Ceramics, 2nd ed.,
John Wiley and Sons, New York.

33. loffe, A.E. Regel, A.R. (1960) Progress in semiconductors, Vo14, 237, A.F. Gibbon ed.,
Heywood Co. Ltd London.
DIFFERENTIAL REFLECfOMETRY AND OTHER OPTICAL TECHNIQUES IN
CORROSION RESEARCH

R.E. HUMMEL
Department of Materials Science
and Engineering
University of Florida
Gainesville, Florida

ABSTRACf. Optical techniques for the study of corrosion phenomena, such as differential
retlectometry and ellipsometry are discussed. The experimental results presented include the
corrosion of various metals such as copper, the passivation of nickel and nickel-copper alloys
in aqueous environments, and the dezincification of a-brass.

Contents

1. Introduction
2. Interaction of Light with Matter
3. Measurement of the Optical Constants
4. Thickness of Corrosion Films
5. Spectroscopic Differential Retlectometry
6. Sample Preparation
7. Principles of Differential Retlection Spectra
8. Electrochemical Oxidation of Copper
9. Free Corrosion of Copper
10. Estimate of Film Thickness from Differential Retlectograrns
11. Corrosion by Annealing in Dry Oxygen
12. Passivation of Nickel in Aqueous Solutions
13. Dezincification
14. Conclusions
15. Acknowledgements
16. References

241
M. G. S. Ferreira arul C. A. Melerulres (eds.),
Electrochemical arul Optical Techniques for the Study and Monitoring of Metallic Corrosion, 241-283.
© 1991 Kluwer Academic Publishers.
242

1. Introduction

Optical techniques have considerably broadened the spectrum of methods that are useful for
the study of corrosion phenomena. The concept is simple enough: light penetrates the
surface of metals only 50 to 100 atomic layers, that is, about 100 to 200 A Some corrosion
films which form on these metals are more transparent than metals, others may be less. In
any event the principal interaction of light with many solids occurs at its very surface. Thus,
the information which is obtained from the reflected light stems mostly from the corrosion
layer itself and only to a lesser extent from the underlying substrate. This will become clearer
in the next chapter.
The interaction of matter with light is different from that involving x-rays which
penetrate a metal much deeper, that is, characteristically 1 to 50 j!m. X-ray spectroscopy is
therefore less sensitive to thin corrosion products.
There is a second characteristic of optical techniques which makes them so powerful:
the investigations can be performed in situ, that is, the formation of corrosion products can
be observed while they develop.
Corrosion engineers and electrochemists are generally not too familiar with the
principles of optics, that is, with the electromagnetic theory of light. The present contribution
aims towards bridging this gap from the side of physics; in other words, it attempts to help
corrosion engineers in their appreciation and assessment of optical techniques.
Below, various optical methods will be discussed, among them differential
reflectometry which has been proved to be a useful technique, particularly for studying
electrochemical corrosion concomitant with electrochemical polarization measurements.
Experimental results will be presented emphasizing the corrosion behavior of copper,
tungsten, molybdenum and chromium. Further, the passivation of nickel and nickel-copper
alloys in various aqueous environments will be discussed. Finally, optical studies involving
dezincification of a-brass are summarized.

2. Interaction of Light with Matter

If light, that is, electromagnetic radiation meets a boundary which separates vacuum (or air)
from any type of matter, such as a metal, the light intensity becomes smaller the further the
light penetrates into the substance (Fig. 1). This process is called "damping" of the light.
Metals damp the intensity of light in a relatively short distance whereas glass and many
"transparent" corrosion products such as aluminum oxide allow the light to penetrate much
deeper. The characteristic penetration depth, W, is defined to be that distance at which the
intensity, I, of the light wave which travels through a material has decreased to lie or 37%
of its original value, 1 that is when
0 ,

1 (1)
e

The characteristic penetration depth is proportional to the wavelength of the light,

1, and inversely proportional to the "damping constant", k:
 243

l. (2)
W=-
4ttk

Characteristic values for the penetration depth and the damping constant are given in Table
1.

Table 1: Characteristic penetration depth, W, damping constant, k, index of

refraction, n, and reflectivity, R, for some materials at l. = 600 nm.
Material Water Glass Gold Aluminum

W (cm) 32 29 1.5 x 10-0 7.96 X 10-7

k 1.4 X 10-7 1.5 X 10"7 3.2 6.0
n 1.33 1.5 0.21 0.97
R (%) 2 4 92 90.3

Materials which strongly damp the light, such as metals, reflect most of the incoming
light. On the other hand, materials which have a small value for "k", such as glass or many
oxide films transmit most of the light; they have a very small reflectivity, see Table 1.

-Vacuum -r~-- Metal - - - -

Figure 1: Modulated light wave. The amplitude decreases exponentially.

The reflectivity, R, (that is, the ratio between reflected intensity and incoming
intensity) is connected with the damping constant, k, through Beer's equation

R = (n-lt+k2 (3)
(n + It + k 2

where
244

C_
n =-- (4)
clMd

is the index of refraction of the material, that is, the ratio between the speed of light in
vacuum and the speed of light in the medium. (It should be noted that equations (3) and (4)
are defined for the air/metal interface. The equations become more involved when the light
enters the metal through a corrosion layer, see below.) The parameters "n" and "k" (and
sometimes "R") are called the "optical constants". They depend on the wavelength of the
light (and often as well on the angle of incidence).
Equation (2) is obtained from the wave equation

(5)

Damped amplitude Undamped wave

by making use of (1) and by knowing that the intensity I equals the square of the electric
field strength, g;«U = 27rc/"J.. is the angular frequency of the light, t is the time and z is the
penetration distance into the matter. Equation (5) can be derived from the well-known
Maxwell equations which form the basis for the electromagnetic theory.) By inspecting the
wave equation (5) we note that the right side consists of two parts. The one to the very right
describes the propagation of the light wave in a sinusoidal manner. The other factor on the
left controls the decrease of the amplitude in the material. It is deduced from equation (5)
that this decrease in intensity of the light is more pronounced the larger the damping
constant and the further the light penetrates into the matter (see Figure 1).
Taking the above mentioned facts into consideration, it becomes now evident why
optical techniques probe predominantly the surface of a material, and why very thin layers
of an oxide have frequently such a large influence on the optical constants. This effect
becomes detrimental when measuring the bulk optical constants while assuming a "clean"
surface. However, it is a definite advantage in corrosion research where the properties of
monolayers are often of interest.
One more important point should be added. The optical constants nand k are not
independent variables. Instead, they are interrelated by the following equation:

n=n-ik (6)

Here, ft is the complex index of refraction. This implies that n is the real part and
k is the imaginary part of ft. The (complex) index of refraction is connected with the
(complex) dielectric constant through
(7)

where
 245

(8)

Combining (6), (7), and (8) yields

(9)

Equating individually the real and imaginary parts in (9) provides

(10)

and
(11)

e2 is often called the absorption product or briefly the "absorption". The reflectivity is
connected with e 1 and e2 in a rather involved equation which can be derived by combining
equations (3), (10) and (11):

R {~+ 1 - J2 (~ +e
= l)
(12)
Jei + e; + 1 + J2 (Jei + e; + e l)

3. Measurement of the Optical Constants

When linearly polarized light (that is light whose electric vcctor vibrates only in one direction
perpendicular to the beam) impinges at an angle a on a metal, then the reflected light is
generally elliptically polarized, see Figure 2. In elliptically polarized light the length and
direction of the light vector (that is, the electric field strength, g') alters periodically.
Specifically the tip of the light vector moves along a path which resembles the thread of an
elliptically shaped screw. The projection of the tip of this light vector onto the x-y plane is
an ellipse (Fig. 2).
Elliptically polarized light can be thought to be composed of two mutually
perpendicular, linearly polarized waves, which have a phase difference, fl, in between them.
The analysis of the elliptically polarized light yields two parameters, the azimuth, 1\1" (see
below) and the just mentioned phase difference, fl. From fl and 1\1r the optical constants n
and k can be calculated by applying rather complicated equations which are derived from the
already mentioned Maxwell equations. In corrosion research, however, this calculation is
generally not done. Instead, for convenience, the measured parameters 1\1r and fl, and their
variation during the corrosion process are reported. We shall elucidate this in more detail
below.
246

plane of
Incidence

Figure 2: Reflection of linearly polarized light on a metal surface (from Ref. [4]).
(Note: the terms "plane polarized light" and "linearly polarized light" are used
interchangeably. )

Let us return to Figure 2 in which linearly polarized light whose vibrational direction
is conventionally inclined by 45° towards a plane which is perpendicular to the plane of
incidence is shown. This angle is called the azimuth, "' .. in contrast to the azimuth of the
reflected light, "'" which is defined as

(13)

(see Figure 2) where ~p and ~ are the parallel and perpendicular components of the
reflected electric field strength, I g'1. For the actual measurement of "', and A one needs
a polarizer (consisting of a birefringent material which allows only linearly polarized light to
pass) and a compensator (also consisting of a birefringent material which can be varied in
thickness). This is briefly illustrated in Figure 3, where elliptically polarized light reflected
from a metal is represented by two vectors (representing linearly polarized light) which point
into the x- and the y- direction, respectively. They have a phase difference, A, between them
as mentioned above. By varying the thickness of the (wedge shaped) birefringent material
in the compensator one accomplishes eventually that the light which leaves the compensator
is linearly polarized (Le., A = 0°). The resulting vector, R""" is then tilted by an angle V,
against the normal of the plane of incidence. One determines V, by turning the polarizer
(which is commonly called "analyzer") to a position at which its axis is perpendicular to R"".
In short, A and V, are measured by simultaneously altering the thickness of the compensator
and rotating the analyzer until no light leaves the analyzer. This cumbersome and time
consuming method is simplified in modern ellipsometers by employing photoelectric
 247

Analyzer

Compensator

from metal

Figure 3: Vector diagram of light which was reflected from a metal surface (from Ref.
[5]).

detection devices and computer controlled servos which accomplish the rotation of the optical
components. Most commercial ellipsometers are equipped with a monochromatic light source
(laser) which allows one to simplify the compensator to a "quarter wave plate" made for the
wavelength of the laser which has to be rotated for the measurement of IJ.. Details on this
and on the theory of optical properties can be found in the literature [1-10].

4. Thickness of Corrosion Films

One of the major aims of ellipsometry measurements is to determine the thickness of

corrosion products which form on a metal or alloy. This involves application of the "two layer
formulas" and the knowledge of a number of experimental parameters. We shall discuss this
only briefly here and leave a more detailed description of ellipsometry to the contribution by
Dr. Smialowska. In short, the equations derived by Drude and Tronstadt [11, 12] for
"transparent, non-absorbing" surface films of up to 100 A thickness on a metal substrate have
been known since the turn of this century. Their complicated and involved nature is no
longer a major obstacle when employing personal computers. The Drude equations contain
the angle of incidence, the wavelength of light, the optical properties nand k for the metal
as well as for the film, and the measured parameters 1jr and !J.. for both the "clean" metal and
the metaVoxide combination. (For simplicity, the subscript Or" for the reflected azimuth is
often omitted.) It has been shown that with increasing thickness of the film, the difference
of 1jr between the metaVfilm composite and the 1jr of the bare metal can be considered to
increase in a linear manner, while the difference between the IJ.'s decreases [13]. This simple
relationship is often used in corrosion work.
248

The exact equations for weak absorbing superficial layers involving the complex
optical constants of film and metal substrate have been derived by Va!:fl~ek [14] and in a
more rigorous manner by Urban [15]. These authors took the damping of the amplitude,
caused by the absorption, into consideration. The periodic change of 11 and", as a function
of film thickness calculated for a special case is shown in Figure 4. It can be seen in Figure
4 that a linear relationship between", or 11 and film thickness can be indeed approximated
for small values of d (up to 100 A?), as proposed by Drude and Tronstadt [11,12].
When the exact ellipsometry equations need to be applied for corrosion work the
difficult problem arises in finding reliable values for nand k for the "clean" metal substrate.
This has been attempted in the past in a variety of ways, for example by using metal films
which have been vapor deposited in ultra high vacuum (UHV), and by conducting the optical
measurements in the UHV immediately after deposition. (Note, however, that the optical
properties of thin films are not necessarily those of the bulk material). An alternative
method is to cleave or melt bulk samples in UHV.

(a~
I
~
~

,if

500 roo 1500 iOOO

_I
dlifj-

(b

V'~ ____ ~ _____ ~~~ ____- L____ -=~ ____ ~

"J1-

Figure 4: Calculated azimuth ("') and phase difference (11) for Alp) on aluminum as
a function of film thickness, d (From Ref. [13]).

Knowing the optical constants of the oxide film is often a problem, too. In some
cases, when the corrosion product is known and when it is of homogeneous composition, n
and k values of the bulk oxides can be used, e.g. hematite for oxide films on iron.
In cases where the optical constants of the corrosion product as well as the thickness
of the film are not known, a variety of possible combinations of n, k and d values exist.
(This might even be aggravated by the possibility that the corrosion product may change in
 249

composition with increasing distance from the metal which may cause a gradual variation in
nand k). In any event, the principal handicap of ellipsometry arises from the fact that three
parameters, n, k and d are wanted. They need to be calculated from two measured
parameters V and!J.. One possibility to overcome this problem is by employing a computer
into which likely values for the resulting parameters are fed. The computer then provides
the best fit for the experimental values. There is, however, no unique set of n, k and d
values which results from this approach; instead a band of possible values is obtained which
widens with increasing film thickness (Figure 5). Other possibilities are to measure a second

,0 ---~~~~--------

142

140

f
6. ".
',4
'36

132

, '0 L----Tlno---,.,'0,---,,','0,----.4""'0.------.;'0--------'
THICKNESS, A

Figure 5: Computer calculated V and !J. values as a function of thickness for n = 3.06
and k = 3.67. (Iron in an electrolyte whose index of refraction is 1.33.) [16]

set of !J. and V values at a different angle of incidence, or to measure the film thickness by
an independent method such as interferometry (which fails to be accurate for film thicknesses
below 100 A). The situation is even further complicated when the corrosion film is grown
in an electrolyte and an in situ study is attempted. In this case the optical constants of the
electrolyte need to be known, too. Further, it has to be assumed that nand k of the liquid
do not change during the electrochemical process.
All taken, it becomes evident that alternate optical methods for the study of corrosion
phenomena are needed. One of these methods - differential reflectometry - which provides
different pieces of information on the corrosion products, will be presented in the next
section.

S. Spectroscopic Differential Reflectometry

The differential reflectometer provides a different piece of information than the ellipsometer.
The instrument does not measure the parameters !J. and Vr nor the optical constants nand
k. Instead, a kind of an optical absorption spectrum is obtained which is characteristic for
each corrosion product. In essence, a corrosion film is identified by its characteristic energies
of certain absorption peaks, similarly as known from x-ray spectra. (The difference from x-ray
spectra is, however, that the information stems from the surface layer owing to the smaller
energy of the probing photons, as mentioned already in the Introduction.)
250

The differential reflectometer measures the normalized difference in reflectivity

between two similar specimens (which are mounted side by side) as a function of the
wavelength of the light. In corrosion research one sample is usually a clean metal (reference)
while the other is the same metal which is, however, covered with a thin film of a corrosion
product. As explained above, light interacts, as a rule, completely with a transparent or semi-
transparent film. Thus, owing to the differential nature of the technique, the effect of the
underlying metal substrate is essentially subtracted out, if the corrosion film is fairly
transparent and thin enough. The resulting optical spectrum represents, then, mainly the
properties of the film alone with only minor contributions by the substrate. The differential
reflectometer utilizes unpolarized "white" light (which is emitted for example by a high
pressure xenon source). This light passes through a scanning monochromator and is then
alternately deflected to one or the other sample by means of a vibrating mirror (Fig. 6). The

Light Source

Pholomul, iplie r

Figure 6: Schematic diagram of the differential reflectometer [17,18]. For clarity the
angle of incidence is shown larger than actually used in the instrument.

light probe is comprised of a fine line parallel to the boundary, separating the two samples
which moves over the two specimens at near normal incidence. The total area scanned is
generally 2x4mm 2• The mirror oscillates with a convenient frequency, for example with 50
or 60 Hz. The light which is reflected by the two samples falls on a light sensitive device and
is then electronically processed to yield .iRJR where .iR=Rt-R z is the difference in
reflectivity between the two specimens and R = (R t + R z)/2 is the average reflectivity.
Measuring R t and R z at the same time and forming the ratio between .iR and R eliminates
possible fluctuations of the line voltage and eradicates intensity variations in the spectral
output of the light source, the spectral sensitivity of the detector, and the spectral
reflectivities of the mirrors. Utilizing a normalized difference in reflectivity also eliminates
spectral variations of the window of a corrosion cell and possible spectral variations of an
electrolyte. Thus, in situ investigations can be conveniently obtained without adverse effects
 251

caused by any media which are in the path of the light beam before and after it reaches the
samples. In essence, the only signal which is eventually recorded by the differential
reflectometer stems from that medium which is different. This is in the present case the
corrosion product on one of the metal substrates.
The output signal from the electronic system of the differential reflectometer, that
is, 4R!R is fed to the y-axis of an x-y recorder. A potentiometer attached to the scanning
gear of the monochromator delivers a dc voltage proportional to the wavelength. Thus, a
differential reflectogram, that is, a plot of 4R!R versus wavelength, is generated
automatically. A scan between 200 and 800nm (1.6 to 6.0 eV) requires from one to three
minutes, depending on the scanning speed selected. The sensitivity in observing the
normalized difference in reflectivity is better than one-hundredth of a percent. In short,
differential reflectometry is fast, sensitive, non-destructive, and does not require a vacuum.
In cases where the evolution of a certain corrosion product needs to be monitored as a
function of time, only one wavelength may be used. In this case 4R!R is obtained within a
fraction of a second. Another acceleration of the data taking process can be achieved by
utilizing white light and a multi-channel analyzer.

6. Sample Preparation

For electrochemical corrosion studies, a metal disc is divided into two equal sized parts. The
samples are then mounted in metal-Iographic resin and electrically insulated from each other
with an appropriate thin sheet between them. Two insulated wires provide connections
between the samples and two potentiostats. This arrangement enables one to apply a
different potential to each sub-sample. The specimen pair is then mechanically polished
down to 1",m diamond paste on felt cloth, rinsed, and subsequently immersed in an
electrolyte contained in an electrochemical corrosion cell, Fig. 7.
Before the actual corrosion experiment, both sub-samples are held for a short time
about 50 mV anodic to the reversible hydrogen potential. Such a potential was found to be
sufficiently cathodic to prohibit film formation prior to the start of the corrosion experiment
and also facilitates some reduction of any oxides which might have formed during the time
which elapsed between mechanical polishing and insertion into the electrolyte. On the other
hand, the potential is sufficiently high to prevent Hz evolution which could disturb the optical
measurements. Thus, both specimens are initially in the same condition.
During the corrosion experiment, one sample half is kept at the above mentioned
protective potential (reference) while a selected more anodic potential is applied to the other
one. The evolution of a corrosion product can then be studied in situ by recording a series
of differential reflectograms which are taken after certain time intervals. Experimental results
obtained in this way will be presented in subsequent sections.
It should be added in passing that the corrosion cell window which was used in our
investigations as well as the electrolyte become gradually opaque for wavelengths smaller than
238 nm. This slightly restricts the useful wavelength range of the differential reflectometer
for in situ corrosion experiments.
For high temperature oxidation studies, an entire metal slab is heated in an oxygen-
rich atmosphere. After cooling, the corrosion product from one part of the specimen surface
is removed by a "half polishing" technique [20]. This way, a corroded and an uncorroded
252

Figure 7: Schematic diagram of a corrosion cell used for in situ differential reflection
studies [19].

specimen is conveniently obtained one next to the other. The sample pair is inserted into
the differential reflectometer without making use of a corrosion cell.
For free corrosion at ambient environments and at room temperature the half
polishing technique explained above can be again utilized. An alternative method has been
additionally employed in some instances. For this preparation method the entire metal is at
first mechanically polished, then one-half of the specimen is covered with a protective lacquer
while the other part is allowed to develop a corrosion film. After a certain amount of time,
the lacquer is peeled off which results, as before, in a corroded and an essentially uncorroded
surface next to each other. If a transparent lacquer (e.g. collodion) is chosen, then it is
possible to scan the light across the samples without removing the lacquer. There are,
however, limitations to this latter method due to a possible spectral dependence of the
absorption of the cover layer.
It should be added that all three preparation methods lead under equal conditions to
differential reflection spectra which show identical peak energies. The only differences in the
reflectograms are found in the strengths of the peak intensities, which indicates that various
thicknesses of the corrosion films must have been present. It will be shown in the next
section that the peak heights can indeed be utilized to determine the thickness increase of
a corrosion product.
The question often arises whether or not the strain which is introduced by mechanical
polishing, however small it may be, would cause spurious peaks in a differential reflectogram.
To answer this, one has to realize that both samples are polished simultaneously and thus
have the same amount of surface damage. Owing to the differential technique, all possible
related effects are therefore eliminated in a differential reflectogram. Nevertheless, for
 253

reasons of precaution, one may briefly electropolish both samples after mechanical polishing,
that is, before the corrosion experiment.
The situation is, however, slightly different in cases where surface roughening occurs
during the corrosion process on one of the samples only. There, a small effect is expected
which may cause a slight variation in the peak heights (not the peak energies) in a differential
reflectogram. Detailed studies on this have not yet been carried out so far. It should be
mentioned, however, that the peak energies in a differential reflectogram are the main
characteristic parameters which are used to identify a specific corrosion product rather than
the peak heights. (This will become clearer in the next section.) Thus, surface roughening
is not considered to be of major importance in differential reflectometry.
Before describing actual experiments, a brief introduction pertaining to the origin of
the differential reflection spectra is given now.

7. Principles of DilTerential ReDection Spectroscopy

If a photon of sufficiently large energy impinges on a solid, it may transfer its energy, E, to
an electron which, in turn, is excited across the Fermi level, E F, into an allowed, unfilled state.
This is schematically shown in Figure 8a where a transition from a filled d-band to an unfilled
p-band is indicated by a solid arrow. In metals this transfer occurs predominantly by
conserving momentum (no change in wave number). It is therefore called a "direct interband
transition". Oxide films are sometimes semiconducting. In semiconductors "indirect interband
transitions" are possible, which implies that phonons (lattice vibration quanta) are additionally
involved in the transition process (Fig. 8b). Indirect interband transitions may be disregarded

(a) E (b)

EF-O_._.-.- _._0

P
d

w L r

Figure 8: Schematic representation of electron bands. (a) direct interband transitions

in copper; (b) indirect interband transitions in a semiconductor. The abscissa in both graphs
is the wave number (which is proportional to the momentum).
254

for the interpretation of metal spectra because they are generally weaker than direct
transitions by two or three orders of magnitude. They are only observed in the absence of
direct transitions.
In any event, interband transitions are responsible for the peaks in differential
reflectograms [21]. Since each material has a unique electron band structure, one would
correctly expect that the energies for a series of interband transition peaks would also be
unique. In other words, an experimental differential reflectogram is seen as a "finger print"
of a given material, similarly as known from x-ray or Auger electron spectra.
The question then arises why a spectrum of the difference of two reflectivities is
needed, that is, why the reflection spectrum of one sample alone would not suffice? To
answer this, we have to realize that conventional reflectivity curves possess in general a broad
structure, as schematically depicted in Fig. 9. (This is due to the fact that optical interband
transitions take place between broad energy bands rather than distinct energy levels, as
known from x-ray spectra). By measuring the difference between two nearly identical
reflection curves, sharp peaks are obtained (Fig. 9). In essence a first derivative-like

R2
/
--,
R ,
\
\
\
.rl\
I \ \

r
,, \
\
i . \
\ ,,
,
I
~ R ,,/" ;
i
.,."".i ""
..... . -....~...... .......

E-
Figure 9: Schematic representation of the reflectivity spectra of two nearly identical
solids (R\ and R z). The difference spectrum ~R of the two solids is also shown. (Note: the
change in R is grossly exaggerated.)

spectrum of a conventional reflection spectrum is automatically obtained by differential

reflectometry.
Differential reflectometry belongs to a family of techniques which is called
"modulation spectroscopy". Modulation techniques emphasize transitions at "critical points"
in the band structure (Fig. 8a) and discriminate the observed electron transitions from an
essentially featureless background.
Differential reflcctometry can be best understood by considering "compositional
modulation". Here, a change in the band structure is obtained by small additions of solute
 255

atoms to a base metal. For example, substituting some monovalent copper atoms through
bivalent zinc causes the Fermi energy to shift to a higher energy due to the extra electrons
which are added to the matrix, see Fig. 10. In addition the individual electron bands may
change slightly as indicated in Fig. 10. This, in turn, causes a change of the interband
transition energies for the alloy. When the difference in the reflectivities between the two
alloys of different composition is periodically measured, peaks in the differential reflectogram
are obtained, as described above, and as depicted in Fig. 11.

\
In summary, any external perturbation to the electron band structure will manifest
itself by interband transition peaks in a differential reflectogram.

---------

~ \\1

r 81
If:l~

Figure 10: Part of a calculated electron band diagram for a pure material (e.g., copper,
solid lines) and for a dilute copper-based alloy (dotted lines). Two interband transition
energies, Er, are also depicted.
256

A c
(%)

llf=5=;==;::,=:;=2::,,0:;::,==r==2-:::;.,5==;=~==r=~;:::==*~~ E(eV)
BOO 700 600 500 400 300 200 ~(nm)

Figure 11: Experimental differential reflectogram obtained by scanning the light of the
DR between pure copper and a copper-based alloy.

8. Electrochemical Oxidation of Copper

We start our discussion of experimental results with the oxidation of copper since copper has
fairly well characterized corrosion products whose electronic band structure [22] and optical
constants [23] are known for some cases and whose calculated [24] and experimental [25]
Pourbaix diagrams have been established. The theoretical equilibrium potential - pH
diagram for the system copper-water (Fig. 12) suggests essentially two oxidation products,
depending on pH and potential: cuprous oxide (Cu 20), cupric oxide (CuO) and possibly
cupric hydroxide, Cu(OHh also called hydrated cupric oxide. The latter is speculated to co-
exist with Cup in a limited region of the Pourbaix diagram [25,26].
Fig. 13a depicts an experimental differential reflectogram of a copper specimen pair
whose one half was potentiostatically held in the Cup region of the Pourbaix diagram (18
hours at -250 mV(SCE) and pH 9) while the other half was maintained 50 mV anodic to the
reversible hydrogen potential to prevent film formation as well as hydrogen evolution. The
resulting differential reflectogram is characterized by several peaks of which the most
prominent one is situated near 380 nm, that is, at 3.25 eV.
In order to elucidate whether or not the corrosion cell window or the electrolyte may
have an influence on the peak structure of the aforementioned corrosion product, the
sample pair was removed from the solution, rinsed in triply distilled water and dried. Then
a second, ex situ reflectogram was obtained (Fig. 13b). The peak positions are observed to
be nearly identical in both cases. It is therefore concluded that the electrolyte and the
window do not change the reflectograms and that the present film does not undergo a
compositional alteration upon removal from the solution.
For identification of a given corrosion product, as represented for example by Fig. 13,
several methods can be employed. One of them involves a combined ESCA-AES analysis
 257

t 7000
I
~
-S500
~
lI.J'

-500

Figure 12: Calculated equilibrium Pourbaix diagram for copper in chloride-free solution
[24] considering the species Cu, Cup and CuO. All metal-containing ionic species are at 10'"
activity.

(i.e., electron spectroscopy for chemical analysis in conjunction with Auger electron
spectroscopy). This was done in the present case applying magnesium Ka x-rays [27]. The
resulting data clearly identify the present corrosion product as CUzO. A second method of
identification is accomplished by comparing the absorption peaks in Fig. 13 with interband
transition energies deduced from calculated electron band structures. This has been also
done [19] with the same result as above. The differential reflectogram depicted in Fig. 13
can therefore be linked with confidence to CUzO. This is to be expected since the oxide
formation was conducted in the Cup region of the Pourbaix diagram.
It is of further interest to know whether or not x-ray radiation or an insertion of the
present sample into the vacuum of the ESCA chamber would alter the film composition. To
investigate this, a third differential reflectogram was taken after the ESCA experiment as
shown in Fig. 13c. No changes in peak energies have been observed which indicates that in
the case of CUzO the ESCA analysis does not lead to any alteration of the corrosion product.
This result cannot be expected, of course, for all possible corrosion products, particularly not
for hydroxides. A check as described above in Fig. 13c has therefore to be conducted in each
individual case.
The second principal corrosion product of copper, CuO, was obtained by holding
copper in a buffered solution at +200 mV(SCE) and pH = 9.2 for 18 hours (Fig. 12). The
resulting differential reflectogram (Fig. 14) shows only shoulders around 560 and 360 nm
superimposed on a steady increase in aR!R towards smaller wavelengths. (The shoulder at
258

l1R
R

10%

800 700 500 400

A(nm)

Figure 13: Differential reflectograms of Cup obtained in (a) in situ (pH 9.0, -250
mV(SCE), 18 hours); (b) same as in (a) above, but measured outside of corrosion cell; (c)
same as in (b) above, but after ESCA analysis. The zero lines of the individual curves have
been shifted for clarity. (The electrolyte of this and subsequent experiments consisted of 143
ml 0.7 H 3B03 mixed with 117 ml 0.5 M NaOH.) The insert gives the wavelengths of the
peaks. The dotted line emphasizes the smaller peaks.

560 nm observed for both Cu and CUzO is probably caused by the fundamental absorption
edge of copper which is located at this wave-length as can be seen in Fig. 11). ESCA analysis
performed on this film, similar CuO [27]. Likewise, a comparison of an experimental with
a calculated differential reflectogram obtained by making use of the optical constants for CuO
clearly identifies Fig. 14 to be characteristic for cupric oxide [19]. As before, in situ, ex situ,
and differential reflectograms after ESCA analysis are identical [27] indicating that CUO is
not affected by drying and by insertion into the ESCA vacuum chamber.
The third corrosion product which is believed to form on copper in aqueous solution
is more difficult to obtain and to identify. Cu(OH)z is said to be a metastable species which
coexists initially with Cup and is reported to have a blue color [24,26]. Verink and Lee [28]
found, while determining potentiokinetically an experimental Pourbaix diagram for the
copper-water system, a secondary passivation event in the CUzO region and ascribed this to
the formation of Cu(OH)z. Shanley et al [19] found, when holding copper in aqueous
solutions at -200mV(SCE) and pH 9, an initial decrease in differential reflectivity, before the
peaks, characteristic for CUzO, eventually emerged after longer potentiostating times. This
decrease in ARm, particularly in the blue region, was attributed to the formation of cupric
 259

[(cfll-
16 2 3 4 55

Z2t!Y
I

D~---

800 500 200

-::Ifnml

Figure 14: Differential reflectogram of copper held potentiostatically in the CuO region
of the Pourbaix diagram (+200 mV(SCE); pH = 9.2; 18 hours) [19].

hydroxide. Finnegan [29] duplicated these experimental conditions (-232 mV(SCE); pH 9.2),
but used a Cu-9.2 at. % Zn alloy for the experiment which he potentiostated in a buffered
solution. (The electrolyte was stirred and continuously purged with hydrogen gas). Fig. 15
depicts the resulting reflectogram which shows a decrease in differential reflectivity towards
smaller wavelengths and a pronounced valley near 276 nm. The corrosion product exhibited
a light blue color which may be related to the valley in the reflectogram in the blue region.
No positive identification of this corrosion product can be made at this time because of the
fact that an electron band diagram for Cu(OH)z is not yet available and ESCA studies are
not possible due to the requirement that the specimen has to be removed from the
electrolyte and inserted into a vacuum. This procedure has been found to destroy Cu(OH)z.
In order to construct experimental Pourbaix diagrams by optical means, the
wavelength of the probing light of the differential reflectometer is maintained at a constant
value (for example, 385 nm in the case of copper). Then, the potential on one of the
specimen halves is slowly increased from the immunity region while toR/R is recorded
continuously (Fig. 16). At certain potentials (E) the toRJR versus E curve eventually exhibits
a change in slope which is attributed to the crossing of a border between two fields of
stability. Repetition of this scan for various values of pH leads to the boundary lines of
interest. These lines, are, however, not as distinct as shown in the equilibrium diagram of Fig.
12. Instead, a broad transition area is observed which suggests the coexistence of two
neighboring species near the boundary.
We demonstrate this now for a specific case, namely, for copper held at pH 12 in a
0.15 N NazS04 solution. At the beginning of the experiment, a potential of -900 mV is
applied to both sample halves. Then the potential on one of them is slowly raised (about 100
260

flR
R
0

(%)
465

40

276
80L-__~____~____~__~~__~=-~~
800 700 600 500 400 300 200
}.. (nm)

Figure 15: Differential reflectogram of Cu-9.2 at % Zn, held at -232 mV(SCE) and pH
9.2. The suspected corrosion product is Cu(OH)2.

mY/min) until oxygen evolution occurs. At that point the potential scan is reversed, that is,
the potential on the sample is gradually lowered to its original value. The result is depicted
in Fig. 16a. We note that ~RJ1{ starts to increase at about -500 mV(SCE). The rise is
is initially small, but accelerates with increasing potential. At about -325 mV(SCE) the rise
slows down, peaks eventually near -150 mV(SCE) and then reaches a minimum near
+20mV(SCE). The first potential region between about -500 and -325 mV(SCE) is
ascribed to the formation of Cu 20 which agrees reasonably well with the equilibrium Pourbaix
diagram. Starting near -325 mV(SCE) the growth of Cup appears to be reduced or Cup
is converted into another species, that is, CuD. This process is initially slow but picks up in
momentum the higher the applied potential is. A thickening of the CuD film causes finally
an increase in ~R!R at positive potentials.
When reversing the scan, CuD remains stable far into the Cu20 region, that is, until
about -400 mV(SCE) at which potential ~R!R starts to further increase, probably due to the
formation of Cup. It is not until -800 mV(SCE), however, before the corrosion film is
dissolved and both sample halves have nearly the same reflectivity. It is learned from Fig. 16a
that the relatively large scanning speed used in this experiment prevented a sharp delineation
between the individual areas of stability.
The nature of the hysteresis-type behavior during the decrease in potential, seen in
Fig. 16a, particularly the large value for ~RJ1{ between -400 and -700 mV(SCE) demands
some further exploration. The differential reflectometer is ideally suited to identify the
species involved. Copper was held for 16 hours in the CuD region of the Pourbaix diagram
(+500 mV(SCE» in a chloride-free solution of pH 10. The differential reflectogram
obtained this way (Fig. 17a) shows, as expected, the CuD characteristics. Subsequently, the
potential on the sample was decreased to the Cup region (-225 mV(SCE), pH to). After
 261

('I> )
30

20

10
(.,
o-F"'-----
-800 -1100 -400 -200 0
ECmV(SCEI)
J

... 0.8

o
-0.2 (b)

-0.4

-800 -1100 -400 -200 '200 <400

ECmV(S CEll

Figure 16: Potentiokinetic scans of (a) reflectivity and (b) electrochemical polarization
for copper in .15N NaZS04 solution at pH 12 (1 = 385 nm). Scanning speed approximately
100 mY/min.

20 hours some Cu20 characteristics (for example the 380 nm peak) are observed which are
superimposed on a strong CuO background, (Fig. 17b). After a total potentiostating time of
88 hours, an even stronger 380 nm peak is seen which is still superimposed on an upward
trend of AR!R known to be typical for CuO (Fig. 17c). It can be concluded from these
results that under the chosen conditions, Cup develops on top of an already existing CuO
layer without dissolving the latter to a major extent. This explains the large value of ARJR
in the respective potential region seen in Fig. 16a.
For comparison, Fig. 16b depicts a potentiokinetic polarization curve which was taken
concomitantly with the differential reflection scan. A striking correlation between the
behavior of this and the optical curve is noted.
In order to further investigate the coexistence of various corrosion products, an
experiment was conducted in which copper was potentiostated near the Cu~/CuP
boundary line. Fig. 18 depicts a differential reflectogram for which copper was held for 18
hours at -SO mY(SCE) and pH 9.2, that is, near the CuzO/CuO co-existence line. It is
observed that the primary corrosion product is Cu20, as indicated by the multiple peak
structure, comparable to that in Fig. 13. The rising slope, particularly between 300 and 200
nm, suggests the presence of some CuD.
262

'00 t>Qv ,', ~.:. zoo

Figure 17: Differential reflectogram of copper (a) potentiostated in the CuO region of
the Pourbaix diagram (+500 mV (SCE), pH 10, 16 hours); (b) same sample after the
potential has been reduced to -225 mV (SCE) (20 hours); (c) same as in (b) but
potentiostated for 88 hours. The curves have been shifted for clarity.

30

10

BOO 700 400 300

>., om

Figure 18: Differential retlectogram of copper which was held in the CuO region near
the CuP/CuO equilibrium co-existence line (-50 mV (SCE), pH 9.2, 18 hours).

Copper held briefly in the CuO region near the Cup area always exhibits initially a
CUzO component in the differential reflectogram, probably due to one or several monolayers
of Cu20 at the copper interface. With increasing polarization time, the reflectograrns
gradually lose the Cup peaks and show more of the CuO characteristics. We shall return
to this point in a later section.
 263

The kinetics of the formation of a certain corrosion product can be studied in real
time by recording a series of subsequent differential reflectograms. As an example, the
emergence of CUzO is depicted. Fig. 19 displays a sequence of reflectograms obtained for
copper which was held for various times in a buffered electrolyte of pH 9. After 10 minutes
(0.17 hours) the characteristic features of CUzO are already noticeable. At the beginning, the
peak height near 380 nm grows rapidly. The increase eventually slows down as the
potentiostating times become longer. The growth kinetics obey a logarithmic relationship as
can be seen in Fig. 20 in which the peak height at 380 nm is plotted versus time. These
experimental results suggest that the signal strength in a differential reflectogram is related
to the film thickness.

[(eJl)-
16 2 ] 4 ,- 6

I 70%

800 600 400 200

~},(nml

Figure 19: Differential reflectograms depicting the in situ evolution of Cup on a copper
substrate in a buffered electrolyte of pH 9 (-200 mV (SCE)). The individual curves are
shifted for clarity.
264

Figure 20: Growth kinetics of Cup on copper as deduced from the peak heights at 380
nm in Figure 19.

9. Free Corrosion of Copper

The most common type of corrosion occurs when materials are exposed to the atmosphere
or to water. For the identification of the corrosion products which form in air, one half of
the surface of a copper slab is covered with a protective lacquer after polishing. This film
is peeled off after the sample has been exposed to laboratory environment for a certain
amount of time. Fig. 21a shows that after one day, the resulting differential refiectogram
yields the multiple peak structure characteristic for Cup. The upward bend at its tail at low
wavelengths indicates, however, that some CuO is additionally present (compare to Figures
13 and 14). With increasing time of exposure to air, the refiectograms lose successively the
Cup peaks and tend to display the pattern typical for CuO (Figs. 21b and c). This result
indicates that the initial corrosion product on copper which forms in air is Cu 20 and not
CuO, as the copper-oxygen phase diagram predicts [30]. To explain this, one needs to know
that CuO tends to reduce to Cup in the presence of metallic copper [31].
Differential refiectograms taken on copper which was immersed into triply distilled
water display again some mixed oxides with Cup being the predominant species, see Fig. 22.
(There is little difference in these refiectograms whether oxygen is continuously streamed
through the water or not.)
These few examples vividly demonstrate the capabilities of the differential
refiectometer to discriminate between various corrosion products. We turn now to a different
subject and discuss how to estimate the thickness of a corrosion film from differential
refiectograms.
 265

f
5·4~~ _ _ __

lL-____~~----~~----~
600 400 200

X, nm

Figure 21: Differential reflectograms obtained for copper which was exposed to a
laboratory environment for (a) 1 day; (b) 4 days and; (c) 11 days [32]. The curves are shifted
for clarity.

)., nm

Figure 22: Differential reflectogram of copper which was immersed for 18 hours in
distilled water through which oxygen was bubbled. One half of the copper slab was protected
during corrosion by a lacquer [32].

10. Estimate of Film Thickness from DitTerential ReDectograms

The thickness dependence of the normalized difference in reflectivity can be calculated,

applying the (rigorous) two layer formulas discussed above, if the optical constants of
substrate and corrosion product are known. This has been done for euo on copper,
immersed in an aqueous solution [15,19]. The resulting differential reflectivity is shown in
266

Fig. 23 for a specific wavelength. The curve displays a steep initial increase in ~R!R for
layers up to about 400 A, followed by strong interference-caused oscillations which gradually
damp out. (Interference effects cease to occur when the film thickness is large enough to

60

20

o 2000 3000 4COJ 5000

Film thickness. A

Figure 23: Calculated differential reflectivity as a function of film thickness for CuO on
a copper substrate in aqueous solution (1 = 600 nm, neu = 0.17; Kcu = 3.07; Ilcuo = 2.6; kcuo
::: 0.445; nmo ::: 1.33) [19].

prevent the light from reaching the metal/film interface.) The film thickness of greatest
interest for corrosion-optics is in the initial range in which interference peaks do not yet
occur. The calculated ~R!R for the first 200 A is depicted in Fig. 24 for CuO on Cu. This
graph demonstrates that, in essence, the thickness of CuO can be obtained from the signal
strength in a differential reflectogram. As an example, the film represented by Fig. 14 is
determined by utilizing Fig. 24 to be about 5 A thick. This result is well within the generally
accepted range. Similar curves as the one shown in Fig. 24 can be calculated whenever the
optical properties of film and substrate are known. As a second example, the thickness
dependence of ~R!R for NiO on a nickel substrate is included in Fig. 24.
The thicknesses calculated this way should be taken with a grain of salt, however. An
inherent uncertainty in these values stems from the fact that the optical constants of the
corrosion products, which form as a thin film, may not be exactly the same as those of the
bulk. This was already emphasized when discussing ellipsometry. The question then arises
how a certain error in the optical constants would change the estimated film thickness when
using differential reflectometry. A calculation shows that a 10% error in the index of
refraction changes the film thickness likewise bi' about 10%. Staying with the above
mentioned example, that is, a film thickness of 5 A: a 10% error in d is considerably less
than a monolayer which is too small to be of any concern. An error in the damping constant,
k, has an even smaller influence on the estimated film thickness: a 10% change in k varies
d by only 4%. In conclusion, even though the differential reflectometer is not meant to be
an instrument for the measurement of film thicknesses, an estimate of d can nevertheless be
obtained with reasonable accuracy from the signal strength in a differential reflectogram. In
 267

100
f.!
R
( ... ) 80
I :
60 ------ .----1- ~----I- -,-- -,---.
: : Mid !
40 +--:-_--+_+__i_-l_\~

20 -~--------:;".,

O~~~-r,-~~~~-r~~~~-r-r~
o 50 100 150 200
o
Film thlckn... A

Figure 24: Enlarged portion of Figure 23 for small film thicknesses (l = 600 nm, CuO
on Cu in aqueous solution) [19]. (Note: a similar curve published in [18], Figure 31 has a
printing error. The unit on the abscissa should read "A".) For comparison, AR.IR: for NiO
on Ni is also depicted (l = 550 nm)

cases where interference peaks occur, the film thickness can be calculated from the
wavelengths of these peaks [20].

11. Corrosion by Annealing in Dry Oxygen

Oxidation at elevated or high temperatures is quite common in practice. In order to conduct

the appropriate experiments under controlled conditions, heating in dry oxygen is advised.
After the annealing, the oxidation product on half of the sample needs to be polished off to
obtain the necessary configuration required for the differential reflectometer. The
differential reflectogram for copper which was heated at 106°C for 8 minutes in dry oxygen
has the characteristics of CuO similarly as shown in Fig. 14 [20] in agreement with the
copper-oxygen phase diagram [30]. Annealing at higher temperatures results in reflectograms
similarly as depicted in Fig. 13 [20] which indicates formation of Cup, again in accord with
the copper-oxygen phase diagram [30]. (It should be mentioned, however, that the Cup
reflectogram was observed already after heating at 144°C, that is, substantially below the
reported critical temperature of 375°C. This confirms earlier observations [30] that CuO
tends to decompose under certain conditions into CU20 and O2.)
In Figures 25-27, differential reflectograms are depicted which have been obtained
by heating tungsten, molybdenum and chromium as specified in the respective figure captions.
Characteristic interband transition peaks are observed which can be utilized for the
268

identification of these corrosion products. In the case of tungsten, where a calculated

electron band diagram for W03 exists, the experimentally observed peak structure agrees with
the calculations.

~{%J

,,

800 700 600 500 400 300 200

A{nm]

I I I I-
1.6 2 2.5 4 5 6
E leV]

Figure 25: Differential reflectogram of tungsten trioxide on a tungsten substrate obtained

by heating W at 251°C for 20 minutes in dry oxygen [20].

AR
1i" I"J
15

10

800 700 600 500 400 300 200

~- Alnm]
I' I I
1.6 2.0 2.5 3 4 5 6
E{.VJ~-

Figure 26: Differential reflectogram of molybdenum oxide obtained by heating Mo for

20 minutes at 244°C in dry oxygen [20].
 269

10

o+-----------------~

8DD 700 600 500 400 300 200

-A.(nrn)
I I I I I I I
1.6 2.0 2.5 3 4 5 6

E t'V!-

Figure 27: Differential reflectogram of chromium oxide obtained by heating Cr at 510°C

for 20 minutes in dry oxygen.

12. Passivation of Nickel in Aqueous Solutions

The nature of the protective films which form on nickel in aqueous solutions has been the
subject of a large number of scientific papers, (most of which are quoted in Ref. [27 and 33]).
Despite all this research effort, disagreement still exists between different authors. For
example, it has been suggested that the film which forms in acid electrolytes is either
Ni(OH)2, NiO, NiOOH, Ni30 •. Ni0 2, non- stoichiometric NiO, or chemisorbed oxygen which
forms via the NiO,d.' NiOH+, or NiOOH intermediates. A slight majority of investigators
claim that in neutral solutions NiO is the passivating film, while others suggest that NiOOH,
Ni30., Ni0 2 or the interconversions of these compounds exist. Finally, most investigators
believe that the passive film in alkaline electrolytes consists of Ni(OH)2 or probably of Ni30.,
NiP3 or NiOOH. It cannot be the purpose of this paper to decide authoritatively between
these various suggestions solely based on differential reflectometry, particularly as long as
electron band diagrams for all above mentioned species have not yet been calculated. (Once
this has been accomplished, however, the peak energies in differential reflectograms might
give an independent basis for one or the other assignment).
However, differential reflectometry does provide the unique opportunity to follow
in situ the interplay between various species which form when potential and pH are altered.
This will be emphasized in the following paragraphs. A recently revised nickel-water
Pourbaix diagram is depicted in Fig. 28 for reference purposes [35].
By varying pH and potential, three types of differential reflectograms have been
found. The reflectogram observed in most cases is depicted in Fig. 29. It is distinguished by
270

~~~------------------~

NiOOH

+1

NI+2 r- _ 0.5

_ lj(o~)~ - 0
- - I I -0 5
-~~i(OH)" •
~ '---~

Ni
w
()
C/)

>

o 4 8 12 16
pH

Figure 28: Nickel-water EMF-pH diagram at 25°C [35]. All metal-containing ionic
species are at 10-6 activity.

l>R
R

8CXJ 500 400 300 200

X(nm)

Figure 29: Differential reflectogram obtained typically by potentiostating nickel between-

600 and -200 mV (SCE) and in the pH 8-10 range.
 271

a steady increase of AR/R towards smaller wavelengths. (The cut-off near 238 nm is caused
by the gradual lack of transparency of the electrolyte and the corrosion cell window towards
shorter wavelengths.) The film represented by Fig. 29 has been analyzed by ESCA [34]. The
resulting 1s oxygen and Ni 2P3/2 peaks have been compared with literature data on nickel
compounds [36-39]. From this it is concluded that the film in question is Ni(OH)2'
MacDougall strongly objects to this interpretation and claims instead that non-stoichiometric
NiO, which supposedly has an identical XPS spectrum as Ni(OH)z, would be the only
passivating species which occurs on nickel [40].
The second type of differential reflectogram which has been observed in this context
is characterized by a peak near 345 nm, superimposed on a steady rise of AR/R towards
smaller wavelengths, as known from Fig. 29. The differential reflectogram depicted in Fig.
30, thus, suggests a co-existence of two corrosion products. This type of reflectogram is

aR
R

20 ...

800 700 600 500 400 300 200

A(nm)

Figure 30: Differential reflectogram of nickel annealed in (non-dried) oxygen at 460"C

for 3 minutes [33]. Note that a similar reflectogram (but with a weaker peak at 345 nm) is
observed when nickel is potentiostated at +300 mV (SCE) and pH 11.75 or similar potentials
and pH values

obtained, for example, by potentiostating nickel between zero and +300 mV(SCE) at pH
11.75 or by heating nickel shortly at 460°C in oxygen. The ESCA spectra near the oxygen
1s and Ni 2P3/2 peaks both indicate that a superposition of two closely spaced peaks exists
which can be computer deconvoluted. The resulting maxima are characteristic for NiO and
Ni(OH)z, as seen in Figure 31 [33]. This suggests that a mixed film, consisting of Ni(OH)z and
NiO may form under certain conditions. The formation of mixed oxides during passivation
is not unusual. It has been reported, for example, also for iron [44].
The third film type observed in differential reflectometry of nickel can be only
obtained in situ at high potentials, in solutions with pH ~ 8, and if a film as represented by
272

Fig. 29 pre-existed on the nickel. The reflectogram is distinguished by a broad peak near 640
nm which, by growing stronger, causes ARJR near 238 nm to decrease (Fig. 32). This

100

-;;; -;;;
.., 1::
...
!:!
...
50

8.E.
EnerQY (eV) (b)

Figure 31: Dissolution of (a) Ni 2p3f2 and (b) oxygen 1s ESCA peaks representing the
nickel film whose relfectogram is depicted in Figure 30 [33]. B.B. = binding energy.

llR a : 30 min. at 550 mY.c8

b : (a) plu. 2 min •• t 900 mY.ce
R c : Ca) plu. 12 hr. 8t 900 mYlce

~ ____ c

10-" b

800 700 600 500 400 300 200

A(nm)

Figure 32: Differential reflectograms obtained on nickel in sulfate solution at pH 8. The

potentials are given in the figure [33]. The curves are shifted for clarity.
 273

suggests that a conversion reaction takes place. The peak near 640 nm rapidly disappears
when the sample is removed from the electrolyte or if the potential on the working electrode
is reduced. Since this film is only stable in situ, a positive identification by ESCA or similar
ex situ techniques is not yet possible. It has been suggested by several investigators [41-44]
that Ni(OH)2 can be oxidized to NiOOH at high potentials via a deprotonation reaction:

Ni(OH)2 + OR" - NiOOH + H 20 + e'

Thus, for the time being, the reflectograms in Fig. 32b and 32c are designated as
"NiOOH". This also agrees with the Pourbaix diagram depicted in Fig. 28.
It is noted that the "peak" near 640 nm is rather broad in contrast to the relatively
sharp peaks observed in other differential reflectograms. This type of broad structure has
been observed before on amorphous copper-zirconium alloys. It is therefore suggested that
the corrosion products represented by Fig. 32b and c are non-crystalline (i.e., possibly a gel).
A potentiokinetic scan between -900 and +500 mY is shown in Fig. 33 for which the
wavelength of the differential reflectometer was fIXed at 330 nm. It is observed that ..1RJR

~
II
(%)
15

10

-800 -600 -400 -200 +200 +400

J E[mV(SCE)]

[P~2J
0.04

0.03

0.02

0.01

o
-0.01 (b)

-800 -600 -400

I
-200 b 400
E[mV(SCe)]

Figure 33: Potentiokinetic scans of (a) differential reflectivity, and (b) electrochemical
polarization for nickel in 0.15 N Na2S0 4 solution at pH 11.75 (l. = 330 nm). Scanning speed
approximately 100 mY/min.
274

begins to rise near -720 mV(SCE) indicating the start of film formation. This agrees with the
Pourbaix diagram depicted in Fig. 28. (The electrochemical polarization starts to decrease
at this potential which also indicates formation of a passive film.) From then on, aR/R, that
is, essentially the film thickness, rises gradually with only occasional small changes in slope.
(Ellipsometrically determined changes in film thickness arrive at the same result [45].) The
electrochemical polarization eventually increases rapidly near 450 mY, (that is, in the region
of oxygen evolution) while aR(R: stays essentially constant. The optical data suggest that any
new species which may form at higher potentials develops via a transformation reaction of
an already existing species.
When reducing the potential, aR/R does not become smaller. This suggests that the
protective film remains on nickel under the existing conditions, that is, in alkaline solutions.
(This result is in agreement with the literature [46].) Actually, Fig. 33 seems to suggest that
the film thickness grows even slightly more, as long as a potential > -700 mV is applied.
The thickness of the corrosion films can be estimated with the aid of Fig. 24. These
thicknesses vary, depending on the experimental conditions between 5 and 60 A (See "Note"
in Fig. 24).
Having thus laid the groundwork for the distinction between three types of
differential reflectograms it is now possible to investigate the passivation of nickel in various
pH ranges. This has been done [34] by exposing nickel to 1 N Na 2S0 4 solutions of pH 4,8
and 12, which have been either de-aerated by purging them with purified Nz,
or left open to the atmosphere. Concurrent with the differential reflectograms, electrical
measurements have been taken in 50 mV increments after two minutes of stabilization times.
For brevity only a few of the resulting graphs are depicted below. (It should be re-
emphasized that the optical reference as well as the working electrode are initially in the
same condition and at the same cathodic potential. Thus, a possible pre-passive film which
might have formed after polishing and during the transfer to the corrosion cell is initially
present on both sample halves. Therefore, any difference in aR(R: essentially represents a
difference in film thickness between the reference and the working electrode).

The experimental results can be divided into two groups:

1) Films on Ni which form in a de-aerated or open solution of pH 4 or in a de-aerated

solution of pH 8 show, with only minor modifications, the same behavior (Fig. 34). The
reflectograms of these films are exclusively of the type as shown in Fig. 29. The differential
reflectivity at 238 nm, that is, essentially the film thickness increases with increasing potential,
but eventually levels off to a constant value (Fig. 35). The current density peaks near the
potential of maximum film growth rate. The film thickness, as estimated from aR/R, is
substantially larger if grown in de-aerated solutions of pH 4 and 8 than that on a film
exposed to an open solution of pH 4 [34].
 275

(0) ,- 150 mVISCEI

I b) - 50 m'JISCE)
( c) 350 mVISCE I

800 700 6C() 500

A (nrnl

Figure 34: Differential reflectograms obtained on nickel in a de-aerated 1 N NazS04

solution of pH 4. The potentiostating time at each potential is 2 minutes. The curves are
shifted for clarity [34].

350

300
80%
~-
E 250
u E
c:
<! 60% aJ
200
3-
i
>- '"
(\J

150
~ 40%10::
"0 "-
0::
10 100 <l
'"t::J
U
50

~~~ ____
·200
-L~===c==:==~====t===~o
o 200 4CC 800
PotentloL (mVISCEI]

Figure 35: Polarization and differential reflectivity at 238 nm for Ni in de-aerated 1 N

Na ZS04 solution of pH 4 [34].

2) Films on nickel, potentiostated in a solution of pH 8 open, and pH 12 open or de-aerated

yield (with minor modifications) reflectograms as depicted in Fig. 36. The results suggest a
steady transition between the above mentioned three curve types when increasing the
276

potential. The current density under these conditions rises steadily as potentials are
increased, similarly as shown in Fig. 33b. The passivating films grown in alkaline solutions
are substantially thinner (about one sixth) than those formed at lower values of pH.

(0) -450 mV(SCE)

(b)· -50 mV(SCE)
(cI 250 mVISCE)
I d)' 450 mVISCE)
Ie) 700 mV(SCEI

60%

40%

20%

700 600 500 400 300 200

" Inm)

Figure 36: Differential reflectograms obtained for nickel in an open solution of pH 8

[34]. Note the transitions from the curve types shown in Figures 29, 30 and 32 when the
potential is increased.

In summary, differential reflectometry vividly demonstrates that the composition of

the passivating film on nickel, its thickness and its mode of formation are different for various
values of pH and potential and whether or not the solution contained air. Details are given
in Ref. [34].
Staying with the topic of passivation for another moment, it is of course of interest
to know how certain nickel-based alloys, such as the Monels, behave under similar
conditions. Such a study involving Ni -30 wt % Cu has been done [47] yielding quite similar
results as for pure nickel. As a rule, the protective film thickness for the alloy has been
found to be larger than for pure nickel, except in de-aerated solutions of pH 4 and 8.
Another observation was that the alloy passivates at more anodic potentials than pure nickel.
The respective reflectograms, etc., are not shown here to avoid reader's fatigue. One new
point should be emphasized nevertheless. A mixed film consisting of Cup and Ni(OH)2 has
been found for pH 4, open solution and for pH 8 de-aerated solution, both at low potentials.
This result is displayed here because of its uniqueness. The experimentally found
reflectogram depicted in Fig. 37d can be interpreted to be the superposition of the curves for
Cup (Fig. 13) and Ni(OH)2 (or nonstoichiometric NiO for that matter) as shown in Fig. 29.
It should also be of interest that some de-alloying of copper has been observed in the special
case of de-aerated solutions of pH 4 at anodic potentials. The details are contained in Ref
[47].
Finally, a study of nickel passivation in solutions containing Cl· ions has been
performed. At pH 12 no difference in the film formation process nor in film stability is
 277

AR
R

800 700 800 500 400 300 200

A(nm)

Figure 37: The graphic addition of differential reflectograms for (a) CUzO and (b)
Ni(OH)z results in a reflectogram (c) that is very similar to that obtained from a Ni-30 wt.
% Cu electrode (d) held at -50 mV(SCE) in an open solution of pH 4 [47].

observed when chloride- containing and chloride-free solutions are compared. However, Cl"
ions shift the onset of passivation to more anodic potentials in solutions with pH 4 or 8 [48].
The most interesting behavior is seen in Fig. 38 in which the potential-dependence of
polarization and differential reflectivity are compared for chloride-free and chloride-
containing electrolytes at pH 4, open solution. It is noted that for cathodic potentials,
I1RJlt initially decreases in both cases, probably due to the dissolution of a pre-existing film
on the working electrode which might have formed after polishing and during the transfer
to the corrosion cell. Figure 38 also suggests that the renewed formation of a protective film
starts near -40 mV(SCE) in Cl"environment compared to -100 mV(SCE) for the chloride-
free solution. Of particular interest is that immediately after passivation, that is, after the
current density has returned to a very small value, j rises again in the presence of Cl-ions.
This rise is accompanied by a steady decrease in differential reflectivity. Both observations
suggest depassivation, that is, film thinning, or growth of pits (or a combination of both
mechanisms). Indeed, it has been visually observed that upon removal from the electrolyte,
the nickel surface appears to be severely pitted. This concludes the presentation of results
which have been contributed to the rather complex topic of passivation of nickel by utilizing
differential reflectometry. We turn now to a new subject.
278

4200

~2400
~
o
<:1800
~

0%

-400 -200 o 200 400 600 800

Potential (mV. c .)

Figure 38: Polarization and differential reflectivity data (at 238 nm) for nickel in pH 4
open solution, with and without CI" ions [48].

13. Dezincification

Common yellow brasses (for example copper-30% zinc alloys) are distinguished by a higher
hardness and higher resistance to impingement attack than pure copper. These favorable
properties are reduced when some zinc is lost. Based on visual inspection and on tensile
strength measurements, it has been generally concluded that in copper alloys containing less
than 15 at. % Zn (red brass) essentially no dezincification would take place [49, 50]. Others
dispute this contention. A surface sensitive technique, such as differential reflectometry, is
well capable to decide whether or not the alloy composition on the surface has changed. For
this, the reflectivities of a reference alloy and a possibly de-alloyed sample are compared.
(Both samples have initially the same solute content.) Any difference in composition
between the two samples yields a differential reflectogram which is similar to that in Fig.
11. Peak "A" is known to shift linearly with the difference in zinc content [51]. Thus, from
this change in interband transition energy, the surface composition can be obtained. In
addition, the depth of the dezincified layer can be deduced from the peak height in a
differential reflectogram. This is demonstrated in Fig. 39 in which a series of differential
reflectograms are depicted as taken on an initial Cu -9.2% Zn alloy which has undergone
electrochemical treatment at -350 mV(SCE) in a vacuum de-aerated solution of pH 4 [52].
The resulting peak structure is typical for copper-based alloys involving different solute
concentrations. From the position of peak A, a zinc loss of 3.8 at. % is deduced. Similar
experimental differential reflectograms for other alloys indicate that under the same
electrochemical conditions, a Cu-21.4 at. % zinc alloy loses 4.4 at. % Zn while a Cu-53 at.
 279

5 6'

37
(If--_ _-

'"'--1_---'---_-.11
800 (;00 ·/00 200

Figure 39: Differential refiectograms of a Cu-9.2 at % Zn alloy. One sample half was
potentiostated in a pH 4 solution at -350 mV (SCE). The other half remained unexposed
to the electrolyte (ex situ) [52].

% Zn alloy has lost only 2.6 at. % of the solute. Thus, the amount of dezincification depends
on the initial solute content. Fig. 39 also demonstrates how the de-alloyed layer grows with
increasing time of exposure to the electrolyte, which is indicated by the height increase of
peak A By comparing the growth kinetics of the just mentioned alloys, it becomes evident
that yellow brass dezincifies 16 times faster than red brass. In other words, prolonged
corrosive attack changes only the depth of the dezincified film while the composition of this
layer stays essentially constant once initial de-alloying has taken place [52].
One more interesting observation should be added. If a Cu-9.2 at. % Zn alloy is
potentiostated for 72 hours at the same pH and potential as above, but in an open cell, that
is, oxygen-containing solution, then the differential refiectogram has the characteristics of
a mixed Cu 20lCuO film indicating little or no alloy peaks (Fig. 40).
Fontana and Greene [49] stress the point that the presence of oxygen increases the
rate of attack and that dezincified areas show in the presence of oxygen 90 to 95% copper
with some of it present as copper oxide. However, below this copper-oxide layer the original
alloy seems to be relatively well preserved as suggested by the absence of alloy peaks.
280

12
10
8
6
4
2
o L-_~_---,_ _~_-,::--=----:::-::-:--~;:-
BOO 700 600 500 400 300 200
).,. lnm)

Figure 40: Differential reflectogram of a Cu-9.2 at. % Zn alloy potentiostated at -350

mV(SCE) in an open solution of pH 4 for 72 hours (ex situ) [52].

14. Conclusions

The aim of this paper was to demonstrate the immense capabilities which optical techniques
can provide to supplement and augment the body of knowledge in corrosion science. In
particular, passivation and the initial stages of corrosion are the fields in which optical
methods are most useful, especially when the films need to be studied in situ. The best
known instrument for optical investigations is the ellipsometer, which has been used for
almost one hundred years. For that reason, the relatively new method of differential
reflectometry, which is still comparatively little known among corrosion scientists, and which
has certain advantages over ellipsometry has been emphasized here. It has been shown that
this instrument is well capable to identify corrosion products in situ (once their
characteristic reflectograms or their band structures are known). It can also elucidate the
kinetics of film formation and the thickness of these films. The author would be pleased if
this paper stimulates further research with the differential reflectometer.

15. Acknowledgements

I am deeply indebted to Professor E. D. Verink, Jr. for his stimulating interest in these
studies and his encouragement. Thanks are also due to my former graduate students (c. w.
Shanley, R. J. Andrews, J. B. Andrews, R. J. Smith, F. K Urban, and J. E. Finnegan) who
contributed their talent and endurance to the referenced work. The financial assistance by
the National Science Foundation is likewise gratefully acknowledged.
 281

16. References

1. F. Abele's Ed., "Optical Properties of Solids", North-Holland Pub!. Comp., Amsterdam

(1972).

2. M. Born and E. Wolf, "Principles of Optics", 3rd ed., Pergamon Press, Oxford (1965).

3. O.S. Heavens, "Optical Properties of Thin Solid Films", Academic Press, New York
(1955).

4. R. E. Hummel, "Optische Eigenschaften von Metallen und Legierungen", Springer-

Verlag Berlin, Heidelberg, New York, (1971).

5. R E. Hummel, "Electronic Properties of Materials", Springer Verlag Berlin,

Heidelberg, New York, (1985).

6. W.E.J. Neal and R.W. Fane, "Ellipsometry and its Application to Surface
Examination", J. of Phys. E: Scientific Instruments Q, 409, (1973).

7. E. Passaglia, RR Stromberg and J. Kruger, Eds "Ellipsometry in the Measurement

of Surfaces and Thin Films", Symposium Proceedings, Washington, D.C. (1963),
National Bureau of Standards, Misc. Public 256 (1964).

8. B.D. Seraphin ed. "Optical Properties of Solids - New Developments", North Holland
Publ. Compo Amsterdam (1976).

9. A Val:f~ek, "Optics of Thin Films", North-Holland Pub!. Compo Amsterdam (1960).

10. F. Wooten, "Optical Properties of Solids", Academic Press, New York (1972).

11. P. Drude, Wiedeman Ann. Phys 2Q, 884 (1889).

12. L. Tronstad, "Optische Untersuchungen zur Frage der Passivitat des Eisens und
Stahls", Kg!. Norske Videnskab. Selskabs Skrifter, Trondheim (1931).

13. A VaMrek in "Ellipsometry in the Measurement of Surfaces and Thin Films" (see
Ref. 7), p.25.

14. A Val:frek, Opt. Spectroscopy, 11. 128 (1961).

15. F. Urban III, App!. Surface Science, 33/34, 934 (1988).

16. J. Kruger, in Ref. [7], p.131.

17. J. Alfaro Holbrook and R. E. Hummel, Rev. Sci. Instrum, 44, 463 (1973).

18. RE. Hummel, phys. stat. sol. (a), Th 11 (1983).

282

19. e.W. Shanley, R E. Hummel, and E.D. Verink, Jr, Corrosion Sci. 20,467 (1980).

20. F.K Urban III, RE. Hummel, and E.D. Verink, Jr., Corrosion Sci. 22, 647 (1982).

21. RE. Hummel and R Enderlein, Phys. Rev. B ~ 1529 (1984).

22. J.P. Dahl and Ae. Switendick, J. Phys. Chern. Solids 27, 931 (1966).

23. P.e. Ladelfe, A W. Czanderna, and lR Biegen, Thin Solid Films 10, 403, (1972).

24. M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, First English

Edition: Pergamon Press, Oxford (1966).

25. T.S. Lee III, Master's Thesis, University of Florida (1972).

26. AG. Akimov, I.L. Rozenfeld and M.G. Astafev, Soviet Electro. Chern. 12,546 (1976)
and 13, 1279 (1978).

27. RJ. Smith, Dissertation, University of Florida (1984).

28. E.D. Verink, Jr. and T.S. Lee III, Proceedings of the Third Int. Congress on Marine
Corrosion and Fouling, Gaithersburg, Maryland, 241 (1972).

29. J.E. Finnegan, Dissertation, University of Florida (1980).

30. M. Hansen, Constitution of Binary Alloys, McGraw Hill, New York (1958).

31. J. Kruger, J. Electrochem. Soc.1Q2, 847 (1%9).

32. C.W. Shanley, RE. Hummel and E.D. Verink, Jr., Corrosion Science 20, 481, (1980).

33. RE. Hummel, R.J. Smith and E.D. Verink, JR., Corrosion Science 27, 803 (1987).

34. RJ. Smith, RE. Hummel and J.R Ambrose, Corrosion Science 'J1, 815 (1987).

35. D.e. Silverman, Corrosion - NACE n, 546 (1981), and "Electro-chemical Techniques
for Corrosion Engineering", R. Baboian, ed., p. 117, NACE, Houston, (1987).

36. N.S. McIntyre and M.G. Cook, Anal. Chern. 47, 2208 (1975).

37. N.S. McIntyre, T.E. Rummery, M.G. Cook and D. Owen, J. Electrochem. Soc. 123,
1164 (1976).

38. J. Ali, PhD Dissertation, University of Florida (1983).

39. T.L. Barr, J. Phys. Chern. 82, 1801 (1987).

 283

40. B. MacDougall, D. F. Mitchell and M. J. Graham, Corrosion ~ 85 (1982).

41. E. Jones and W.F.K Wynne-Jones, Trans. Faraday Soc. g 1260 (1956).

42. P.D. Lukovtsev, Proceedings of the 4th Conference on Electrochemistry of the

Academy of Sciences, Moscow (1959).

43. KE. Heusler and K Schoner, Ber. Bunsenges. Phys. Chern. 71.. 885 (1973).

44. H. Kaesche "Die Korrosion der Metalle" 2nd Ed. Springer-Verlag Berlin, Heidelberg,
New York (1979).

45. N. Sato and K Kudo, Electrochim. acta.12. 461 (1974).

46. B. MacDougall, J. Electrochem. Soc. ill. 789 (1980).

47. R.E. Hummel and R.J. Smith, Corrosion Science ~ 279 (1988).

48. R.E. Hummel and R.J. Smith, to be published.

49. M.G. Fontana and N.D. Greene, "Corrosion Engineering", McGraw Hill, New York
(1978).

50. H. Uhlig, "Corrosion and Corrosion Control", John Wiley and Sons, New York
(1971).

51. R.J. Nastasi-Andrews and R.E. Hummel, Phys. Rev. B ~ 4314 (1977).

52. J.E. Finnegan, R.E. Hummel and E.D. Verink, Jr., Corrosion TI. 256 (1981).
ELLIPSOMETRY IN STUDIES OF METALLIC CORROSION AND OXIDATION

Z. Szklarska-Smialowska and Rengan Krishnakumar

Fontana Corrosion Center
Department of Materials Science and Engineering
The Ohio State University
Columbus, Ohio 43210

ABSTRACT

The sensitivity and the in-situ applicability of ellipsometry

have led to its widespread use in corrosion research. This article
introduces the fundamental principles and measurement techniques
involved in ellipsometry. A critical evaluation of the advantages and
shortcomings of the technique is presented. The generic nature of the
information obtained, and a detailed account of some of the studies
reported in literature are discussed. The latter is subdivided into
the various forms of corrosion to which ellipsometry has been applied.
Studies on characterization of passive films have been classified
with respect to the metal. A section on the use of ellipsometry to
characterize the oxidation behavior of materials exposed to oxygen is
also presented.

1. INTRODUCTION

The kinetics of most corrosion and oxidation phenomena in

materials is determined by the nature of the reaction product film that
forms on the interface between the reactive media involved. It is
therefore essential to completely understand the characteristics of the
film to exercise control over the corrosion or oxidation process. The
nature of the reaction product film ranges from the compact protective
passive film formed on chromium in aqueous environment at room
temperature, to the non-protective oxide film on pure iron under
similar exposure conditions. The dramatic influence of the surface
film on the kinetics of metal dissolution triggered the quest for
understanding the nature of these f'jlms in the early twentieth century.
Ellipsometry is an optical technique that was first applied to
study surface films in 1935 [1]. This technique has since been
used extensively, principally in corrosion studies, due to its high
285
M. G. S. Ferreira and C. A. Melendres (eds.),
Electrochemical and Optical Techniquesfor the Study and Monitoring of Metallic Corrosion, 285-354.
© 1991 Kluwer Academic Publishers.
286

Figure 1: An instantaneous 'snapshot' of a plane-polarized light wave

showing the electric (E) and magnetic (8) field vectors along a
particular ray. The wave is moving to the right. [Reference: R.
Resnick and D. Halliday, Physics Part II, John Wiley &Sons, Inc., p.
1149 (1968). p.1149 1968).

sensitivity and ability to be used in-situ. The fundamental theory

behind ellipsometry for film free surfaces was developed by Fresnel in
the early nineteenth century. Following this, in 1890, Paul Drude
expanded the theory to film covered surfaces. Since the first
application of ellipsometry in 1935 [1] to characterize surface films,
significant refinements on the ellipsometer have taken place and the
progress in instrumentation is continuing even at this time.
Simultaneously, computational techniques to analyze the ellipsometric
data have also evolved, since the early manual calculations that could
only employ the approximate form of the equations proposed by Drude.
With the advent of high speed digital computers, it has become possible
to solve for the optical and thickness parameters of surface films by
employing the, full form of the Drude equations. It has also become
possible to do it in a dynamic manner as the surface reactions proceed.
This has been achieved by suitably combining advances in ellipsometer
instrumentation, high speed data acquisition, and computational
techniques.
2. THEORY

In order to fully comprehend the principles of ellipsometry, it

is essential to understand the wave nature of light. Visible light is
 287

an electromagnetic radiation composed of electric and magnetic field

vectors oscillating perpendicular to the direction of light
propagation. The electric and magnetic field vectors are perpendicular
to each other at any instant of time and at any point in space, and lie
on a plane whose normal lies along the direction of propagation of the
light. Due to the mutual orthogonality of the electric and magnetic
field oscillations, specifying the nature of anyone of the fields
completely defines the light wave. Therefore, in all the discussions
that follow, only the electric field vector (E) will be considered. To
completely specify a light wave in vacuum, one has to specify the
amplitude, phase, frequency, wavelength and direction of propagation of
the light wave. The Illustration of the fundamental principles of
ellipsometry is best done by considering applications that employ light
of a single wavelength (monochromatic light). The ensuing discussion
will therefore assume 'light' to mean 'monochromatic light'. The
direction of oscillation of the E vector is the polarization direction.
The light from a monochromator comprises of E vectors oscillating in
all possible directions normal to the propagation direction, in a
totally random fashion. This is called randomly polarized or
unpolarized light.
When light is incident on an interface between two media of
different refractive indices, a part of it gets reflected and a part of
it gets refracted. The relative intensities of the two light waves is
determined by the nature of the particular interface that causes such
reflection and refraction. Ellipsometry deals only with reflection,
and only this process will be discussed in this article. The above
interactions may result in a change in amplitude, phase, direction of
propagation, polarization, and velocity of the light wave. All such
changes can be quantitatively predicted [page 284, reference 2].
Ellipsometry utilizes the correlation between the change in
polarization of the E vector upon reflection, and the physical nature
of the two media that make up the interface, in order to determine the
interface characteristics. As it is impossible to detect changes in
polarization if the incident light wave is randomly polarized,
ellipsometry employs linearly polarized light. As the name implies,
linearly polarized light is one whose E vector oscillates along a line
at any point in space as a function of time. The tip of the E vector
at different points in space at any particular time instant, lies in
one single plane that also contains the direction of propagation of the
light (Figure 1). In order to understand the change in polarization
upon reflection, the plane of incidence has to be introduced. This is
the plane that is defined by the incident light wave propagation vector
and the normal to the reflecting interface. The component of the E
vector that lies in this plane is called the P component, and that
which is normal to the plane of incidence is called the S component.
Light that is linearly polarized in any direction can be resolved into
the P and S components, and this is always done due to the following
fact. When light polarized in either the S or P direction is incident
on an optical interface, it undergoes reflection without any change in
288

in polarization, although there could be changes in the amplitude

and/or phase of the wave. What is to be noted here, and which forms
the basis of ellipsometry, is that the change in phase of the wave
caused by reflection is invariably different for the Sand P
components. Therefore, if an incident monochromatic light wave has its
Sand P components in phase with each other (which is necessarily true
for linearly polarized light), upon reflection, the Sand P components
will be out of phase. Whenever this happens, the resulting state in
the most general case, is elliptically polarized. The name elliptical
polarization arises due to the fact that, at any point in space the
tip of the E vector traverses an ellipse as a function of time. The
orientation of the ellipse and its ellipticity is related to the
physical nature of the film from which the reflection took place. The
theory that governs this relation has been rigorously developed [page
284, reference 2].
The material of interest will be referred to as the substrate
and the medium surrounding it (electrolyte in a corrosion application)
will be referred to as the ambient medium. Theoretical relations that
are relevant to reflection from a film-free substrate will be discussed
first, following which a filmed substrate will be considered. There
are two basic relations that relate the reflected, refracted, and
incident E vectors. These are the Snell's law and the Fresnel complex
amplitude reflection co-efficients [page 284, reference 2].
Fresnel com(!lex am(!litude refection co-efficients:

Erp N1cos4>o - Nocos4>l (1)

r
E.
lp p N1cos4>o + NocoS4>}

Ers N1cos4>o - N}cos4>} (2)

= r
Eis s Nocos4>o + NIcos4>}

Snell's law:
Nosin4>o = N}sin4>l (3)

No and NI are the complex refractive index of the ambient medium and
the substrate respectively; each is written as
N = n - ik (4)
where,
n index of refraction
k = extinction coefficient or the absorptivity of the film.
 289

~o = angle of incidence (Fig. 2)

~I = angle of refraction (Fig. 2)
E = complex amplitude of the E vector
the subscripts s,p stand for Sand P components, while the subscripts
r,i stand for the reflected and incident wave, respectively.
The index of refraction of a medium is defined as the ratio of
the velocity of light in vacuum to the velocity of light in the medium.
The complex refractive index is also the square root of the dielectric
constant of the medium.
The complex amplitude of the E vector carries information about
the amplitude and the phase of the wave. For instance, the reflection
co-efficient for the P component, r p' can written as follows.

(5)

where
Irpl = ratio of the amplitude of the P component of the E vector in the
reflected to that in the incident wave.
0p = phase shift in the P component of the reflected wave relative to
the incident wave, caused by the reflection process.
The experimentally determined parameter in ellipsometry is the
ratio of the complex Fresnel amplitude reflection coefficients for the
P and S components, p.
Irpl ei(op - is)
p = ~ (6)
~
This is usually written as

p = tan ~ei~ (7)

where ~ and ~ are the instrument output data. The above equations can
be re-written in the form,
2
NI = NO sin¢o [ I +~ 2
(I+e) tan ¢o ]1/2 (8)

This equation can be used to determine the complex index of refraction

of the substrate (N I ) from that of the ambient medium (No) after
experimentally determining the value p at the angle of incidence ¢o
290

MEDIUM (0)
MEDIUM (I)

Figure 2: Oblique reflect ion and transmission of a plane wave at the

planar interf ace between two semi-infinite media 0 and 1. "P" and
are axes parall el and perpe ndicu lar to the plane of incide nce "s"
respectively [po 271, reference 2]

FLM (t)

SUBSTRATE (2)

Figure 3: Obliq~e reflect ion and transmission of a

amblen~ (O)-fl lm {l)-Su bstrat e (2) System withplane wave by an
parall el-pla ne
boundarles. [po 283. reference 2]
 291

The above equations were developed by Fresnel for a bare

substrate and later Drude used them to develop his relations for a film
covered substrate. The complex amplitude reflection coefficient for
reflection from a film covered substrate is given by
-i2,8
R = r OIp + r I2p e (9)
p

(10)

(11 )

r OI = complex Fresnel amplitude reflection co-efficient for reflection

from the interface between medium 0 and I (Fig. 3).
r 12 = same as above but at the interface between media 1 and 2.
The subscripts p and s have the same meaning as used earlier.
The r 01p ' r 01s ' r l2p ' r l2s ' values can be obtained from
relations 1 through 3 by using the incident and refracted angles in the
appropriate media. Rp and Rs can be written as

Rp = IRpl eif.p (12)

Rs = IRsl eif.s (l3)

The ratio of Rp to Rs is given by

p Rp IRpl ei (f. p - f. s ) (l4)

Rs llQ
This is the experimentally determined parameter in ellipsometry. It
can also be written as

p = tan~ eif. (IS)

292

where
tan.h'I' = IRpl and A
II = "rp - "rs
A A (16)

TRJ
The functional dependence of I:J. and w on the system parameters can be
symbolically written as

( 17)

where A = wavelength of light

In the most general case No' NI , N2, the index of refraction of
the ambient medium, film and substrate respectively, are all complex,
and hence p is a function of 9 variables. However, the ambient medium
is usually transparent and hence No is real. Also, in most studies
the substrate refractive index is evaluated before any film formation.
Moreover, the angle of incidence 80 and the wavelength of light being
used, A, are known parameters. So the system unknowns are the 2
parameters (n,k) in the film refractive index NI , and the film
thickness d1. The problem in most studies then boils down to
evaluating 3 unknowns from 2 experimentally determined parameters (I:J.,
!/J). Clearly this cannot be done with one observation of p. This 1S
one of the drawbacks of the technique and ways of circumventing it are
addressed in a later section.
During the time when Drude developed his rigorous equations, he
was unable to solve them in an explicit manner due to the absence of
powerful computational techniques. Therefore, Drude expanded the
exponential terms in a series form and used only the first term to
arrive at the following approximate equations [page 10, reference 3].

I d1
I:J. - I:J.' -A (1 - - ) - (18)
N2 A
1

2(1{1 - !/J'} 8(1 - C) (1 - ~ )-±d (19)

NI
where I:J.', 1{1' correspond to the film free values; A, 8 and Care
functions of the angle of incidence and index of refraction of the
substrate. The above equations are valid only when the film thickness
is very small compared to the wavelength of the light being used, i.e.
when d/ A «1.
The approximate form of the Drude equations is now being used
mainly to get a quick qualitative measure of film thickness and
refractive index rather than for quantitative studies.
 293

Although seldom applied to corrosion studies, the theoretical

interpretation of ellipsometric measurements to obtain sub-monolayer
surface coverage with reaction products will have to be mentioned here
for completeness. Kruger has reviewed the various theoretical models
[4a] that have been developed to take advantage of the excellent
sensitivity inherent to ellipsometry for the detection of sub-monolayer
films. One of these models, for example, relates the fractional
coverage of the substrate with the film to an effective refractive
index for the interface region, and subsequently employs the Drude
equations presented above.
3. EXPERIMENTAL PROCEDURE AND INSTRUMENTATION

A brief discussion of the experimental techniques used to

measure the ~ and ~ values will now be presented. It is appropriate
at this point to define ellipsometry. Rothen [page 7, reference 3]
defines it as the art of measuring and analyzing the elliptical
polarization of light. The basic experimental method that has been
historically employed to accomplish this is the null-detection method.
In the simplest sense, this involves placing a compensator and an
analyzer in the path of the elliptically polarized light. The
compensator introduces a known phase difference between the Sand P
components of the reflected wave. This phase difference is varied by
the investigator until the wave that emerges from the compensator is
linearly polarized. In other words, the compensator eliminates the
phase difference between the Sand P components, by imposing a phase
change that exactly compensates for such a change caused by the
reflecting surface. As the phase difference imposed by the compensator
is a known quantity, the ~ due to reflection can be determined. The
direction of polarization of the resulting linearly polarized light is
determined by an analyzer. In order to achieve this, the analyzer is
capable of continuous rotation, and at any particular angle transmits
only the component of light that is polarized at that angle.
Therefore, when the analyzer's transmission axis is normal to the
polarization direction, no light passes through; this is called the
null point (Figure 4). Once the polarization direction is determined,
the tangent of the angle that the polarization direction makes with the
plane of incidence is the tan ~ that appears in the equation for p.
This assumes that the incident light is linearly polarized at 45° to
the plane of incidence. Clearly, the procedure just outlined is time
consuming (typically requires a few minutes) and tedious. Moreover,
visual detection of the null point is subject to error. It is
therefore nearly impossible to conduct accurate dynamic measurements of
a continuously changing surface film. In order to obviate these
difficulties, automatic ellipsometers have been developed that
facilitate accurate tracking of dynamic surface phenomena in situ, as
they occur. An extensive account of all the automatic ellipsometers
and their principle of operation is provided in the text of Azzam and
Bashara [page 405, reference 2]. A brief description of one, the
294

rotating ana"lyzer ellipsometer, will be outlined here. An illustration

of its operation is presented in Fig. 5. In this instrument, an
angular encoder keeps track of the instantaneous angle (azimuth) of the
analyzer as it continously rotates at about 50 revolutions per second.
Corresponding to each of these angles a photodetector determines the
light intensity in that direction. By correlating information from the
angular encoder and the photodetector, the light intensity as a
function of angular position around the direction of light propagation
is determined. This completely defines the ellipse of the reflected
wave polarization, from which the ~ and ~ information can then be
extracted. This procedure for collecting a set of data points takes
about twenty milliseconds and the instrumentation is being improved to
achieve higher accuracy and speed.
4. ADVANTAGES OF ELLIPSOMETRY
Two important characteristics of ellipsometry make it an
extremely valuable technique for surface analytical studies. These are
its high sensitivity and its in situ applicability.
Ellipsometry can be employed to detect sub-monolayer film
coverage of the substrate surface due to its high sensitivity which
arises from two factors(S):
a. Measurements in ellipsometry involve the determination of
the ratio of the amplitude of the P-component of the E vector
to that of the S-component, and the relative phase difference between
the 2 components. Since these are relative measurements, any errors
introduced by fluctuation of the light source (or other sources of
amplitude modulation that have an identical effect on both sand p
components) cancel out.
2. The other reason for the sensitivity arises from the fact
that the measurements involve angular quantities. The rotating
analyzer ellipsometer, for example, measures the light intensity in
the reflected wave as a function of the angle from the plane of
incidence. This, when combined with the fact that angular measurements
can be made with high resolution, contributes to the sensitivity of the
technique.
The other valuable characteristic of the technique is its in
situ applicability. Due to the fact that visible light is not
absorbed by the ambient medium in a corrosion or oxidation
situation, all ellipsometric studies can be conducted in situ. This is
unlike the other powerful surface analytical techniques such as AES,
XPS, SIMS, etc. These techniques suffer from the drawback that the
sample must be placed in a high vacuum before the study can be
conducted. This is thought to alter the nature of passive films,
especially when they are hydrated [6].
Raman spectroscopy and Mossbauer spectroscopy are two other in
situ techniques that are currently being employed. These two
techniques however, have a much lower sensitivity than ellipsometry.
For instance in Mossbauer spectroscopy, the sampling depth is 3000
A (6); Raman spectroscopy (7) requires at least 10 monolayers of film
 295

Figure 4: Illust ration of the null-p oint, by having the analyz er

transmission direct ion normal to the polari zation direct ion of the
and D. Hallid ay, physic s part
incide nt light. [Refer ence: R. Resnick
II, John Wiley & Sons, Inc., p. 1151 (1968)J

Digit,,1
<ign,,1
(256)

.~ (
Da~
(256) I
pulses
~ b ...... ,.'
9 Sm,,11

WIIIL-'9 :I c()~:e~er I
9
Encoder digital
outputs computer
IBM 1110

L- __ ~
Slart pulse ( I)

Figure 5: Schem atic illust ration of the operat2] ion of an automatic

rotatin g-anal yzer ellipso meter. [po 413, referen ce
296

material for detection. Differential Reflectometry, Extended X-ray

Absorption Fine Structure (EXAFS), Internal Reflection Infra-red
Spectroscopy (IRIS) and Electron Spin Resonance (ESR) are other in situ
techniques that are currently being employed for surface analytical
studies.

5. LIMITATIONS OF ELLIPSOMETRY
The extensive application of ellipsometry to precisely
determine the thickness, and to identify the structural and physical
nature of surface films has been tempered by a few inherent problems
associated with the technique.
a. The theoretical equation of Fresnel &Drude that model the
reflection process is only applicable to a planar isotropic 3-
dimensional film formed on a substrate. There are two common
experimental conditions that violate this assumption. 1) Surface
roughness in the substrate or at the substrate-film interface. 2) The
situation where the surface film formed has a gradient in its
refractive index or is a film of two or more uniform layers, each
having a different refractive index. The variations in refractive
index along the film thickness can arise due to either compositional or
structural variation. The real part of the refractive index of a
surface film increases with the density of the film.
Several techniques have been employed by investigators to
circumvent such non-ideal conditions. The presence of surface
roughness can be modeled by replacing the roughened layer by an
optically equivalent, homogeneous hypothetical film with sharp
boundaries [4a]. The refractive index of such a film can be determined
by an effective medium approximation. This uses the refractive indices
of the substrate and the film and the volume fraction of the two phases
to arrive at an effective refractive index for an optically equivalent
film. The equations of reflection can then be applied and subsequent
analysis carried out in a manner identical to that for a smooth
interface. The problem of a gradient in the optical constant along the
film thickness can be modeled by subdividing the film into several
discrete homogeneous layers, each of which has a different (but
smoothly varying) optical constant [4a, 4b].
b. Another significant problem associated with ellipsometry is
the indeterminate set of equations in the case of an absorbing film
(non-zero 'k'). There are 3 unknown parameters for such a film (ie,
the real and complex part of the refractive index of the film and the
film thickness) and only 2 experimentally determined parameters (~ and
~), thus rendering a unique solution impossible. This problem has
invoked a rather extensive array of solutions. Three popular ones
are:
(i) Independent film thickness calulation by employing other
techniques like coulometry.
(ii) Assumption of a non-absorbing film (k=o)
 297

(iii) Obtaining ~ and ~ values as the film grows, for

continuously increasing thickness. This set of data is
compared to a model that assumes the uniform growth of a
homogeneous film [46].
Cahan [8] has discussed the limitations of the above techniques
and a few others. A discussion of the utilization of the intensity
information in the reflected light wave, to narrow down the possible
solutions for the film properties, is also presented. The widespread
usage of this 3-parameter ellipsometry is primarily limited at present
by the instability of the common light sources.
c. The uncertainty in the optical constants of different
materials poses another problem in the precise identification of
surface films. Variations in the refractive index of materials with
grain size, sample size, surface contamination, surface stress and
strain have been documented in literature [page 418, reference 2]. A
good example of the manifestation of such variations is the scatter in
the literature data for the refractive index of pure aluminum.
These problems make it imperative to use a supporting
experimental technique to corroborate the conclusions drawn from
ellipsometry. Coulometry is sometimes employed for this purpose [page
252, reference 4a]. Differences between coulometry and ellipsometry
results, however, could arise due to the following factors:
1. Surface roughness
2. Film hydration
3. Incorrect values for the cation oxidation state in the film and/or
density of the film.
4. Film dissolution or metal dissolution through the film.
Other techniques that are employed include x-ray specular
reflectivity [9], impedence spectroscopy [10], electron-optical
techniques [16] etc. It should be pointed out that most of the
problems mentioned are not serious when ellipsometry is employed only
to provide qualitative trends during film formation and growth. Hence
ellipsometry is a very handy tool to study film growth kinetics.

6. DATA INTERPRETATION
Investigators have employed ellipsometry to study corrosion
processes both as a qualitative and quantitative tool. Qualitative
studies identify trends in the experimental parameters with the
kinetics of film formation or dissolution. Such studies detect the
occurrence of a film conversion process or provide insights into the
growth mechanisms. For instance, the field assisted growth mechanism
of oxide films can be detected by a linear variation in ~ with
potential, since ~ is linearly proportional to film thickness at small
values of the film thickness. Quantitative studies, under ideal
conditions, can help identify the composition of the film and to
determine the thickness of the film with precision. This latter
procedure is employed extensively by the semiconductor industry to
298

obtain accurate information about the effect of process variables on

the thickness of oxide overlayers.
Qualitative techniques do not require an exact knowledge of the
optical constants of the substrate. However, such techniques cannot
make use of the excellent resolution provided by ellipsometry and the
possibility of film identification in-situ. Quantitative techniques
suffer from all the drawbacks associated with the uncertainty in the
values of the relevant optical constants. When calibrated with a
complementary technique, however, they provide valuable quantitative
information.
It is worthwhile to document the qualitative conclusions drawn
by various investigators from the trends in ~ and ~ values observed
during the oxidation or dissolution of surface films. It should be
pointed out that although these conclusions provide a good physical
feel for the experimental parameters, they are not universally valid,
and should be understood in the context of the particular experimental
conditions.
(a) A decrease in ~ at a constant ~ has been observed [11]
upon potential cycling of Cr in 85% H3P0 4 . This is attributed to a
film conversion process where a film with a high absorptivity (k) is
formed at a nearly constant thickness (~ is inversely related to
absorptivity).
(b) A simultaneous decrease in both ~ and ~ is attributed to
the development of surface roughness in the metal/film interface with
simultaneous growth of a transparent surface layer in an aluminum-NaCl
solution (O.IN) system [12].
(c) A reduction in ~ (while ~ decreases continuously) is
attributed to a change in the oxidation state of the film in an Ni-O.lM
KOH system [13].
(d) If the ~ and intensity variations are very small during
passivation (with a decrease in ~), it is likely that the film formed
is transparent [14].
(e) The ~ vs ~ curve during film growth will lie in a closed
loop, if: 1) the film is transparent (k=O), and 2) the film is
homogeneous with a constant refractive index [2].
The larger the size of the loop formed, the higher is the
corresponding refractive index of the film and the greater its density.
(f) The higher the conductivity of a material, the higher is
the concentration and mobility of charge carriers that can absorb the
energy contained in a light wave. Therefore, such materials will cause
a rapid damping of light and hence will have a high 'k' value [15].
(g) Large microcrystals and a rough substrate exhibit an
apparent increase in 'k' for aluminum with respect to a very fine
structure [16]. This could be due to the fact that the free electrons
in the grain boundary are less mobile than in the grains and hence
grain boundaries cause a reduction in conductivity. Since larger
microcrystals have lesser grain boundary area per unit surface
area, they might have a higher 'k' compared to smaller crystals.
 299

(h) At small values of film thicknesses, a linear ~ vs ~ curve

(decreasing ~ and increasing ~) obtained during film growth could mean
that an oxide film with a constant refractive index is growing [13].
Review papers on ellipsometic studies as relevant to corrosion
have been written by Kruger [3,4a], Paik [17], and Greef [18]. A list
of references on ellipsometry, as applied to corrosion and
electrochemistry has been compiled by Hayfield [19]. A representative,
but not exhaustive, example of recent studies will be discussed here.

7. MECHANISM OF FORMATION AND STRUCTURE OF ANODIC OXIDE FILMS.

7.1. KINETICS OF PASSIVE FILM FORMATION
Ellipsometry is unique in allowing the kinetics of passive
films to be studied in situ, as they rapidly grow in contact with an
electrolyte. Modern automatic ellipsometers enable measurements to be
made in less than O.OSsec. after the anodic current is turned on.
The majority of studies on the kinetics of anodic film
formation has been conducted on iron in borate buffer solutions. The
results show that the anodic oxide films grow in several stages
[20,21], and that the final film thickness depends upon many factors
such as solution composition, metal composition, potential, etc. The
film formed is often composed of two layers differing in chemical
composition and/or structure.
In a study by Zakroczymski et al. [22], simultaneous
ellipsometric and electrochemical measurements were made on the
kinetics of film formation on iron in 0.05 MNaOH solution; two stages
of film growth were identified. During the first stage, which lasted
for 2 sec, the anodic current was consumed only for film growth,
whereas in the second stage (times longer than 2 sec) release of iron
cations to the electrolyte occurred as well. This conclusion was based
on measurements of the electric charge consumed for film growth in
comparison with the ellipsometric results. Fig. 6 shows changes in the
total charge density (Qtotal) as a function of time during film growth
at different applied potentials. Qfilm is the charge density consumed
only for film growth. The film thickness calculated from the charge
consumed during the first passivation stage was less than that
determined ellipsometrically (cf. Figs. 7 and 8). This difference is
thought to be caused by a simultaneous increase in both the film
thickness and the density as passivation progressed. It was presumed
that the film density changes from an initial value, characterizing a
liquid phase, to a final value, characterizing the solid phase oxide
film. The increase in film density is probably related to the fact
that the quantity of water occluded in the oxide film decreases with
time. Changes in the calculated average film density with time are
shown in Fig. 9. The chemical composition of the passive film formed
in 0.05 MNaOH during the second passivation stage corresponds to
300

.E 150

.....
(..)

>-
! 100
~
~
.2
(..)
50

01 10 10() 1000
lime,s

Figure 6: Changes in current density with time [Reference 22]

Eo o.05M NoOH, 298K

£::-o.950V__ «111'/, -E:KHC
E:;
llt FeOOH
.. , ................................ .
EO, 0,~5O
c rl'lm
L!"

z
<> ...... ._.--2.~~
~4
....
i
.>L
:/' .,.,..."."..--------
...
"",-'- -~

...§
~ 3 " ,,/
:/ /' -0,3
:1/
il: 2 .1/
1/
:./
....~
." ""

'6,01 1.0 10 100 1000

Time,s
Figure 7: Film thickness calculated from the electric charge vs time
at different potentials [Reference 22]
 301

6 O.05MNoOH,298K
.. '
EO ·0.550

. . .' .' .
~ 5 .... .. ' .. ' .'

.:. ........ ••••.••• ......-----

----
0.250
.",,-...-
'0. 4
aUerootive ... -:::;,... __. - - . - - . -O.O~
j \ .
~., 0°. '/0
--' ..... ___ ---

.... .///
~ 3 .0° ./ ".,.,.. ....... - - -

..... ,
S , : /
ii: 2
; .... //
'j//
Z
~DI 0.1 1.0 10 100 1000
Time,s

Figure 8: Film thickness evaluated from ellipsometric data vs time at

different potentials [Reference 22]

5,-------,-------~------,_------,_------~

~4.280/cmJ _
--- ---- --- - -- .......-;-;;-;-.;-; ... ~:"';':':t~-
"e
~
4
.. '
.... '\'\'"
\ 0.250
00 : 0.550

,:.: a05MNoOH,298K
£.-o.950V_ «IV/I. £:"'E
...... :~:...
______________________ !~C~~~~~_

O.~----~~------~------~------~----~~
001 0.1 1.0 10 100 1000
Time,s
Figure 9: The average film density as a function of time [Reference
22]
302

FeOOH. The film density is constant and the complex index of

refraction is N=I.84 - 0.138i, at A=5461 ft.
7.2 IRON
There are many papers [23-30] that report studies on the
passivity of iron by ellipsometry, often together with other techniques
such as coulometry, chemical analysis of the solution, A-C impedence,
AES, etc. These studies often showed contradictory results. There
seems to be no agreement regarding the exact nature of the passive film
on iron, even that formed in borate solutions where the majority of

.5 31.5
8
~oo............. ", )

e
0
tP~ 31.4

0
0- .......
!!I
~1 31.39
i
-1.50
ISO
TIllE SEC
300
0
45OI2a.8
~ 24.8 25.8
PIUIIIZER !Eli
31.2

aV·1

Figure 10: Galvanostatic oxidation-reduction cycles for iron in

deaerated borate solution with (1) and without (2,3) electrolyte change
after passivation. A potentiostatic interval was maintained after
passivation in expts. 1 and 2 but not in 3 [31].

passivation experiments have been carried out. The inconsistency may

be due to the complexity of the processes involved, and limitations of
the techniques used. As was emphasized by Ord [31], the choice of
experimental conditions can also determine the nature and thickness of
the film formed on the metal surface.
Ord [31] has given an example of ellipsometric measurements of
passive film formation on iron in borate solution, which was followed
by film reduction either with or without changing the electrolyte in
which passivation occured. If the electrolyte were not changed,
ferrous ions formed during passivation would be present prior to
raising the potential above that of FeOOH deposition. Fig 10 compares
 303

the electrochemical and ellipsometric results of such experiments. An

iron electrode was passivated at 1mA/cm2 to a potential of -0.825 V
(relative to Hg/HgS0 4 electrode) and held at this potential until the
current density fell below 2 ~A/cm2 (point A). The electrolyte was
changed in experiment 1 but not in experiments 2 and 3. Subsequently,
an oxidation current of 10 ~A/cm2 was applied, and then the current
reversed when the potential reached 0.3 V (point B). In experiment 3,
the passivation was performed at -0.825 V until the current density
reached 20 ~A/cm2. In each of these galvanostatic oxidation-reduction
cycles, different

Table 1: Optical Constants, "2' and Increase Rate of Thickness With

Potential, k, for Passive Films Formed on Steel in 1.0 kmol.m- 3
Na 2S0 4 ' pH 6.0 [32].

Passive Film Transpassive Film

Specimen
Composition N2 k Inm V-I
10Ni Steel 2.8-0.4i 2.2
6Cr-10Ni Steel 2.8-0.4i 2.3
10Cr-IONi Steel 2.8-0.3i 1.9
14Cr-IONi Steel 2.4-0 4i 1.7 2.9-0.4i 1.4
18Cr-IONi Steel 2.5-0.4i 1.6 2.9-0.4i 1.5
27Cr-IONi Steel 2.4-0.4i 1.5 2.9-0.4i 1.8

ellipsometric results were obtained as shown on the right side of Fig.

10. were obtained; this indicates differences in the thickness of the
films.
7.3. ALLOYING ELEMENTS IN STEEL
In several studies, ellipsometry has been used to determine the
effect of alloying elements on the passivation of steel.
Sugimoto and Matsuda [32] have measured the optical constants
and the thickness of passive films formed on steel containing 10% Ni
and 6 to 27% Cr in sulfate solution of pH 6.0 and 0.2. Film thickness
obtained at pH 6 was always greater than at pH 0.2. Film thickness
304

decreased with increasing Cr content. As shown in Table 1, the real

part of the refractive index of the passive film is higher at lower Cr
concentrations which suggests a higher iron oxide content in the film.
Comparison of the results on the effect of Cr content on the thickness
of films formed on the surface of austenitic [32] and ferritic [33]
steels (Fig. 11) has lead to the conclusion that under similar
experimental conditions, thinner films are formed on austenite than on
ferrite. The difference in the NiO content of the film might be the
reason for thi s.
Results reported by Totir [34] are in agreement with those
presented above, ie increased Ni/Fe ratios cause the passive film
thickness to decrease. For example, in ellipsometric measurements made
in boric acid-borax solution of pH B.5, the following film thicknesses
were obtained:
pure iron: 4.3 nm
85% Fe 15% Ni: 4.4 nm
70% Fe 30% Ni: 3.3 nm
pure Ni: 1.3 nm

4D~~~~~-'~-'---'---'''
(a)l.okmo~· -Naa50. (pH 1-0), O.SV
I I I
3.5"---01::--t--- --<>-- IC, .141411

- - 4 - IC,-IONi It4lel

Ur--t-

2.0 0l-~5L~~.-.)~~20;::~25;=!~30l-1
Cr conlentl _I'f.

2.5~-:-1~--,---,----r---r---,-,
(b) 0.5kmo1·m'-l!zso. I pH 0.2 ), O.SY

1.0

Figure 11: Relation between passive film thickness, d and Cr content

relations for xCr steels and xCr-l0Ni steels at O.SV in 1.0 Kmol.m- 3
Na 2S04, pH 6.0 (a) and 0.5 Kmol.m -3 H2S0 4, pH 0.2 (b) [References 32
and 33].
 305

In spite of the very important role of minor elements

(especially carbon) and impurities (sulfur, phosphorus) in processes
such as pitting, intergranular corrosion, stress corrosion cracking of
various steels, etc., relatively little attention has been paid to
study their effect on passivity. Results of two studies are presented
here.
Kozlowski and Flis [35] have made ellipsometric and
electrochemical measurements on iron and its alloys, with 0.14 to 1.87%
phosphorus, in borate solution of pH 8.5. Some of their results are
shown in Figs. 12 and 13. They found that with increasing P content,
both the critical current density for passivation and the passive
current are increased; this suggests that the density of oxide films
formed on Fe-P alloys is less than that on pure Fe. The formation of
less protective films in the presence of phosphorus is explained by
the incorporation of phosphorous oxidation products into the film.

005M H:#'l3+CIH2M tbM·pHl5·1mY..-1

1-Fe
2-Fe-Q14P
3-Fe-O.33P
4-Fe-Q51P
5-Fe-tlOP
~-t82P

1D

Figure 12: Anodic polarization curves for Fe and Fe-P alloys measured
at a scan rate of 1 mv/sec [35].

In another study on Fe-C alloys (0-0.7% C) by Flis et al. [36],

it was found that the properties of passive films formed in 0.05 N KOH
solution are affected by both the content and distribution of carbon in
the alloy. Low and medium carbon content (up to about 0.4% C) have
been shown to increase the rate of anodic dissolution and decrease the
rate of passivation, whereas higher carbon contents resulted in the
opposite effect. This behavior is illustrated in Fig. 14, which shows
polarization curves measured for furnace cooled Fe-C alloys. Fig. 15
presents changes in the optical constants nand k of anodic films as a
function of potential for alloys of different structures. The higher
306

refractive indices obtained for high carbon alloys indicate that

thinner and more protective films are formed on their surfaces,
compared to low-carbon alloys. Fig. 16 gives the film thickness as a
function of potential for alloys with different carbon content. The
beneficial effect of high carbon content on the protective properties
of passive films may be attributed to the initially very fast metal
dissolution accompanied by the formation of a residue of carbon. This
appears to promote a rapid growth of a thin, adherent and highly
protective film of magnetite on the metal surface.
7.4. NICKEL:
Visscher and Barendrecht [13] studied the oxidation of nickel
in 0.1 MKOH for potential use in alkaline batteries. The substrate
refractive index was obtained for a cathodically polarized Ni surface
(N = 1.887 - 3.45 i at 5461 A). The optical data obtained on an anodic
sweep at 0.5 mv/sec to I V (RHE) was identified with the growth of a

QOSM HJB:>:r0012M NoAD?pHa5

08V(SCE)
6 /
" , ' ..... ..

E
c

~
~3
u
x
t-
;[
...J
;;:2

10 100 1000
TIME,s

Figure 13: Film thickness vs polarization time at 0.8V (SCE). Stages

of the film growth are denoted by I - IV [35] ..
 307

a non-absorbing film with n = 2.45. This value did not vary with
potential and was attributed to a hydrated nickel oxide film (NiO
xH 20). The simultaneous decrease in both ~ and ~ at E>IV(RHE) was
attributed to a further oxidation of Ni 2+ to a trivalent state. The
difference between film thickness predicted by coulometry and
ellipsometry was attributed to water entrained in the film.
The determination of n, k and thickness of the film can be
solved numerically by simultaneous measurements of the reflectance (R)
and e1lipsometric parameters ~ and~. Measurements of this kind were
performed on nickel in borate solution (pH 8) by Ohtsuka et al [37] and
in KOH solution by Paik and Smialowska [38]. In both cases, it was
found that the optical parameters increased with increasing anodic
potential, which indicates that the film composition varies with
potential. This is contrary to the conclusions by Visscher et al.,
[13], who state that the refractive index does not vary with potential.
Ohtsuka et al [37] explained this fact by the formation of a
nonstoichiometric film, whereas Paik and Smialowska [38] claim

O.OSM KOH
o.'2mVos-1 C
0.' 0/0

--..- ..- 0.36 fwrit. - PIIXlile

------' 0.118 ftrrit.- ptarlltt
-- 0.03 f .....itt
N Q3 _. __.- 0.70 Widmansttltttn
'~
<
0.2

0.1

o 0.4
POTENTIAL • VH

Figure 14: Polarization curves for furnace cooled Fe-C alloys [36].

that the film composition is close to Ni(OH)2 which is partially

dehydrated to NiO. The increase in n with potential indicates that the
film density increases, while the increase in k implies that the
electronic conductivity increases with increasing potential.
308

7.5. COPPER
Ord, De Smet and Huang [39] have investigated the oxidation and
reduction of a Cu single crystal in deaerated 0.1 MNa 2C03 . This was
part of a study aimed at understanding the mechanism of electrochemical
oxidation and reduction in solid state films. Dynamic ellipsometry and
coulometry were employed simultaneously with galvanosatic cycling.
Figure 17 illustrates the potential-time and the optical signal
behavior upon cycling. A cathodic current of 18 ~A/cm2 was applied
from points A through C in the figure, followed by an anodic current of
the same magnitude from C through F. At F, another current reversal
was applied, to bring the potential to M. The polarizer angle used in
this study decreases by an amount proportional to the film thickness
and the analyzer angle changes in a way that depends on the refractive
index of the film. Regions DE, EF, GH and KL (more clearly seen in
Fig. 18) have been identified as film conversion processes, involving a
change in the oxidation state of at least a part of the film. Based on
galvanostatic cycling terminated at various anodic potentials (Fig.
19), the composition of the oxide at the anodic end point and the
sequence of reduction processes were identified. It can be seen from

O.05""'1C(JH

,#*.-._..,,-,_.-._'
2.0 • ....._-.-. . .

1. ~.- .......'
c
1.1 .M.

1. 0

Q.3

Q2

Ql

-0.8 -0.4 0 Q4
POTENTIAL 0 VH

Figure 15: Changes in the refractive index, nand K, of anodic films

as a function of potential for Fe-C alloys of different structures
[Reference 36].
 309

the figure that, when the anodic part of the cycle is terminated at
relatively low potentials, only one cathodic wave is required to bring
the optical signal back to the original values before oxidation. The
film formed at the anodic end point in this cycle is therefore
completely reduced and is believed to be a homogeneous Cu 20 layer. On
cycling to more positive potentials, two anodic and cathodic waves are
observed. The second anodic wave is attributed to further oxidation of
the outer Cu 20 layer to CuO. The first cathodic wave is due to the
reduction of the outer layer to Cu 20 while the inner layer remains
unchanged. Reduction of the inner Cu 20 layer is believed to occur in
the second cathodic wave, while the outer layer remains unchanged.
This is supported by :
1. The charge passed during the first cathodic wave is identical to
the difference in the total charge passed between the anodic and
cathodic processes. This difference arises due to the fact that the
outer Cu 20 layer remains unreacted at the end of the cycle.

0.05 M KOH
lh
6

til
til
III

.
Z
'"
~5
%
~
;:;:

-0.8 -0.4 0 0.4 0.8

POTENTIAL • VII

Figure 16: Film thickness as a function of potential for Fe-C alloys

of different structures [36].

2. At the end of the first cathodic wave, the corresponding optical

signal shows that the reaction products remained on the electrode (Fig.
19). The reduction of the inner layer, while the outer layer remains
intact, is claimeed to be possible due to the fact that the porosity in
the outer layer keeps the inner layer in contact with the electrolyte
throughout the reduction process. The final stage of the reduction
process is the deposition of copper on the outer surface of the film.
Ord, et al. [39], have also grown the oxide film by repeated cycling
310

POLARIZER degrees
15.5 17 18.5
·~------~~------~~--------r-----~--'44.6
</J
Q)
Q)
H
0.
Q)

'"
1.2 44.1 a:
UJ
N
>-
-'
<t:
Z
~ <t:
'5 .2 43.6
>

,\(
..J
~
Z
.JJ
f- -
a
"-

-1 .8 '
0 150 300 450 600
TIME seconds

Figure 17: A galvanostatic cycle at 18 ~A/cm2 in unstirred pH

12 carbonate electrolyte. [Reference 39]

POLARIZER degrees
13.1 14.1

44.1 ffi
N
>-
-'
<
z
-<

.
'6>
.3 43.6

-'
-< -.5
;:::
zW
I-
0
CL

120 240 360

TIME seconds

Figure 18: The seventh galvanostatic cycle from the sequence in figure
20. Cu in deaerated 0.1 MNa 2Co 3 [Reference 39]
 311

or:;;;--:-,~--~~--",,:::r---2;9OIf8.4

i
.,.
3
o

900

rigure 19: A sequence of galvanostatic cycles at 18 pA/cm 2 with

progressively higher anodic end points [Reference 39J

65

..
:
60

166
1!1
>60
;i
~
CT Cl

f1 f'

40
5 15 25 30
POLARIZER degrees

Figure 20: Optical end-point data from a sequence of cycles with end
points at f & K in figure 17. The numbers by the letters indicate the
cycle number at which the data was obtained. The curves shown are
calculated growth curves for films with the indicated refract;~e
indices (Reference 39J
312

(with the cathodic end pOint at K in Fig. 17), from the 4.6 nm thick
film produced during the first cycle, to a 53 nm thick film in the 43rd
cycle. The optical data obtained during such film growth is identified
with a film of refractive index N = 2.2 - 0.17 i at the anodic end
point, and N = 2.1 - 0.09 i at the end of the second stage of the
reduction process (Fig. 20). It has also been reported that the GH and
EF segments of the optical data become shortened upon stirring by argon
flow. It was concluded that the film conversion process involved in
these stages proceeded by a dissolution precipitation mechanism.
7.6. ALUMINUM
Allen's study [40] on aluminum and vapor deposited thin films
has resulted in nand k values for pure aluminum. It has also helped
gain a quantitative insight into the reactivity of aluminum with oxygen
at 10- 9 torr vacuum. It is clear that any study which aims to directly
quantify the optical constants of pure aluminum has to be conducted at
high vacuum « 10- 7 torr). The author employed a combination of Auger
electron spectrocopy (AES), argon ion sputtering and ellipsometry to
conduct this study. The samples used were mechanically polished,
cleaned in organic solvents, placed in the multi-purpose chamber and
evacuated to 10- 9 torr. Subsequent AES analysis indicated an air
o
formed (between polishing and evacuation) oxide film of 10 A thickness.
This oxide film was removed by argon ion sputtering. The sputtering was
continued until the oxygen peak on the AES spectrum disappeared.
Ellipsometric investigations were then conducted (at 10- 9 torr) and
evaluation of the 6 and ~ values indicated nand k to be 1.566 and
o
7.938 respectively at A=6328 A. Another interesting conclusion from
this study is the rate of contamination of pure aluminum by oxygen at
10- 9 torr. The authors reported the re-occurance of the oxygen peak in
the AES spectrum after only 15 minutes following completion of the in
situ cleaning process. The oxide layer formed at this pressure
o
saturates at 2.5 A thickness. Allen et al. [40] attributed the scatter
in the optical constants of aluminum found in the literature to varying
levels of oxygen contamination on the surface (as most of the studies
were done at comparatively higher pressures of 10- 4 to 10- 6 torr). The
effect of such contamination is to reduce the apparent values of the
optical constants as evidenced by the author's evaluation of nand k
for contaminated samples (Table 2). From the magnitude of the change
reported by the author, however, it seems unlikely that such
contamination can account for the relatively large scatter documented
in literature.
Greef et al. [41] have taken an air formed oxide coated
aluminum as the reference state and studied the subsequent anodic film
growth and dissolution kinetics. The 6 and ~ signal from such
 313

oxidized surface was considered to be the bare surface signal for model
development. The authors claim that such a procedure does not in any
way affect the qualitative conclusions drawn. Although this study does
not provide any quantitative data, it does provide valuable insights
into the growth and dissolution kinetics of the oxide film on aluminum
under the experimental conditions employed. Aluminum rods (99.999%)
were mechanically polished to 0.05 ~m and immersed in NaOH (0.1, 1.0,
4.0 Mconcentration). Ellipsometric studies were conducted at
different anodic potentials in the passive range and for different
electrolyte concentrations. It was found that the refractive index of
the oxide film increased with increasing anodic potential and
decreasing electrolyte concentration (Figs. 21 and 22). As the
refractive index of the oxide film is directly proportional to its
compactness, the above observations yield qualitative information on
the variation of compactness with polarization potential and
electrolyte concentration. The authors then reported the results of an
experiment conducted at a specific anodic potential in O.lM NaOH and
fitted a model to the data obtained. The model used a composite oxide
film with an inner compact layer (higher refractive index) and an outer
porous layer (lower refractive index).

Table 2 Effect of Oxide Surface Contamination on calculated Optical

Constants (37).

Measured film Measured Measured

thickness (A) C (0) ~ (0) n k

o 41.84 145.31 1.566 7.938

3.5 41.83 144.72 1.522 7.802
6.2 41.83 144.26 1.486 7.698

This is illustrated in Fig. 23. Greef et al. [41] also studied the
dissolution characteristics of the oxide film by polarizing the
specimens either at an anodic potential in the active dissolution range
or at open circuit. This was done after a passive film had formed.
Two distinct dissolution kinetics were observed depending on the
electrode potential during dissolution.
a) On exposure at open circuit potential, the ~ and ~ values returned
to the reference state values (obtained before passivation). The path
followed by the ~ vs ~ curve during dissolution was, however, different
from that followed during film growth (Fig. 24). The authors fitted
this data with a model that requires simultaneous dissolution of both
the inner and outer layers of the oxide film. A graphic illustration
of the growth and dissolution characteristics predicted by the model is
presented in Fig. 25.
314

210r----------------------------------,

130 Bar• •
I
Substrat. •
/

90
39 41 43 45 47
PSl/dtg

Figure 21: Family of anodization curves at the indicated growth

potentials. Al exposed to 0.1 MNaOH. Each run was carried out with
a freshly polished electrode [Reference 41]

205~--------------------------------~

Figure 22: Family of growth curves at 2.5 V, for NaOH Concentrations

of O.lM (A), 1.OM (B), and 4.0M (C) [Reference 41]
 315

220-r------------~--1--::~~::--...,

185 ~ .,.
........1 indR ,,: ,:
thic:lcr-.
0'9' Y
0°'
o ,
,,
"5
,
"
!
rim
Growth

80+----r---r----,-----r---~
36 38 40 42
PSl/d~

Figure 23: Comparison of the duplex model for film growth (dashed
line) superimposed, upon the experimental curve (open circle) and the
uniform film model curve (bold line). The inset shows the schematic
of the duplex model [Reference 41]

210r------------------------.

130

90 4S
37 39 41 43
PSl/d~
Figure 24: Growth curves for 2.SV in O.lM NaOH interrupted at
different film thicknesses (G) and their resultant dissolution curves
at open circuit (0) [Reference 41]
316

1.51
",8
'"
REFJW:TIVE ",c
INDEX
'"
1.34

",D

().o

'An
Figure 25: A three-dimensional plot indicating how the refractive
index of the film varies from A (the initial anodic pulse) to 0 (the
dissolution limit) with both time and film thickness [Reference 41]
17Sr---------------------------------~

115

B5
34 38 40 42
PSI'",

Figure 26: A family of curves for growth at 2.SV (G) interrupted at

different stages to show the dissolution signal at -O.5V (O)
[Reference 41]
 317

2) When dissolution kinetics was studied by polarizing the passive

specimen to potentials in the active dissolution range, the 6 and ~
values traced a loop before returning to the reference point (Fig. 26).
The loop size was proportional to the steady state dissolution rate at
that potential. At high anodic potentials, the loop size was small due
to reduced steady state dissolution caused by film formation. At open
circuit and at cathodic potentials, the loop size was either negligible
or absent due to the low steady state dissolution rates (as predicted
by Butler-Volmer kinetics). At intermediate potentials, in the active
range, the loop size was the largest. The authors interpreted this
loop formation as due to parallel dissolution of the metal substrate
along with the film and the consequent surface roughening. They
claimed that following such surface roughening, there was again a
preferential dissolution of the substrate that smoothens out the
surface and returns the optical values to their reference state. The
substrate surface roughening was associated with a drop in the k value.
The variation in k is illustrated graphically in Fig. 27.
Dell'oca et al. [42] conducted a similar study on pure aluminum,
and traced the variations in 6 and ~ with anodization at constant
current density in O.6M H3P04 electrolyte. They have quantitatively
determined the film thickness at which the porous film formation
commenced. This thickness was determined from the point at which the
6, ~ curve deviated from a reference curve which was obtained in an
c
\
\

A
\
\

.~
&5I
~o

Figure 27: Three-dimensional plot showing how the extinction

coefficient varies with time and the rough layer thickness during the
process of growth (A to B) and dissolution (B to C) of the rough layer
[41].
318

electrolyte (3% tartaric acid) that is known to impart a compact oxide

film on aluminum. The authors were able to model the trends in the ~
~ values only by requiring both the porous layer and the compact layer
to grow simultaneously. While Greef et al. [41] propose that the
compact layer stops growing after the formation of the porous layer,
Dell'oca et al. [42] claim that after the formation of the porous film,
the compact layer contributes upto 20% of the total film growth rate.
Dell'oca also obtained the voltage at which pore initiation occurs
(from voltage vs. thickness curves and from the thickness at which pore
initiation occurs). It was noted that for all current densities
investigated, this voltage corresponds to about 80% of the peak voltage
obtained at that particular current density .

.~~----~.~----~.~----~.~~
!'SI/",

Figure 28: Ellipsometry results on tantalum anodized in dilute

phosphoric acid. As oxidation proceeds, the optical parameters change
as indicated from the bare surface signal (+). The second cycle is
indicated by dotted circles. The next cycle (crosses) lies essentially
on the first cycle [43].
 319

7.7. TANTALUM:
Dell'oca and Young [43] investigated the nature of the surface
film formed on anodization of tantalum in dilute phosphoric acid.
Their experimental ~ and ~ data, obtained ex-situ, formed a closed loop
after two cycles. They attributed this to the growth of a film with
two layers. An inner transparent (K = 0) layer with n = 2.21, at A =
5461 A, and an outer transparent layer with n = 2.14, growing at equal
rates, were used to model the experimental data (Fig. 28). The
authors, however, have not ruled out the possibility of modeling the
same data by a homogeneous surface film with a continuous variation in
'n' across the thickness. The substrate refractive index used in this
study (3.3 - 2.24 i) is an apparent value, corresponding to the coverge
o
of a 50 A thick film. Dell'oca et al. also pointed out the possibility
of determining the cation and anion transport numbers based on the
ellipsometrically determined relative growth rates of the two layers.
7.8. MAGNESIUM-ALUMINUM ALLOY (Magnox; 0.8 wt.% Al. 0.005% Be rest Mgl
In an attempt to understand the excellent corrosion resistance
of Magnox Al 80 alloy in alkaline solutions, Greef, Harvey, and
Tyfield [44], have conducted an ellipsometric study of this system.
This alloy is used for uranium fuel cladding. The spent fuel elements
are kept immersed in NaOH solution (pH> 11), until the short-lived
radioactive products decay. Greef, et al. followed

Figure 29: Cyclic voltamrnograms of zinc in 0.20M LiC10 4 . Four

consecutive scans from -I.4V to more positive potentials are shown.
The electrode was allowed to stabilize before each run [14].
320

the formation of a film in deaerated 0.01 MNaOH and fitted the optical
data obtained with a model consisting of an inner transparent barrier
o
layer of MgO (n = 1.70 at 5461 A) and an outer porous layer of
refractive index 1.40. They reported that the thickness of the barrier
layer cannot exceed 5 nm. This conclusion, however, seems to stem from
the fact that the outer layer refractive index is fixed at 1.4; an
arbitrary value. A higher thickness for the barrier layer is possible
with a more porous outer layer. In air saturated electrolyte, the
substrate was found to undergo roughening, and the optical data is
modeled employing Bruggeman's effective medium approximation for an
inverse cone configuration of the metal-film layer.

(0)

.
-III ,.........:........~.---
.
, .............-..,
',.
-a•• ,
" ..
-a.s .~

-12.5
"
. '.
-12.1 ' .
-a•• '.
' ..
-a.,
..... -"
"

-II
'.....-....................-.-.......
,

~1 ~
• -I.. -1.11 -I.. -us -1.11 -us -1.11
11/ (b)
'l.I

lI.r

II.'

II,'

II••

-I"
E/v
11.1
-J.. -us -... -us
«

-us -1.11
Figure 30: Ellipsometric parameters associated with the CV of figure
29. In (a) and (b) the values of ~ &~ are plotted vs potential [14].
 321

7.9. ZINC
Hamnet and Mortimer [14] studied the oxidation behavior of Zn by
simultaneous cyclic vo1tammetry (Fig. 29) and ellipsometry (Fig. 30) in
neutral lithium chlorate. The substrate refractive index (3.105-6.047i
o
at 6328 A) was calculated from the optical data obtained at a cathodic
potential of -1.4 V(SCE). Hamnet and Mortimer assumed the oxide film
to be transparent, based on the observation of very slight variation in
~ and the reflected light intensity with film growth. They model the
optical data at the anodic end point (in the passive
o
region), by a film with n = 3.19 and thickness = 3.51 A. The thickness
was shown to conform to coulometric oredictions and was attributed to a

lal 89 91

Ihl 82 86 90 <l @eiJ

Figure 31: Comparison between measured and calculated (-) values of
ellipsometric parameters during the anodic oxidation of Sn-Ni alloy in
O.lM KOH (circular points correspond to Sn-Ni and triangular points
correspond to Ni 3Sn 2 and Ni 3Sn 4). (a) E=0.2V (curve 1, film refractive
index = 1.38-0.38i; curve 2, film refractive index = 1.38-0.04i; (b)
E=0.7V (curve 1, film refractive index = 1.4-0.075i; curve 2, film
refractive index = 1.4-0.080i) [45].
322

monolayer of ZnO. The difference in the refractive index with bulk ZnO
(n ZnO = 2.0) was attributed to the difference between thin film and
bulk properties. On subsequent scans, the anodic limit was extended to
more positive potentials, where a significant dissolution current was
observed. However, even at these potentials it was reported that the
optical signal remain unchanged from the values observed inside the
passive range. Hamnet and Mortimer [14] concluded that dissolution
proceeded with the film intact, via a place exchange mechanism.
7.10. TIN-NICKEL ALLOY:
Pokulska et al. [45] have investigated the oxidation behavior of
Sn-Ni (50 atom%) alloy coatings in 0.1 H KOH. The substrate refractive
index obtained upon cathodic reduction at -0.9 V(SCE) was N - 2.66 -
2.77 i with A = 6328 A. The specimen was subsequently polarized to 2.0
V(SCE) and the resulting film growth followed ellipsometrically. The ~
and ~ data obtained (Fig. 31) were fitted to a model of a film with
refractive index varying from n = 1.38 - 0.038 i initially at 20 nm
film thickness, to one with refractive index N = 1.38 - .04 i at 60 nm
film thickness. The linear variation of film thickness with logarithm
34 -
TI /0.1 M Sulp/1Jric Acid

33 <~Open Circuit t 100 min

161· /
32 "'" -K'-..:..-cothodIC Reduction
16 ' , I • 10 pA cm-2
1"""'0',
31
0.
• .~ I~ 0
,
...." 30 10 0 ~

/~
~

29
"I' • 6 "'-0')-. ~I
-?'to 0
28 '0 (1 "'" -,,
'O~I

27--:;;----I--;-_-1-~~-L--.l..---L_...Lj
72 16 80 84 88
t./d~.

Figure 32: Dynamic ellipsometric data on the dissolution of the

anodized film formed on titanium, in a 0.1 mol/dm3 H2S0 4 solution.
The anodization was done in a pH 6.9 phosphate buffer [46].
Theoretical fit is given by the bold line.

of time was attributed to a place exchange mechanism of oxide growth.

 323

8. EFFECT OF HYDROGEN ON ANODIC OXIDE FILMS

Ohtsuka et al. [46] studied the influence of hydrogen
incorporation in Ti0 2 films using simultaneous coulometry and
ellipsometry. This was done in view of earlier reports that hydrogen
incorporation could make the film less protective. Two distinct
reduction phenomena were identified by Ohtsuka et al, one in an acidic
medium and the other in a neutral medium. The anodized film formed in
phosphate buffer was shown to undergo thinning with reductive
dissolution of Ti ions {Ti{IV} to Ti{II}} in 0.1 MH2S04. The study
o
identified the film refractive index as N = 2.1 - 0.03i (at 5461 A) by
fitting the dynamic ellipsometric data obtained during the dissolution
process (Fig. 32). The substrate refractive index obtained from the
literature was N = 2.7 -2.9i (at 5461 A). The other reductive
phenomenon was shown to occur in a phosphate buffer solution (pH 6.9)
upon cathodic polarization. Sharp, reversible changes in fl and 1/1 were
reported, which closely followed the changes in applied potential (Fig.
33). The magnitude of the change was proportional to the applied

Ti 1 Phosphate Soln.. pH 6.9

d·27.5nm

o:e 0 l-eothoclic _ _ Open _

'Ii
ILl
~ _2 Reduction Circuit ~
.. E·-O.95V ,- - -- 0.0 ~
b--4 I
I
I
/E-
>
-0.4~

~~:~'
-10

i 36 - , -,.,/,;;;::---- 74 t
:;: 0--0--0_ ....

34 _f---._J\o_~_ 70 ~
32 0 6 12 18 24 30 66
Time. t IloZs

Figure 33: Change in cathodic current and ellipsometric parameters

during potentiostatic reduction at E=-0.95V followed by open circuit
decay in phosphate buffer solution at pH 6.9. The original film was
anodically formed by potentiostatic oxidation for Ih at E=5.65V [46].
324

potential, and was attributed to a reversible proton reduction and

incorporation as atomic hydrogen in the anodic film. Ohtsuka et al.
used the thickness obtained prior to dissolution in the acidic medium,
for determining the ellipsometric data obtained from the neutral medium
(with the hypothesis that the film thickness remains unchanged
throughout the hydrogen incorporation process). They obtained nand k
values for the film, as a function of cathodic polarization potential.
They found that n decreases while k increases with cathodic
polarization. All data were taken after a 30 min. exposure (Fig. 34).
The decrease in n was attributed to the reduced polarizability of the
02- ions due to the -O-H bonding, and the increase in k to the
absorption level created in the band gap of the anodic oxide, by the
oxidation of H to H+.
De Smet and Ord [47] made extensive use of ellipsometry to
understand the electrochemical reduction (coloring) and subsequent

Ti I Phosphate Soln.. pH 6.9

2.2
I c ·0.5h 0" d '27.5 nm

c oe j ••d·I&.Onm

-
Ie
2.1
~
---n
k-
0.4 ...

t ,;
0 .....

.~ ~Y"

. ,""
2.0
/" f
.2 -
.~ c
i 1.9
~~~,o .2
~
i .~

no,.H··'T
0.0;
1.& w
(;
..
.....
0
0.3 •
.2 0.2
i
°•
!
0.1 Eh
l/o
O. _0_0'/.
0 -0.4 -0.& -1.2
PoIentlal. E IV VI. RHE

Figure 34: Change in composition and refractive index of the anodic

oxide film on titanium during potentiostatic reduction for O.Sh in
neutral phosphate buffer solution as a function of cathodic potential
[46].
 325

oxidation (bleaching) processes in the anodic oxide of molybdenum.

This study was a part of an investigation on the role of hydrogen in
solid-state electrochemical processes. Studies were conducted in
acetic acid with 0.02M sodium borate and 2% water. Two stages, each in
the coloring and bleaching, are identified in the study, based on
sharp transitions in the optical signals (Fig. 35) during galvanostatic
cycling. Each of the stage was identified with hydrogen incorporation,
evolution, or oxidation from the film. De Smet and Ord also obtained
the film refractive index at the end of each of these stage by
conducting the coloring-bleaching cycle on films of different thickness
(Fig. 36). Each of these films was formed by terminating the anodic
part of the galvanostatic cycling at different potentials. It is to be
noted that the substrate refractive index (4.1-3.86i at 6328 A) was
calculated from the optical signal obtained at open circuit
(corresponding to a 'nominally' clean surface). The study also used
coulometry to arrive at the film composition upon reduction. Fig. 37
is a plot of the charge passed during the different cathodic/anodic
stages, following the initial film formation, as a function of the film
thickness. The film thickness was obtained from a theoretical fit to
Fig. 36. The charge required to bleach (Qa) the film (to bring it back
to its state before coloring) for different values of the coloring
charge (Qc) has been determined. The results are shown in Fig. 38 as a
ratio of these charges to the charge passed for anodic film formation
(Qg)' The anodic (bleaching) charge can be seen to increase linearly
with the cathodic charge upto Qc/Qg = 0.50. De Smet et al. have
concluded from Figures 37 and 38 that, the completion of stage I of the
cathodic process occurs at the point where the slope changes in Fig.
38. At higher values of Qc/Qg , all cathodic charge is consumed in the
hydrogen evolution process, and does not contribute to hydrogen
incorporation in the film (coloring not produced in this stage).
However, it is stated that some of the evolved hydrogen is trapped in
the film. From the ratio of the charge consumed during the first stage
of the cathodic process to that consumed for film formation, at low
coloring current densities, the film was postulated to consist of an 0-
O-H structure in the colored state. At higher coloring current
densities, the film is believed to consist of mixed phases of O-O-H
and 0-H 2• It is usual practice to use ellipsometry to study the
variation in thickness of a homogeneous film with time, potential, or
other external variable. In this study, however, ellipsometry is also
used to study the mechanism of a film conversion process, wherein the
film refractive index changes. De Smet et al. have done theoretical
fitting to the optical data obtained dynamically in the transition
state of the film, during the conversion process. Based on this they
identified the first stage of the coloring process to proceed
heterogeneously, from the film-electrolyte (F/E) interface to the
metal-film (M/F) interface (Fig. 39). This was done by comparing the
326

POLARIZER degrees
-~5 ________~____~~'~5______~2~5______-1.~

40 15
I
...,
....>
~

l
<
30
<f' 25
c o)

;~: '~
~ (,I
o
c.. _ 5 ....:.----3~_':~,====6;O~'~===~~9:;:;O;:O----1~2CC
L_

TIME seconds
Figure 35: Upper part: Polarizer, ,analyzer nu1,l settings for t~e
growth, cathodic conversion, and anod1c reconVerS1?n of the anodlc
oxide, Lower part: Potential vs time, At A an anod1c current of 1.74
mA/cm 2 is applied to the nominally clean electrode, At B the current
is lowered to 174 pA/cm 2; at C it is switched to -174 pA/cm 2 and at E
it is switched back to 174 pA/cm 2 , At F the film continues its anodic
growth [Reference 47]
90

E2
(I)
Q)
~60
0>
Q)
",

a:
w
N
>
..J
D3
~30
« DI A

E5
E4

D5 D4

-°25 20 65 110 155

POLARIZER degrees
Figure 36: Initial points (CI-C5), inflection pOints (01-05), and end
points (EI-ES) for a sequence of cathodic coloring cycles applied at
progressively greater film thickness. The C, 0, and E points have the
same relative positions as in figure 35. The curves fitted to the data
are for growth of films with refractive index values 2.15-0 for the
anodic film, 2.07-0.491 for the film at the inflection point, and 2.09-
0.23i for the film at the cathodic end point [Reference 47]
 327

320r-----------------------------------~

240
N
E
....<J
(.)
E
~160
cz:
c
%
(.)

80

" ~ 1~ 1~
THICKNESS nm
Figure 37: The charge passed during anodic oxide film growth (open
circles), the total charge passed on the cathodic cycles, C-E
(squares), the charge to the inflection point, c-o (circles with
horizontal bars), and the charge passed to convert the film back to
the anodic oxide, E-F (circles with vertical bars) [Reference 47]
.75r------,-------r------,-------r-----~

.5
o
01 o
o
"-
til
o

1.25
Qc/Qg
Figure 38: The anodic charge required to convert the film back to the
anodic oxide, Qa vs the cathodic charge used to convert the film, Qc '
both expressed as fractions of the charge required to grow the anodic
oxide to its initial thickness, Qg [Reference 47]
328

experimental points with theoretical curves (Fig. 39). These curves

were calculated for three types of conversion processes:
1) Homogeneous conversion, where the entire film undergoes the
conversion process simultaneously.
2) Heterogeneous conversion, where conversion starts at the F/E
interface and proceeds inward toward the substrate.
3) Heterogeneous conversion, where conversion starts at the M/F
interface and proceeds outward.
Inward sweeping film conversion has been attributed to the high
mobility of hydrogen in the reduced film. Similarly, it was shown that
the bleaching process also proceeds from the F/E to the M/F interface.
This is consistent with the reported field-assisted nature of the
bleaching process; the H+ ion produced by oxidation, moves through the
outer film (from the inner colored film) via a field limited transport
mechanism.

9. INHIBITION:
Only a few studies of organic inhibitors have been reported in
the literature using ellipsometric techniques.
Mrowczynski and Smialowska [48] used ellipsometry to study the
effect of a series of fatty acids on the anodic behavior of iron in
aerated 0.05 Msodium sulfate solution, pH 6 to 12. In the presence of
straight-chain fatty acids containing more than six carbon atoms in the
molecule, corrosion is inhibited when the fatty acid concentration
exceeds a certain critical level, Ccr . At lower concentrations or in
the absence of fatty acids, the corrosion products on the iron surface
exhibit a low refractive index (n = 2.36); above Ccr ' protective oxide
films with n = 2.6 are formed. In the presence of inhibitors the film
is much thinner. Fig. 40 shows the change of film thickness with time
as well as the complex index of refraction obtained for iron in aerated
sodium sulfate solution in the presence of caprylic acid. It was
concluded that the inhibitive effect of fatty acids arises from (a)
blocking the adsorption of sulfate ions on iron, and (b) increasing the
stability and the protective capability of the oxide film during the
anodic process. Reduction of the corrosion rate in aerated sodium
sulfate solution is presumed to result from a synergistic interaction
between the fatty acid and oxygen dissolved in the electrolyte.
Spinelli, et al. [49] studied the influence of benzylamine,
benzotriazole and hexamethylamine on the growth of anodic oxides on
iron in borate buffer solution. The values of optical constants for
the anodic layers did not change appreciably in the presence of
inhibitors. This indicated that the density of the layer did not
change. They also found that after the period of polarization, the
film thickness decreases and the film grows more slowly in the presence
of inhibitors. Figures 41 and 42 show changes in film thickness with
time in borate solution, with and without the presence of inhibitor, at
 329

76~-----.------r------r------~-----'

ffi 40
N
>-
...J
...:
z 03
...:
22

4L-____ ~ ______ ~______ ______ ______...J

~ ~

~ ro M 100 1U 1~
POLARIZER degrees
Figure 39: Polarizer vs analyzer null settings for the initial phase
of the cathodic conversion for initial film thicknesses of 87.7 nm at
C3 and 130.1 nm at C4. C3, C4, D3, and D4 are the corresponding
points from figure 36. The continuous curves are the predictions of
the models fitted to the end point data for homogeneous conversion
(middle curves), heterogeneous conversion with inward sweeping
interface (lower curves) and outward sweeping interface (upper
curves). [Reference 41]
Time,t[min]
10 100
2000

1600
~
"0
~ 1200
III
c
x
'E 800

10 20 30 40 50 00
Time,t [min]

Figure 40: Effect of exposure time on the thickness of films grown on

iron in aerated sodium sulphate solution of pH 6 with and without
addition of caprylic acid [Reference 48]
330

-r-----------------------~
-<
><

30

2 t {Sl

Figure 41: Film thickness vs time for a potentiostatic passivation of

Armco iron at 100 mV (SeE) in borate buffer solution {Reference 49)

-..

--
.ct:
x
30

Figure 42: Film t~ickness vs time for a potentiostatic passivation of

Armco iron at 400 mV (SeE) in borate buffer solution [Reference 49]
 331

two anodic potentials. The film is the thinnest in the presence of

benzotriazole, which means that passivation occurs most easily. The
results obtained are explained by taking into consideration the so
called Hard and Soft Acid and Base (HSAB) principle [50). Hard acids
strongly bind hard bases, while soft acids strongly bind soft bases.
According to this rule, compounds containing 0 or N (hard bases) are
more strongly bound and adsorbed on iron oxide, which contains Fe 3+
(hard acid), than on the clean metal surface.
The passivation of iron in potasium hydrogen phthalate (PHP)
solution (pH 4.5), with and without octylamine was studied using
ellipsometry and electrochemical methods [51). From the anodic
polarization curves, it was found that the anodic current was lower in
the presence of octylamine. The peak corresponding to the
active/passive transition disappeared and the corrosion potential was
more noble. Hence, octylamine facilitates passivation by adsorption on
the metal surface.
The thickness and kinetics of film growth were calculated from
results of simultaneous electrochemical and ellipsometric measurements.
The film thickness calculated from coulometry did not agree with that
obtained from ellipsometry. In the first stage of film growth, and not
too high anodic potentials, the thickness from coulometry was lower
than that obtained from ellipsometry. This effect was explained by
adsorption of organic molecules on the bare metal surface causing the
active metal area to diminish. At this stage, the ellipsometric film
thickness with octylamine, was higher than that obtained without it.
The adsorbed layer of octylamine appears to cause the oxide film
thickness to increase. As the optical parameters of the passive film
were the same in solutions with and without octylamine, it was
suggested that the refractive index of octylamine was close to the
value characterizing the oxide film on iron. It was also found that
after a longer polarization time(after 0.5 s), ie in the second stage
of passivation, film formation and dissolution of the metal or passive
film occurs simultaneously.

10. PITTING CORROSION

Foskin and Kotenev [12) used ellipsometry to study changes in
the oxide film on pure aluminum before and during pitting. This study
is a classic example of a purely qualitative analysis of optical data.
Potentiodynamic cycling at 4 mv/min was done in 0.1 N solution of NaCl.
The potential excursion was carried out into the pitting region and the
associated changes in the surface film were deduced from the trends in
the optical data. Fokin et al. have concluded that film growth occured
initially as the potential was swept anodically until pit initiation.
Subsequently, film thinning was reported to occur with the progress of
pitting. This was attributed to the reduction of excess oxygen,
trapped in the hydroxyl film, during the pit repassivation process. At
more anodic potentials, where the pit repassivation processes
332

terminate, roughening of the M/F interface was reported to occur. The

optical data that led to the above conclusions are shown in Fig. 43.
Fluctuations in the reflected light intensity were attributed to
pitting and the consequent surface roughness.
Sugimoto, et al. [52] made microscopic ellipsometric
observations of the spatial distribution of film thickness on ISCr-SNi
stainless steel, ie within each grain before pit initiation and after
formation of the first pit. The film thickness was found to vary from
site to site. Pits nucleated at sites where a small thickness
difference of 0.2 to 0.5 nm occurred. However, it was not clear
whether pits nucleated at the thinner or thicker areas. It appears,
therefore, that there is no correlation between film thickness and
susceptibility to pit nucleation. At the early stage of pit growth it
was found that film thinning occured around each pit and not on the

Figure 43: Potential dependence of current (1), ellipsometric

parameters, and increment of film thickness for aluminum in an NaCl
solution [12].

surface, as was suggested by Fokin et al. [12]. A secondary pit was

seen to form at the boundary between the film thinning area and the
 333

surrounding. Sugimoto et al. [52] did not explain the reason for the
nucleation of the first pit at specific sites on the alloy surface. As
far as secondary pits are concerned, they attributed their initiation
to such factors as microcracks resulting from differences in volume,
chemical composition, hydration, eletrostriction and interfacial
tension; all these lead to mechanical breakdown of the film.

11. ATMOSPHERIC CORROSION

Valuable information on the mechanism of atmospheric corrosion
can be obtained by ellipsometry. For example, Eichhorn and Forker [53]
have used ellipsometry to study the initial stages of atmospheric
corrosion by measuring the water film growth on iron and steel covered
with oxide films formed in air. For this purpose, a mixture of air and

4.0r------------------.

r
d..,o
~
nPo•=2.68!O.04,d".. =5, 3nm

'"
100%

'/-l:::::!::::=::::::::::==::::::::==::: :::z I
I
'00 JOO
..L_
m,n

Figure 44: Change in water film thickness (d) with exposure time (t)
as a function of relative humidity (rh); T=298K [53].

water vapor, of known relative humidity (RH), was let through the
ellipsometric cell over an iron or steel specimen. It was found that
334

at RH < 80%, only one or two layers of water molecules were formed on
the surface of all the materials studied (taking 0.31 nm as the
diameter of H20 molecule). At RH = 100%, there were 10 to 12 layers of
adsorbed water molecules. A plot of the water film thickness vs RH is
shown in Fig. 44. Complete desorption of water in the early stage is
influenced by the quality of the air-formed film (which could contain
different amounts of OH species) and also by the RH. The thickest
water film, at constant RH, was observed on samples with primary oxide
layers containing the highest amount of OH species, as identified by
low refractive indices and high film thickness values. The authors
conclude that the adsorptive bonding of water molecules to OH groups
occured through hydrogen bridges. Only at very high RH values did the
adsorbed water gradually become a typical liquid phase with properties
of compact solutions into which reducible species from the atmosphere,
and metal ions from the underlying substrate could be dissolved. When
this occured, the essential conditions for electrochemical corrosion
are created.
Reinhard and Kunzelmann [54] studied the interaction of two
vapor-phase inhibitors (VPI), dicyclohexyl ammonium nitrite (Dichan)
and dicyclohexylamine (Dichamine) with iron using ellipsometry. They
measured the change in thickness of water layers on iron covered with a
pre-formed film (primary oxide film). A flow of either dry or humid
air saturated with Dichan or Dichamine was maintained over the test
specimen at a constant rate. Figures 45 and 46 show the variation in
refractive index and thickness of the adsorbed layers as a function of
time. The effect of Dichan differed from that of Dichamine. In the
case where iron covered with a water-free primary oxide was exposed to
moist air saturated with Dichan, the formation of hydroxyl groups on
the metal surface (as a result of the dissociative chemisorption of
water molecules) was impeded and Dichan was directly adsorbed on the
primary oxide. The resulting increase in layer thickness was minimal.
The effect of Dichan as VPI was mainly due to its adsorption on the
primary oxide layer, thus making the metal surface hydrophobic. Dichan
as VPI was only effective at saturation in moist air. It lost its
influence as soon as it was removed from moist air. It also desorbs
when the moist air was replaced by air free of inhibitor. It was
concluded that inhibition of corrosion by Dichan was due to its
interaction with hydroxyl groups in the passive film.

12. STRESS CORROSION CRACKING

Smialowska and Gust [55] studied the stress corrosion cracking
(SCC) and pitting of austenitic stainless steel in MgC1 2 solutions.
The occurrence of these processes and their extent depend on MgC1 2
concentration, temperature, potential and stress. If the steel surface
is in its passive state under critical environmental conditions,
pitting takes place; in the presence of sufficiently high tensile
 335

Q3

r '2
Q1

so 70 90 110
t/mjn~

Figure 45: Change of the refractive index n, of adsorption layer on

iron in dependence on the RH and the sort of VPI (evaluation according
to the single-layer model). Denotations of the curve cf. Figure 46
[Reference 54]

I
10 30 50 ---=7=--0---9~0-- 110

Figure 46: Change of the thickness d, of adsorption layer on iron in

dependence on RH and the sort of VPI. ma(p) pure moist air, RHzIOO%;
rna (i) moist air saturated by VPI; da (p) pure dry air, (RH) = 0%; da
(i) dry air saturated by VPI; (evaluation according to the single
layer model). I-pure air; II-air containing DICHAN, III-air
containing DICHAMIN [Reference 54]
336

stresses, cracks may nucleate at the bottom of the pits. This was
found to occur in 5% MgC1 2 solution at 40-90 C and in a more
concentrated solution of 35% MgC1 2 at 40 and 60 C. In contact with hot
35% MgC1 2solutions, stress corrosion cracks initiated on the apparently
smooth metal surface.
SCC of austenitic stainless steel, mostly transgranular, has
also been found to occur in 5-10 N H2S0 4 + 0.5-1 N NaCl [56] and in 1-3
N HC1 [57, 58] at room temperature. Ellipsometry was used to find out
if the preexisting air-formed film on the steel surface changes its
properties under above environmental conditions. It was shown [59]
that upon exposure of steel to 5% MgC1 2 at 40-90 C or to 35% MgC1 2 at
60 C, the original air-formed, highly protective oxide film did not
undergo any significant deep modification. The refractive index of the
film varied from 2.3 to 2.6.

Table 3: Optical Parameters of Passive Steel and Surface

Films Produced by Immersion of Steel in
Various Solutions(}) (reference 59).
Solution Temp. C

5% MgCl 60 2.27(1-1.67i)
5% MgC1 2 90 2.27(1-1.67i)
35% MgC1 2 60 2. 61 (1-1. 50 i )
35% MgC1 2 90 2 . 61 ( 1-1. 50 i ) 1.46(-0.00Ii)(2)
O.IN HCl 20 2.46(1-1.59i) 1. 90 (1- 0 . 3i)
IN HCl 20 2.46(1-1.59i) 1. 75(1-0.3i)
3N HCl 20 2.46(1-1.59i) 1.55(1-0.3i)
5N H2S0 4 + 0.5N NaCl 20 2.46(1-1.59i) 1.65(1-0.3i)

(l)Wave length 546 nm; n3 (I-K 3i) = complex refractive index of passive
steel surface; n2(I-K 2i) = complex refractive index of surface
films.
(2)
n2 +- 0.04, K2 +- 0.001.
 337

In contact with 35% MgC1 2 at 90 C, the previous oxide film transformed

into a layer with a refractive index of 1.4 to 1.9. (Table 3). These
low values suggest that the layer is composed of salts and
oxyhydroxides which exhibit moderately protective properties and hinder
the passivation process. When the whole surface is covered with such a
layer, cracks nucleate on the smooth metal surface.
A film with similar properties is thought to exist at the bottom
of corrosion pits and therefore cracks often initiate at pits. Hence,
SCC of the anodic type occurred only when a protective passive film
changes into a moderately protective one.

13. DELAMINATION OF PAINTS

Ellipsometry was used by Ritter and Kruger [60] for a model study
of the corrosion behavior of painted metal surfaces. They constructed
a special electrolytic cell which allowed simultaneous ellipsometric,
electrochemical, and pH measurements to be made in the solution under
the coating, during prolonged exposure to chloride-containing
environments. Transparent organic coatings on iron were used to
simulate a painted metal surface. Some results of these
measurements are shown in Fig. 47. On the basis of these results, a
model was presented, according to which delamination of the paint
occured in several stages. In the first stage, oxygen, water, sodium
cations and chloride anions permeate the coating and pitting takes
place. In the second stage, as the pH attains a value of about 14, a
sharp drop of 6 occurs, and iron dissolves in the form of HFe0 2-. In
the third stage, deterioration of the organic coating occurs. With the
use of this technique, the effects of inhibitive pigments [60] and
chelating inhibitors [61] on the delamination of paints were studied.

14. MICROSCOPIC AND SPECTROSCOPIC ELLIPSOMETRY

A significant progress in the use of optical techniques to study
surface phenomenon was made in the last two decades by the development
of spectroscopic and microscopic ellipsometry.
Microscopic ellipsometry was used by Sugimoto and Matsuda [62] for
the determination of optical parameters and thickness of the films
formed on a and 1 grains of 22Cr-10Ni austeno-ferritic stainless steel
in Na 2S0 4 solution of pH 6. The optical parameters of the oxide free
surface were obtained by removing the oxide film in a 0.5% bromine-
methanol solution. They found different optical constants and
different film thickness on a (ferrite) and 1 (austenite) grains, both
in the passive and transpassive regions of potential. The thickness of
the film on a grains was larger than that on 1 grains (Fig. 48).
Significant differences in film thickness was found to exist at
alfa/gamma phase boundaries. The authors analyzed the composition of
338

Jl m

13

12

II
pH
10

J9
!
j8

12

'if
(deg)

25
rriJ
SHE
400

I(min)

Figure 47: Variation of pH, potential (SHE) and ellipsometric

parameters with time for Fe coated with collodion in 0.05N NaCl
[60].
 339

40

- CI01IdIn

3.0

....
E
c

2.0

Figure 48: Variation in thickness (d) of the passive and transpassive

films as a function of potential (E) for individual grains of ferrite
and austenite and for a macroscopic area containing a mixture of
austenite and ferrite grains [62].
the grain by Electron Probe Micro Analysis (EPMA) and measured anodic
polarization curves on alloys of the same composition as a and ~
grains. They found that the corrosion rate in both active and passive
states of ~ simulated alloy was higher than on the a alloy. It is
known from the work by Manning, et al. [63] that pits in chloride
solution initiate at the alfa/gamma phase boundaries in 304 Land 308 L
steels.
Hanekamp et al. [64] employed dynamic and spectroscopic
ellipsometry to study the initial oxidation (under ultra high vacuum)
of Ni(llO) exposed to oxygen. The complex dielectric constant of the
clean Ni surface was determined as a function of wavelength between
o 0
4000 A and 8000 A (Fig. 49). The pure nickel substrate is obtained by
argon ion bombardment. The difference between the values obtained and
the literature values was attributed to: surface contamination,
physical differences between single crystal and polycrystalline
samples, and differences between thin film and bulk characteristics.
Dynamic ellipsometry was then employed to study the change in ~ upon
340

~,

~
~r
,
•
~
.. • .. ~TH (N1r' .~
,. .

• .• •

..
.. • .. .. ..
\PI'OQ..ENClH Uti)
"
.111
,. .
Figure 49: The components of the complex dielectric function (a) e
and (b) e" as a function of wavelength. The calculated values for the
clean surface are presented by solid lines, and for the oxygen exposed
surface by dashed lines. The data points are values reported in
literature [64].

oxidation. Hanekamp et al. identified the oxygen chemisorption and

oxide nucleation stages from differences in the rate of change of ~
with exposure. The point in Fig. 50, where a steep increase in ~
occurs, was attributed to the transition from chemisorption to oxide
formation. The change in ~ on saturation (upon oxidation of the
samples) is shown in Fig.51 as a function of wavelength. Also shown in
the figure is a theoretical fit to the data (obtained by employing the
effective medium approximation of Bruggeman) for a surface film
composed of 90% NiO and 10% Ni. The value of refractive index of NiO
employed in the calculation was not indicated.
In an attempt to understand the enhanced catalytic activity of Pt-
Cr alloys on the oxygen reduction reaction, Gottesfeld et al. [11]
studied the oxidation behavior of these alloys. Ellipsometric data
obtained on potential multicycling in 0.05 MH2S04 were fitted to a
surface film with n = 1.429 - 1.445 i at A = 5500 A (Fig. 52). The
 341

=rr--------------------------------------,
•

•
a

a~~~~~~~--~~~
'!1.1l -1.0 -G.1l 0.0 0.1l 1.0 1.1l 2.0 2.1
LOG E>CPOSI.H: (LOG L)

-8
Figure 50: Change in "delta" upon oxygen exposure at 1.3 X 10 Torr
as a function of sample temperature. Annealing temperature 570K
[Reference 64]

~--------------------------------~:

Figure 51: Change in ellipsometric parameters, calculated for layers

of NiO and Ni at room temperature on clean Ni (110). The numbers in
the figure indicate the volume fraction of NiO. (X) measured values
after 20L oxygen exposure at room temperature [Reference 64]
342

~O.4 - SUBSTRATE: n ~ 2.881; k ~ 3.651

~~

FILM: n ~ 1.429; k ~ 1.445 4.5

-0.8
~
;>.

"
-1.2

-1.6

-2.0
-34 -28 -22 -16 -~ 10
dl>/o

Figure 52: Experimental ellipsometric parameters (points) and

theoretical fit (solid curve)to a model assuming uniform film growth,
for the surface layer develops on Pt O. 2crO. 8 alloy cycled between 0.04
amd 1.40V (OHE) in O.05M H2S0 4, The number adjacent to each
experimental point is the optically evaluated thickness of the surface
layer in nm [11].

In an attempt to understand the enhanced catalytic activity of Pt-

Cr alloys on the oxygen reduction reaction, Gottesfeld et al. [11]
studied the oxidation behavior of these alloys. Ellipsometric data
obtained on potential multicycling in 0.05 MH2S04 were fitted to a
surface film with n = 1.429 - 1.445 i at A = 5500 A (Fig. 52). The
corresponding thickness obtained was compared with that calculated from
experiments at 11 other wavelengths; the variation was reported to be
only about 2%. Moreover, the authors were also able to model the
complex dielectric constant obtained at different wavelengths with a
mixed phase containing both metallic and film material. A 30% volume
fraction of the metallic phase and a film refractive index of n = 1.36
- 0.006 i (allowed to vary freely in the model) was shown to provide a
good fit to the experimental data (Fig. 53). The theoretical
calculations were made employing the Maxwell-Garnett model for the
mixed phase. Gottesfeld et al. [11] attributed the enhanced catalytic
activity to the porous surface layer and hence the increased surface
area of exposure. It can be seen from fig. 53 that no fit was obtained
for the experiments conducted in 85% H3P04 . This was attributed to
substantial amounts of phosphate contained in the film. In the above
 343

•
II

-2 • -4

-4
..:...:.-._._;..,----
~'Cr ••....zc.u -'~';::::-I
....
-I
~Pt
..... *""""
,.X...... ............

••••••••••.••••••
-12

-I _II
101 "I 121 •• _ .. JII
AI.

Figure 53: Wavelength dependence of the complex dielectric constant

for bulk Pt, bulk Cr, bulk Pt o . 2 CrO. 8 ' and the surface layer (S.L)
formed on Pt O. 2 Cr O•8 as a result of potential cycling up to I.4V in
O.05H H2S0 4, and up to I.6V in 85% H3P0 4. The solid curves for the
surface layer formed in the H2S04 correspond to a fit of the data with
effective medium model, assuming a composite film with two phases where
the metallic phase (host) has optical properties of bulk Pt O. 2CrO. 8 and
occupies 30% of the volume [11].

cycling experiments, the anodic scans extended into the transpassive

range. ~ and ~ data were reported to remain invariant with cycling
within the passive range. Gottesfeld et al. observed a decrease in ~
at a nearly constant ~, and attributed this to the anodic formation of
Cr0 2, which is a highly absorbing film. This phase is believed to be
responsible for the increased catalytic activity of Pt-Cr alloys.

15. GASEOUS OXIDATION OF MATERIALS:

Ellipsometric studies of gaseous oxidation have been carried out
mainly by the electronics industry [65, 66, 67]. Sato [65] studied the
influence of process variables on the thickness of IrOx films formed on
plasma oxidation of Ir-C films deposited on a glass substrate. The
344

90
80
70
w 60
w
Q:
~
w 50
,
Q

40
<
I-
...J
w
Q
30
20
10
0
0 10 20 30 40 50
PSI/DECREE
Figure 54: Ell ipsometry results for plasma etching of Ir-C films in
oxygen. The numbers denote the etching time in minutes [Reference 65]

I~----~------~----~----~----~
o 10 20 30 50

". /.OEGIt£E
Figure 55: Calculated ellipsometry curves for IrOx film on Ir-C,
assuming thickness of Ir-C film to be infinite. The complex
refractive indices n-ik, and n2 -ik 2 refer to IrOx and Ir-C
respectively [Reference 65]
 345

100
90
80
w 70
w
n::
l:)
w 60
0
"<
f-
..J
50
w 40
0
30
20
10
a 10 20 30 40
PSI/DEGREE

Figure 56: Calculated ellipsometry curves for IrOx films on Ir-C film
with finite thickness dZ' The solid curve is for n1-ik 1 = 1.5-0.2i,
the dotted curve is for n1-ik 1 = 1.5 - O.li [Reference 65)

100

~ .
90 20 Ii

w
80 . 1. 3,.1 0 -~ Pc
2. 7,.H:-~ '"c
w a
n::
l:)
w 70 • :. 3,.: :-. Po
0 ~ :. :.::- Cl e
"....<
..J
SOt
I
00

~:~
W

Q.
0 50f-

::l
0. •
~0
___ e Q 0
0- 0

10 2C 3: :.:
"5: :::~~::::

Figure 57: Ellipsometry results for plasma oxidation of Ir-C films.

The numbers indicate the oxygen pressure. The solid symbols indicate
the initial values at t=O [Reference 65]
346

300~--------------'-----------'

300

240

~180

120

o 10 20 30 40 50 00 70 80 90
Psi

Figure 58: The ellipsometric parameters for all of the oxide films.
The open squares are the measured data and the solid line is the
calculated trend using the indicated film free values and index of
refraction [Reference 67]

2~r---------------------------~

1500

1000

500

o
V
2 6 8 10

Oxidation Time (min)

Figure 59: Thickness, measured with ellipsometry plotted as a

function of the square root of the oxidation time, for three different
oxidation temperatures [Reference 67J
 347

potential application of iridium oxide films in electrochromic displays

(due to its short response time) prompted this study. Carbon was used
in the substrate since it was reported earlier that it causes the
formation of thicker oxide films at steady-state. The refractive index
of the Ir-C film calculated from optical data obtained prior to
oxidation (at 10- 3 Pal was 2.4-1.3i at A • 6328 A. The data obtained
upon oxidation is presented in fig 54 and compared with theoretically
evaluated maps of 8 and ~ for various n, k and thickness values for the
oxide film (Fig. 55). The film was determined to have a refractive
index of I.S-0.2i. The decrease in ~ observed in Fig. 21 was
attributed to a reduction in the thickness of the underlying Ir-C film.
Sato arrived at this conclusion by comparing the optical data with
theoretically calculated values of 8 & ~, upon variation in the
thickness of the Ir-C and the oxide films (Fig. 56). It was reported
that the steady-state thickness of the oxide film occured when ~
started to decrease. The experimentally determined 8 and ~ values,
upon plasma oxidation at different oxygen pressures, are presented in
Fig. 57. Based on these results, it was concluded that the maximum
steady-state thickness occured at an oxygen pressure of 5x10- 3 Pa. The
thickness ranged from SOnm to IIOnm. The thicker film on Ir-C,
compared to that on Ir (of the order of 10nm), was attributed to the
selective sputtering away of C (confirmed by XPS). This resulted in
increased porosity and exposed area of the surface layer.

SOO-C 4~ 400'C 3~
106

10- 12

...
.,;
105
c 10·13 u

JE
c
:I E
• l:-
c
0 u
u ~ 104
~
•
II: 10. 14

loJ
0.0012 0.0013 0.001. 0.0015 0.0011

Irremperelure("K)

Figure 60: Arrhenius plot of the rate constants for the formation of
the oxide on 90 wt.% Ti-W alloy [67].
348

The sensitivity of metal-oxide-metal tunnel diodes depends strongly

on the thickness of the oxide layer which forms the potential barrier
for tunneling electrons. Inoue et al [66] studied the effect of oxygen
pressure on the film thickness on nickel upon plasma oxidation. The
substrate refractive index used was 2.06-3.89i at A = 6500 A. The
source for this value was not reported. The steady-state oxide film
thickness and the film refractive index (4-5) was calculated from the
optical data by assuming K = O. The deviation from the literature
value (2.2 for NiO single crystal) was attributed to the
polycrystallinity of the sample and to the influence of the glass
substrate (on which the 2500 A thick Ni film was deposited). The
thickness determined from ellipsometry was confirmed by ion micro-
analysis. The film thickness was proportional to the oxygen pressure
and ranged from 70 A to 90 A, for oxygen pressures of 4.5xl0- 4 to
5.3xl0 -4 Pa.
Tompkins et al [67] studied the oxidation of Ti-W (90 wt% Ti)
films. This material is used as a barrier between Al and Si, with Al
and Au as a secondary mask. The optical data obtained upon oxidation
of the sample is shown in Fig. 58. The substrate refractive index
(2.84-3.08i at 6328 A) was calculated from the first set of data points
indicated in the figure. Although this did not correspond to the pure
substrate value, the conclusions regarding the film thickness were
corroborated by combined AES and sputter profiling. The film
refractive index obtained from a theoretical fit (2.25-0.21i) is
believed to be accurate to within 5% for 'n' and within 10% for 'k'.
Variations beyond this range caused a visual lack of fit to the
experimental data. The variation in the film thickness with the square
root of time was shown to be linear (Fig. 59), and hence it was
concluded that the oxidation followed parabolic kinetics. The
activation energy for the oxidation process was also determined by
plotting the slope of the lines in fig. 59 (proportional to the rate
constant) against liT, where T is the sample temperature (Fig. 60).
The slope of the line in Fig. 60 yielded an activation energy for
oxidation of 33.9 Kcal/Mole.
Langevoort et al [68] investigated the effects of surface cold
work on the oxidation behavior of Incolloy 800 H under UHV conditions.
The oxygen pressures (PO) were in the range 10- 3 to 10 2 Pa and the
2
sample temperature varied from 300K to 600K. Surface cold work was
imparted by Ar+ ion bombardment (energy of 2 to 4 KeV, intensity of
0.08 to 1.5 ~a/cm2). The extent of oxidation was inferred from the
variation in 6. It was found that the extent of oxidation increased
with both Ar+ ion energy and intensity, provided that the specimen
exposure conditions (PO and temperature) were right. For example, at
2
 349

sufficiently low Po ' the film thickness would be less than the cold
2
worked layer under the lowest energy and intensity of Ar+ ions. Hence,
further increases in the latter parameters would not produce any
increase in the oxide film thickness. A similar argument holds for a
high Po and for the extremes in temperature. An increase in the
2
extent of oxidation was attributed to an increase in the concentration
of point defects due to Ar+ ion bombardment.

16. CONCLUDING REMARKS

This article is intended to provide an awareness of the
principles and broad applicability of ellipsometry to corrosion
research. To attain this objective, a background of the technique was
provided, followed by a detailed discussion of a wide variety of
applications. It is to be emphasized that ellipsometry is a valuable
tool when used in conjunction with other optical, electrochemical, or
electron optical techniques. It is, however, also important to realize
its limitations.

ADDENDUM
Notable contributions to instrumentation and data analysis
were made since the submission of this article for publication. Cohn
et a1. [69] have presented a technique (Dynamic Imaging
Microe1lipsometry) to rapidly scan the sample to obtain a two-
dimensional image of the surface by employing several optical
detectors. This technique was developed to gain further insights into
the nature of changes taking place in passive films associated with
localized corrosion phenomenon. De Smet and Ord [70] have reported the
application of the simplex algorithm for an efficient calculation of
the film refractive index. Starting with a reasonable 'guess' for the
refractive index, the algorithm progressively determines 'better'
values that minimize the difference between the calculated and
experimental optical parameters. A procedure for quantitative
estimation of the goodness of fit is also presented. Gottesfe1d
[71]has written a comprehensive review of the principles and
applications of ellipsometry to electrochemistry.
350

REFERENCES
1. L. Tronstad, Trans. Faraday Soc., Vol 31, pp. 1151 (1935).
2. R.M.A. Azzam and N.M. Bashara, Ellipsometry and Polarized Light,
North-Holland, Amsterdam (1977).
3. A. Rothen, Ellipsometry in the Measurement of Surfaces and Thin
Films-Symposium Proceedings; Passaglia, Stromberg and Kruger-Eds;
NBS Misc. Publ. #256, (1963).
4. (a) J. Kruger, Advances in Electrochemistry and Electrochemical
Engineering, P.Delahay and C.W. Tobias-(Eds), Vol. 9, p. 227
(1973). (b) D. E. Aspnes, Thin solid films, Vol. 89, p. 249
(1982)
5. R.H. Muller, Advances in Electrochemistry and Electrochemical
Engineering, Delahay and C.W. Tobias-(Eds), Vol. 9, p.167 (1973).
6. M.E. Brett, K.M. Parkin and M.J. Graham, J. of Electrochem. Soc.,
Vol. 133, p.2031 (1986).
7. N. Iwasaki, Y. Sasaki and Y. Nishina, J. of Electrochem. Soc.,
Vol. 135, p. 2531 (1988).
8. B.D. Cahan, Passivity of Metals and Semiconductors, M. Froment-
(Ed), Elsevier Science Publishers B.Y., Amsterdam, p. 187 (1983).
9. P.B. Barna, Z. Bodo, G. Gergeley, P. Croce, J. Adam and P. Jakab,
Thin Solid Films, Yolo 120, p. 249 (1984).
10. J. Rishpon, I. Reshef and S. Gottesfeld, Passivity of Metals and
Semiconductors, M. Froment- (Ed), Elsevier Science Publishers
B.V., Amsterdam, p. 205 (1983).
11. S. Gottesfeld, M.T. Paffet, A. Redondo, J. Electroanal. Chem., Vol
205, p.163 (1986).
12. M.N. Fokin and V.A. Kotenev, Zashchita Metallov, Vol. 24, p.lll
(1988).
13. W. Visscher and E. Barendrecht, Surface Science, Vol.135, p.436
(1983).
14. A. Hamnet and R.J. Mortimer, J. Electroanal. Chem., Vol. 234,
p.185 (1987).
 351

15. F. A. Jenkins and H. E. White, Fundamentals of Optics, McGraw Hill

Book Company, Inc. (1957).
16. P.B. Barna, Z. Bodo, G. Gergely and O. Szigethy Vacuum, Vol. 33,
p.93 (1983).
17. W.K. Paik, M.T.P. International Reviews of Science,
J. O/M Bockris- (Ed), Vol.6, Electrochemistry, Butterworths
(1973).
18. R. Greef, Comprehensive Treatise on Electrochemistry, R.E. White,
J. O/M Bockris, P.E. Conway, E. Yeager - (Eds), Vol. 8, Plenum
(1984).
19. P.C.S. Hayfield, A.T. Kuhn, Techniques in Electrochemistry,
Corrosion and Metal Finishing - A Handbook, A.T. Kuhn - (Ed), J.
Wiley &Sons (1987).
20. J. Kruger, J.P. Calvert, J. Electrochem. Soc, Vol. 114, p.43
(1967).
21. Z. Szklarska-Smialowska, W. Kozlowski, J. Electrochem. Soc. Vol.
131, p.234 (1984).
22. T. Zakroczymski, Chwei-Jer Fan, Z. Szklarska-Smialowska, J.
Electrochem. Soc. Vol. 132, p. 2862 (1985).
23. J. O/M. Bockris, M.A. Genshaw, V. Brusic and H. Wroblowa,
Electrochim. Acta, Vol. 16, p.1859 (1971).
24. A.G. Akimov, N.J. Andreeva, I.L. Rozenfeld, Elektrokhimiya, Vol.
15, p.1888 (1979).
25. B.D. Cahan and C.T. Chen, "Proceedings of an International
Symposium Honoring Professor H.H. Uhlig on his 75th Birthday,"
Corrosion and Corroion Protection, Proceedings Vol. 80-81, R.P.
Frankenthal - (Ed), The Electrochem. Soc., p.37.
26. F.C. Ho and J.L. Ord, J. Electrochem. Soc., Vol. 119, p.139
(1972).
27. J.L. Ord and O.J. de Smet, J. Electrochem. Soc., Vol. 123, p.1876
(1976).
28. N. Sato, K. Kudo and T. Noda, Z. Phys. chern. N.F., Vol 98, p.271
(1975).
29. R. Nishimura, K. Kudo and N. Sato, Surf. Sci., Vol. 96, p.413
(1980).
352

30. S. Juanto, J.O. lerbino, M.J. Miguez, J.R. Vilche, A.J. Arvia,
Electrochimica Acta, Vol. 32, p.1743 (1987).
31. J.L. Ord, Passivity of Metal and Semiconductors, M. Froment -
(Ed), Elsevier Science Publishers, p.95 (1983).
32. K. Sugimoto, S. Matsuda, Conference Proceedings, Stainless Steel
1987, York U.K. 14-16 Sept. 1987, The Institute of Metals, Carlton
House Terrace, London, (1988).
33. K. Sugimoto and S. Matsuda, Mater. Sci. Eng., Vol. 42, p.181
(1980).
34. N.D. Totir, 10th International Congress on Metallic Corrosion,
Vol. IV, p.3849 (1987) Madras.
35. W. Kozlowski and J. Flis, Corr. Sci., Vol. 28, p.789 (1988).
36. J. Flis, H. Oranowska, I. Szklarska-Smialowska, to be published.
37. T. Ohtsuka, K. Azumi, N. Sato, Passivity of Metals and
Semiconductors, M. Froment - (Ed), Elsevier Science Publishers, p.
79 (1983).
38. Woon-Kie Paik, I. Szklarska-Smialowska, Surface Sci, Vol. 96,
p.401 (1980).
39. J.L. Ord, D.J. Desmet, and I.S. Huang, J. Electrochem. Soc., Vol.
134, p.826 (1987).
40. T.H. Allen, J. Vac. Sci., Vol. 13, p.112 (1976).
41. R. Greef and C.F.W. Norman, J. Electrochem. Soc., Vol. 132, p.2362
(1985).
42. C.J. Dell'Oca, P.J. Fleming, J. Electrochem. Soc., Vol. 123,
p.1487 (1976).
43. C.J. Dell'oca, and L. Young, Surface Science, Vol. 16, p.331
(1969).
44. R. Greef, P.R. Harvey, and S.P. Tyfield, Materials Science Forum,
Vol. 8, p.395 (1986).
45. J. Pakulska, A.A. Wronkowska, A. Wronkowski, Thin Solid Films,
Vol. 150, p.115 (1987).
46. T. Ohtsuka, M. Masuda, and N. Sato, J. Electrochem. Soc., Vol.
134, p.2406 (1987).
 353

47. D.J. DeSmet, and J.l. Ord, J. Electrochem. Soc., Vol. 134, p.1734
(1987).
48. G. Mrowczynski, Z. Szklarska-Smialowska, J. of Applied
Electrochemistry, Vol. 9, p.20l (1979).
49. P. Spinelli, R. Fratesi and G. Roventi, Werkstoffe u. Korrosion,
Vol. 34, p.161 (1983).
50. R.G. Pearson, J. Am. Chern. Soc., Vol. 85, p.3533 (1963); J. Chern.
Education, Vol. 45, p.581, 643 (1968).
51. H. W. Shen and Z.Szklarska-Smialowska, to be published.
52. K. Sugimoto, S. Matsuda, Y. Ogivara, K. Kitamura, J. Electrochem.
Soc., Vol. 132, p.1791 (1985).
53. K.J. Eichhorn and W. Forker, Corrosion Sci., Vol. 28, p.745
(1988).
54. G. Reinhard, U. Kunzelmann, Werkstoffe u. Korrosion, Vol. 38,
p.ll0 (1987).
55. Z. Szklarska Smialowska and J. Gust, Corr. Sci, Vol. 19, p.753
(1979).
56. S.J. Acello and N.D. Greene, Corrosion, Vol. 18, p.286 (1962).
57. N.A. Nielsen, Corrosion, Vol. 27, p.173 (1971).
58. G. Bianchi, F. Mazza, S. Torchio, Carr. Sci., Vol. 13, p.177
(1978).
59. Z. Szklarska-Smialowska and N. lukomski, Corrosion, Vol. 34, p.177
(1979).
60. J.J. Ritter and J. Kruger, Organic Coatings; Science and
Technology, Vol. 6, p.309 (1982).
61. J.J. Ritter, Journal of Coating Technology, Vol.56, No. 714, p.55
(1989).
62. K. Sugimoto and S. Matsuda, J. Electrochem. Soc., Vol. 115, p.2323
(1983).
63. P.E. Manning, C.E. lyman, D.J. Duquette, J. Electrochem. Soc.,
Vol. 132, p.2323 (1983).
64. l. J. Hanekamp, S. J. H. Brader, G.A. Bootsma, Surface Science,
Vol. 135, p.383 (1983).
354

65. Y. Sato, Japanese J. of Appl. Phys., Vol. 25, p.189 (1986)

66. N. Inoue. and Y. Yasuoka, Vacuum, Vol. 34, p.687 (1984)
67. H.G. Tompkins and S. Lytdle, J. Appl. Phys., Vol. 64, p.3269
(1988)
68. J.C. Langeroort, D.C. Heslinga, L.J. Hanekamp, A. Van Silfhout, L.
Lefferts and P.J. Gellings, Reactivity of Solids, part-A, Elsevier
Science publishers, p.209 (1985)
69. R.F. Cohn, J.W. Wagner, and J. Kruger, J. Electrochem. Soc.,
Vol. 135, p.l033 (1988).
70. D.J.De Smet and J.L. Ord, J. Electrochem. Soc., Vol. 136, p.2841
(1989)
71. S. Gottesfeld, Electroanalytical Chemistry, A series of advances,
A.J. Bard-(Ed), p.143 (1989)
LASER RAMAN SPECTROSCOPY: PRINCIPLES AND APPLICATIONS TO CORROSION
STUDIES

C. A. Melendres
Materials Science and Chemical Technology Divisions
Argonne National Laboratory
Argonne, Illinois 60439
U.S.A.

ABSTRACT. Laser Raman Spectroscopy (LRS) has become an important tool

for the in-situ characterization of corrosion films on metal surfaces.
We introduce here the experimental techniques involved in applying LRS
to electrochemical systems, and survey the literature for examples of
studies on corrosion and passivation phenomena. Laser Raman
spectroelectrochemical studies of corrosion and film formation on such
metals as Pb, Ag, Fe, Ni, Co, Cr, Ti, Au, Zn, stainless steel, etc. in
various solution conditions are reviewed. Results on studies of the
structure of corrosion inhibitors are also presented. Further
extension of the technique to corrosion studies in high-temperature and
high-pressure aqueous environments is demonstrated.

1. INTRODUCTION

Laser Raman Spectroscopy (LRS) has become an important tool for the
study of electrochemical systems and processes because of the molecular
specific information that it pr-ovides on the structure of the solidI
solution interface in situ. 1 ,2,3,4 The spectrum of the inelastically
scattered light from a sample contains information related to the
various normal modes of vibration of the system and constitutes a
fingerprint by which the material can be identified. Such specific
structural information nicely supplements the kinetic data contained in
the conventional current-potential curve and is essential for arriving
at a detailed interpretation of the mechanism of electron-transfer
reactions. The technique also nicely complements vibrational
information obtained by infrared (IR) and electron energy loss (EEL)
spectroscopies. Numerous applications of LRS to electrochemistry have
been made since Fleiscl~ann and Hendra 3 first reported use of the
technique to study the composition of solid phases electrochemically
formed on mercury. The study of anodically formed surface phases and
corrosion films continues to be an important application of the
technique; work in this area since 1973 is surveyed here. Theory and
instrumentation are discussed briefly. We deal mostly with practical
applications in order to illustrate the power of the technique. We
also point out the limitations and problems in its use to study
355
M_ G. S. Ferreira and C. A. Me/endres (eds.).
Electrochemical and Optical Techniques/or the Study and Monitoring 0/ Metallic Corrosion. 355-388.
© 1991 Kluwer Academic Publishers.
356

interfacial electrochemical phenomena. Examples are taken from the

literature and from results of work in the author's laboratory. The
illustrations are not meant to be all inclusive, rather just of
sufficient detail to be helpful to the beginner in the field.

Z. THEORY

When light of a single frequency Vo strikes a sample (Fig. la), the

scattered radiation is found to consist of a number of discrete fre-
quencies which can be grouped into three categories, i.e. Vo - vvib,
V o ' and Vo + Vvib, where vvib is the frequency of a normal vibration 1 ,5
of the material. Raman spectroscopy consists in the measurement and
analysis of the inelastically scattered light of frequency Vo - vvib,
Vo + vvib, etc. The set of spectral lines of lower frequency than v o '
i.e., Vo - vvib, are called the Stokes lines, while those with higher
frequencies, Vo + vvib are referred to as anti-Stokes lines. The
elastically scattered light of frequency Vo is termed the Raleigh line.
The change in energy states that accompany the scattering process is
schematically illustrated in Fig. lb, where a molecule is excited from
the electronic state (with vibrational levels v = 0, 1) to a virtual
higher electronic state. The corresponding Raman spectrum is shown
schematically in Fig. lc. From an analysis of such Raman spectra, it
is possible to extract molecular electronic structural information from
a calculation of the set of fundamental vibrational frequencies, vvib'
The intensity of the scattered Raman light has been derived and
shown to be 2

(1)

where 10 is the incident light intensity, (dazz/dr) is the change of

molecular polarizability a zz with vibrational coordinate r, and k is a
constant. It is important to note the dependence of the intensity on
the fourth power of the frequency of the scattered light. Likewise,
the essential requirement (selection rule) for observation of the Raman
effect is that the polarizability of the molecule change with vibra-
tion, i.e., da/dr + O.

3. EXPERIMENTAL ASPECTS

The instrumentation required for LRS consists simply of a light source

for exciting the sample, an analyzer to determine the wavelengths of
the scattered light, and a detector to measure the intensity of the
scattered radiation. The sample is contained in an electrochemical
cell of suitable design. Commonly used sources of high intensity and
coherent light are lasers, e.g., Ar+ or Kr+ gas lasers, which provide
discrete wavelengths in the visible region from about 450 to 680 nm.
Dye lasers provide tunability over a wider range of wavelengths.
Analysis of the frequency distribution of the scattered light is done
 357

Vo + Z'vib
( ANTI- STOKES)

l'o - V.ib
(a) (STOKES)

---------- -,- - - VIRTUAL

- - - -,- r- -- r- -- STATES

t Va
Va Va
l'o
?; l'o- Z{ib 1'0 1.'0 Va+Vvib
0::
W
Z
W

V= I
IVvib
V=O
(b)

STOKES
J1 ANTI-STOKES

Is

Va
V-

(C)

Figure 1. (a) Light scattering from a material; (b) Schematic energy

level diagram illustrating transitions that give rise to Raman spectra;
(c) Sample spectrum corresponding to (b).
358

with single-, double-, or triple-stage monochromators, which are out-

fitted with appropriate gratings and mirrors for dispersion of the
incoming collected light beam. Choice of monochromators is often made
in conjunction with the detector used. The latter can be a photomulti-
plier tube for single-channel photon counting or maybe one of a number
of multichannel optical detectors, such as a diode array, silicon
intensified target, or a silicon photon imaging detector. Further
details on the subject of Raman instrumentation are given in a number
of recent reviews 6 ,7; we will only touch on the basic principles here.
The ultimate aim of instrument design is to maximize the detection of
the inelastically scattered light. The magnitude of the problem may be
appreciated by realizing that this fraction is only about 10- 6 that of
the incident light intensity, 1 0 , The scattered light intensity, Iscat
from molecules on an electrode surface is given by,

(2)

where N is the number density of molecules on the surface per cm 2 and

au/ao is the differential cross section of the molecule for Raman scat-
tering, usually in the range of 10- 31 cm 2 /steradian. The amount of
light detected by a spectrometer system is

Idet = I scat x L x t x (O/4il) x R (3 )

where L is the path length of the laser beam on the sample, t the time
for data collection, 0 the solid angle subtended by the detector and R
the instrument response. The latter is the amount of light detected
divided by the amount entering the spectrometer. From equations (2)
and (3), it follows that the detected light intensity is

Idet = k' x N x t x R x (au/oo) (4 )

where the proportionality constants have been lumped into a single one,
k'. Equation (4) shows how the Raman signal may be increased by in-
creasing the density of molecules on the surface, the time for data
accumulation, and by improvement in the instrument response, R. The
scattering cross section may be increased by making use of the
resonance Raman effect. 2 Veirs 8 has compared the instrument response
for a double monochromator fitted with a photomultiplier tube detector,
a triple spectrometer equipped with an intensified photo-diode array,
and a single monochromator with an imaging photomultiplier tube. He
found R for these systems to be in the ratio of 1:13:300.
An experimental setup employed in the author's laboratory for
in situ LRS of electrochemical systems is shown in Fig. 2. The three-
compartment spectroelectrochemical cell is fitted with a special high-
purity quartz window inclined at an angle of 60° with respect to the
horizontal. This is nearly the optimum angle for maximum scattering
intensity for most metals as shown by Campion. 4 The working electrode
is brought to within 0.5 mm of this window to minimize the solution
volume being traversed by the laser beam striking the electrode
surface. The scattered radiation is collected at an angle of 90° with
respect to the incident beam direction and then analyzed in the double
 359

monochromator. The light intensity is measured by a photomultiplier

tube detector together with associated photon-counting electronics.

POTENllOSTAT

REFIRENCE WORKING
ELECTROOE ELECTROOE
(tl

DOUBLE
MONOCHROMATOR

LASER BEAM
QUARIT II1NDOW

Figure 2. Experimental set-up for laser Raman spectroelectrochemical

studies.

Of important consideration in the LRS experiments are the laser

frequency for exciting the sample and the incident beam power. The
latter is so chosen to be sufficiently intense to obtain a reasonably
good signal without causing composition or structural changes in the
sample. For a laser beam of 2 mm diameter incident on a sample, a
power level of 50 to 150 mW is often adequate. The choice of excita-
tion frequency is usually dictated by the laser sources available.
Commonly used sources are Ar+ and Kr+ lasers as stated previously. It
is of great advantage to select a laser exciting frequency which is
coincident with the electronic absorption spectrum of the material;
this allows for signal enhancement via the resonance Raman effect.

4. APPLICATIONS TO THE STUDY OF CORROSION AND PASSIVATION FILMS ON

METALS

The determination of the structure and composition of corrosion films

on metals constitutes one of the most important practical application
of laser Raman spectroelectrochemistry. An understanding of the rela-
tionships between composition and physico-electrochemical properties of
360

passive films is essential to the technological development of methods

for controlling the degradation of materials via corrosive mechanisms.
Over a dozen metals have been studied b~ LRS under a variety of solu-
tion environments. Among these are A7' Au 10 Pb,11,12,13 Hg,3
Cu,14,15 Co,16 Fe,17-l9 Ni,16,20,2l,4 zn,22 and Fe-Cr alloys.19,23 We
review here results of studies on these systems and in the process
illustrate the kind of information obtainable by the Raman technique.

4.1. Lead and Mercury

The corrosion films on lead and mercury were among the first to be
studied by LRS for good reasons. High atomic number elements combine
to form large molecules with highly polarizable electronic structures
and hence have relatively large cross sections for Raman scattering.
Raman spectral intensities are, therefore, high and detection sensitiv-
ities are better than for molecules composed of lower atomic number
elements.
Fleischman and his group3 first studied the formation of solid
phases on a Hg surface by LRS. By constant potential anodization in
HCl and HBr solutions, they produced the surface phases Hg2C12 and
Hg2Br2 respectively, which yielded characteristic Raman bands at 41,
168, and 273 cm- 1 for the chloride and 36, 135, and 218 cm- 1 for the
bromide. HyO was prepared from NaOH solution and showed bands at 330
and 567 cm- .
Corrosion and film formation on Pb have been studied in aqueous
chloride 12 and sulfate 11 ,13 solutions. Reid, Cooney, et al. 12 investi-
gated Pb in chloride media in order to verify the phases predicted by
the Pourbaix diagram. They suspected the existence of the polymeric
phase [Pb(OH)ln'Cln which did not appear in the diagram. They studied
films formed at different pH's as a function of applied potential in
order to confirm their theory; they then proposed the inclusion of the
phase [Pb(OH)ln'Cln in the construction of the Pourbaix diagram of the
Pb/H20 system. Investigations with a similar objective were undertaken
by Thibeau et al. 11 ,24 who found that the composition of surface films
on Pb, which was corroded in 0.1 M sulfate, were not in complete agree-
ment with the predictions of the calculated potential-pH diagram. The
Pourbaix diagram 25 did not predict the formation of PbO under any con-
dition; however, they obtained spectra of this phase at various poten-
tials in acid and neutral solutions and at potentials above the immun-
ity region in basic solutions. Pourbaix diagrams of Pb in the presence
of S04= have been constructed by Thibeau et al. 24 and McMahon et al. 26
for sulfate activities of 0.1 and 6.4 X 10- 4 , respectively. The latter
construction is presented in Fig. 3. As can be seen, both sulfate and
oxide surface phases are predicted to exist depending on the electrode
potential and solution pH. We studied the corrosion of Pb in 0.05 M
H2S04 solution (pH=1.6) initially13; a cyclic voltammogram is sh\n./I1 in
Fig. 4a. Typical active dissolution and passivation behavior is
observed with the various regions indicated in the figure. The
Pourbaix diagram predicts that controlled potential anodization at 0.16
V vs SHE (-0.45 V vs Hg/Hg2S04) should lead to the formation of PbS04
on the surface. The Raman spectrum observed in-situ (Fig. 4b) shows
the most intense bands of PbS04' Assignment of these bands was done by
a comparison with the spectrum of a standard sample known to be PbS04
 361

D••

0.6

0.'

PbSO.
> 0.2
ui ~.

:
:I:
Z
~
-'
< PbO
;:
Z
·0.2 ~
W
I-
"
0
IL ~O."

Po

·O.(i

-0.8

12 16

pH

Figure 3. Pourbaix diagram of the Pb-OHZO system at Z5°C for a sulfate

activity of 6.4 x 10- 4 (from reference 16).

(Fig. 4c). Confirmation of this assignment was carried out by X-ray

diffraction analysis of the surface film. Moreover, from a theoretical
analysis of the number and groups of vibrational frequencies observed,
one can also surmise that the three bands observed correspond to two of
the four modes of vibration of the 504= group that belong to different
symmetry classifications 5 ,Z7. The 504= anion is tetrahedral with Td
symmetry. Theory predicts the appearance of four groups of bands in
the Raman spectrum of 504= compounds. From Fig. 4c it is evident that
four such groups of bands are indeed observed. The splitting of the
degenerate bands VZ, V3, and V4 is apparent and is brought about by
the internal electric fields in the crystal that the ion experiences.
Further anodic oxidation of the lead electrode at a higher ~oten­
tial should produce a new phase, e.g., P-PbOZ in acid solutions. 8 On
stepping the potential from 0.16 to 1.95 V vs SHE in 0.05 M HZS04 in
our experiment, we observed the white PbS04 layer to convert to a black
coating which was identified by XRD as p-PbOz. However, the Raman
spectrum we observed was irreproducible and depended on the laser power
incident on the sample. There seemed to be an induction time of a few
minutes following laser irradiation before a spectrum could be ubserv-
ed. Similar observations were made on a standard p-PbOZ sample. Rota-
tion of the sample caused some of the peaks to disappear or to shift in
frequency. On removal of the sample from the spectrometer, we found
peach colored spots where the laser beam had impinged on the sample.
It is evident that the laser beam had been changing the sample at the
point of impact, causing some transformation of the p-PbOZ to other
forms of Pb compounds and leading to erroneous results. This is a good
362

~. (a)

(b)

(c)

Figure 4. (a) Cyclic voltammogram of Pb in 0.05 M H2S04 at 25°C (scan

rate = 10 mV/sec); (b) In-situ Raman spectrum of anodic corrosion film
on Pb at 0.16 V vs SHE; (c) Raman spectrum of standard PbS04; Ar+
laser, 488 nm line, Power ~ 280 mW, s = sensitivity scale.

illustration of a problem associated with the study of dark and strong-

ly light absorbing samples by the Raman technique. Sample heating by
the laser beam can lead to changes in chemical composition or phase
transformation of strongly light absorbing materials. It is, there-
fore, necessary to employ as Iowa laser power as possible consistent
with obtaining a reasonable Raman signal. It is also important to
obtain spectra at different laser power levels in order to ensure that
no composition or phase change has occurred.

4.2. Silver and Gold

Silver and gold have figured prominently in surface enhanced Raman

spectroscopy (SERS) investigations where they are used as substrates
for structural studies of absorbed molecules as well as for investiga-
tions into the mechanism of surface enhancement. 29 Kotz et al.,9 as
well as Hamilton and co_workers,ls investigated Ag in alkaline
solutions while work in our laboratory was carried out in acid sulfate
 363

solutions. 14 In NaOH, the cyclic voltammogram of Ag exhibits two

anodic waves at about 0.4 and 0.6 V vs Hg/HgO correspondin~ to the
formation of surface layers of AgZO and AgO, respectively. AgO was
identified by a strong band at 430 cm- l with a sensitivity of about 10
monolayers. On the other hand, AgZO could not be detected because of
laser photodecomposition which led to the formation of Ag0 9
In 0.05 M HZS04 solution,14 anodic oxidation occurs at about 0.3 V
vs Hg/HgZS04 resulting in the formation of a AgZS04 film on the elec-
trode surface (Fig. 5). Raman spectra as a function of potential are
shown in Fig. 6; the bands at 450, 60a, 960 and 1130 cm- 1 are aTain
characteristic of S04=, while the low frequency band at NZZ9 cm- is
probably due to vibration of the whole S04= group with respect to the
Ag surface or to Ag+ adatoms on the surface. Conversion of the AgZS04
film to Ag70aS03 (which is a AgO phase containing some S04= impurity)
has been observed by X-ray diffraction analysis (XRD) following anodi-
zation at about 1.3 V.
Gao et al. 30 studied gold extensively as an ubiquitous substrate
for obtaining intense SERS. They prescribed a series of oxidation-
reduction cycles at well defined potential limits in 0.1 M KCl that can
be used to generate optimum surface conditions for obtaining strong
signals. Gold is, moreover, interesting because of the relatively wide
electrochemical 'window' available for investigating electron transfer
reactions. The surface is nonetheless oxidized at fairly anodic
potentials, and the structural changes have been studied by DeSilvestro

30

".,0 20

"'z
e '" 10

~
w
or
or
=>u
u C -10
o
:r
5 -20

Figure 5. Cyclic voltammogram of Ag in 0.05 M H2S04 at 25°C. Scan

rate = 5 mV/sec, Electrode area = 0.9 cm Z .
364

o
~
(e)
" ~

(d)
..
0

~~ 5 (c) 0

D--(-~~ ( 0)

o
RAMAN SHIFT, em-I

Figure 6. In-situ Raman spectra of Ag anodized at various potentials

in 0.05 M H2S04; (a) Open circuit (OCV) = -O'Oi
V vs Hg/Hg2S04, (b)
0.18 V, (c) 0.55 V, (d) 0.75 V, (e) 1.30 V; Ar laser, 514.5 nm line,
Power = 80 mW.

and Weaver. 31 The latter work exploited the power of SERS to obtain
information on the initial stages of film formation and growth in
aqueous solutions. The authors found that during the anodic oxidation
of Au in 1 M KOH, two Raman bands are observed; one (VI) occurs at
around 400-430 cm- l which appears at about -0.25 V vs SCE, and another
band v2 from 0.2 to 0.25 V (the so-called "pre-oxide" formation region
of the current-potential, i-E, curve). Applying a more positive
potential (NO.2 to 0.25 V) corresponding to the beginning of the
voltammetric "anodic oxide" region resulted in the disappearance of vI
and the appearance of a new band, V2, at N540 cm- 1 . These bands were
assigned to Au-OH vibrations based on their frequency shifts in similar
experiments carried out using deuterated water as solvent. The model
of anodic surface oxidation of noble metals has been proposed 32 to
consist of two steps: 1) initial adsorption of °
or OH, followed by 2)
place exchange with the surface metal atoms along with further electron
transfer to form the "oxide film." DeSilvestro and Weaver assigned the
vI band to reversibly adsorbed OH- in step 1 and V2 to the "oxide film"
formed in step 2. The Vz frequency of 540 to 580 cm- 1 however, appears
to be lower than that due to surface AuZ03 (N650 cm- 1 ) as observed by
Electron Energy Loss Spectroscopy (EELS) and for bulk Au(OH)3 as
obtained by Raman Spectroscopy (N635 cm- 1 ). The charge associated with
the oxide film appears to be equivalent to or even less than one mono-
layer.

4.3. Copper and Cobalt

Copper also exhibits the SER effect; it is, moreover, of great interest
industrially as a material for heat exchanger tubes and its corrosion
behavior has been characterized electrochemically in a wide variety of
aqueous solution environments. Raman spectroscopic studies had been
 365

carried out by Hamilton et al. 15 in alkaline solution; we have

undertaken investigations in sulfate. 14 The voltammogram15 of Cu in
0.1 M NaOH shows anodic waves at about -0.40 V and 0.0 V vs SCE with
corresponding cathodic reduction waves at about -0.7 and -0.8 V,
respectively. Using an optical multichannel detector, Hamilton et al.
monitored the initial formation of CuZO near 0.05 V with Ar+ laser
excitation using the 488 nm line. For potentials more anodic than
about 600 mV a broad band appeared between 500 and 650 cm- 1 This band
was assigned to Cu(OH)Z' The formation of CuZO at -367 mV was evident
from the spectra (band at 649 cm- 1 ) after passing 6.8 mC/cm Z of ch§rge
leading the authors to conclude that they could detect a film N40 A
thick. They also found the Raman scattering intensity to be
proportional to the film thickness. The authors did not observe
scattering from the CuZO film using laser excitation with the 647.1 nm
line. Since CuZO has an optical absorption edge close to the 488 nm
line of the Ar+ laser being used they concluded that the detection of
the CuZO was by a resonance Raman effect.
The cyclic voltammogram of Cu in 1 M NaZS04 is shown in Fig. 7a.
Cu does not completely passivate in sulfate solution. 14 Film formation
occurs, however, and the spectrum in-situ at O.Z V is shown in Fig. 7b.
The bands at 145, Z14, and 644 cm- 1 for the red-orange film formed are
assigned to CuZO, the presence of which is confirmed by XRD analysis.
Interestingly, a CuZS04 film was not observed, this compound being
rather soluble in the solution.
We have also examined the composition of surface films on Co in
0.05 M NaOH solution. 16 Previous spectroscopic investigations have
consisted mainly of ellipsometry33,34 and Mossbauer spectroscopy.35,36
At pH 11.9 in 0.05 M NaOH, the Pourbaix diagram of the Co-HZO system
predicts the formation of the phases Co(OH)Z, C0304, Co(OH)3, and CoOZ
with increasing potential. 25 The current-potential curve shows an
anodic wave at about -0.5 V vs Hg/HgO assigned to the formation of CoO
and two other waves between 0 and 0.5 V which had been assigned by Sato
et al. 33 ,34 to the formation of Co(OH)3 and C0304' Our laser Raman
investigation of this system did not yield any spectrum in the primary
passive region at about -o.z V but showed a multiplet of bands in the
secondary passive region at NO.45 V. In-situ measurements showed weak
bands at 475, 515, 587, and 690 cm- 1 Comparison of these lines with
the spectra of standard samples of CoO, C0304 and Co(OH)Z led us to
conclude that the secondary passive film consists of a mixture of CoO
and C0304' This is an ingreement with Sato's finding that a bilayer
film of CoO/C0304 oxide exists at the Co surface in the potential range
that we have studied.

4.4. Titanium and Zinc

Ohtsuka, Guo and Sat0 37 studied the anodic oxide films on titanium at
various potentials both in neutral phosphate and acidic sulfate solu-
tions. Following anodic oxidation for one hour to form the film at
constant potentials of 5.6 and 10.6 V vs RHE in neutral phosphate solu-
tion (0.1 M, pH 6.9), the electrodes were examined in-situ at 1.60 V to
avoid oxygen evolution. Raman bands at 145, 400, 515 and 640 cm- 1 were
observed which correspond to the anatase form of TiOZ' Similar bands
were observed in 0.1 M H2S04 solution after anodizing for one hour at
366

300

200

'"E 100

-100
-0.5 0 0.5 1.0

(c)

.
#
#

>-
I-
Ui

.,...
z
I-

(b)
m
:! ~
~
q
;;

BOO 700 600 500 400 300 200 100 0

RAMAN SHIFT. c:m- I

Figure 7. (a) Cyclic voltammogram of Cu in 1 M Na2S04; Scan rate = 5

mY/sec, Electrode area = 1.3 cm 2 ; (b) In-situ Raman spectrum of anodic
corrosion film on Cu at 0.2 V; (c) Ex-situ spectrum; Ar+ laser, 488 nm
line. Power = 40-100 mW.

8.1 and 10.1 V then holding at 1.6 V vs RHE. They were unable to
observe Raman bands below 4 V although the film thickness measured
about 13 nm. They hypothesized that the film is still amorphous at
this potential but transforms into a more crystalline structure at
higher potentials. Thus they begin to see Raman bands at 5.6 V where
they also observed the electrode surface to have a crater-like morphol-
ogy. Strong bands due to the anatase modification were observed at 10
V where the oxide film appears to have a ripple-like morphology.
Hugot Le-Goff also reported work on thin Ti02 films which she
studied by interference enhanced Raman spectroscopy.38.39 She found
that films grown by anodic oxidation of Ti consisted of two different
crystallographic forms that depended on the formation potent.ial.
Films prepared at potentials higher than 50 V were composed of anatase
while those formed at lower potentials were microcrystalline,
consisting of aggregates of 20 to 25 atoms each having the local order
of anatase. The spectra of these two types were different; the anat.ase
spectrum showed bands at 398. 520 and 625 em-I. while the microcrystals
showed bands at 350 and 630 cm- 1 The author found that an
 367

electropolished Ti sample left in air for a long period of time

developed a film whose spectrum was the same as the thin
microcrystalline film. The author also took advantage of enhancement
of the Raman signal by interference effects in the film which varied as
a function of thickness; a 10 to zo times enhancement was obtained.
More recently Felske and Plieth 40 obtained in-situ Raman spectra
of oxide films electrochemically formed on Ti in 0.5 M HZS04' They
observed that below the formation potential of 10 V, the spectra were
clearly due to rutile; at higher potentials, however, the spectra were
neither characteristic of rutile nor anatase. Upon electrode emersion
and drying of the film, spectra due to anatase were evident. An
identical spectrum was obtained when the oxide film was formed in HN03'
In all cases, the films appeared to be crystalline. The authors felt
that differences in their results with those of others may be due to
the film formation and ageing conditions.
The corrosion and passivation of zinc have been studied mainly by
electrochemical techniques. (41) The composition of the passive film
has been disputed to be either ZnO or Zn(OH)Z' A dual layer has been
proposed (4Z,43) consisting of a loose floculent film of Zn(OH)Z which
precipitates in the prepassive region together with a more compact Zno
layer which is formed directly by a solid state mechanism of surface
oxidation. Hugot Le-Goff and co-workers ZZ ) carried out an "in-situ"
Raman spectroscopy study of the passivation of zinc in alkaline
solutions. They concluded that ZnO was the main constituent of the
passive layer on zinc throughout the potential range of their
investigations. They found no trace of Zn(OH) although they believed
that the passive layer had a dual structure. They suggested two
different mechanisms for the formation of the passive layer depending
on the potential i.e., (a) passivation induced by ZnO precipitation
from a supersaturated zincate solution, at potentials near the
dissolution region, and (b) passivation due to the growth of a ZnO
layer by a solid-state mechanism, in the passive potential region.

4.5. Nickel and Iron

Nickel and iron have been most extensively studied electrochemically

for their corrosion characteristics because of their industrial impor-
tance. Their electrochemical properties are also of great interest for
the operation of Ni-Fe and other batteries. We studied the anodic
corrosion of Ni in alkaline and near-neutral solutions 16 while
Delichere, Hugot Le-Goff, and Yu 44 reported work in acid. DeSilvestro,
Corrigan and Weaver 45 took advantage of the SERS effect on Au to study
the oxidation and reduction of electrodeposited Ni(OH)Z films on a Au
substrate. The oxidation of Ni in near-neutral and alkaline solutions
proceeds in two stages, the first leading to the formation of Ni+ Z and
the second to the formation of a higher valency state commonly thought
of as Ni+ 3 The former leads to a very thin film of Ni(OH)Z in base
and most probably to NiO in acid. No Raman spectroscopy study has
definitely identified the surface film on Ni in alkali in the passive
region but UV-visible reflectance measurements by Hahn et al. 45 have
been interpreted in terms of the species Ni(OH)Z' The film appears to
be at best a few monolayers thick. DeSilvestro et al.,45 obtained the
Raman spectrum of five equivalent monolayers of Ni(OH)Z cathodically
368

deposited on a roughened Au surface by employing the SERS technique.

Using the 647.1 nm line of a Kr+ laser for excitation, they found a
single band at 455 cm- 1 which they identified as the Ni-OH stretching
mode. Oxidation of the Ni(OH)Z gave rise to two bands at 480 and 560
cm- 1 which they identified as Ni-O (rather than Ni-OH) vibrations due
to the absence of a deuterium isotope effect. The oxidized form could
only be detected in thicker layers using 488 and 514.5 nm excitation
leading the authors to believe that the enhancement was by the reson-
ance Raman effect. The further oxidation of Ni(OH)Z on Ni surfaces, in
near-neutral and alkaline solutions, occurs in the transpassive poten-
tial region close to 0z evolution. We have studied the structure of
this higher oxide form by LRS and XRD techniques and have assigned the
two bands at 480 and 560 cm- I to a structure consisting of layers of
NiOz with intercalated HZO and ions such as K+, Na+, OH-, H+, etc.
between the layers. 16 ,47 We found that the spectra of various prepara-
tions 37 of this higher oxide, which have been referred to in the
literature as NiOOH, Niz03, Ni304 hydrates, and NiOz, yield the same
vibrational spectrum (Fig. 8) and have very similar X-ray powder
patterns (Fig. 9). We concluded that these various forms of higher
nickel oxide have the same layered NiOZ structure and differ only in
the degree of structural disorder of the NiOz slabs, the nature and
amount of intercalated ions and HZO molecules. Such a structure is
depicted in Fig. 10. We propose that the higher oxide form of Ni be
called "hydrous NiO z " instead of NiOOH as has been done for many years.
The maximum valency of Ni in such a compound is +4, although the actual
material prepared in the laboratory, most probably, consists of a
mixture of Ni,+z Ni,+3 and Ni+ 4 . UV-visible spectroscopic evidence for
the existence of Ni+ 4 has been presented by Corrigan and Knight. 50
The composition of corrosion films on Ni in acid solutions has
been studied by Delichere, et al. 44 and also in our laboratory.ZO,5Z
The former investigators identified the passive film on Ni in 1 M HzS04
solution to be NiO; in the region of secondary passivity, they assumed
that Niz03 is present. We have also obtained a spectrum of the surface
film on Ni anodized in the transpassive region in 0.05 M HzS04Z0 but
have not been able to assign it unequivocally to any oxide of Ni. How
ever, in concentrated HzS04 solution, our results 5Z indicate that the
corrosion film formed in the passive region is composed of NiS04·HZO.
This assignment was confirmed by XRD studies.
A number of studies have dealt with the in-situ identification of
the passive film on iron by LRS. 17 ,18,38,53 Thibeau et al. 54 obtained
the spectrum of a number of iron oxides which they anticipated to be
the corrosion products of iron. Their group also studied the oxidation
of Fe304 on iron and steel surfaces. 55 Nazri 53 studied extensively the
dissolution and film formation on iron in borate solutions using cyclic
voltammetry (CV) and rotating ring disc electrode (RRDE) techniques.
He used LRS in an attempt to identify the passive film on Fe. The film
showed several bands at low frequency (ZOO-800 cm- l ) which did not
match those of the known oxides or oxyhydroxides of iron. Froelicher 18
et al. employed very fast oxidation-reduction cycles (scan rate = 6
v/sec) in order to increase the thickness of oxide films on iron in a
numb~r of solutions from 0.5 M HZS04 to I M KOH. Thick films (NIOO to
ZOO A) were also produced by polarization maintained for a long time
 369

.......M SHI"T. c.-·

Figure 8. Raman spectra of various preparations of the higher oxide

form of Ni (taken from powders pressed with KBr into a pellet); Ar+
laser, 514.5 nm line, Power ~30-100 mW. (a) P-NiOOH (after Glemser)48;
(b) 1-NiOOH (after Glemser); (c) NiZ03eZHZO (after Cairns)49; (d)
Ni304exHZO (after Cairns); (e) NiOZ (purchased from Alfa).

I~
'" Ie)
. d
6~ 55 45 35 2.5 15

~ 55 45 35 2!i 15

~
.. ..
::1
65
•
1
OS

.,d
Z5
"

•• .. .. •
,d
OS Z5

., .,1 •• I
.. ..
DIFFRACTION ANGLE. 21
25
(dl~tU:1)

Figure 9. X-ray powder diffraction patterns of higher nickel oxide

preparations. (a) p-NiOOH (prepared following Glemser)48;
(b) 1-NiOOH (prepared following Glemser); (c) NiZ03eZHZO (prepared
following Cairns)49; (d) Ni304-xHZO (prepared following Cairns);
(e) NiZ03-xHZO (Aldrich,* NaCl lines)
370

0 0 0 0 0
I I I 1 I
Ni Ni Ni Ni !'Ii
I I I I
0 0 0
HOB H+ OH- Na.+
0 0 0- - -
1 I (
N Ni Ni !'Ii Ni
I 1 J I (
0 0 0 0

Figure 10. Probable structure of "hydrous Ni02" film.

(~2 days). The films were thus artificially thicker than those of the
passive films. For an iron sample polarized for 60 hours in 0.5 M
H2S04 solution, they obtained a Raman spectrum which corresponded to a
mixture of Fe203 and Fe304. Dlinnwald, Lossy, and Otto 17 appear to have
been the first to make a definite in-situ LRS identification of the
composition of the passive film on Fe in KOH solution. They observed
bands at 543 and 674 cm- 1 in-situ but only after cycling the electrode
17 times from -1.3 to 0.5 V vs SCE at a scan rate of I mY/sec. These
bands correspond to those of Fe304. Continued cycling of the electrode
(over 20 cycles) resulted in the appearance of a reddish brown "rusty"
coating which yielded bands at 251, 303, 390, 482, 552, 682, and 1003
cm- 1 . This set of bands resembled closely those for a-FeOOH. The
coating could be washed off by a strong water jet, after which the
sample appeared metallic again. The spectrum taken after washing
corresponded to that of Fe304 once again. This is taken as evidence
that the a-FeOOH is an overlayer on top of the Fe304 which Dunnwald and
co-workers believed to constitute the passive film. The existence of
such an overlayer comes about due to the oxidation of Fe+ 2 species in
the solution and simultaneous deposition on the electrode. Such a
mechanism has been proposed by Nazri and Yeager. 56 It is also
worthwhile to note here that Dlinnwald et al. found that the same
spectrum of Fe304 is obtained when the sample is allowed to stand in
ambient air for one day. Thus, they concluded that exposure to gaseous
02 was not critical in their eXQeriments.
Hugot Le-Goff and Palotta 38 ,39 reported results of in-situ LRS of
Fe in 1 M H2S04. Following 8 hours of anodic polarization, they found
a slight increase in the reflectance of the electrode surface and were
able to obtain a spectrum which corresponded to a mixture of a-Fe203
and Fe304. The film thickness was measured by nuclear microanalysis to
be about 10 nm. The authors equated the induction time needed to
observe a spectrum to the time needed to age the film to the point
where the "¥el structure developed into a mOLe crystalline fnr-TI)'
H

Thanos 9 took advantage of the SER effect exhibited hy small All,

particles to study the corrosion pLoducts of Fe and Fe-9~Cr alluy. He
electrodeposited Ag on pre-passivated Fe-Cr alloy electrodes. He found
that thicker corrosion products on Fe could be formed in HN03 by what
he termed a reductive precipitation at -0.8 V vs SeE. Fe304 was iden-
tified at stationary electrodes while Fe(OH)2 was found on rotated
ones. He observed a broadening of the Fe304 band at 680 cm- 1 upon
 371

oxidation of Fe(OH)Z which he attributed to the formation of higher

defect oxides and hydroxides.
We have also studied the corrosion behavior of iron in phosphate
solutions 57 from pH 1.6 to 11.9. Cyclic voltammetry measurements
showed mostly active dissolution in acid solutions of 0.1 M H3P04 and
NaHZP04; complete passivity is observed in alkaline 0.1 M NaZHP04 and
Na3P04 (Fig. 11). The composition of the corrosion films formed in the
latter solutions was determined by L~S to consist mainlr of Fe304
(bands at 556 and 680 cm- 1 ) with P04- (944 and 1068 cm- ) incorporated
into the lattice structure of the film (Fig. 1Z). In acidic 0.1 M
NaHZP04, multilayer film formation occurs. Open circuit corrosion
leads to the formation of a surface film whose in situ spectra are
shown in Figs. 13a and 13b after two hours and two days of standing,
respectively. The bands at 10Z0 and 1346 cm- l are due to material on
the surface. The film was examined microscopically and found to
consist of two layers: a very thin greyish-white outer layer and a
bluish undercoating, which is adherent to the electrode surface. A
composite Raman spectrum of the duplex film is shown in Fig. 13c while
that of the bluish inner layer is shown in Fig. 13d. X-ray diffraction
analysis of the latter shows the structure to be the same as that of
the mineral vivianite or Fe3(P04)Z·8HZO. The outer layer could not be
analyzed by XRD, but we believe that it is the phase Fe(HZP04)Z, in
line with predictions of the Pourbaix diagram.
Finally, in a related study Keiser et al. 58 ,59 studied the ambient
oxidation of Fe304 on iron and steel surfaces, as well as the electro-
chemical reduction of a, 7, 0, and amorphous FeOOH in aqueous sulfate
solution. They found that Fe304 was oxidized to 7-FeOOH which convert-
ed to a-FeOOH after several weeks. The electrochemical reduction of
amorphous 7- and O-FeOOH on weathering steel substrates produced Fe304'

.~
-2.0 -I.:a -1.0 -0" -0.1

~
. 0

(d)
0-
x
~ -1.0 -0.3
~
-1.0 -0.' 0 0.' 0.'

~t)

0.1 ......

(e) 1
T
-LO -0." 0 0.' -LO -0.. 0 0.'
POT[HTlAL,. •• se( (vou,1

Figure 11. Cyclic voltammograms of iron in aerated phosphate solu-

tions; (a) 0.1 M H3P04 (pH 1.6); (b) 0.1 M NaHZP04 (pH 4.4); (c) 1:1
volume mixture of 0.10 M NaZHP04 and 0.16 M NaHZP04 (pH 6.4); (d) 0.1 M
NaZHP04 (pH 8.7); (e) 0.1 M Na3P04 (pH 11.9); All scan rate = 10
mV I sec.
372

,..
.....
I-

Z
I-
Z

2~O 500 1000 1500 z~oo

Figure 12. Raman spectrum of corrosion film on iron in 0.1 M Na3P04

(pH 11.9) after potential hold at 0.4 V for 19 hours.

4.6. Metal Alloys

A number of papers have dealt with the LRS studr of surface films pro-
duced by atmospheric oxidation of metal alloys5 ,54,60,61; very little
work, however, has been done to identif~ the corrosion films formed in
aqueous solution environments. Thanos l attempted to obtain in-situ
spectrum of Fe-9ZCr alloy without much success. However, following

(e)

(d)

2!10 :500 1000 1300 2000 "0 ~oo 1000 1>00 2000
" ..... ,," SHifT (c ...·,)

Fi~ure 13. Raman spectra of U'[Cl1S iun films on iU1ll in (!.1 H N~H:,P()4'
Ar laser, 514.5 nm line, Power ~50 to 150 mW.
 373

electrodeposition of Ag on the prepassivated electrode (to take advan-

tage of the SER effect), he was able to observe Raman active bands of
thin films of oxides and hydroxides on the passivated alloy. A band
due to Fe304 was observed as in the case of Fe which became broad at
potentials greater than about 0 to 0.2 V vs SCE. Such broadening was
explained by the author to be due to lattice disorder because of random
distribution of Fe+ 2 and Fe+ 3 ions.
Thierry et al. 23 also studied the anodic corrosion films formed on
single crystals of Fe-Cr and Fe-Cr-Mo alloys using LRS in combination
with X-ray Photoelectron Spectroscopy (XPS) and nuclear microanalysis.
Films were formed in the passive and transpassive regions in 1 M KOH.
The former had a thicknp~s in the range of 5 to 6 nm and was determined
most likely to be a mixed a-(Fe,Cr)203 phase. The transpassive film
was found to be 5 to 25 nm thick and was formed in various stages. The
latter stages were identified to include the formation of 5 nm of a-
(Fe,Cr)203, followed by 10 nm of Fe-Cr spinel, and a top layer of 10 nm
7-FeOOH. They found no significant differences in spectra from corro-
sion films formed on Fe-Cr and Fe-Cr-Mo single crystals.
A study of Ni-lO% Mo alloy has been reported by Delichere et al. 44
Polarization of the alloy in 1 M H2S04 solution showed the existence of
a NiO film over a more extended range than was the case of pure Ni in
the same solution.

5. RAMAN SPECTROSCOPY STUDIES OF CORROSION INHIBITORS

The inhibition of corrosion through the addition of small quantities of

organic and inorganic substances to the solution is of great practical
importance. Considerable research activity goes on in this field 62 but
few have taken advantage of the potential of LRS in elucidating the
mechanism of corrosion inhibition. A number of theories have been pro-
posed to explain the action of inorganic inhibitors, and these theories
most probably apply to organic ones also. Among these are the chemical
reaction theory, electrochemical polarization theory, adsorption
theory, and electrostatic polarization theory.63 Whereas one mode of
action may be dominant in a certain case, it is not always a straight-
forward process to delineate this and it is likely that, quite often,
several of the mechanisms are ogerative to various extents.
Kester, Furtak, and Bevolo 4 first utilized the SER effect to
study the structure of benzotriazole (BTA) on Cu. They obtained spec-
tra of BTA at different potentials but were unable to assign the
observed frequencies to the various normal modes of vibration of the
molecules. They established that the observed signals were due to SER
scattering. They also observed a hysteresis in the Raman intensity as
a function of applied potential which they attributed to a bonding
rearrangement of the molecules in the protective surface layer. They
further noted that the structure of the surface complex of BTA with Cu
was altered by laser illumination during anodization which resuitpd in
better protective abilit~ of the coating.
Thierry and Leygraf 5 studied the adsorption of 1, 2, 4 triazole
and imidazole on microlithographically prepared Cu surfaces at the cor-
rosion potential. They compared their spectra with those on eu sur-
faces subjected to oxidation-reduction cycles (ORCs) and found the
spectra to be very similar, suggesting that the ORC had no influence on
the orientation or possible dissociation of the molecules on Cu. The
374

spectra were also similar to those in ethanol, indicating that the

molecules did not undergo any dissociation upon adsorption. They were
able to obtain an intense spectrum of the 1, Z, 4 tr~azole on micro-
lithogra~hed Cu (with particle size of ZOOo to 3000 A and periodicity
of 5500 A) at the corrosion potential of -150 mV vs SCE in 0.1 M NaCl;
they did not, however, observe a spectrum on a smooth Cu surface which
was polished with diamond paste. This demonstrated that an ORC is not
necessary to obtain SERS but microscopic surface roughness appears to
be essential.
Loo et al. 66 also studied the adsorption of some imidazoles on CU
and Ag electrodes in 0.1 M KCl (pH 7.7). They found that whereas the
Raman spectrum of frE'E' imioA7.nlp E'xhibits complex spectral features in
the region between Z500 and 3100 cm- 1 due to intermolecular hydrogen
bonding, the SER spectrum does not. Thus, they concluded that the
imidazole is adsorbed on Cu and Ag surfaces through its nitrogen atom
with the five membered ring perpendicular to the substrate; otherwise,
hydrogen bonding would occur for flatly oriented molecules giving rise
to a more complex spectrum. Lo0 67 also studied the adsorption of
thiourea on Cu and Ag electrodes in 0.1 M KCl solutions. The solution
spectrum of thiourea, SC(NHZ)Z shows bands at 409, 479, 733, 1096, and
1404 cm- 1 . Upon adsorption of the molecule on a Ag or Cu surface, the
733 cm- 1 band showed a large red shift of N30 cm- 1 and exhibited a
large reduction in relative intensity. This band was assigned to a C-S
stretching frequency; since it was the most strongly affected band, Loo
concluded that the thiourea is bonded to the Cu or Ag surface via the S
atom. X-ray, 1R, and Raman studies suggest that the electronic struc-
ture of the thiourea may be represented by equal contributions from the
resonance hybrid structures, a, b, and c, shown below (Fig. 14).

/NHZ )) fZH+ /
NHZ

-S-C

'"
S=C -S-C
\ '\
NHZ NHZ NHZ+

(a) (b) (c)

Fig. 14 - Resonance Structures of Thiourea

Loo also found a low frequency vibration at ZZ1 cm- l for thiourea
adsorbed on Ag and one at ZZ8 cm- l for adsorption on Cu; these, he
attributed to the metal adatom-thiourea complex vibration rather than
to Ag-Cl or Cu-Cl complex, respectively. The metal thiourea complex
was proposed to be AgSC(NHzlZ+Cl- and CuSC(NHZlZ+Cl-.
Rubin, Gutz, and co-workers 68 also studied the adsorption of ben-
zotriazole (BTA) on a Cu electrode as a function of potential in dif-
ferent chloride solutions at various pHs. They concluded that the
protective film on Cu in the presence of BTA is copper (I) benzotriazo-
late or CuC6H4N3 [Cu(I)BTA]. The material is polymeric with the struc-
ture shown in Fig. 15.
 375

Fig. 15 - Structure of Surface Film on Cu in the presence of

Benzotriazo1e

The CuBTA polymer film grows from a ~hemisorbed monolayer to form a

protective layer which is 3000-4000 A thick. The growth of such thick
films involves transport of Cu ions through the layer. The Cu ions are
then thought to react with benzotriazolate anions at the solid/liquid
interface to produce more polymer.
Thierry and Leygraf 69 compared the inhibitor properties of five
structurally related N-heterocyclic molecules towards the corrosion of
Cu in NaCl solutions. Polarization resistance measurements in 0.01 and
0.1 M NaCl solutions showed that the percentage inhibition efficiency
increases in the order mercapto-benzimidazole (MCBIA» benzimidazole
(BIA» methylbenzimidazole (MBIA» hydroxybenzimidazole (HBIA). Raman
spectral measurements showed all the molecules to be adsorbed on the Cu
surface without dissociation. For BTA and BIA, they found significant-
ly larger differences in frequency shifts for vibrational modes or1g1-
nating from the triazole or imidazole part of the molecule than for
modes originating from the benzene part. They interpreted this to mean
that adsorption of these molecules occurs via the triazole and imidaz-
ole part and most likely through interaction of all three nitrogen or
carbon atoms with the Cu substrate. MCBIA, HBIA, and MBIA, on the
other hand, showed no preferential enhancement in intensity or frequen-
cy shift originating from the imidazole part; therefore, they are
oriented differently than BTA or BIA. AC impedance measurements were
used to obtain double layer capacitances of Cu in 0.1 M NaCl with added
5 mM inhibitors. The measured capacitances decreased in the order BIA<
BTA< MCBIA; the values also decreased with increasing anodic potential
in the range of -700 to -100 mV vs SCE. Thus, the more efficient
inhibitors exhibited lower double layer capacitance which the authors
attributed to a thicker polymer layer. The difference in inhibition
efficiency of the five molecules investigated was explained by the
number and type of functional groups or atoms in the triazole Ol- imid-
azole ring that interact with the Cu. This interaction was thought tn
involve N-N-N and N-C-N in the case of BTA and BIA, Nand S from the N-
Hand S-H groups in the case of MCBIA, and only N from the N-H group in
the case of HBIA and MBIA. From the behavior of the molecules studied,
the inhibiting efficiency of various atoms in functional groups follows
the order S > N > o.
376

Ohsawa and Suetaka 70 studied the intensity enhancement and spec-

tral change in the Raman scattering of mercaptobenzotriazole (MCBTA)
adsorbed on a Ag electrode from 0.1 M NaZS04 solution. They concluded
that MCBTA is adsorbed on Ag in the form of ionized thiol. They ob-
served dramatic changes in Raman intensity in changing the concentra-
tion of MCBTA at 1 x 10- 8 M and 1 x 10- 6 M. They ascribed the former
to a change in orientation of adsorbed molecules and the latter to the
formation of a multilayered film. They found the largest intensity
enhancement to be for the species adsorbed in normal orientation with
respect to the surface. The addition of r-, Br-, or Cl- resulted in
the conversion of the thiol species to the thione form. They believe
that this conversion and thp nrrupation of adsorbed sites by halide
ions is related to the effect of halide in neutralizing the protective
action of MCBA.

6. HIGH TEMPERATURE RAMAN SPECTROELECTROCHEMISTRY

The extension of LRS techniques for in-situ measurements in high-tem-

perature/high,-pressure (HTHP) aqueous solution environments is of con-
siderable practical interest. Such conditions are encountered in a
number of applications, such as, studies on the stress corrosion crack-
ing of metals used in nuclear reactor coolant systems, thermodynamic
processes in geothermal environments, general and localized corrosion
in industrial boiler water and condenser systems, etc. These facili-
ties normally operate at temperatures above ZOOoC and pressures over
1000 psi. LRS measurements in-situ on corrosion films produced under
such environmental conditions require the construction of special appa-
ratus;71,7Z a spectro-electrochemical cell for HTHP application is
shown in Fig. 16. The cell body (A) is made of titanium for corrosion
resistance and has dimensions of 5.7 cm diameter by 7.6 cm long. It is

SECTION FRONT

A • Titanium Cell Body H • Heaters

B . Counter Electrode Port 51 • Sapphire Window for
C • Reference Electrode and Laser Beam Inlet
Electrolyte Exit Port 52' Sapphire Window for
o . Electrolyte Inlet Scattered Raman Light
E • Light Exit or Thermocouple Port T • Tellon Sheath on
F • Viewing Port Electrode Lead
G . Shaft Seal . W . Working Electrode

Figure 16. Cell for in-silu laser Raman spectroscopy and electrochemi-
cal studies in high-temperature/high-pressure aqueous solution
environments.
 377

bolted to the end flanges "X" and "yo with Teflon O-ring seals between
the parts. Provision is made to admit the laser beam through a port
fitted with a 0.95 cm dia. by 1.11 cm thick sapphire window (51); the
inelastically scattered light is collected through a second sapphire
window (5Z) which is Z.Z2 cm dia. by 0.95 cm thick. The corrosion
sample is normally prepared in the form of a 0.47 cm dia. disc and is
used as the working electrode (W). In one study,26,71 for example,
lead was deposited on a nickel disc by splat cooling from a molten lead
bath. The electrode connection is made through a 0.16 cm nickel rod
sheathed in heat-shrinkable Teflon (T); it is held inside a 0.635 cm
cylindrical shaft which can be moved back and forth or rotated through
the use of a especially designed shaft seal (G). These movements are
necessary for proper alignment of the electrode surface with respect to
the laser beam. The counder electrode is introduced through either
port B or E; we have used a loop consisting of several turns of Pt.
wire. Either one of ports B or E is fitted with a thermocouple for
temperature measurement. A silver-silver sulfate (Ag/AgZ504) reference
electrode is introduced through port C. This electrode is constructed
in a manner similar to the Ag/AgCI quasi-reference described by
Danielson. 73 It has a potential of 0.041 V with respect to Hg/HgZS04
at 25°C. We have calibrated it by comparison with a Ag/AgCI reference
up to a temperature of Z90oC.
The cell is heated using two "Calrod" heater cartridges (H) which
are imbedded in the cell body. Two thermocouples are similarly mount-
ed. All seals (black shading in Fig. 16) are made of Teflon; metal-
reinforced Teflon rings, obtained commercially (Bal-Seal Engineering
Co., CAl, are used for the shaft-seal and to seal the sapphire windows.
A viewing port (F) fitted with a sapphire window is also provided to
facilitate positioning of the working electrode and the cell inside the
illumination chamber of the spectrometer.
The electrolyte solution is refreshed in order to prevent buildup
of corrosion products inside the cell; it flows through the cell,
entering at port D and exiting at C. The recirculation system is shown
in Fig. 17. The electrolyte solution is prepared in a stainless steel
reservoir (R) and deoxygenated by sparging with high purity nitrogen.

"[CT~O·[1..tCT"OCl4[WICAl CHI..
1..0" PIIUSU'U;
RnUl.&TOIII "('Ul.I.TO"

'''[SSUIU "ELI"
(10,.1, ,

: GAS -'"''
D
Z

,..
;' S S TAn
01

UCI( ,,,tssuJt[ .EGUI,. ... TO..

un AT 1100 ,.,,)
I 1II0 ... ' .... 1.[
'"T
I
... ________ ~-rl
r-srA .. O

\"') r~
T.
INSTRUMENTS

Figure 17. High-temperature/high-pressure spectroelectrochemical

facility.
378

The solution is then pumped into the cell using a high-pressure pump
(P) after passing through a Ti block preheater (Q). The effluent from
the cell is cooled by passing it through a heat exchanger (M) and then
discarded into the drain. The spectro-electrochemical cell and Ti
block preheater are mounted on a movable cart fitted with a screw drive
system (N) which allows positioning of the cell in the illumination
chamber of the Raman spectrometer. The electrochemical equipment and
Raman instrument used have been described previously16 and are shown
schematically in Fig. 2.
Initial testing of the HTHP facility was undertaken through a
study of the anodic corrosion of lead in 7 X 10- 4 M NaZS04 (i.e., 100
ppm NaZS04 solution, RH
= 6.1). The choice of this system was based on
our previous studies 1 which showed that the corrosion of lead results
in the formation of a surface layer of PbS04 that has a large cross
section for Raman scattering. Other studies 74 also indicated that the
PbS04 film is stable at high temperatures. The low concentration (100
ppm) of NaZS04 was designed to simulate the level of impurities often
present in reactor coolant water. Fig. 18 shows cyclic voltammograms
of Pb in 7 X 10- 4 M Na2S04 (02-free) at Z5°C and Z80oC. It is evident
that typical active dissolution-passivation behavior is observed at
both temperatures; the voltammetric behavior is basically the same at
intermediate temperatures. Reversibility is apparent from the presence
of the corresponding cathodic wave. Fig. 19 shows Raman spectra
obtained in-situ at Z5 and Z800C for the passive film formed by con-
stant potential anodization at -0.10 V vs Ag/AgZS04' Assuming that all
the charge under the cathodic wave in Fig. 16 is due to the reduction
of the passive film, we estimated the film thickness to be on the order
of 1-3 #m. Based on the observed vibrational spectrum, the anodic
corrosion film is identified as PbS04, consistent with the spectrum of
this material observed in our previous work. 13 The same spectra are
obtained upon anodization at intermediate temperatures. We observed no

280'C

.• 25'C

~a

'OT[NTlAl YI ~,/A<lZSO•• V

Figure 18. Cyclic voltammogram of Pb in 7 x 10- 4 M NaZS04 at Z5° and

280°C. Scan rate ~ 1 mY/sec.
 379

Z
-1

'"Z ...
'"-1 .,...
-<
25 ·C

o
RAMAN SHIFT, em-I

Figure 19. In-situ Raman spectra of passive film on Pb anodized at

-0.1 V vs Ag/AgZS04'

'"

,..
I-
U>

...
Z

I-
~~~-~
::

50 'c

o zoo 400 600

RAM AN SHIFT. em-I

Figure ZO. In-situ Raman spectra of corrosion films on Pb formed at

open circuit.
380

significant shift in band frequencies but found a steady decrease in

signal-to-noise ratio as the temperature was raised from 20 to 280°C.
We have also observed the formation of a surface film on corrosion
at the open-circuit voltage (OCV) of about -0.96 V (Fig. 20). The
Raman s~ectrum showed bands due to S04= vibrations, as well as one at
147 cm- , which we attributed to Pb-O stretching. X-ray diffraction
study of corrosion films scraped off the electrode surface indicated
that the film consisted of the phase 3PbO·PbS04·H20 (or Pb403S04·H20).
This film persisted up to about 110°C. Above this temperature we found
the phase PbO·PbS04 (or Pb z OS04) to predominate as exemplified by the
bands at 143 and 972 cm- 1 (Fig. 20).
Following successful demonstration of the operation of the HTHP
facility, a study was subsequently conducted on the corrosion and pas-
sivation behavior of Ni. an important component of stainless steel and
other alloys used in reactor coolant systems. Fig. 21 shows cyclic
voltammograms if Ni in 7 X 10- 4 M Na2S04 solution as a function of tem-
perature. Both primary (A) and secondary (B) passivation peaks are
clearly evident at temperatures below 120°C. Passivity breaks do~~
above this temperature and the metal actively corrodes. The hysteresis
in the current-potential curve above 200°C is interpreted to indicate
the occurrence of pitting corrosion. Fig. 22 shows in-situ Raman spec-
tra obtained by potentiostating the specimen in the secondary passiva-
tion region (about 0.55 V vs Ag/Ag2S04)' The spectra observed corres-
pond to that of "hydrous Ni02" as we have shown in previous works. 20 • 47

Figure 21. Cyclic voltammograms of Ni in 7 x 10- 4 M NaZS04 as a

function of temperature. Scan rate = 10 mY/sec.
 381

RAMAH SHIFT, t:m~1

Figure ZZ. In-situ Raman spectra of secondary passive film on Ni in 7

x 10- 4 M NaZS04 at various temperatures.
° Attempts to
The film thickness is estimated to be about 100 to ZOO A.
obtain a spectrum of the film in the primary passivation region were
not successful, probably because of the thinness of the film and its
low Raman scattering cross-section.
We have demonstrated that it is possible to fruitfully undertake
in-situ laser Raman spectro-electrochemical investigations of corrosion
and passivation phenomena on metals in aqueous environments at high
temperature and pressure. It is possible to extend such measurements
to the critical point of water (374°C) or even to supercritical condi-
tions by using appropriate seals. The cell could be modified slightly
for UV-visible reflection spectroscopy. Moreover, by replacing the
sapphire window (SZ) with a metal disc counter electrode, a more uni-
form current distribution could be achieved in purely electrode-kinetic
measurements using fast transient techniques. The working electrode
must also be modified accordingly for such studies. We have, for
example, carried out AC impedance studies of Ni in 7xlO- 4 M NaZS04 at
temperatures up to ZOOoC with the present system.

7. LIMITATIONS OF THE RAMAN SPECTROSCOPIC TECHNIQUE

While we have highlighted the advantages of the LRS technique for stud-
ies in corrosion and electrocatalysis, it is also important to point
out the limitations. There is plenty of room for improvement in detec-
tion sensitivity in order to enable the study of corrosion and
passivation films at near monolayer levels. This is inherently
382

difficult for corrosion films because the metal dissolution process

itself causes roughening of the metal surface and thus increases the
background of diffuse scattered light. For oxide and hydroxide films,
the low frequency of the metal-oxygen vibration (generally less than
1000 cm- l and hence closer to the Raleigh line) compounds the problem
of retrieving a weak signal from a large background. Moreover, metal
oxides have also relatively low cross sections for Raman scattering.
Black and dark colored samples generally have a high absorption
coefficient for visible light and hence exhibit weak Raman scattering.
Sample heating by the incident laser can be minimized by spinning the
sample or through the use of other similar techniques. It is also very
important experimentally to check that the Raman spectra obtained are
not due to the decomposition products of photoreactive samples. One
way of doing this is to obtain spectra at different laser power levels
and to use the lowest power possible on the sample.
The laser beam used to probe the sample can cause other structural
and compositional changes on the sample, e.g. the drying of normally
hydrous surface films, sample decomposition, etc. The need to deter-
mine film compositions in-situ cannot be overemphasized and spectra
obtained ex-situ should always be interpreted with caution. 40 ,S7
Another problem encountered in the use of LRS for the identifica-
tion of corrosion films on metals is the lack of suitable standard com-
pounds for comparing the spectrum of the unknown film. Theoretical
methods for the analysis of vibrational spectra are not sufficiently
advanced to allow the assignment of spectral bands to definite
compounds from first principle calculations. 76 It is often necessary
to synthesize reference compounds based on an educated guess of what
may be reasonably expected on the basis of thermodynamic considera-
tions. Tedious chemical manipulations can be required. It is,
moreover, necessary to check the composition of the standard using x-
ray diffraction and elemental chemical analysis.

8. FUTURE OUTLOOK

We have given here a survey of the applications of Raman spectroscopy

in conjunction with electrochemical studies in aqueous corrosion. The
survey is by no means complete and will undoubtedly become quickly
outdated. The author anticipates more and more use of the technique in
electrochemistry. New developments in detector technology and advances
in other aspects of Raman instrumentation, such as better
monochromators and filters, will enhance sensitivity to the region of
monolayer and submonolayer quantities of corrosion films and
inhibitors. Such capability has already been demonstrated for the
adsorption of cyanopyridine on Rh. 7S The phenomenon of SERS will
continue to be applied to appropriate systems using Au. Ag. and Cu
electrodes. The development of Fourier transform Raman spectrometers
should bring about significant reduction in the cost of Raman
facilities and hence encourage wider use of the technique. The
combination of FTIR and Raman spectroscopy in a single instrument is
becoming popular and would represent a powerful capability for
electrochemical and surface science laboratories.
 383

9. ACKNOWLEDGEMENT

The author acknowledges support of his research by the U.S. Department

of Energy, Office of Basic Energy Sciences, Division of Materials
Science, under contract W-31-109-Eng-38. He also thanks his colleagues
and students, among them w. Ruther, J. J. McMahon, S. Xu, X. Feng,
W. Paden, C. Rios, W. Walczak, T. Tipton, N. Camillone III, L. Pena-
Cordova, R. Grubbs, and J. Solis-Fernandez, for their help in the
conduct of laboratory experiments. The continued encouragement of V.
A. Maroni and D. M. Gruen at Argonne National Laboratory is gratefully
acknowledged.
384

10. REFERENCES

1. M. C. Tobin. Laser Raman Spectroscopy, Wiley Interscience, New

York (1971).

2. P. R. Carey, Biochemical Applications of Raman and Resonance

Raman Spectroscopies, Academic Press, NY (1982).

3. M. Fleischmann, P. J. Hendra and A. J. Mcquillan, J. Chern. Soc.

Chern. Communications (1973), 80.

4. A. Campion, Vi~rational Spectroscopy of Molecules on Surfaces,

J. T. Yates Jr. and T. E. Madey, editors, Plenum Press, NY,
(1987). p. 345.

5. K. Nakamoto, Infrared and Raman Spectra of Inorgan~c and

Coordination Compounds, Third Edition John Wiley, NY (1978)

6. R. K. chang, in NATO Advanced Study Institute Series:

Spectroscopic and Diffraction Technigues in Interfacial
Electrochemistry, C. Gutierrez and C. A. Melendres, (editors),
Kluwer Publishers, in press.

7. A. Campion and W. H. Woodruff, Anal. Chern. 59, 1299A (1987).

8. D. K. Veirs, V. K. F. Chia, and G. M. Rosenblatt, Applied Optics,

26, 3530 (1987).

9. R. Kotz and E. Yeager, J. Electroananl. Chern. 111, 105 (1980).

10. J. DeSilvestro and M. J. Weaver, J. Electroananl. Chern. 209, 377

(1986).

11. R. J. Thibeau, C. W. Brown, A. Z. Goldfarb and R. H Heidersbach,

J. Electrochem. Soc. 127, 37 (1980).

12. E. S. Reid, R. P Cooney, P. J. Hendra and M. Fleischman, J.

Electroananl. Chern. 80, 405 (1977).

13 R. Varma, C. A. Melendres and N. P. Yao, J. E1ectroanan1. Chern.

207, 1416 (1980).

14. C. A. Melendres, S. Xu and B. Tani, J. Electroananl. Chern. 162,

343 (1984).

15. J. C. Hamilton, J. C. Farmer, and R. J. Anderson. J. Electrochem.

Soc. 133, 739 (1986).

16. C. A. Melendres and S. Xu, J. Electrochem. Soc. 131, 2239 (1984).

17. J. Dunnwa1d, R. Lossy and A. Otto in Passivity of Metals and

Semiconductors, M. Froment (editor), Elsevier Science Publishers,
Amsterdam (1983) p. 107.
 385

18. M. Froelicher, A. Hugot Le-Goff, C. Pallotta, et al. - ibid. p.

101.

19. I.C.G. Thanos, Electrochimica Acta 31, 811 (1986).

20. K. Carr, S. Baer and C. A. Melendres, Corrosion 42, 307 (1986).

21. J. DeSi1vestro, D. A. Corrigan and M. J. Weaver, J. Electrochem.

Soc. 35, 885 (1988).

22. A. Hugot Le-Goff, S. Joiret, B. Saidani and R. Wiart, J.

E1ectroana1. Chem. 263, 127 (1989).

23. D. Thierry, D. Persson, C. Leygraf, et.al., J. Electrochem. Soc.

132, 305 (1985).

24. R. J. Thibeau, C. W. Brown, et. al., J. Electrochem. Soc. 127,

1913 (1980).

25. M. Pourbaix 'Atlas D'equilibres Electrochimiques' , Gauthiers-

Villars, Paris (1963) p. 489.

26. J. J. McMahon, W. Ruther, and C. A. Melendres, J. Electrochem.

Soc. 135, 557 (1988).

27. W. G. Fateley, F. R. Dollish, et.al Infrared and Raman Selection

Rules for Molecular and Lattice Vibrations: The Correlation
Method, Wiley Interscience, New York (1972).

28. J. Burbank, A. C. Simon and E. Willihnganz in Advances in

Electrochem. and Electrochem. Enginnering, edited by C. W. Tobias
and P. Delahay, vol. ~, (1971) p. 157.

29. R. K. Chang and T. E. Furtak, editors, Surface Enhanced Raman

Scattering, Plenum Press, NY (1982).

30. P. Gao, M. L. Patterson, et al., Langmuir, !, 173 (1985).

31. J. DeSilvestro and M. J. Weaver, J. E1ectroanal. Chem. 209, 377

(1986).

32. B. E. Conway and H. Angerstein-Kozlowska, NBS Publication 455,

107 (1976).

33. N. Sato and T. Ohtsuka, J. Electrochem. Soc. 125. 1735 (1978).

34. K. Kudo, N. Sato and T. Ohtsuka in 'Passivity of Metals', R. P.

Frankenthal and J. Kruger, (editors) the Electrochem. Soc.
monograph series, Princeton, New Jersey (1978) p. 918.

35. H. Leidheiser Jr, G. W. Simmons and E. Kellerman, J. Electrochem.

Soc. 120, 1516 (1973).
386

36. G. W. Simmons, E. Kellerman and H. Leidheiser Jr, J. Electrochem.

Soc. 123, 1276 (1976).

37. T. Ohtsuka, J. Guo, and N. Sato, J. Electrochem. Soc. !1l, 2473

(1986).

38. A. Hugot Le-Goff and C. Pallotta in Proceedings of the Tenth

International Conference on Raman Spectroscopy, W. L. Peticolas
and B. Hudson, (editors) University of Oregon Press, Eugene,
Oregon (1986) p. 6-16.

39. A. Hugot Le-Goff and C. Pallotta, J. E1ectrochem. Soc. 132, 2805

(1985) .

40. A. Felske and W. J. Plieth, Electrochimica Acta 34, 75 (1989).

41. J. McBreen and E. J. Cairns, in H. Gerischer and C. W. Tobias,

(eds); Advances in Electrochemistry and Electrochemical
Engineering, Vol II, P. 298, John Wiley, N.Y. 1978.

42. R. W. Powers and M. W. Breiter, J. Electrochem. Soc. 116, 719

(1969) .

43. M. C. H. McKubre and D. D. MacDonald, J. Electrochem. Soc. 128,

525 (1981).

44. P. Delichere, A. Hugot Le-Goff and N. Yu "Proceedings of the

Tenth International Conference on Raman S ectrosco ,W. L.
Peticolas and B. Hudson, (editors)
Eugene, Oregon (1986) p. 5-45.

45. J. DeSilvestro, D. A. Corrigan and M. J. Weaver, J. Phys. Chem.

90, 6408 (1986).

46. F. Hahn, B. Beden, J. J. Croissant and C. Lamy, Electrochimica

Acta 31, 335 (1986).

47. C. A. Melendres, W. Paden, B. Tani and W. Walczak, J.

Electrochem. Soc. 134, 762 (1987).

48. O. Glemser and J. Einerhand, Z. Anorg. Chem. 261, 26 (1950; Z.

Elektrochemie 54, 302 (1954).

49. R. W. Cairns and E. Ott, J. Amer. chem. Soc. 55, 527 (1933); 55,
534 (1973).

50. D. A. Corrigan and S. L. Knight, J. Electrochem. Soc. 136, 613

(1989) .

51. R. Heidersbach, C. Brown, R. Thibeau and A. Goldfarb, Technical

Report No.4, Contract No. NOOOI4-76-C-0889, Office of Navel
Research, Metallurgy Program - code 471, Aug. 1978.
 387

52. C. A. Melendres and B. S. Tani, J. Electrochem. Soc. 133, 1059

(1986).

53. G. Nazri, Ph.D. Thesis, Case Western Reserve University, 1981.

(University Microfilms #3267, Ann Arbor, Michigan).

54. R. J. Thibeau, C. W. Brown and R. H. Heidersbach, Applied

Spectroscopy 32, 532 (1978).

55. J. T. Keiser, C. W. Brown and R. H. Heidersbach, Corrosion, 38,

352 (1982).

56. G. Nazri and E. Yeager, Extended Abstracts, Electrochem. Soc.

Meeting, San Francisco, CA., May 8-13, 1983; Abstract No. 54.

57. C. A. Melendres, N. Camillone III and T. Tipton, Electrochimica

Acta., 34, 281 (1989).

58. J. T. Keiser, C. W. Brown and R. H. Heidersbach, Corrosion, 38,

357 (1982).

59. J. T. Keiser, C. W. Brown and R. H. Heidersbach, Proceedings of

the Second International Conference on Environmental Degradation
of Materials, Virginia Polytechnic Institute and State
University, Blacksburg, VA, Sept. 22-23, 1981, p. 43.

60. D. J. Gardiner, M. Bowden and P. R. Graves, Phil. Trans. Roy.

Soc. Lond. A320, 295 (2986).

61. R. L. Farrow, R. E. Benner, A. S. Nagelberg and P. L. Mattern,

Thin Solid Films, 73, 353 (1980).

62. See for example, Proceedings of the Sixth European Symposium on

Corrosion Inhibitors, Ferrara, Italy, Sept. 16-20, 1985, vols. I
& II.
63. G. H. Cartledge, Corrosion 18, 316t (1962).

64. J. J. Kester, T. E. Furtak, and A. J. Bevolo, J. Electrochem.

Soc., 129, 1716 (1982).

65. D. Thierry and C. Leygraf, J. Electrochem. Soc. 133, 2236 (1986).

66. B. H. Loo, Y. G. Lee and A. Etlage in Proceedings of the Ninth

International Conference on Raman Spectroscopy, Tokyo, Japan,
Aug. 27-Sept. I, 1984, p. 538.

67. B. H. Loo, Chern. Phys. Letters 89, 346 (1982).

68. J. Rubin, I. G. R. Gutz, O. Sala and W. J. Orville-Thomas, J. of

Molecular Structure, 100, 571 (1983).
388

69. D. Thierry and C. Leygraf, J. Electrochem. Soc. 132, 1009 (1985).

70. M. Ohsawa and W. Suetaka, J. Electron Spec. and Related Phenomena

30, 221 (1983).

71. C. A. Melendres, J. J. McMahon and W. Ruther, J. Electroananl.

chern. 208, 175 (1986).

72. W. Ruther, J. J. McMahon and C. A. Melendres, in Surfaces,

Inhibition and Passivation - Proceedings of an International
Symposium, E. McCafferty and R. J. Brodd, (editors) vol. 86-7,
The Electrochem. Soc. Inc., Princeton, New Jersey (1986).

73. M. J. Danielson, Corrosion 35, 200 (1979).

74. M. Lepeintre, C. Mahieu and J. Monjou. C. R. Acad. Sci. Paris

261. 3389 (1965).

75. C. Shannon and A. Campion. J. Phys. Chern. 92. 1385 (1988).

76. E. B. Wilson Jr., J. C. Decius and P. C. Cross, Molecular

Vibrations. McGraw Hill, NY (1955).
USE OF SURFACE-ENHANCED RAMAN SPECTROSCOPY IN STUDIES OF
ELECTRODE-ELECTROLYTE INTERFACES

T. M. DEVINE
Department ofMaterials Science andMineral Engineering
University o/California
Berkeley, CA 94720, USA

Abstract. This paper describes the use of Raman Scattering (RS) and, in particular, surface
enhanced Raman Scattering (SERS) for the identification of atomic and molecular species
absorbed on the surfaces of metals immersed in aqueous electrolytes. In addition, examples are
given of the use of RS and SERS in .in situ studies of corrosion product films and passive films
fortmed on metals. The overall intent of the article is to indicate how to analyze Raman spectra.

1. Introduction

Because of the strong absorption of infrared radiation by water, IR investigation of the interface
formed by an electrode and an aqueous solution must employ cells with thin-film
electrolytes. [1 J Because water is highly transparent to optical radiation, Raman scattering
experiments using incident radiation in the visible range can be used to study the structures and
chemistries of bulk aqueous solutions and of electrode-electrolyte interfaces located within
macroscopic volumes of electrolytes. Furthennore, Raman spectroscopy may be capable of
detecting the same vibrational modes of a species as is IR spectroscopy. However, as is
discussed below, should the scattering species possess a center of symmetry, its Raman and IR
spectra will be complementary. Consequently, the Raman spectrum of a substance can provide
valuable infonnation about its vibrational modes. The major drawback of Raman spectroscopy
is that it is an inherently weak process. Typically the intensity of Raman scattered radiation is
on the order of 10-7 times the intensity of the incident radiation[2-4J. As a result Raman
spectroscopy generally requires a large amount of the scattering substance and/or incident
radiation of high intensity. Obviously, the weakness of the effect makes the detection of the
radiation that is Raman scattered by the small (monolayer) quantity of material present at an
electrode surface highly unlikely. Nevertheless, as discussed in Part II of this paper, it is
possible to detect radiation that is Raman scattered by submonolayer amounts of material.
Before describing the results of these Raman scattering experiments, the phenomena
responsible for Raman scattering will first be described. The overall intent of the article is to
infonn the reader how to analyze Raman spectra.

389
M. G. S. Ferreira and C. A. Melendres (eds.).
Electrachemical and Optical Techniques/or the Study and Monitoring 0/ Metallic Corrosion. 389-437.
© 1991 Kluwer Academic Publishers.
390

2. Theory of Raman Scattering[4-6]

2.l. CLASSIC DESCRIPTION

Paralleling the discussion of IR absorption in the preceding paper, the origin of Raman
scattering will first be considered by treating the process as a classical phenomenon. A
charged particle such as an electron will generate an electromagnetic wave when it vibrates.
The frequency of the wave will match that of the vibrating charge. The electric field associated
with an electrically-neutral particle will be proportional to its polarization p. When a molecule

is exposed to an electric field e its charges will be polarized an amount proportional to the
applied electric field.

(1)

The constant of proportionality a. is the polarizability of the molecule.

...a = ...a o + (dQ
aa ) '"
j 'lj

o (2)

0.0 is the polarizability of the molecule when each atom occupies its equilibrium
position. Qj is the jth nonnal coordinate.
If the molecule is vibrating, then the variation of the nonnal coordinate with time can be
described by
Q j = QioCOS (Om t
(3)

For a molecule exposed to a time-varying electric field of the fonn

(4)

its polarization (1) will be given by

P = rno + (:~J Q) . eocos (Oi t

o (5)
 391

Since the electric field associated with an electromagnetic wave that is generated by a
vibrating molecule is proportional to the polarization of the molecule, (6) indicates that in
general three different waves will be scattered. The first has a frequency which matches that of
the incident radiation. The other two waves have frequencies that are shifted from that of the
incident radiation. These are inelastically scattered, and are referred to as Raman scattered
radiation. Note that the magnitude of the shift in frequency of the Raman scattered radiation is
± com' the vibrational frequency of the scattering molecule. Thus, the Raman spectrum can
serve as a fingerprint for identifying unknown substances.
Note also that in order to have Raman scattered radiation, there needs to be at least one
component of the polarizability tensor whose derivative with respect to the normal coordinate of
the molecule has a non-zero value. The criterion of non-zero differential polarizability for
Raman activity may be compared to the criterion of changing dipole moment with molecular
vibration for infrared activity.

(:~J ;to
o for Raman activity

(;;) ;to
J 0 for infrared activity (8)

In the discussion of infrared absorption it was noted that whether or not a particular
vibrational mode of a molecule met the criterion for infrared activity (8) could sometimes be
determined by inspection for those cases in which the geometry of the molecule was simple.
Whether or not a vibrational mode of a molecule satisfies (7) is usually more difficult to
determine by inspection even in those cases in which the geometry of the molecule is simple.
However, it is worth recalling that the polarizability of a molecule is in part a function of its
volume. Hence, any vibrational activity which changes the volume of the molecule may satisfy
(7).
392

Figure 1 compares the infrared and Raman activities of the 4 normal vibrational modes (3N-
5==4 for N==3) for the linear molecule CO 2 , Again, for a simple structure such as that of CO 2'
the infrared activity can be ascertained by considering whether or not the vibrational modes
cause a change in the dipole moment The Raman activity can be approximately ascertained by
considering whether or not the volume of the molecule changes during the vibration. Note that
the modes that are IR-active are Raman-inactive and vice-versa. This is true of all molecules
with a center of symmetry.
Just as in the case of infrared absorption, Raman scattering involves the interaction of
electromagnetic radiation and matter. Consequently, Raman scattering is a quantum mechanical
process that can well be described by the formalism of time-dependent perturbation theory just
as was the case of infrared absorption.

IR-ACTIVE RAMAN ACTIVE

o
~ ......... ~ .....--..-

0 0 YES NO

0
...... ~ ~

0
......
o
.......... ~

YES NO

9 9
t
0 NO
VES

(0 (0 g YES NO
XBL 904-1241

Figure l. Infared and Raman activities of the normal modes of CO2 .

 393

2.2. SEMI CLASSICAL DESCRIPTION AND THE SELECfION RULES FOR RAMAN
ACTIVITY

The Hamiltonian for a system consisting of a molecule exposed to an electromagnetic field is

aq, a2q , .
H = Ho + 'h:, q, - 4,P i £ - ~R ij ax. ax. - higher order terms
1 1 I,) 1 J (9)

Ze = net charge on the molecule,$ (x,y,z) = potential at (x,y,z) and HO =Hamiltonian

of the molecule when $ = O. Neglecting higher order tenns, the Hamiltonian for a neutral
molecule may be approximated by

(10)

(11)

This represents a semiclassical or or partial quantum mechanical treatment as the molecule is

described by quantum mechanics but the incident radiation is treated classically, i.e., light is
described by E = Eosin ro t rather than by photons.
As in the analysis of infrared absorption, the question of whether or not the incident radiation
will induce a vibrational transition that will be Raman active is detennined by the use of time
dependent perturbation theory. Substituting (10) and (11) into (47) of the previous paper[7]and
integrating gives

_ H 'vg { exp [i(ro -rogv)t] -1 exp [ - i(ro -rogv)t] -I}

am(t) - h .(
lro-rogv ) + .(
lro+rogv ) (l~

(13)

where H' is the time dependent portion of H and, from (10) is given by P E.
Thus, a vibrational mode will be Raman active if H'vg has a non-zero value. In other words
394

a vibrational mode will be Raman active if the integrand 'ifv ex 'If g spans A 1.
As an example consider the three normal modes of vibration of H 20. These have the
symmetries A l' A 1 and B 1.[7] The polarizability of the molecule contains terms such as x 2,
y2, z2, xy, yz and xz. The character table for C 2v indicates thatx 2, y2 andz 2 each span A l'
Since the ground state also spans A I' the two vibrational modes with symmetry A 1 will be
Raman active. The remaining vibrational mode of symmetry B 1 will be Raman active only if a
spans B l' The character table for C 2v indicates that xz spans B l' Hence, all three vibrational
modes of water may be Raman active.
The preceding paper[7] on infrared spectroscopy described the use of group theory to identify
the symmetries of the vibrational modes of a molecule. It was also pointed out that the
frequencies and amplitudes of the vibrations of a molecule were most conveniently
characterized in terms of the normal modes of vibration. The discussion there centered on the
behavior of CO, a molecule whose structure is so simple that its single vibrational mode is
determined by inspection. In this paper, the discussion of experimental results will focus on
the Raman spectra of water. Considering its importance in corrosion reactions, much
information about the mechanism of dissolution may be obtained from information about the
concentration and structure of water at the surface of a corroding metal. Because the structure
of H 2 0 is more complicated than that of CO, its vibrational modes are not immediately
apparent. Therefore, in keeping with the tutorial nature of the paper a procedure for
determining the frequencies and amplitudes of the normal modes of more complicated
molecular structures will be described. Together, the discussions in the two papers will
provide a complete deSCription of the analytical treatment of the vibrational motion of a
molecule. Hopefully, this will enable the reader to determine the frequencies and amplitudes
associated with the normal modes of vibration of a molecule as well as enable the reader to
appreciate what information the vibrational spectra of species present at an interface can provide
with respect to the orientation of the species as well as the bonding within and between such
species.

3. Determination of the Number and Symmetries of Normal Modes of

Vibration [8-11]

The motion of a molecule consisting of N atoms may be completely specified by 3N

coordinates. The location of each atom is specified by the x, y, and z values of its position
measured with respect to some origin. If the only interest is in the "internal" motions of a
molecule, i.e., the motion of each atom with respect to its equilibrium position in the molecule,
then the number of coordinates required to specify the motion of interest is 3N - 6 (for a non-
linear molecule) or 3N - 5 (for a linear molecule). The reduction in the number of coordinates
is achieved by removing the three coordinates required to specify the rigid body translation of
 395

the molecule and the three coordinates required to specify the rigid body rotation of a non-linear
molecule. Two coordinates are required to specify the rigid-body rotation of a linear molecule.
As an example, consider the AB molecule used in the treatment of a molecule as a harmonic
oscillator c.f., author's previous paper[7]. In this case 3N - 5 = 1. One coordinate, the length
of the AB bond, is required to specify the internal motion of the AB molecule. In
H20, 3N - 6 = 3. Three coordinates, e.g., the lengths of the two O-H bonds and the H-O-H
angle, completely specify the internal motion.
A normal vibrational mode is an internal mode of the molecule in which all atoms vibrate
about their equilibrium positions with the same frequency, pass through their equilibrium
positions at the same time and reach their maximum displacements from their eqUilibrium
positions at the same time. In other words, a normal mode consists of the simple harmonic
oscillation of the normal coordinates, Qi' The normal coordinate Qi describes the displacement
of each atom from its equilibrium position during the i th vibrational mode. The entire trajectory
of a given atom occurs along a single straight line which passes through the eqUilibrium
position of the atom. The Q i are linearly related to the mass-weighted Cartesian displacement
coordinates of the atoms.

3N

Q i = ~)ikqk
q=1 (14)

The lik are chosen such that the expressions for the internal kinetic and potential energies of the
molecule are given by

3N-6. 2
2T= L,Q i
;=1 (15)

3N-6

and 2V = ~)iQi
i=l (16)

The usefulness of the normal coordinates stems in part from the simplicity of the expressions
for kinetic and potential energies in (14) and (15). In addition, the normal coordinates form an
orthogonal basis and any vibration can be represented by a linear combination of the normal

°
modes. Thus, the analyses of the vibrational spectra of a substance is greatly facilitated by
knowledge of the normal modes of the species. As an illustration, the normal modes of H2
will now be identified.
The process for identifying the normal vibrational modes of a molecule begins with a
396

determination of the symmetry of the molecule. The point group symmetry of H 20 is C 2v

Using the mass weighted Cartesian displacement coordinates of each atom as a basis, and
applying each symmetry operation of C 2v to the basis results in a 9x9 matrix representation of
each symmetry operation. In fact, this was illustrated in the previous paper.[7] Next, the
reducible, 9x9 matrix representations are reduced to block diagonal form through the use of
equation (94) of the previous chapter. The final result provided in equation (96) of reference
[7] is

(17)

That is, the representation of C2v provided by the basis consisting of the mass weighted
Cartesian displacement coordinates is equivalent to the sum of the irreps (note: "irrep" stands
for "irreducible representation") listed in the last equation. The words "equivalent to" in the
previous sentence mean that the multiplication of the 9x9 matrix representing a symmetry
operation R in the point group C2v and the mass weighted Cartesian displacement basis will
yield the same product as the multiplication of the 9x9 ~ matrix representation of the
symmetry operator (i.e., that consisting of the block diagonal matrix

D(A'\R)
D(AJ(R)

D( AJ(R)

D(A,)CR) 0
o D(BJ(R)

D(B'\R)

D(BJ(R)

D(B')CR)

D(B')(R)
(18)
 397
A,
with the mass weighted Cartesian displacement basis where D (R) =mauix representation
of the operator R in the irrep A 1. The importance of this equivalence is that since the mass
weighted Cartesian coordinates (14) are linearly related to the nonnal mode coordinates both
coordinate systems span the same symmetry species. That is, r(Qi)= 3A 1+A2+3Bl +2B2.
Proof of this important relationship is given in Appendix I.
The value of knowing r<Qi) is as follows. Since the action of a symmetry operator of a
point group on a non-degenerate nonnal coordinate must simply reproduce that nonnal
coordinate (i.e., cannot yield a sum of the initial nonnal coordinate and others -- otherwise the
potential energy of the molecule would change as a result of the action of the symmetry
operation -- see Appendix II) then each non-degenerate nonnal coordinate must exhibit the
symmetry of one of the I-dimensional irreps contained in r<Qi). Similarly, since the
application of a symmetry operation on a degenerate nonnal coordinate may yield a combination
of this degenerate coordinate plus other coordinates which share the degeneracy without
causing a change in the potential energy of the molecule, then each n-degenerate normal
coordinate must exhibit the symmetry of one of the n-dimensional irreps contained in r(Qi).
Schematically, this may be represented by

(19)

Block-diagonalized Nonnal coordinates

form of symmetry
operator

That is, each non-degenerate normal coordinate serves as a basis for one, one-dimensional
irrep. Each n-degenerate normal coordinate combines with the other (n-l) coordinates which
share the degeneracy to serve as a basis for an n-dimensional irrep.
398

In the case of H 20, for example, the fully reduced representation consists (as shown in (18))
of a sum of 9 one-dimensional irreps. Each is spanned by a normal coordinate. Three normal
coordinates will describe the rigid body translation of the molecule and three will describe the
rigid body rotation. The character table for C2v indicates that the coordinates describing the
rigid body translation span the irreps AI' Bland B 2' The coordinates describing the rigid
body translation span the irreps A 2 , Bland B2' Consequently, the remaining three irreps, 2A 1
+ B2 , are spanned by the normal vibrational modes of H2 0. Again, since each is a one-
dimensional irrep, each is spanned by a single, non-degenerate vibrational mode. Thus, H20,
has three normal vibrational modes (in agreement with 3N-6), two of which span A 1 and the
other spans B 2'

4. Specification of the Normal Modes

The next task is to identify the frequencies and displacement vectors of these three normal
modes. It turns out that a description of the vibrational modes of a polyatomic molecule is
more readily obtained by solving Lagrange's equation of motion rather than Newton's. Of
course, the two approaches provide an identical description. Lagrange's equation of motion is

J!..aL _ aL - 0
dt ar 1
Or.-
1 (20)

where L =KE - PE and r i =i th coordinate of the molecule. Thus, to solve (20) the kinetic
(KE) and potential energies (PE) of the molecule must be expressed as functions of position of
the atoms in the molecule.

While Cartesian coordinates may at first seem to provide a convenient system for describing
the position of atoms within a molecule, Cartesian coordinates do not necessarily provide
readily analyzable descriptions of vibrational motion and, hence, are not usually employed. In
particular the quantities

may not be readily associated with simple physical processes such as the

stretching of a particular bond. Instead, in general,

 399

[ ~J
d~id~
j 0 is the force restoring the molecule to its equilibrium configuration following a

displacement which is a combination of changes in bond lengths and changes in bond angles.
It is difficult to associate changes in this term with specific changes in the geometry of the
molecule.
Consequently, "internal coordinates" such as in the case of H 2 0 the stretching of the two 0-H
bonds, r 1 and r2' and the bending of the intramolecular angle (H -0- H), r 3' represent a better
choice of coordinates. In this case the potential energy of the system is written as

21)

d2y )
where F ij = ( ill idr i
o (22)

Therefore,
Y = -2I-F-+
r r
- (23)

where r = column vector whose components are the 3N-6 internal coordinates r' = row
vector whose components are the 3N-6 internal E = (3N-6) x (3N-6) matrix whose elements
are the coefficients in the expression for the potential energy. Written out in matrix form, the
above expression becomes, for the case of H20

(24)

In Appendix III it is demonstrated that the kinetic energy expressed as a function of the
internal coordinates is

1 .,~-l.
T= zr ~ r (25)
400

where the elements of the matrix Q are

G ij =LIl!B ikB jk
k=1 (i.j = 1.2 •...• 3N-6) (26)

where J.lk=I/mk and Bik are the direction cosines relating the internal coordinate ri to the
Cartesian displacement vectors ~.

3N

r i = LBikSik
k=1 (27)

Substituting (21) and (25) into Lagrange's equation gives

d d [I( -I)
..,,--. ] - ~I .r.)] =0
rlT L"2F 1.. r) 1
otdr k "2 G i r.r.
j') or k

{(\:. -Fr =0
:. i - G.Fr' =0 (28)

Assuming a solution to (28) of the form

(29)

and substituting into (28) gives

(30)

The solution to (30) is non-trivial provided that

det (A. ! + Q .1:) = 0 (31)

 401

Solving (31) for A. provides the frequencies of the nonnal vibrational modes. Since for the
case of H2 0 both Q and E are non-diagonal 3x3 matrices, solving (31) requires a fair amount
of algebra. If specifying the vibrational modes of the molecule of interest required the use of N
internal coordinates, then (31) would represent a homogeneous Nth order polynomial equation.
1b simplify the solution to (31) a new coordinate system is introduced. This coordinate system
....
is made use of the symmetry coordinates, t i. The symmetry coordinates fonn a basis for each
of the irreps that are spanned by the nonnal vibrational modes of the molecule. As proved in
Wilson ".,[8] the symmetry coordinates result in a factorization of the matrices Q. and E
which greatly eases the solution of (31). This will be illustrated by solving the secular
detenninant for H2 O.
Since the vibrational modes of H 20 span the irreps 2A 1 + B I' the symmetry coordinates for
°
H2 may be obtained by applying the projection operator for A I and B 1 to the internal
coordinates for H2 O. As shown in Appendix IV, applying the projection operator of an irrep to
one component of the basis of the irrep will generate the sum of all the components of the basis
of the irrep. Since, in the case of H 20 the irreps are one-dimensional, the basis will be one-
dimensional. Hence applying the projection operator to one part of the basis will generate the
entire basis function. This will serve as one of the symmetry coordinates.
The projection operators are applied to the internal coordinates because these are thought to be
part of the basis for the irreps A I and B I'

p (1) = projection operator for the I th irrep

(32)

where dt = dimension of the representation

h = number of symmetry operators in the point group
R = symmetry operator
X(l)(R) =character of the matrix of the symmetry operator R in the
llh irrep therefore,

P<A,) =t LX(A')(R)* R =t[l. E + I· Cz + I· 0" + 1· 0",]

R (33)

Similarly,
402

(B) 1
P I
4 1· E - 1· C 2 + 1·
= -[ 0' u - 1· 0' u' ]
(34)

Therefore,

(35)

(36)

Likewise,

(37)

and
P{AJ . f
3
= f
3 (38)

(39)

Where rl and r2 are the internal coordinates describing the two O-H bonds and r3 is the
internal coordinate describing the H-O-H intramolecular bond angle. The last equation
indicates that the H-O-H angle is not part of the basis for the irrep B I' Thus, the symmetry
coordinates, after normalizing, are

(40)

fl - r 2
t ----
2- v'2 (41)
 403

(42)

Consequently, the symmetry coordinates tj are linearly related to the internal coordinates Tj by

the matrix ll.

(43)

(44)

where

.lI=
[* "* 01
I
Tz7z
o 0
I 0
1 (45)

Similarly

-1..... -+
r=.ll t =.l[ t (46)

T.=
J
(U- I ) ji
t.
1
(47)

where

TzTz
I I 01
~= [ fi-* 0 =1I
o 0 1

Note the ll-I ll =!

404

From (46), the following is true: (48)

Therefore, substituting (47) and (48) into the equation for potential energy (23) gives:

(49)

where

(50)

F Il +F 22 +2F I2 F Il -F 22 F 13 + F 23
2 2 Vi
F l1 - F22 F II + F 22 - 2F 12
= 2 2
0
F 13 +F 23
0 F33
..Ji (51)

Since F 11=F22= force constant for the O-H bond and F13=F23= force constant linking
changes in potential energy to the change in the force between 0 and H due to changes in the
H-O-H angle, UPU' may be written as

(52)
 405

Similarly,

r'.Gr = t'lIill1't (53)

where

(54)

Therefore, the secular equation becomes

(55)

where
gll=G II +G I2 fll=FII+FI2
g 13 = "\1'2 G 13 and f 13 = "\1'2F 13
g22 = G ll - G I2 f 22 =F lI -F I2
g33=G 33 f 33 = F 33

{56)
406

(57)

(58)

(59)

The next task is to identify the values of g ij and fkl so that the numerical values of Am may be
obtained. Starting with the matrix .E. the elements F ij are defined as

F =
ij
;iV)
(drJ;lr j
(60)
o

F
11 (~~) j 0
= Second derivative of V with respect to one O-H bond

F22= (02VJ
or/ 0

F - ( 0
12-
\r)
or or
j 3
= Change in the force between 0 and H due to changes in the
o

H-O-H angle etc.

 407

XBL 904-1240

Figure 2. Internal coordinates of H2 0.

The elements of the Q. matrix are readily obtained once the elements B ij are detennined. For
the case of water the Bij are obtained from figure 2.
Internal coordinate #l=r 1 =change in length of O-H bond

cos (2
= q1 6) + 2
-n - -2 -6 - 3
q cos2 (n- - -6) - q cos-e
q COS2 242 (61)

(62)

(63)
408

B =
r=('t- n 0
cos '2
0
e
- COs(~ - f)
cos-( 2"-"2
n e)
-cos 1-
e
0
-cos ~ - cos(~ - f) cos !!.
o 1
'J
cos 1 - cos (-!} - ;)
0
.
-sm };
e
-cos ~ -cos(l- -~)
r r r r

(64)

Since G ij = LIl
k
kB ik B jk
(65)

GIl == LllkB ikB jk == IlIBl12 + B 122] + 1l 0[B13 2 + B 14 2]+

k
IlIB l/ + B 16 J

(66)

Similarly,

(67)

- sin ell
r '-0

(68)
 409

_ _ II sin e
- 1'"0 r

(69)

=llo+Il H (70)

21l H 211 0
= -r2- + -(I-cos
r2 e)
(71)

-Ilosin e
IlH+llo lloCOS e r
- Ilosin e
G 1).. = llocOS e IlH+ Ilo r
- Ilosin e -Ilosin e 21l H 21lo
- - + --(I- cos e)
r r r2 r2 (72)

Substituting into equation (54) gives

410

IlH + lloO + cos 8) 0 r

UGU= 0 IlH+llo(1-cos8) o
- V2llosin 8 21l H 21lo
r o - r2 + -r20 - cos 8)

(73)
The values of the parameters in the above equation are

J.l.H 5.975xl023 gm-l

J.l.o 3.764xl022 gm-I

r O.85xlO-8 an

e = 104.5 0

Assuming that the force constants Fij =Fij 0y' the F matrix (52) becomes

(74)

(aZy]
F 11 = --2 = kOH ==
dr 1 0 force constant for the O-H bond

(flV]
F33= --2
ar 3 0
=kHH==
force constant for the H-H intramolecular interaction
 411

Substituting the above two matrices into the secular detenninant gives

(75)

(76)

From Colthup ~ al.[5] the value of kOH is estimated to be 5x10 5 dyne/cm and kHH/r 2 is

estimated to be 5x104 dyne/cm. Substituting into the previous equation gives

1.= 3. 22xlO 29 S-2

At = 5. 677xlOl4 S-l
:. U =3012 em-I

Substituting the appropriate values of the parameters into the quadratic equation gives

(77)

29 2
1.= 1.8875 x10 29 ± 1. 243x10 29 = { .3.1305XlO 29sec
0.6445 xlO sec 2 (78)
412

'A,'1 = {
5.595x10 14 sec 1
2.5387 XlOl4 sec I (79)

:. U = {2968cm-1
1347cm -I (80)

Thble 1 compares the calculated and measured values of the vibrational modes for H 2 0. The

calculated values can be used to identify the peaks observed in the experimentally obtained
spectrum.[12]

TABlE 1

U~(em-1) UJn~lIs-<em -1 ) ~ (sec- 2 )

1347 1595 0.904 x 10 29

2968 3642 4.713 x 10 29
3012 3756 5.013 x 10 29

In addition, the experimental data may now be employed to calculate the values ofkOH and
k HH. For example, substituting

into (76) gives

kOH = 7.78 x 105 dyne/em

Equations (35) and (42) link the symmetry coordinates with the irreps they span. Thus t 1
forms the basis of A l' t2 is the basis for Bland t3 is the basis for B l' Thus the symmetry
exhibited by each of the vibrational modes listed in Table 1 may be identified. These are listed
in Table 2 below.
 413

TABLE 2

Characteristic
Freqyency Ccm~}

3642

3756

1595

The only task remaining is to specify the precise changes in geometry of H20
associated with each of the 3 nonnal modes of vibration. This is achieved by first detennining
the eigenvectors of (28). A matrix M is then fonned by each eigenvector of (28) serving as a
row in M Matrix multiplication of M and the internal coordinates provides the nonnal
coordinates, Qi.

The eigenvectors of (28) are obtained by substituting each of the 3 eigenvalues Ak

listed in Table I into (28) and solving for A ik. This additional arithmetic is perfonned in the
text by Graybeal [13] which may be consulted for the quantitative desCription of the
displacements of each atom during the 3 vibrational modes. These results are graphically
summarized in figure 3.
414

XBL 904-1239

Figure 3. Nonnal modes of vibration of H2 0

5. Use of SERS in Corrosion Studies

5.1. INTRODUcrION

The chapter by C.A. Melendres in this volume addresses the use of conventional Raman
spectroscopy in studies of corrosion phenomena. In this chapter, the discussion will
emphasize the use of Surface Enhanced Raman Scattering (SERS) in corrosion investigations.
As mentioned in the introduction to this paper, the intensity of Raman scattered radiation is so
low that a large quantity of the scattering substance is generally required to produce a
measurable signal. Nevertheless, because of the phenomenon of SERS, it is possible to obtain
the Raman spectra from submonolayer quantities of material present on the surface of certain
metals.[14,15] A great deal of effort, both experimental and theoretical, has been devoted to
understanding the mechanism(s) of SERS. These studies have been well summarized in a
 415

number of reviews and little attention will be given here to the mechanistic aspects of
SERS.[l6-27] Suffice it to say that there are two major hypotheses. In one, the enhancement
mechanism is opemtive only within a short distance from the metal surface and involves charge
transfer between the metal and adsorbate. In the other model, the mechanism of enhancement
is thought to extend over longer distances from the metal surface and is based on image and/or
electromagnetic effects. The short range effect is thought to require atomic scale roughness.
The charge transfer occurs between the adsorbate and active sites (adions) on the surface. The
long-range electromagnetic effect requires microscopic roughness features which permit
coupling between the incident radiation and surface plasmon-polaritons. Strong experimental
evidence in support of each of these enhancement mechanisms exists and some suggest that
both may be opemtive, to greater or lesser extents, depending on the chemistry and structure of
the metal-solution system. To date SERS in aqueous solutions has only been observed from
species located on the surfaces of electrodes of silver, copper and gold. At first glance this
would appear to preclude the use of SERS in studies of the corrosion of alloys of more
engineering significance, such as ferrous alloys. However, as described below, it is possible
to obtain in-situ Raman spectm from thin films and adsorbed species located on the surfaces of
any metal immersed in an aqueous solution by first partially coating the metal of interest with a
thin deposit of silver.

.,.
Iii.
T/OC T/OC
-I•

,...••
10
so
30 s.
10
-I.

.'.
T/OC T/oC
-I•
10
I"

.
10 .0
M so
10 '0
-10

3900 lIOO 3700 3600 3500 3400 3300 3200 3100 3000 2900 2100

Wovenumber/cm- I

Figure 4a. Isotropic (Ia) and anisotropic (II') Raman spectra of water as a function
of tempemture [29] in the region of the OH stretch mode. The intensity of the
Raman scattered mdiation is plotted on the vertical axis.
416

o~~~==~~~~~~==;=~=---~~~~o
2800 2400 2000 1600 1200 800 400 o
Wavenumber lem- I

Figure 4b. Isotropic (Ia) and anisotropic (Iy) Raman spectra of water in the region
of the vibrational, translational and bending modes. [29]

5.2. CONVENTIONAL AND SURFACE ENHANCED RAMAN SCATIERING BY WA1ER

In spite of the large number of infrared and Raman spectroscopic studies of water, [28,29] its
structure is still a subject of controversy.[30] This is illustrative of the simple fact that
calculating the vibrational spectrum of a known structure is easier that calculating an unknown
structure from its vibrational spectrum. A discussion of the points of disagreement concerning
the structure of water is beyond the scope of the present paper. Rather, only the more obvious
features of the Raman spectrum will be considered to illustrate the use of Raman spectroscopy
in providing information about the structure of a substance.
 417

The nonnal mode analysis of H 2 0 perfonned in part lost of a symmetric stretch mode VI,

an antisymmetric stretch mode V2 , and a bending mode V 3. The observed vibrational Raman
spectrum of water is consistent with this analysis and is presented in figurse 4(a,b).[28] The 3
peaks in the spectrum are identified by the method employed to generate the calculated values of
the peaks in the Raman spectra presented in Table I. That is, estimates of the intramolecular
force constants lC OH and lC rr are made. From these, values of V i are calculated. These are

then compared to the measured values of viand more accurate approximations of lC OH and

lC HH are then obtained. Note that the values of lC OH and lCHHwhich are back calculated from

the measured values of Vi are still only approximate. This is because of the approximation
incurred in equations (21) and (22) which ignores higher order tenns in the Taylor series
expansion of the potential energy. Omitting the anhannonic tenns means that interactions
between the harmonic modes are not incorporated in the model.
At any rate, the assumed values of lC OH and lC rr allow the first peak to be identified as
resulting from the bending mode of H 2 0 which spans the irrep A 1. The second peak results
from the symmetric stretching of the O-H bonds which spans the irrep AI. Finally, the third
peak results from the antisymmetric stretching of the O-H bonds and the vibrational mode
spans the irrep B 1. In addition to the anti symmetric stretching mode possessing a higher
characteristic frequency than the symmetric stretching mode, the fonner could also be
distinguished from the latter by virtue of symmetry effects. Although not discussed in this
paper, the polarization of Raman scattered radiation is a function of the spherical and
anisotropic components of the polarizability derivative tensor.[6,31] The result is that only
peaks associated with totally symmetric vibrations are polarized. Peaks associated with all
other vibrations are depolarized. Thus, polarization measurements of the Raman scattered
radiation would distinguish the anti symmetric stretching mode from the other 2 modes.
In summary, the nonnal mode analysis perfonned for H20 indicated the number and
symmetries of the Raman active vibrations. By approximating the values of lC OH and KHH , it
was possible to associate each of the 3 peaks in the spectrum with a particular symmetry.
Using the measured values of the characteristic frequencies, it is then possible to back calculate
lC lC
more accurate values for OH and HH. Measurements of the Raman spectrum can then be
used to detect changes in the values of these force constants resulting from, e.g., changes in
temperature or pressure or from the addition of solutes to detennine the influence of these
changes on the structure of water.
One striking characteristic of the Raman spectrum of water is the broad bandwidths of the
peaks. Comparison of the Raman spectra obtained for H 20 in the vapor, liquid, glassy and
crystalline states reveals that the greatest bandwidths occur for liquid H 20. Peaks associated
418

with well-defined vibrations are, in general, quite narrow. The widths of the peaks in the
Raman spectrum of water have been attributed to a large variation from molecule to molecule in
the amount and geometry of hydrogen bonds.[29]
SERS of water has only been observed on silver electrodes immersed in electrolytes
containing halide-ions (other than fluoride). Figure 5 illustrates the spectra observed in the
vicinity of the OH stretch region obtained from silver immersed in 1M KCl.[32] The most
distinctive characteristic of the OH stretch mode is its narrow bandwidth. This suggests the
absence of disorder in the H 2 0 responsible for the scattering. More specifically, the H20 is
not considered to be hydrogen bonded to nearby molecules of H 2 0. In addition, experiments
suggest that not all of the H20 molecules present near the surface of silver contribute to the
SERS spectrum.[33-35] Cycling the potential of silver from -O.lV to 1.0V and then back to-
0.1 V essentially wipes out the SERS of H20. Presumably the H 20 responsible for SERS is
associated with atomic-scale features in the silver surface that are removed by the potential
cycling. The apparent uniqueness of the H20 responsible for SERS minimizes the usefulness
of SERS for characterizing the "typical" molecule of H 20 present in the double layer.
Nevertheless, the uniqueness of this H 2 0 may be advantageous for consideration of corrosion
phenomena. That is, the H2 0 responsible for SERS may be adjacent to "reactive sites" on the
silver surface. In this event, the "SERS active site" might represent silver in a highly activated
state in the overall process of aqueous dissolution. If true, then information about this species
may provide important information about the mechanisms of corrosion. A great deal of work
remains to be performed in this area.

Figure 5. Raman spectra

(in the region of the OH stretch A
of water) of a silver electrode in
IMKCI (A) Polished electrode
(compare to figure 4a); (B)
roughened electrode at -0.2V
(SCE); (C) silver electrode
pushed against cell window so
that it is covered by only a thin- B
layer of solution. [37]

3750 2900
 419

5.4. STUDIES OFINTERFAQALCHEMISTRY

Dorain ~36] studied the SERS spectra of species present on a silver electrode immersed in
buffered solution of l.OM phosphate made from mixtures ofK 3ro 4 • K 2HP04 • KH2ro4 and
H3ro 4. Figure 6 plots the calculated mole fraction of phosphate species as function of pH for
solutions with l.OM total phosphate concentration. Figures 7-9 summarize the results of the
measurements of the SERS spectra of the various phosphate species as a function of the
potential drop at the interface between the silver electrode and the electrolyte. The maximum
SERS intensity of HPO 4- occurred at O.OI2V (SCE) and that for H2ro 4 - occurred at -O.396V
(SCE). Assuming that the relative stabilities of the various phosphate species are the same in
the interfacial region as in bulk solutions. the results of this study can be used to determine the
interfacial pH. This work represents the first measurement of the variation of the interfacial pH
with electrode potential.

H 3 P04
"
/'''-'--''''
\ '
/' 1 '\P03-.'4 .'

z
/
I
:' Hi' 0 4 \, I HPO~ -
I
'-/
I
o ! \ / \ /
~
()
< / 'J \y!
a: I \
, I,
.\i'
u.
I I"~,
I ,, /\
W
..J
o
~
I / \ ,: \
o ,l" ! \\ ...... .'
I \ \

,_
0.0 7.0 14.0

Figure 6. Variation with pH of the calculated mole function of phosphate species in

solutions with l.OM total phosphate concentration.[36]
420

..-
:..............,
> aIQNAL~1
. .
0\
~

t:::
en .: ~

z
w
to-
:...
•:
1--" .
I,

-Z . .i ,/1
I \
Z
" ~~

.:,..:1 ",
., : 8ACtcQROUND
< : 1\ ~
,",".. !I,
.' -" :
~
< .
.:::~....
\~"';r-
~ .
\. :_'
a: .....\......... :I
- -......... 'I
v
--.... --=.....
"
DIFFERENCE

-0.50 0.00 0.50 0.00 -0.50

VOLTAGE (V seE)

Figure 7. SERS inten$ity of the peak associated with HP04 =.[36]

 421

1M K2HP04
a'· ••
PEAK A : ~
; :

...-
SIQNAL-: ~
> ~I

·•·• ..•
~
•
UJ · •

...Z
Z 'ACKGAOUND-i~I"'-\
W
:,.
..... ..,/
:'...'
: I \
\
.
-z ..
:~
:: ••••••• . .....
:
: I
I

...::
:E
•................ I' . . . .,\ \ll o2396V?
: \.'
i..
~,'
~.. 4
. . :" I
:, H PO-
...::
a: -... --....... ...'..
..... •••••••

--
"
, ' •• I
\
- •

'--"\'.,,

-0.50 0.00 0.50 0.00 -0.50

VOLTAGE (VSCE)

Figure 8. SERS intensity of the peak associated with H2ro 4-

422

, .. ,
•
\

,
\
Peak B.B
.
\

HP0 2- \

10
~ P03-
,
\. ..
J:
•\
\
0.8 e\
z
o ~
e
\
i= \
:;)
...J
oCI)
.~
,
~
...J
~ 2

-0.8 -0.4 -0.2 0 0.2 0.4 0.8

Figure 9. Influences of bulk solution pH on the potential at which the maximum in

SERS intensity occurs for H 2FO 4 -, HPO 4-2 and PO4 -3. [36]
 423

5.5. STUDIES OF PASSIVITY

One of the early and perhaps continuing hopes relative to SERS is that the effect would occur
on many metal surfaces. To date, however, SERS on metal electrodes immersed in aqueous
solutions has only been confirmed for silver, copper and gold. To extend the utilization of
SERS to a wider range of metals, "composite" electrodes consisting of silver, copper or gold
and a second metal whose behavior is of interest have been fabricated and studied. For
example, Fleischmann and Tian [37] have shown that the addition of nickel or cobalt ions onto
smooth silver surfaces can result in SERS of species adsorbed onto nickel or cobalt. In
addition, Fleischmann f1 al.[38] and Mengoli tlJll.[39] have demonstrated that SERS of
absorbates present on surfaces of zinc, nickel, cobalt and iron can be detected following the
electrodeposition of thin coatings of these three metals on roughened silver surfaces.
This approach was anticipated by Thanos [40] who, after being unable to obtain
Raman spectra from the passive film formed on iron in 0.02M HNO 3' added 10 -4M of AgNO 3
to the solution with the electrode at -1.OV (SCE). During subsequent oxidation and reduction
of the electrode peaks attributed to iron oxidation products were detected. This technique was
nicely extended by Rubim and Dtinnwald [41] who observed enhanced Raman scattering from
films on iron electrodes which were passivated in O.5M H2SO4' borate buffer (pH 804) or
carbonate/bicarbonate buffer (pH 8.0) and then coated with a discontinuous silver film. The
composition of the passive film varied with the chemistry of the electrolyte. In all cases,
however, a Raman peak occurred at 544 cm -1 and this was attributed to the presence of
Fe(OH)2' It was suggested that the passive film consists of an iron (III) oxy-hydroxide that
forms on top of a layer of Fe(OH) 2 located on the iron surface.
Similarly, Gui and Devine[42]have studied the influence of the amount of silver deposited on
iron electrodes on the enhancement of the Raman spectra of the passive film formed in nitric
acid and in buffered borate (PH 804). The enhancement increased monotonically with the
amount of silver deposited and then reached a maximum value for =35mC/cm 2 of silver.
Larger quantities of silver resulted in a decrease in SERS presumably because of a decrease in
the amount of passive film left uncovered by silver. In addition, it was found that the amount
of Fe +3 species in the film formed in the borate buffer solution increased as the potential of the
iron electrode was changed from -0.1 V to +0.1 V (SCE).
Figures lOa and lOb illustrate the influence of changes in potential on the Raman spectra of
the film formed on iron in the buffered borate solution (pH 804). The electrode of iron was
electrodeposited with a discontinuous layer of silver as described above, inserted into the
buffered borate solution, and then cleaned of surface films by cathodic polarization at -0.9
V(SCE) for 5 minutes. Figure lOa shows the changes that occur in the Raman spectra when
the potential of the electrode is stepped from -0.9V to -OAV and subsequently stepped between
-OAV and +O.IV. At a potential of -4V the peak at 550 cm -1, which is attributed to the
presence of Fe(OH)2 appears quite strong. When the potential is then stepped to +0.1 V the
424

160. 320. 480. 640. 800.

Figure lOa Influence of applied potential on the SERSspectra obtained from the
passive film on iron in buffered borate (pH 8.4). The spectra were
generated sequentially, starting with the lowest which was acquired
immediately after the potential of the iron electrode was stepped from -
0.9Y to -O.4Y. The next spectrum was procured 24 hrs. later (at -O.4Y).
The middle spectrum was measured after the potential of the iron
electrode was stepped from -O.4Y to -0.1 V. The next higher curve is the
spectrum that was generated 24 hrs. later (at +0.1 V). The top most
spectrum was obtained after the potential was stepped from +0.1 V to -
OAY. Note, the curves are displaced from each other in the vertical
direction. Consequently, the vertical axis is a plot of relative intensity.
 425

I I ,
150. 320. 480. B40. 800.

Figure lOb. Influence of applied potential on the SERSspectra obtained from the
passive film on iron in buffered borate (PH 8.4). The spectra were
generated sequentially, starting with the lowest which was acquired
immediately after the potential of the iron electrode was stepped from
-0.9V to +0.1 V. The next spectrum was procured 24 hrs. later (at
+0. IV). The middle spectrum was measured after the potential of the
iron electrode was stepped from +0.1 V to -0.4 V. The next higher curve
is the spectrum that was generated 24 hrs.later (at -O.4V). The top most
spectrum was obtained after the potential was stepped from -0.4 V to
+0.1 V. Note, the curves are displaced from each other in the vertical
direction. Consequently, the vertical axis is a plot of relative intensity.
426

intensity of the peak at 550cm -1 diminishes and a new, strong peak appears at 250cm -1. Based

on the Raman spectra obtained from numerous oxides and oxyhydroxides ofFe++ and Fe+++,
this new peak is attributed to the presence of Fe +3 in the passive fIlm. When the potential is

next stepped from +O.IV to -OAV, the peak at 250 cm- 1 disappears and that at 550 cm- 1
increases in intensity. These changes in the passive film with changes in potential occur
quickly (i.e., on the order of:S; several minutes) and further changes do not occur in the film
over the next 24 hours.
Figure lOb indicates the changes that occur in the Raman spectra when the potential of the
electrode is stepped from -0.9V to +0.1 V (rather than -OAV as in figure lOa) and subsequently
stepped between +0.1 V and -OAV. The results are practically identical to those in figure lOa.
Collectively, the results in figures lOa and lOb indicate that the passive film formed at +0.1 V
contains iron in the +2 and +3 oxidation states, while the film formed at -OAV appears to be
free of Fe+3 and to consist of Fe(OH)2. This result is quite consistent with earlier
ellipsometric and coulombic reduction experiments on the passive film on iron. Such studies
have suggested a dual-structure to fIlms formed at high anodic potentials and a single structure
for films formed at low passivating potentials.
It is worth noting that the thicknesses of the passive films formed on iron in borate buffer for
which SERS spectra were obtained are as thin as 20A according to coulombic measurements
and ellipsometric measurements by other workers.[ 43-46] This indicates the ability of SERS to
measure .in-siIl! the structures ofvery thin films. Studies are now underway to examine the role
of specific anions on the formation and breakdown of passive films formed on iron and iron-
chromium binary alloys.

References

1. Foley J. K., Korzeniewski C., Daschbach 1. L., and Pons S. (1986) 'Infrared Vibrational
Spectroscopy of the Electrode-solution interface', Electroanalytical Chemistry, v.14, A.
J. Bard, editor.

2. Steger, E. (1970) 'Raman spectroscopy with Poor Scatterers', Raman Spectroscopy vo1. 2,
H.A. SzymanSki, ed., p.175.

3. Barnett, G.P. and Albrecht, A.C.(1970) 'Comments on the Derivation of the Dispersion
Equation for Molecules', Raman Spectroscopy vo1.2, QJ;2..d1. p.207.
 427

4. Bhagavantam, S. (1942) Scattering of Light and the Raman Effect, Chemical

Publishing Co.Brooklyn, NY.

5. Colthup N.B., Daly L.H., and Wiberley S.E. (1964) Introduction to Infrared and
Raman Spectroscopy, Academic Press, NY.

6. Long, D.A. (1977) Raman Spectroscopy, McGraw-Hill, London.

7. Devine, T.M. (present Proceedings) 'Infrared Spectroscopy at Electrode-Electrolyte

Interface'.

8. Wilson E.B., Decius J.C., and Cross P.C. (1955) Molecular Vibrations, McGraw-
Hill,NY.

9. Cotton, F.A. (1971) Chemical Applications of Group Theory, 2nd Edition, J. Wiley
&Sons,NY.

10. Atkins, P.W. (1983) Molecular Quantum Mechanics, 2nd Edition, Oxford University
Press,Oxford.

11. Graybeal, J.D. (1988) Molecular Spectroscopy, McGraw-Hill, NY.

12. Scherer, J. R. (1979) 'The Vibrational Spectroscopy of water', Advances in Infrared and
Raman Spectroscopy, v.5, p.149.

13. Graybeal, J. D. (1988) Molecular Spectroscopy, op.cit., p. 539.

14. Fleischmann M., Hendra P. J., and McQuillan A. J. (1974) 'Raman Spectra of Pyridine
Adsorbed at a silver Electrode', Chern. Phys. Lett., 26, 163.

15. Jeanmaire, D. L. and Van Duyne, R. P. (1977) 'Surface Raman Spectroelectrochemistry',

J. Electroanal. Chern., 84, 1 .

16. VanDuyne, R. P. (1979) 'Laser Excitation of Raman Scattering from Adsorbed Molecules
on Electrode surfaces', Chemical and Biochemical Applications of Lasers,
Moore, C. B., editor, vA, p.lO!' Academic Press, NY

17. Furtak, T. E. and Reyes, J. (1980) 'A Critical Analysis of Theoretical Molecules for the
Giant Raman Effect from Adsorbed Molecules', Surf.Sci., 93, 351.
428

18. Otto, A. (1980) 'Surfaced Enhanced Raman Scattering (SERS), What Do We Know?',
App. Surf. Sci., 6, 309.

19. Creighton, J. A. (1980) 'Raman Spectroscopy of Adsorbates at Metal Surfaces', Springer

Series in Chemical Physics v. IS, Vibrational Spectroscopy of Adsorbates, R. F.
Willis, editor, p. 145.

20. Kerker, M. (1981) 'Colloid Spectroscopy', Pure Appl. Chern., 53.

21. Otto, A. (1983) Light Scattering in Solids v.4, Topics in Applied Physics, M.
Cardona and G. Guntherodt, editors, Springer, Berlin.

22. Furtak, T. E. (1983)'Current understanding of the Mechanism of Surface Enhanced Raman

Scattering', J. E1ectroanal. Chern., 150, 375 .

23. Pockrand, 1. (1984) Surface Enhanced Raman Vibrational Studies at Solid/Gas

Interfaces, Springer Tracts in Modem Physics, vol. 104, Springer-Verlag NY.

24. Otto A., Billman J., Eickmans J., Ertiirlc D., and Pettenkorfer C. (1984) The "Adatom
Model" of SERS (Surface Enhanced Raman Scattering): The Present Status', Surf.
Sci., 138, 319.

25. Efrima, S. (1979) 'Surface Enhanced Raman Scattering (SERS)" Modern Aspects of
Electrochemistry, no. 16, B. E. Conway, R.E.White and J. O'M. Bockris, editors,
vA, p.lOl.

26. Chang, R. K. (1987) 'Surface-Enhanced Raman Scattering at Electrodes: A Status

Report', Ber Bunsenges. Phys. Chern., 91, 296 .

27. Chang, R. K. and Furtak, T. E. (1982) Surface Enhanced Raman Scattering, Plenum
Press, NY.

28. Walrafen, G. E. (1972) 'Raman and Infrared Spectral Investigations of Water Structure',
Water, A Comprehensive Treatise, Felix Franks, editor, v.l, p.151.

29. Scherer, J. R. (1978) Advances in Infrared and Raman Spectroscopy, v.5, p.200, op.cit.

30. Frank, H.S. Water, (1972) 'Structual Models', A Comprehensive Treatise, v.l, p.515,
op.cit.
 429

31. Woodward, L. A. (1970) 'Vibrational Rules of Selection and Polarization: Their Practical
Uses and limitations', Raman Spectroscopy, Theory and Practice, v.2.

32. Fleischmann M., Hendra P.J., Hilll.R., and Pemble M. E. (1981) 'Enhanced Raman
Spectra from Species Formed by the Condensation of Halide Ions and Water Molecules
on Silver Electrodes', 1. Electroanal. Chern., 117, 243 .

33. Pettinger B., Philpott M. and Gordon, J. (1981) 'Contribution of Specifically Adsorbed
Ions, Water, and Impurites to the Surface Enhanced Raman Spectroscopy (SERS) of
Ag Electrodes', J.Chem. Phys., 74, 934.

34. Fleischmann, M. and Hill, I. R. (1982) 'Electrochemical Effects', Surface Enhanced

Raman Scattering, op.cit.

35. Macomber S. H., Furtak T. E.,and Devine T. M. (1982) 'Enhanced Raman

Characterization of Adsorbed Water at the Electrochemical Double Layer on Silver' ,
Surf. Sci. , 122, 556.

36. Dorain P. B., Von Raben K. U, and. Chang R. K. (1984) 'Voltage Dependece of the pH at
the Interface of a Ag Electrode: A SERS Measurement of a Phosphate-Phosphoric Acid
System', Surf. Sci, 148, 439 .

37. Fleischmann M. and Tian, Z. Q. (1987) The Induction of SERS on Smooth Ag by the
Deposition of Ni and Co', 1. Electroanal. Chern., 217, 411.

38. Fleischmann M., Tian Z. Q., and Li L. J. (1987) 'Raman Spectroscopy of Adsorbate on
Thin Film Electrodes Deposited on Silver Substrates', 1. Electroanal, Chern., 217,
397.

39. Mengoli G., Musiani M., Fleischmann M., Mao B., and Tian Z. Q. (1987) 'Enhanced
Raman Scattering from Iron Electrodes', E1ectrochim Acta, 32, 1239.

40. Thanos, I. C. G. (1986) 'In-Site Raman and Other Studies of Electrochenically Oxidized
Iron and Iron-9% Chromium Alloy', Electrochim. Acta, 31, 811.

41. Rubim, 1. C. and Diinnwald, 1. (1989) 'Enhanced Raman Scattering from Passive Films
on Silver-coated Iron Electrodes', J. Electroanal. Chern., 258, 327.
430

42. Gui, 1. and Devine, T. M. (submitted to J.ofElectrochem. Sci.) 'Obtaining SERS from
the Passive Film on Iron'.

43. Nagayama, M. and Cohen, M. (1962) 'The Anodic Oxidazation ofIron in a Neutral
Solution', J. Electrochem. Soc., 109,781.

44. Bockris J. O'M., Genshaw M.A., Brusic V,. and Wroblowa H. (1971) 'The
mechanism of the Passivation of Iron in Neutral Solutions: An Ellipsometricy and
Coulometric Investigation', Electrochim. Acta, 16, 1859.

45. Frankenthal, R.P. (1971) 'Passivation of Irom in Borate Buffer Solution', Electrochim.
Acta, 16, 1845 .

46. Kruger, 1. (1989) 'The nature of the Passive Film on Iron and Ferrous Alloys', Corr. Sci.,
29, 149.

6. Appendix I

6.1. SIMILARITIES OF REPRESENTXI'IONS FORMED BY LINEARLY RELATED

BASES

In this section it will be demonstrated that when the normal mode coordinates and the mass-
weighted Cartesian displacement coordinates serve as bases for a point group they produce
equivalent representations. This results from the fact both coordinates are linearly related.
Consider 2 sets of bases, Xi and Yj which are linearly related.

(I - 1)

Yj = L
1
(a-I) ji Xi
(I - 2)

Under the action of the symmetry operation R, Xi is transformed to X'j

(I - 3)

The symbol S denotes a vector (i.e., a Nxl or lxN matrix) and the symbol S denotes a NxN
matrix.
 431

In tenns of individual matrix elements, the previous equation becomes

CI - 4)

The character of the symmetry operator R in the basis x is given by

3N

X[R(x)] = Lr ii
(I - 5)
i=l

Similarly, under the action of the symmetry operation R, y j is transfonned to y'k

Y'=RY (I - 6)

In tenns of individual matrix elements, the previous equation becomes

(I - 7)

The relationship between Y' and X' is obtained from (1-2):

Y
k
= LI (a-I) X'
kl I (I - 8)

Substituting (1-4) and (I-I) into (1-8) gives

(I - 9)

Comparing (I-9) to (I-7) indicates that

432

~ rli~
Pk'J = (a-I) kl~ a ..
~ 1J
i j (I - 10)

(I - 11)

Therefore,

(I - 12)

Since (a-I) kl and aikare numbers (not matrices), their order of appearance in (1-12) is

(a -I) klaik = 0 i1
unimportant. Furthermore,.

Thus, (I-12) becomes

3N

x[ R( y)] =0 i1 r Ii =L r II
(I - 13)
1=1

Comparing (I-13) to (1-5) indicates that

X[R(x)] = X[R(y)] (I - 14)

Thus, the characters of 2 representations formed from bases related to each other by a linear
transformation are equal.

7. Appendix II

7.1 PROOF THAT NON-DEGENERATE NORMAL MODES ARE BASES FOR 1

DIMENSIONAL IRREPS OF THE SYMMETRY GROUP

The action of the symmetry operation R on the non-degenerate normal mode coordinate Qk can
 433

not result in a mixture of Qk and some other normal coordinate Qj" If it did, then the potential
energy of the molecule would change as a result of the symmetry operation and this can not
occur. That is

(II - n

(II - 2)

where all the Qk are assumed to be non-degenerate. Taken together, (II-I) and (11-2) indicate
that

(II-3)

That is, the action of the symmetry operation on a non-degenerate normal coordinate must not
result in a mixture of non-degenerate normal coordinates. Thus, a non-degenerate normal
coordinate acting as a basis produces a one dimensional (i.e., irreducible) representation of the
symmetry group. By similar reasoning it is concluded that d-fold degenerate normal modes
may be mixed together by the action of a symmetry operation. However, they may not be
mixed together with any other normal mode. Thus, d-fold degenerate normal modes provide a
d-dimensional irrep of the symmetry group.

8. Appendix III
8.1. KINETIC ENERGY IN TERMS OF THE G MATRIX
Here it is demonstrated that the kinetic energy may be expressed as a function of the Matrix Q
whose elements are

G ij = L Jl~BikB jk(i,j,= 1,2, .... ,3N - 6)

(III - 1)
k-I

Bik are the direction cosines relating the internal coordinates ri to the Cartesian displacement

vectors ~k
434

3N

ri = L Bik~k
k=1 (III - 2)

In matrix form
(III - 3)

The symbol S denotes a vector (Le., a Nxl or lxN matrix) and the symbol S denotes a NxN
matrix.
The kinetic energy of a molecule associated with its internal modes of motion is given by
1-'.
T = "2 q 'q (III - 4)

mt ~(j) h _
Where weighted Cartesian displacement vector: (J) 1 Il i

q =lX3 matrix and q' =transpose of q=3Xl matrix.

Since (III - 5)

(III - 4) may be rewritten as

T
="21P"- -P (III - 6)

Using (II1-3), the kinetic energy maybe expressed in terms of internal coordinates

_ aT _ aT a fj a ~i _
Pi - a' - a' . a . -PjBjilli
qi a ~j qirj (III - 7)

(e.g.)P i = L PjB jllll

j=1 (III - 8)

(III - 9)
 435

In tenns of matrix notation, (III-7) will be written as

p;:::C'p (m - 10)

where Cij=Bij~
where 12=3NXl matrix (i.e., column matrix)
~=3NXN
P=NXI matrix (i.e" column matrix)
The transpose of (III-lO) is
IY;::: p:c (III-H)

where P' =IX3N matrix (i.e., row matrix)

P'=IXN matrix (i.e., row matrix)

C=NX3N matrix
Substituting (III-10) and (III -11) into (III-6) gives

T;:::.l P' C C' Ii

2 lXN N){ 3N 3NXN N){ 1 (m - 12)

The lower case lettering represents the no, of rows x the no. of columns for each matrix.
In terms of matrix elements, the last equation becomes
T;::: l/2PiP/>ij (III - 13)

Where the Pi are the linear momenta associated with the internal motion referred to as
Cartesian coordinates -

(III - 14)

(III - 15)

N ote th at P.p[:;c PP I 8 il th· . general,0rthogonal.

J J
·
J smce e mtern al coord·mates are not, m
Comparing (III -15) to (III -1) indicates that
T;::: .lP'GP
2--- (III - 16)

Since
436

. dr i
r - ----
aT
i - dt - ap
i , the matrix I is

(III - 17)

- -17
Therefore p= G r (III - 18)

and (III - 19)

Substituting (III-IS) and (lII-l9) into (lIl-16) gives

17' -17
T = "2r.G. r (III - 20)

9. Appendix IV
9.1. DETERMINATION OF THE BASIS FUNcrIONS FOR AN IRREP
The problem which is addressed here is the detennination of a basis for an irrep of a molecule.
This is achieved through the use of a projection operator as described below.
Consider a set of n orthonogal functions

(N - 1)

These, for example, might be atomic orbitals for the atoms composing the molecule. The
action of any operator R in the symmetry group of the molecule may be described by

n
i ~ i·
R0. = £.. 0 b r'ab
b=! (N - 2)

Where r~b =element in the a row and b column of the i th representation of the operation R.

Multiply the above equation by the number [ri.. bJ and sum each side of the equation over all

operations in the group:

 437

Applying the GOT to the last equation gives

(N - 4)

Thus, the coefficients multiplying the functions 0 ~ are all zero except that multiplying 0 i.,

Even this coefficient will also be zero unless i=j and b=b'.
For convenience, the projection operator is defined:

(IV-5)

Equations (IV-3) and (IV-5) may now be combined to yield

(IV-6)

Thus, when pia' b' is applied to the function 0 ~ , only if i==j and b=b', will the result be non-

zero. That is, only if 0~ is or contains 0

i
b, will the results be non-zero. In other words,

. . operator
the proJecuon pia'b' extracts f rom the fu ncuon
. 0'b, the fu nctlOn
. 0 J
a' •

In particular

(IV-7)

Consequently, if the diagonal elements of the matrix representation of each of the symmetry
operations R are known for the j th irrep, then (IV -7) can be used to generate from a knowledge
of just one function of the basis j, the sum of all of the functions which together constitute the
basis of the jth representation. This approach is particularly powerful if the j th representation is
one dimensional. Then the projection operator acting on a part of the basis function will
generate the entire basis function.
INFRARED SPECTROSCOPY AT ELECTRODE·ELECTROLYTE
INTERFACES

T.M.DEVINE
Department ofMaterials Science and Mineral Engineering
University of California
Berkeley, CA 94720, USA

Abstract.

The fundamental aspects of vibrational spectroscopy are reviewed. A classical and

partial quantum mechanical description is given of the vibrational modes of polyatomic
molecules along with an introduction to the use of Group Theory in identifying vibrational
modes that would be active to infrared absorption and Raman scattering. Examples are given
of the use of infrared spectroscopy to identify in·.si1Y the adsorbates and films that fonn on
metal surfaces immersed in aqueous solutions.

I. Introduction

Most researchers studying corrosion phenomena would agree on the value of in-.siJ;y
techniques that are capable of providing structural and/or chemical information about the
interface between a metal and an aqueous solution. Ellipsometry(l-6), infrared (IR)
absorption(7-l0), Raman scattering(8-13), optical second harmonic generation (SHG)(l4),
Mossbauer spectroscopy(l5) and SEXAFS(l6) are some of the teChniques that have been
applied to the in-.si1Y investigation of the interface between a metal and an aqueous solution. In
spite of the wealth of information that they provide, the cost of equipment and, in some cases,
the lack of knowledge of the fundamental prinCiples underlying the technique and, in other
cases the difficulty in interpreting the results of measurements, have all combined to limit the
use of the techniques and the understanding of the results.

In this sequence of two papers two of the techniques which can perhaps most readily
provide in .si1Y information about the structure and chemistry of the metal-solution interface,
infrared absorption and Raman scattering, will be discussed. The two articles will illustrate the
fundamentals underlying the techniques, the experimental procedures employed, the
experimental and theoretical limitations of the techniques, and the types of infonnation which
have been obtained and their interpretations.

Each paper is divided into two parts. The first describes several aspects of the
analysis of vibrational spectra. The present paper, for example, provides an introductory-level
treatment of the interaction of electromagnetic radiation with a molecule. This is followed by a
description of the use of group theory to identify the symmetries of the vibrational modes that
are infrared active. The formalism is also applicable to the determination of the symmetries of
the vibrational modes that are Raman active and is so used in the second paper. The
experimental results of IR spectroscopy discussed in Part II of the present paper emphasizes
the data that has been obtained from CO adsorbed at interfaces. This relatively simple molecule
exhibits a correspondingly simple IR spectra.

Part I of the second paper illustrates a general procedure for identifying the normal
modes of vibration of a polyatomic molecule. For the purposes of illustration, the technique is
used to specify the vibrational modes of H 2 0. Part II describes the Raman spectra obtained
from films formed on metals immersed in aqueous solutions and also from species located in
the double layer. In combination, both papers provide the corrosion scientist with (i) a
439
M. G. S. Ferreira and C. A. Melendres (eds.),
EJectrocMmicaJ and Optical Techniquesfor the Study and Monitoring of Metallic Corrosion, 439-483.
© 1991 Kluwer Academic Publishers.
440

reasonably complete description of a procedure for identifying the infrared and Raman active
vibrational modes of a molecule, and (ii) examples of information of species present at metal-
solution interfaces obtained by IR and RS.
IR absorption and Raman scattering are used as tools for identifying the composition
of substances, for studying the symmetry of the scattering or absorbing substances, for
determining the frequency of a particular vibrational mode and for investigating intramolecular
and intermolecular bonding. At the risk of stating the obvious, IR absorption and Raman
scattering accomplish these tasks because of the influence which electromagnetic radiation has
in inducing certain vibrational, rotational and electronic transitions in species. The theoretical
treatment begins with an analysis of the effect of electromagnetic radiation that is incident on a
molecule. For the molecule to absorb radiation it must undergo an electronic, rotational or
vibrational transition. The present study will focus on vibrational transitions. The objective is
to reveal which vibrational modes are IR-active and and which are Raman-active. The first task
then is to identify the different vibrational modes of a molecule. A harmonic oscillator is used
to model the behavior of simple molecules. A brief, classical description of a harmonic
oscillator is followed by the quantum mechanical description. Through the use of normal
coordinates, vibrational behavior of polyatomic molecules is shown to consist of harmonic
oscillations of normal modes. For these to be IR-active, the radiation must cause a change in
the electric dipole moment of the molecule and the transition must be from one quantum
vibrational state to a second which is immediately adjacent (higher or lower) in energy. For
molecules such as diatomic molecules that are simple in structure, it is usually possible to
determine by inspection whether these two selection rules are satisfied. However, for more
complicated polyatomic molecules this is often impossible. In such cases Group Theory is
employed to determine whether or not transition to a particular vibrational mode may be
triggered by the absorption of radiation. This formalism is both elegant and powerful. Its use
is illustrated by the determination of the symmetry of each of the vibrational modes of water.
I I. TheoQ' of Infrared Absorption and Raman Scattering ( 17 -21)

2.1. Qassical and OUantum Mechanical Description ofHarmonic Oscillator

The discussion of the analytical treatment of IR absorption and Raman scattering

begins with the harmonic oscillator. This type of motion is familiar to most investigators and
consequently represents a good system to analyze. In addition, as is demonstrated later on, the
use of normal mode coordinates permits the treatment of the motions of very complicated
molecules with the formalism developed in the analysis of the harmonic oscillator.

Classically, a vibrating diatomic molecule AB depicted in figure 1 may be treated as 2

k = force constant; ~ = displacement

Figure 1 Vibrating diatomic molecule AB.

 441

masses, m A and m B' connected by a spring of stiffness k. Since our interest is in the
vibrations of the molecule, that is, in the displacement, x, of A and B relative to each other, we
may factor out of the equation of motion the movement of the center of mass of the molecule.
Then, the internal motion of the system is characterized by

p2
E tota! = K.E.+P.E.= 21l + V(~).
(1)
where

(2)

Vex) = potential energy of the molecule when the bond length is different from its
equilibrium value by an amount x and

P =linear momentum =Il( ~; )

The force restoring the bond length to its equilibrium value (x=O) is given by

(3)

Assuming that x is small, the above equation can be approximated by

F= -kx (4)

The potential energy is then obtained from the relationship

()V( ~)
F=---
a~ (5)

The potential energy associated with the relative location of A and B within the
molecule is then given by

(6)

Therefore, the total energy of the molecule given in (1) becomes

p2 1 2
E total = 21l + 2k~
(1)

In a conservative system

dE total =0
dt
442

(8)

Hence,

or
Il d~ d 2~ + k~ d~ = 0
dt dt2 dt
thus
d2~ k
-=-~
dt2 Il (9)
A solution to equation(9) is

(10)

The diatomic molecule thus vibrates at a frequency of,jf. Substituting (10)

into (1) gives

(11)

Note that while there is a restriction on the vibrational frequency of the diatomic
molecule, the classical model does not limit the possible values of the total energy of the
system. In fact, experiments indicate that this classical description cannot account for all of the
observations of the behavior of molecules. Consequently, we are compclled to seek another
formalism. The quantum mechanical approach, whose fundamental principles are summarized
in Appendix I, is able to successfully explain the results of experiments which the classical
approach cannot. For pragmatic reasons thc classical model is abandoned in favor of thc more
successful quantum mechanical treatment. It is worth noting that the specific goals of the
following discussion are: (1) identification of the vibrational modes of a polyatomic molecule,
and (2) determination of whether or not a givcn vibrational mode is IR -and/or Raman-active.

The analysis begins with the determination of the vibrational modes of a diatomic
molecule AB. The quantum mechanical description of the diatomic molecule AB is given by
Schrodinger's equation.

(time independent) H(r) y i(r) = EYi(r) (l2a)

(time dependent)

where: H::: Hamiltonian of the system

Yi =time independent wave function of the system in state i
Ii ::: time dependent wave function of the system in state i
q =energy of the system in state i
 443

The Hamiltonian includes operators for the electronic, vibrational, rotational and
translational energies of the molecule. These energies can differ considerably in magnitude so
that it is possible to write H as a simple sum of each of these energies without considering their
interactive effects(22):

(13)

With the Hamiltonian expressed as a sum, the wave function solution to Schrodinger's
equation (12) can be expressed as a product

Y = Yel Yvibr Y rot Ytrans (14)

Consequently, the different components of the energy of the molecule can be

considered separately, e.g.,

(time independent) Hvibr(x)Yvibr(x) = Evibr(x)Yvibr(x)

(15a)

(time dependent) Hvibr(x,t)vibr(x,t) = i h aYvibr(x,t)/at

(15b)

For a diatomic molecule AB, (l5a) becomes

h2-d+ 2 1 2}v
[- -211 d~ - 2k~
2 vibr
=E vibr 'Ijf vibr
(16a)

The solution of (l6a) will identify the vibrational modes of the harmonic oscillator
which is the immediate goal of this section. However, before doing so, a brief digression will
be taken to consider the time dependent Schrodinger equation (15b). This digression will pay
dividends in the treatment of the interaction of radiation with a molecule by the time dependent
perturbation theory .

In the event that the potential energy is not a function of time, then (15b) becomes

HJ~)'¥(~,t) = iha~;,t)
(l6b)

Therefore, we can factor 'I'(x,t) = 'I'(x)0(t) (17)

thus
444

0(t)HJ~)'P(~) = ih'P(~)aWr(t)

HJ~)'P(~) ih a0(ti
'P(~) = 0(ti at (18)

The left side is a function only of x and the right side is a function only of t. For the
equality to hold for all values of x and t, both sides must equal the same constant

Hv(~)'I'(~) = C
'I'(~) (19a)

i h d0(t)
---=C
0( t) at (19b)

Solving the last equation give

0(t) = Cexp (-i~t)
(20)
where C = constant

The value of c is obtained by realizing that (I5a) and (I5b) must be consistent, that is,
substituting (17) and (20) into (I6b) must give (I5a). With these changes (I6b) becomes

(21)

(22)

Comparing (22) with (I5a) indicates that c =E. Thus, the time dependent fonn of any
wavefunction, when the potential energy of the molecule is independent of time is given by:

'P(~,t) = 'I'(~)exp ( - i ~ t) (23)

Use will be made of this result later when the influence of electromagnetic radiation on
ththeory.

We now return to the identification of the vibrational modes of a diatomic molecule.

The time independent Schrodinger equation describes the case in which the diatomic molecule
AB is not subjected to any external fields, and (15a) becomes
 445

[- -211 2 -k
h2-d~d+ 1 2} =E vibr 'V vibr
2 2 ~ vibr
(24)

The solutions to (24) are given by

x2
'V(~) = N v H v(x)exp ( - "2) (25)

1
"2
where x = (llhO)) ~

Hv(X) = Hermite polynomials

1
"2

Nv = nnnn>lintion """""t = C· vI, X t)

v Hv(x)

°
1
2
1
2x
4x2 -2
3 8x3 -12x
4 16x 4-48x2 + 12
5 32x 5-160x 3+120y
6 64x 6-48Ox 4+ nOx 2-120

By =(v+1!2)hw v=O,1,2 ... (26)

The vibrational modes of the harmonic oscillator are given by (25) and the energies of
the harmonic oscillator as a function of the vibrational mode are given by (26). Note that in
contrast to the classical treatment, the quantum mechanical treatment limits the possible energy
values of the vibrating diatomic molecule. Further discussion of the vibrational modes of the
harmonic oscillator will be delayed until after an analysis of the vibrational modes of a
polyatomic molecule. The analysis will illustrate that the modes of a polyatomic molecule can
be treated using the formalism developed above to describe the modes of a harmonic oscillator.
446

2.2. Ouantum Mechanical Description of the Vibrational Modes of a Polyatomic Molecule

Consider now a polyatomic molecule of N atoms where N ~ 3. Once again, the goal
is to identify the vibrational modes and to determine which are IR active and which are Raman
active. Vibrations ofpolyatomic molecules involve all of the atoms. Consequently, the potential
energy of a polyatomic molecule depends on the displacements of all the atoms from their
equilibrium positions:

(27)

Assuming as before that the various xi are small, the potential energy can be expressed
by the first few terms in a Taylor expansion

If the motion of the center of mass of the molecule is ignored, then V(O) =0. In

'.
addItIOn,
(~~) 0
= th . fi . 'l'b . .. .
e restonng orce on the atom at Its eqUl 1 num pOSItion, IS also zero.
Hence, assuming the Xi's are small, the potential energy can be approximated by

V = 1/ 2 ~(
t\ ar~~) ~ i~
1 J
j

o (29)
The sums in (29) are carried out over all 3N displacements of the N atoms. It is
important to remember, however, that not all of the displacements correspond to vibrations. In
fact, 3 of the 3N independent displacements of the 3 N atoms correspond to movement of the
center of mass of the molecule. Three more correspond to changes in orientation of the
molecule about its center of mass (only 2 if the molecule is linear). Therefore, 3N-6
displacements result in changes in the potential energy of the molecule.

Thus, the vibrational Hamiltonian for the polyatomic molecule may be written as

where we have introduced the mass-weighted coordinates

(31)
 447

(32)

Solving Schrodinger's equation for a polyatomic molecule is difficult because of the

existence of interactive terms (i;tj) in the expression for the potential energy (26). It is
possible, however, to carry out a coordinate transformation from qi to Qi, where Q i is a linear
combination of the qi such that the Hamiltonian (30) becomes

(33)

This procedure is illustrated in the chapter on SERS in these Proceedings. In (33)

there are no interactive terms. The coordinates Qi are called normal coordinates. Since the
Hamiltonian consists of a ~ of terms associated with each normal coordinate, the wave
function is represented as a ~ of the wave function for each normal coordinate.

3N-6

'" v = '" v(Q 1)'" v(Q 2)'" V(Q3)"'= II", v( Q;)

i==l
(34)

Substituting (34) into (33) indicates that Schrodinger's equation for a polyatomic

molecule can be subdivided into an equation for each mode:

(35)

This last equation is identical in form to the Schrodinger equation for a harmonic
oscillator of unit mass vibrating along the Q i coordinate. Consequently, the wavefunctions and
energies for the vibrational states of a polyatomic molecule may be obtained by appropriate
modification of (25) and (26):

(36)
where
448
1
"2
Xi=(:i)Q:
and Vi =0,1,2, ...

E= (V i + ~ )hCO i
1
"2
where co.1 = k (37)

Now that the wavefunctions and energies of the various vibrational states of a
polyatomic molecule have been defined, the next task is to determine how the molecule can be
excited from one vibrational mode to another. Thus, if the molecule is in vibrational state i and
it is illuminated by infrared radiation of a certain wavelength, will the molecule be excited into
vibrational state j ?

2.3. Interaction of Molecule with Electromagnetic Radiation

A thorough discussion of the interaction of electromagnetic radiation with a molecule

requires the use of the vector and scalar potentials of Maxwell's equations. Nevertheless, a
very accurate approximation is produced by reducing the problem to the interaction between a
molecule and an electric field that varies sinusoidally with time, e = eosin wt. If the i th
particle in the molecule has a charge ei and is located at ri' then the potential energy of the
particle is increased by

Hi'(t) = ; e· er 1
(38)

Considering all N charged particles that are present in the molecule, the potential
energy of the molecule is changed by the time-varying electric field by the amount

(39)

The complete Hamiltonian of the molecule may be written as

(40)

Schrodinger's equation for the molecule exposed to the time varying electric field is
given by (ISb). Substituting (40) into (ISb) gives

(41)

The wave equations of the Hamiltonian consisting solely of H(x) are the functions
yfx) corresponding to the quantum states i. Their time dependent forms, as shown in (23), are
 449

(42)

In solving (41) we assume that H'(t) can be treated as a perturbation of H(x) and that
the wave functions Y(x,t) of the perturbed molecule can be expressed as time dependent linear
combinations of the time dependent wave functions of the unperturbed molecule.

Substituting (43) into (41) yields

H(~) Lai(t) 'l'i(~) exp ( -

1
* Et)+ H '(t) La i(t)'l' i( ~)exp (-
1
~ Et)

== ihi{~ai(t)'l'l~)exp( - k Et )]
(44)

The last equation reduces to

(45)

Multiplying both sides of (45) by 'l':n (~) and integrating over all space gives

Only one term survives on the right-hand side of (46) because of the orthogonality of

wave functions (see Appendix I). Rearranging gives

da met) i ~ ) . () .
-d-t- == - 11"",a;( t Hmi t exp ( - HO im t)
i (47)

where the following terms have been introduced

Ei -Em
and co.,m == h (48)
450

In the case in which the perturbation H'(t) is weak, it may be assumed that the
molecule is not greatly altered from its unperturbed state. Furthermore, for the case in which
the molecules are all initially in their ground state, g:
a;(O) = O;g.
(49)

Combining (49) with the fact that the molecule is assumed to be only weakly perturbed
results in the approximation:
alt) ",1
ai :# get) '" 0 (50)

Incorporating (50) in (47) yields

cia m ( t) i . () .
-d-t- = - llH mg t exp - (1U> gmt)
(51)

da met)
1b further develop the expression for -d-t-, the sinusoidal time dependency of
H'mg(t) is factored out:

Hmg(t) = Hmgsin U>t = - iH'mg [ exp (iu>t) - exp (- iu>d]

(52)

Inserting (52) into (51) and integrating from t = 0 to time t yields

t

am(t) - am(o) = - ~ fHmg{exp [i(u> - U>gm)t] - exp [- i(u> + U>gm)t]}dt

o ~~

Since

a;(o)=o;g'
__ H'mg{ exp [i( U> - U> gm) t] - 1 exp [- i( U> + U> gm) t] - I}
am(t)- h i(U>-U>gm) + i(U>+U>gm)
(54)

Consequently, (43) and (54) indicate that the probability of the molecule being excited
by the sinusoidal electric field from its ground state g to the state m is proportional to

(55)

In other words, the vibrational mode m of the molecule !lli!Y be infrared active if
Hmg:# 0
On the other hand, the vibrational mode m of the molecule will ~ infrared inactive if
H'mg = 0
 451

2.4. Selection Rules for Infrared Absowtion

In the previous section it was noted that the vibrations of a polyatomic molecule can be
conveniently described in terms of harmonic oscillations when the atomic displacements are
expressed along normal coordinates. A given vibrational mode may be infrared active if
Hmg :FO

where Hmg- f'V·m(~)P-'V g(~)dV (56)

P- is the dipole moment of the molecule. The vibrational Ym and yg are given by
(36), (42), (43) and (54). To further develop (56), and thereby determine which vibrational
modes are IR-active, an expression for P- in terms of the geometry of the molecule is required.
For any given configuration of the molecule, the dipole moment can be expressed as a function
of the displacement of the normal coordinates from their equilibrium values.

(57)

P-o = dipole moment of the molecule when all Qi =0. Substituting (57) into (56) gives

Hmg-P-of'V·m(Q;)'V g(Qi)dV + fL'V:'(Q i {

1
:6.)
1 0
Qi'V g(Qi)dV +...
(58)

The first term on the right hand side of (58) is zero because of the orthogonality of
wave functions. Hence, the infrared activity of the vibrational transition from g to m depends
on the second term which may be rewritten as

(59)

H ence, m. f d .. . tha (.)

rare aCtIvity reqUlres t 1
(:gJ o :F 0 and (11
.. ) th·
e mtegr
al be
non-zero.
That is, for the vibrational transition to be infrared active, (i) the dipole moment of the molecule
must change with displacement, and (ii) (Qi) mg must be non-zero.

For molecules with relatively simple geometries such as C02 it is fairly easy to

(dli)_O
determine whether or not dQ i 0 - • Figure 2 sketches the normal vibrational modes of
452

CD 2. For mode 1, 0
(:gJ = 0
and for mode 2, 2 0
(06 ) ~ 0
• Consequently mode 2 will

be infrared active. Since the vibrational modes are described by normal coordinates, the
vibrational wave functions will be given by (36). Substituting (36) into (59) indicates that
(Qi)mg = d g± 1,g.

--....
.......... -----...... SYMMETRIES

o o
~ ~

o O"g + : Symmetric stretch

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

0 0
~ ~

0 O"u + : Asymmetric stretch

• t • ··

0t 0, 1tu : degenerate bending

0) 0 0) 1tu : degenerate bending

Figure 2. Normal modes of CO2.

 453

For molecules with more complicated geometries than a linear AB2 molecule,
evaluation of (59) by inspection as was done for CO 2 above, is virtually impossible.
Nevertheless arguments based on symmetry can greatly simplify the determination of whether
or not the second term in (58) is non-zero. To this end, the formalism of Group theory will be
introduced.

2.5. Groyp Themy

2.5.1. Overview

The analysis by time dependent perturbation theory of the interaction of

electromagnetic radiation with a molecule indicates that absorption of the radiation may occur
by excitation of a vibrational mode of the molecule if the integral presented in (56) has a non-
zero value. This can be determined quite easily if the symmetry of the integrand is known. A
simple example will illustrate the concept. Consider the two functions sin x and cos x over the
"2
JSin x dx
2!.
range -re/2 :s; x :s; re/2. 2 is readily determined by inspection to be zero. Similarly,
"2
Jcos x dx
It

the integral 2 is quickly determined to be non-zero. Using the formalism of group

theory the range -re/2 S; x S; re/2 would be treated as an object. This object would be
characterized by two symmetry operations, E and cr. E is the identity operation and is a
member of all symmetry groups. The identity operation means "leave all of the components of
the molecule (e.g., nuclei, electrons) unchanged." cr v is a mirror plane lying perpendicular to
the X-axis and containing the point X:O. The operation of crv on cos x leaves the function
unchanged. Consequently. cos x is said to be totally symmetric with respect to the symmetry
group representing the object (Le., the group with symmetry elements E and cr v), The

operation of crv on sin x however. changes the function to a -sin x. Consequently, the integral
of cos x over the same range is non-zero and the integral of sin x over a symmetric range is
zero. This is a particular example of a general rule, if the integrand is totally symmetric, then
the integral over a symmetric range may have a non-zero value; if the integrand is not totally
symmetric. then the integral over a symmetric range will be zero.

Consequently. the infrared activity of a particular vibrational transition can be quickly

determined from a knowledge of the symmetry of the integrand in (56). The technique for
quantitatively characterizing the symmetry of the integrand in (56) will be developed in several
stages. First. more complete definitions of svrnmet'Yoperations and symmetry groups will be
provided. Second. symmetry operations will be shown to be described by matrices. The
process begins by selecting a .basis. which is a set of objects or functions that are transformed
into each other by the symmetry operations. A matrix representation of each symmetry
operation is then formed. Each matrix relates the basis before it is operated on by the symmetry
operator to itself after the operation of the symmetry operator. This collection of matrices
constitutes a representation of the group. Similarly. the symmetry of any function is described
by the set of matrices that indicate how the function changes under the various operations of the
group. The third major step in the development of the formalism of Group theory will be the
demonstration that all of the many possible representations of a group can be expressed in
454

tenns of linear combinations of a fmite number of special representations known as irreducible

representations Gl:!l;ps). Irreps are shown to be of particular importance to the quantum
mechanical applications of Group theory. In particular, if the symmetries of functions and
operators in the integrand in (56) are expressed in tenns of irreps, then the matrix multiplication
that occurs in (56) is readily perfonned with the aid of something called the Great Orthogonality
Theorem. In this manner, whether or not the integral in (56) may have a non-zero value is
detennined almost by inspection. The importance of irreps in the solution of (56) may be
phrased in another fashion. If the symmetries of the individual factors in the integrand in (56)
are known, then the symmetry of the product is readily detennined. Whether or not the integral
in (56) has a non-zero value will then depend on whether or not the symmetry of the integrand
contains the completely symmetric representation. This is detennined by reducing the
symmetry of the integrand to a linear combination of the irreps of the point group in question.
If this contains the A 1 irrep then the integral may have a non-zero value.

In summary, the fonnalism for the application of Group theory to the analysis of
vibrational transitions will be developed in several steps:

(i) definitions and examples of symmetry operators and symmetry groups will be
given;
(ii) matrices will be used to represent symmetry operations;
(iii) reducible and irreducible representations of groups will be fonned;
(iv) symmetry of functions will be illustrated;
(v) symmetry of products will be detennined.

2.5.2. Introduction

A group is defined as a collection of symmetry operations and a rule for multiplying

them together, such that, the product of two members of the group is also a member of the
group. Consider the molecule H 2 0 which belongs to the point group C 2v The symmetry
°
elements in this group are E, C 2, v and 0v'. E is the identity operation. The product of E
and any other symmetry operation R, is equal to R itself:

ER =RE =R (68)
The operation C 2 corresponds to the rotation of the molecule by 180' about an axis
that passes through the oxygen nucleus and bisects the angle between oxygen and the two
hydrogen nuclei. ° °
v is a reflection in the plane of the molecule and v' is a reflection in a
plane perpendicular to the plane of the molecule and containing the 2-fold axis of rotation.

Each symmetry operation corresponds to the movement of the nuclei of the molecule in
such a fashion that after the transfonnation the molecule is indistinguishable from its initial
configuration. e.g., as shown in figure 3, the operation of C2 causes
 455

Figure 3. Application of symmetry operators of the group C2y to a

non-linear molecule, A2B. The 2-fold rotation axis is in the plane of the
page and passes through atom #3. The cry plane of reflection is
parallel to the plane of the page. The cry' plane of reflection is
perpendicular to the plane of the page.
456

The operation of cr v causes

Finally, the operation of cry' causes

1b demonstrate that the symmetry operations listed above constitute a point group, it
must be shown that the product of any 2 of the operations is an operation which is also a
member of the group. e.g. C 2 x cr v (Le., the operation of cry followed by the operation C 2)
causes

Thus, C 2 x cry =C2

Similarly, C 2 x cry' causes

2~1
1~2
3~3
Le., C2 x cry' = E

Finally, cry x cry' causes

Thus cry x cry' = cr v ' .

These products are summarized in the operator multiplication table for H2 0:

E Cz cry cr'
v
E E C2 cry cr'
v
Cz C2 E Cz E
cry cry C2 E cry
cr'
v cr'
v E cr'
v E

In general, the product of two operations AB does not commute, Le.,

[AB-BA]T:#O (69)
 457

Note. however. for the point group C 2v' the order of the multiplication of any two
operations is unimportant.

An important relationship which may exist between two operations is that of similarity.
Tho symmetry operations A and B are said to be similar if

where R is a symmetry operation in the group. None of the symmetry operations in C 2v are
similar. However. in the point group C 3v which contains the symmetry operations E. C 3+' C 3

[c3 (+)j -1 (J C (+) =

v 3
(J
v
'"

Hence (J v and (J v'" are similar. Operations which are related to each other by a
similarity transformation are said to belong to the same ~.

2.5.3. Matrix Representation of Symmetly Operations and Groups

Often it is convenient to divide a group of symmetry operators into classes. The

operators in each class are related to each other by similarity transformations. Each symmetry
operation may be represented by a square matrix. Multiplication of the matrix representing the
symmetry operator by the matrix representing the molecule produces a new matrix which
represents the molecule in the configuration produced by the operation of the symmetry element
on the molecule in its initial configuration. The collection of the matrices for all of the
symmetry operations of the group constitutes a representation of the group.

A representation of a group may be formed by operating each symmetry operation on a

basis. A basis is a collection of objects that are transformed among themselves by the operation
of the symmetry operators of the group. For example. the Is orbitals ofH and the 2s orbital of
° may form a basis for H20): [H'•• H'b' O'.J
The representation ofC2v formed by this basis
is obtained by considering the operations described in Section E.2.

010]
C 2 =[ 100
001

Similarly.

E= [~~ ~] a = [~ ~ ~l and a = [~~ ~].

°° °° °°
v v '

1 1 1 (70)

Another representation of the point group C2v of water is formed by using as a basis
the Cartesian coordinates centered at each of the 3 nuclei in the molecule as illustrated in
458

Yo
::.~-~ .....
Xo ....: ••.......
Za ..........,:
................
~~~:....: ~~~

Ya

Figure 4. Cartesian coordinates associated with each atom in H20.

figure 4:

(71)
 459

000100000
Xa -Xb
Ya -Yb 000010000
Za Za 000001000
Xb -xa 100000000
c2 Yb = -Yb ~ 010000000
Zb Zb 001000000
Xo -xo 000000100
Yo -Yo 000000010
Zo Zo 000000001
(72)

Xb 000100000
Xa
Ya -Yb 000010000
za Za 000001000
Xb Xa 100000000
Yb = -Ya ~ cry = 010000000
Zb Za 001000000
Xo Xo 000000100
Yo Yo 000000010
Zo Zo 000000001
(73)
Xa -Xa 100000000
Ya -Ya
010000000
Za Za (xx)100000
Xb -Xb (xx)o10000
cr'
y Yb = Yb ~ cry'= 00000 1(XX)
Zb Zb (XX)000100
Xo -Xo 000000010
Yo Yo (XX)00Q001
Zo Zo
(74)

Any set of objects or functions which transform among themselves by the operations
of the symmetry group will act as a basis and generate a representation of the group. Consider
one basis which consists of the functions 0i (i =l ..... n). If a new set of functions Yj are
formed by linear combinations of the 0j. then the set Yj will also be a basis for a new
representation of the group. The new representation will be related to the old representation by
a similarity transformation. This means that the character of a symmetry operation (i.e.. the
sum of the diagonal elements of the matrix representation of the symmetry element) in the new
representation will be equal to that in the old representation.

The meaning of the above statements will be demonstrated using H2 0 once again as an
460

example. Consider the basis [(a+b), (a-b),Oj (75)

where: a,b and 0 are the Is orbitals of the 2 hydrogen nuclei and the 2s orbital of oxygen,
respectively.

The representation r(2) of C2v using the basis (75) is

(76)

E[100]
010 [100]
010 100]
001
O"y =
001 0":= [~~
(77)

The relation between the bases in (75) and that composed of the Is orbitals of Hand
the 2s orbital of 0 is given by the transformation matrix c,

c [~aJ = [aa +~ bb] ~ c = [110]

~~~
(78)

adJ' c transpose of CC
transpose of [i;~l
002 1
[ITO]
c- 1 = TcI = Icl = _2 =- -2 110
002 (79)

where cC=cofactor matrix =aij' au =(-1) i+j bnjjl

mij = minor corresponding to removal ofi th row andjth column

Note that c- I c=identity matrix, as required.

Since the basis (75) is related to the basis in (70) by a transformation matrix, the
matrix representation of the operation C 2 using the basis (75) will be related to that using the
basis in (70) by a Similarity transformation.

(80)
 461

~ IT01[01~
c~2) = - ~ 110 100 [110]
110
002 001 001

=-
~~ 002
ITOl
Il~ [0100
10]
001
=- ~200l
~ 02~
002

(81)

As expected, the matrix representation of C2(2) in (81) is identical to that given in

(76). Likewise, the matrices for E, crv and crv'using (75) as basis (and listed in (77) ) could be
generated from those using (70) as basis by a similarity transformation analogous to (80).

i
Note that the character or trace of C 2(2), c[C 2)]= 1, is identical to the character of
C2 (1), C[C2(1)]=1 . This is an example of an important general rule, the characters of 2
matrices that are related by a similarity transformation are identical. The character c of a matrix
is defined as the sum of the diagonal terms. This is an important relationship because later it
will be demonstrated that the characters of matrices representing symmetry operations provide a
great deal of information about the symmetry of the group.

Of the many possible representations of a group, one in particular is of considerable

value. This is known as the block diagonal representation. Here, the non-zero elements of the
matrix lie in square blocks along the body diagonal of the matrix. If each block cannot be
further simplified then the representation is said to be reduced and the individual blocks are
irreducible representations, referred to as irreps. The reduction of a representation will be
demonstrated using water once again as an example. However, before doing so it is necessary
to take a brief digression and develop something referred to as a character table for the point
groupC 2v

2.5.4. Character Tables

In brief, character tables list the characters of the matrices of the symmetry operations
for the irreducible representations of the group. Character tables can be formed with the aid of
a number of orthogonality relations. The principal one is the Great Orthogonality Theorem.
462

where m ij(l)(R) is the element located in the i th row and the j th column of the matrix of the
symmetry operation R in the 1 irrep of the group. h = number of operations in the group and
dF dimension of the representation 1 = number of functions in the basis= number of rows (and
columns) in the matrix representation of any symmetry operation of the group. The discussion
here will be limited to a verbal description of its meaning and a few examples of its validity.
The Great Orthogonality Theorem postulates 3 orthogonal relationships. The product of any
element in a matrix of one irrep and that in a matrix of the same symmetry operation of another
irrep, summed over all symmetry operations of the group, is zero for

(1) different irreps (orthogonality of irreps),

(2) elements in different rows of the same irrep (= orthogonality of rows of elements
of an irrep.)
(3) elements in different columns of the same irrep (= orthogonality of columns of
elements of an irrep.)

The concept of orthogonality is quite familiar to anyone accustomed to the use of

vectors. From that perspective, (82) may be viewed as a vector equation in which mij ( 1 )(R) is
-- (. . 1)
treated as the Rth component of some vector specified by the 3 indices l,i,j: fiR 1,], (82) is
rewritten as
h
"'-- (. '1). --fiR (·.·.1.)-ls:
L..fiR I,J, I,J,
s: s:
- d Uii'U~'Un'
R~ 1
(83)

Pairs of vectors with different values of indices are thus orthogonal. Note that the
d2
vectors are h-dimensional. Each matrix representation of a symmetry operation R contains 1
elements, all of which are orthogonal to each other. The total dimension of the vector
__
fir (i,j,1) d1
2 --
ill R I,J
( . .) Id~
1
is therefore and that of the vector i s . This number, of course,
cannot be greater than the dimension of the space that the vectors occupy which is h. In fact, it
may be demonstrated through (82) that

(84)

Equation(84) is the first oftwo relationships derivable from equation(82) that are used
to develop character tables. The other is obtained by considering a special case of (82) in
which i =j and i' = j'. Equation (82) becomes:

h
'" (1)(R)* fi(l')(R)
L..fiii ··
i i
_ls:
-d un,us: ii '
R 1
(85)
 463

Summing over all possible values of i and i' gives for the left side of (82):

I m/)(R)*m;/l')(R) = IIm~;(R)*Im~~;,(R) = Ix(j)(R)*ll')(R)

i,i',R R i i' R
(86)

The expression in (86) can be further simplified. A class is defined as a subset of a

group and consisting of symmetry operations in which each is related to another by a similarity
transformation. Consequently, the characters of the matrix representations of symmetry
operations within the same class are all equal. Hence, (85) may be written as
In(c)x(l)(c)*x(l')(c) = : On'
c 1

(87)

where the summation over R in (85) has been replaced by the summation over c in
(87) and n(c) = number of symmetry operations in the class C, The vector interpretation of
'1(1) Ix~) = I-v'i1Tc) X(I\C)*
(87) is that the vector given by C C is orthogonal to the vector
(1')
X given
. by

~_(1') ~ ct::\ (1')

£..Xc = £..vn(c)X (c):
C C

(88)

As was mentioned above in the vector interpretation of (82), the number of orthogonal vectors
cannot be greater than the dimension of the space they occupy. Thus, the number of classes,
_(j) d _(11
which is the number of orthogonal components of the vectors X an X of a group,
cannot be greater than the number of values of 1, which is the number of irreps (= the
dimensionality of the vector space). In fact, a more thorough analysis of (82) indicates that the
number of irreps is equal to the number of classes. From this statement and (84) the following
4 facts are obtained. These are used in constructing character tables which list the characters of
the matrices of the symmetry operators in each of the irreps of the point group, The columns of
the table correspond to different symmetry classes (recall that the matrix of each operation in the
same class possesses the same character) and the rows correspond to different irreps,

(1) The squares of the characters of a given symmetry operation summed for each
464

irrep is equal to bln( c).

(2) The sum of the squares of the characters of each symmetry operator in a given irrep
is equal to h.

(3) Rows are orthogonal.

(4) Columns are orthogonal.

As an example, the character table for the point group C 2v will be constructed. As
was demonstrated in section E.2, there are four symmetry operators in C2v and each occupies
its own class. Consequently there are four irreps in C 2v Equation(84) indicates that each
representation is one dimensional:
4

Ld~ = h = 4
1=1 (89)

With this information, the character table for C 2v appears as

C2v E C2 cry cr v '

r(l)
r(2)
r(3)
r(4)

rei) is taken as the identical representation so that each matrix consists of one term with a value
of 1. Now, since all the rows are orthogonal, the remainder of the table can be constructed.

Czv E C2 cry cr'

v

r(1)=A I
r(2)=A 2 -1 -1 (90)

r(3)=B 1 -1 -1
r(4)=B 2 -1 -1

2.5.5. R!.':Ql!!<tion Qf il R!.':l2~~!.':ntation

Sometimes it is desirable to reduce a given representation to a block diagonal form.

However, rather than search for a suitable transformation matrix which would be used to enact
a similarity transformation (that would transform each of the matrix representations of the
symmetry operators to block diagonal form), it is possible to use the character tables and one
additional relationship. The latter is obtained as follows. Since the original representation and
its block-diagonalized form are related by a similarity transformation, the characters of matrices
in each representation are identical. In addition, since each diagonalized block is an irrep, it
may be stated that
 465

X(R) = La lxCI)(R)
I (91)

where, cCR) = character of the matrix of the symmetry operator R in the reducible
representation, c(l)(R) = character of the matrix of the symmetry operator R in the lID
irreducible representation and al = the number of times the lth irrep appears in the block-
diagonalized form of the reducible representation. Since the rows of the character table are
orlhogonal,

LX(1)(R) . X(l')(R) = On.

R (92)

Hence, the value of al in (91) may be obtained by forming a "dot-product". Thus,

multiplying both side of (91) by c(l')CR) and summing over all R gives

Lx(l)(R)'X(l')(R) = LLaIX(I)(R)x(l')(R)
R R I (93)

(94)

By way of illustration, the representation ofC 2v using the basis (71), which consists
of Cartesian coordinates fixed to each nucleus, will be reduced. The matrices of this
representation are given in (72) - (74).

a A, = ± t-'C(R)xCA')(R) =±[x(E).c(A')(E)+X(C,)·x(A')(C,)+x(cr.)·x(A')(cr.)+x(a.. )X(A,) (a .. )]

= ~ [9 . 1 + (- 1)(1) + (1)(1) + (3)(1)] = 3

Similarly ,

a A , = ![(9)0) + (-1)0) + 0)(-1) + (3)(- 1)] = 1

a B, = t[(9)(1) + (-1)( -1) + (1)( - 1) + (3)(1)] =3

(95)
466

Thus, the reduction of the representation is summarized by

(96)

2.5.6. Symmetry of Functions

The symmetry of a function is described by the matrices that indicate how the function
is transformed by the action of the symmetry operators in the point group of interest. In this
section, therefore, the symmetry of a function will be determined by a process that is very
similar to that used to characterize the symmetry of a group. There a basis is first selected, then
a representation is formed and finally the representation is reduced (to a linear combination of
the irreps of the point group). In arriving at the symmetry of a function, the function is taken
as the basis then a representation is formed and finally the representation is reduced. For
example, consider the function x2 and the point group C 2y The representation of C 2v using
x 2 as a basis is:

C2x2 = (_x)2 =HJ 2 =[1]

svx~ (_x)2:=} crv =[1]

Thus, the matrix for each of the symmetry operators in the point group C2v is the one-
dimensional unit matrix. Hence, the character for each is 1. Through the use of (94), or
simply by inspection of the character table for C2Y' one can show that the representation of C2v
with x 2 as basis is the irrep A 1. Consequently, the function X 2 in the point group C 2v is said
to span the irrep A 1.

2.5.7. Direct Products

The ultimate objective of this discussion is the utilization of Group Theory to determine
which integrals of the form of (56) have non-zero values. This goal is nearly satisfied. All that
remains to determine is the symmetry of a product 'Jf a* H'Jfb when the symmetry of each of the
factors is known. This will be achieved through the use of an operation known as the direct
product and which is peculiar to matrices. First, consider the definition of the direct product of
two matrices, A and B.

direct product of A and B = A X B

Upper case X is the symbol used to denote the direct product.

(97)
 467

(98)

(99)

Thus, the direct product results from the product of every element in one matrix with
every element in another. If the dimension of one matrix is d A and that of the other is dB' then
the dimension of AX B is dAd B. One relationship which will be needed later is

(A XB)(CXD) =AC XBD (100)

Next it will be shown that if 'I' (1) and '1'(2) are two bases of representations r(l) and
r(2) , respectively, of a group, then '1'(1) X '1'(2) will also be a basis for the same group and
its representation will be r(l) X r(2). Thus, it is given that

R'I'(l) =M(l)(R) '1'(1) (101)

R'I'(2) = M(2)(R) '1'(2) (102)

The operation of the symmetry operation R on the set of functions '1'(1) is

mathematically represented by the multiplication of the matrix representation of the operation,
~1) (R), and the matrix '1'(1). The collection of the matrices M(l)(R) for the various
operations R produces the representation r(l) of the group. From the definition of what a
symmetry operator does,

R'I'(1)(x.y.z) = 'I'(l)(Rx. Ry. Rz)

'R'I'(1)(x.y.z)'I'(2)(x.y.z) = 'I'(l)(Rx. Ry. RZ)'I'(2)(Rx. Ry. Rz)

(103)

Similarly,
R['I'(1)(x.y,z)X '1'(2) (x,y.z)] = 'I'(1)(Rx. Ry, Rz)X 'I'(2)(Rx. Ry. Rz)

= R (1)'1' (l)(x,y.z) X R(2) '1'(2) (x,y.z)

'R[ '1'(1) X '1'(2)] =R (1) '1'(1) X R (2) '1'(2) (104)

From (90)
468

Equation (105) indicates that when '1'(1) and '1'(2) are bases of the representations
r(1) and r(2) , then '1'(1) and '1'(2) will be the basis for the representation r(l) X r(2).
The next questions are: what is the character of the representation r(l) X r(2) and is
the reduction of the representation r(l) X r(2) expressible in terms of r(l) and r(2). As
illustrated in (99) the direct product of two matrices can be described by

M (I) X M(2) =" £... m

(I)
ij
(2)
m kl
i,j,k,1 (106)

The diagonal elements of the direct-product matrix are those for which i =j and k =1,
simultaneously. The sum of these terms will be the character of the direct product matrix.

i.k k (107)

Thus, the character of the matrix of a direct product is equal to the product of the
characters of the two matrices involved in the direct product.

Whether or not the integral J'I' a'l' bdV has a non-zero value may now be determined by
Group Theory. 'l'a spans rea) and 'l'b spans r(b). 'I' a 'l'b = 'I' a X 'l'b' Therefore, 'l'a'l'b
spans rea) X r(b). Finally, if we have an integral of the form J 'I' a H'I'bdV, it may have a
non-zero value if the integrand spans the irrep A l' Since the Hamiltonian spans A 1 (because
the total energy of the molecule cannot change as a result of the operation of a symmetry
operator of the point group), the integral may have a non-zero value if

(108)

2.5.8. Selection Rules for Infrared Activity

A vibrational mode will be infrared active if it involves a transition between states a
and b and if the integrand 'V .Jl'V b spans A l' As an example, consider H 2 0. The water
molecule has three normal modes of vibration. These have the symmetries A l' A 1 and B l'
The electric dipole moment Jl is given by

IlZ=LZ iez i
(109)
 469

The character table for C 2v indicates that x spans B l' Y spans B2 and z spans A l'
Therefore, for transitions to occur from the ground state (which spans A 1) to one of the three
normal vibrational modes

AIXr(J..L)XAlmust= Al (110)

B 1X r(J..L) X Al must = Al (111)

For (110) to be true r( J..L) must contain A l' Since A 1 is spanned by z and, hence, by
J..Lz the two vibrational modes that span A 1 will be infrared active. For (111) to be true, r( J..L)
must contain B l' Since B 1 is spanned by x and, hence, by J..L x the vibrational mode that spans
B 1 will be infrared active. Thus, all three normal modes of vibration of the water molecule are
infrared active.

III. IR Spectroscopy of Metal-Electrolyte Interfaces

3. 1 . Introduction
The formalism for identifying the vibrational spectra of species that are IR-active was
developed in the previous section. The use of this procedure to extract information concerning
the structure and orientation of species absorbed on a surface will be illustrated in the present
section.

Perhaps the most obvious experimental difficulty encountered in ill-situ measurements

of the IR spectra of species located at the interface between a metal and an electrolyte is how to
eliminate the effect of the bulk solution on the measurement. In particular, how to distinguish
between the spectra originating from species located in the bulk of the electrolyte and the
spectra of interest - that originating from species located at the interface. In addition, since
water strongly absorbs IR radiation, there is the problem in reflection experiments of getting
sufficient intensity of radiation through the bulk solution, to the interface and back out through
the bulk solution.

Sources of infrared radiation may be either monochromatic such as a Nd- Y AG laser

(1.06mm) or continuous such as a ceramic rod heated to a high temperature. IR absorption
experiments require a source that emit a wide-range of infrared wavelengths. The production
of IR radiation is a consequence of two facts: the electronic structure of matter and the
influence which temperature has on the kinetic energy of molecules. As the temperature of a
substance is raised, the average kinetic energy of a molecule is increased. The distribution of
kinetic energy among the molecules follows the Maxwell-Boltzmann distribution law.
Consequently, while most molecules possess the same amount of thermal motion, a few have
kinetic energies which are either much larger or much smaller than the average value. Because
molecules are composed of charged particles, their acceleration and deceleration results in
emission of electromagnetic radiation. Planck was able to describe the energy distribution of a
blackbody radiation when he introduced the quantum hypothesis. This is illustrated in figure 5.
470

/\,
I \ T2
I ..
.,
f; I
I
. ...
... ...
CJ)
z .. ... ...
.. .. ..
I

,,
W I
f-
.... " .
I
Z I
I

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

WAVELENGTH

Figure 5. Intensity VS. wavelength as a function of temperature for

a blackbody.

In order to increase the intensity of the long-wavelength region of the spectrum it is necessary
to raise the temperature of the body. However, as its temperature is raised the average
wavelength is shifted to lower values. Consequently, the production of IR radiation becomes
less efficient as the brightness of the source increases.

1\vo of the more common sources of IR radiation are the Globar and Nemst
filaments.(23). The Globar is a rod of silicon carbide which operates at temperatures on the
order of 1200°C. A Nemst filament is mostly a mixture of rare earth oxides and is operated at
-1900°C. Both experience decreased lifetimes as their temperatures are raised.
 471

Absorption by the bulk solution is reduced by using a thin-layer cell in which the
thickness of the electrolyte is between 1 and 100 mm(24). While a thin layer of electrolytes
does permit a significant amount of IR radiation that is reflected off the surface of the metal
electrode to reach the detector, it is nevertheless still true that absorption by species in the bulk
of the solution overwhelms that caused by species located at the interface. To extract
information about the interfacial region, either (or both) the interface potential drop, d~, across
the electrode-solution interface or the polarization of the incident radiation is modulated. The
experimental results are then reported as difference

~R R(~02) - R(~01)
R= R(~l)
spectrum, e.g.,

where R( ~ J = the intensity of the light reflected from the surface when
~0 = ~ i. The polarization of the incident light is kept constant throughout the experiment.
The technique in which d~ is modulated is referred to as electrochemically modulated
IR spectroscopy. If a dispersive spectrometer is used for data acquisition the technique is given
the acronym EMIRS(25,26). The other technique uses Fourier transform infrared
spectroscopy (FTIR). Its acronym is SNIFTIRS, meaning subtractively normalized interfacial
FTIRS(27-31).

The relative performances ofFT and dispersive spectrometers have been compared by
A.T. Bell who pointed out that a high signal to noise ratio and the simultaneous observation of
all spectral elements are the primary advantages provided by an FT spectrometer(32). These
characteristics ideally suit FT spectroscopy to the study of species adsorbed on a surface

Polarization modulated infrared reflection-absorption spectroscopy uses the acronym

IRRAS(33-36). In this technique the specimen is alternately illuminated by sand p polarized
radiation. As discussed in the next section, s-polarized radiation does not excite any IR modes
of species adsorbed on the surface of a metal. p-polarized radiation can excite IR vibrational
modes in species adsorbed on metal surfaces as well as in species located in the bulk of the
liquid phase. Consequently, the spectra of species located at the surface is obtained from the
difference between the spectra obtained with s and p polarized radiation.

3.2. Influence of the Polarization of Radiation

IR radiation that is s-polarized (i.e., its electric field vector oscillates in a direction
perpendicular to the plane of incidence) undergoes a 180' change in phase when it is reflected
off a metal surface regardless of the angle of incidence. Consequentlr.:, the standing wave
formed from the incident and reflected s-polarized light has a node (lEI = 0) at the metal
surface. Thus, s-polarized radiation will not provide any information about the IR spectra of
species present on the metal surface.

The polarization of p-polarized radiation (*i.e., electric field vector oscillates in a plane
parallel to the plane of incidence) is shifted by reflection off a metal surface. The magnitude of
the change in polarization is a function of the angle of incidence. At grazing incidence the
incident and reflected waves are phase shifted by 1800 and would combine to produce a node at
the metal surface. As the angle of incidence decreases, the difference in polarization between
the incident and reflected waves rapidly decreases. At normal incidence the change in phase is
very small but because the directions of propagation of the incident and reflected waves are
472

now 180 0 apart, a very low intensity once again exists at the metal surface. As the angle of
incidence increases, the intensity of standing wave at the interface increases to a maximum
value that is close to 4 times that of the incident beam at an angle of incidence near 80' (i.e.,
close to grazing). Consequently, p-polarized radiation may provide a good deal of information
about the IR spectra of species present on the metal surface(37).

Both s- and p- polarized radiation will be absorbed equally by randomly oriented

species present in the bulk of the solution. Thus, by modulating the polarization of the incident
radiation between s- and p- polarized, the IR spectra of species located on the metal surface can
be obtained. Even though the intensity ofp-polarized light and hence the amount of absorption
of radiation will be greater for a species that is suitably oriented (see below) at the interface than
for that species located in the bulk of the solution, the measured IR spectrum is dominated by
the species located in the bulk because the amount of species in the bulk is so much greater than
the amount located on the electrode surface for reasonably concentrated solutions. However,
because of the influence which the value of L\~ has on the IR spectra of species present near the
metal surface (see below), the IR absorption properties of species present on the metal surface
may still be measured from the difference of spectra obtained with p-polarized light and at two
different values of L\~ (i.e., potential modulation).

3.3. Influence of Molecular Orientation on Metal Surface

It was mentioned in the previous paragraph that only species which are suitably
oriented on the surface will be infrared active. Recall that the IR selection rules require that the
vibrational mode have a non-zero value of the dipole moment derivative. A molecule that is
vibrating in a mode whose change in dipole moment is parallel to the plane of the surface will
not be active since this mode could not be excited by p-polarized radiation (see below for
exception). Consequently, a mode which is IR active when the species is located in the bulk
solution may become IR inactive if the dipole moment is located parallel to the surface.
Similarly, if the molecule is bonded to the surface, a new absorption peak corresponding to
vibration of the adsorbate against the substrate may be IR active.

It is necessary to point out that even though the vibrational mode may cause a change in
dipole moment that is parallel to the surface, this mode may still be IR active if the molecule is
highly polarizable(38). For example, if vibration in the normal mode whose coordinate Q p is
parallel to the surface results in a change in the substrate-adsorbate bond length, then the Qp
vibrational mode will result in a change in the dipole moment measured normal to the surface.
Thus, the Qp mode may still be IR active even though it is parallel to the surface.

If the Qr displacement vector of the rth normal mode of vibration (which is IR active
in a free molecule) were oriented at some angle other than O' or 90' with respect to the plane of
the interface, then absorption of p-polarized IR radiation would occur to some intermediate
extent. Consequently, the sensitivity of the IR activity of a given vibrational mode to the
orientation of the normal modes with respect to the surface can be used to follow changes in
the orientation of the adsorbed molecule.

3.4. Influence oflnterfacial Potential Drop

The discussion up to this point has described the influence of the incident polarization
on the activity of the IR vibrational modes of adsorbed molecules. Earlier, it was also
mentioned that the IR spectra of molecules on the surface of a metal could also be obtained by
modulation of the interface potential difference, L\~. In general, changes in reflectance of the
metal surface caused by changes in L\~ are the result of four factors(38): (1) changes in the
number of adsorbates at the metal surface and in the bonding between adsorbate and metal, (2)
changes in the migration of anions and cations into or out of the surface region, (3) changes in
 473

the rates of electrochemical charge transfer reactions that either produce or consume soluble
species or films, (4) changes in the reflectivity of the metal surface. The last one is a very
small effect and is usually considered negligible. These four effects must be considered when
interpreting the spectra obtained from potential modulation experiments.

3.5. IR Spectroscopy of CO Absorbed on Metal Surfaces

Before considering examples of IR spectra obtained from species located at the

interface between a metal and an electrolyte, the large amount of data obtained from IR
spectroscopy of CO adsorbed on metallic catalysts will be briefly reviewed.(39) This will
illustrate the type of information about adsorption which can be obtained.
As a free molecule, CO possesses one internal vibrational mode (3N-5=1). This is the
C-O stretching mode which was considered earlier in the discussion of the vibrational modes of
a harmonic oscillator. The reduced mass of C-O is 1.139xlO-23 gm/molecule. The force
constant of the C-O bonds is 1.88x 106 dyne/cm. Hence the vibrational frequency of CO will
be -2155 cm -1. The IR spectra obtained from CO absorbed on a metal surface generally is
considerably different from a single peak at 2155 cm -1 and depends on the identity of the metal
substrate and the atomic structures of its exposed surfaces. For example, when small quantities
of CO are absorbed on the surface of a metal such as nickel, low wavenumber bands (in the
range 1880-1650 cm -1) appear in the IR reflectance spectra.(39) These peaks have been
attributed to CO bonded to several (n) surface atoms. Such CO species are said to be n- fold
bridge bonded to the metal.(40) Presumably the shift in frequency results from weakening of
the C-O bond as a result of bonding of CO to a greater number of metal atoms. As the amount
of CO increases it adsorbs at sites of lower values of n. Absorption peaks in the range 1800-
2000 cm -1 are thought to arise from CO species bridging two metal atoms. This trend of
increasing frequency with decreasing number of bridge bonds is consistent with the CO/Ni
matrix isolation studies of Moskovits and Hulse(41-43) wherein infrared bands were assigned
to complexes of the form NinCO.

With increasing coverage of the metal surface by CO, the amount oflinearly bonded
CO (i.e. bonded to a single metal atom) increases. The band width of absorption peaks
associated with linear-type CO is large (2000-2130 cm- 1 ) due to the variety of one-atom sites
available for absorption. For example, an atom located on the surface in a (111) plane has a
greater number of metal-metal bonds than an atom located in a (100) plane in an fcc metal.
Consequently, in the latter case the strength of the metal-CO bond will be greater and the C-O
bond will be proportionately weaker. Thus the IR absorption associated with the C-O vibration
will occur at a lower frequency.
Summarizing, the frequency of the internal mode of CO increases as the amount of CO
adsorbed on the surface increases because the fraction of CO present in bridged sites
decreases. There are two additional phenomena which can also cause changes in the C-O
vibrational frequencies as the degree of surface coverage changes. The first is an effect of
bonding between CO and the substrate, and the second results from dipole coupling between
adsorbed CO molecules. According to an hypothesis advanced by Blyholder(44), strong
bonding between adsorbed CO and the metal substrate is accompanied by back donation of
electrons from the metal d states into the antibonding 2 IT* orbital of the CO molecule. This
weakens the C-O bond and decreases the frequency of the C-O stretching mode. As the
amount of adsorbate increases, the strength of the bond between CO and the substrate may
decrease resulting in an increase in the C-O vibrational mode with coverage. The dipoles of
adjacent molecules of adsorbed CO may couple and this will be a second cause for changes in
the vibrational spectra with increasing coverage.(45,46) For a linear array of n- molecules,
there will be n normal modes of vibration. Since the intensity of absorption resulting from the
474

Ith nonnal mode is proportional to the square of the change in the dipole moment during the
vibration, the strongest absorption peak will be associated with the nonnal mode in which all
parallel oscillators are in phase. The frequency U 1 of the IR-active mode associated with N
coupled dipoles adsorbed on a metal surface is given by(4)5

where rij =distance of separation of molecules i and j and U 0 = frequency of the internal mode
of an isolated molecule. The equation indicates that the frequency is shifted to progressively
higher values as the number, N, of adsorbed molecules increases. Experiments suggest that
dipole coupling produces a significant shift in frequency of the C-O vibration on platinum
(111).(47)

When the metal substrate on which CO adsorbs is immersed in an aqueous solution

several phenomena that are important to the process of adsorption occur. First, the presence of
the electrolyte gives rise to an interfacial potential drop whose value can be varied by changes in
solution chemistry or by passing electric current into the interface using an external source.
The change in the potential drop at the interface will, in tum, produce changes in the
concentrations of adsorbed species as well as changes in their geometries and orientations.
These effects can be studied and used to provide infonnation about the adsorbate and the nature
of substrate-adsorbate interactions. For example, Kunimatsu ~.(35,36) used polarization
modulated FT-IRRAS to study CO adsorption on Pt in several acidic solutions. The location of
the absorption band shifted to higher frequcncics with increasing potential difference across the
platinum electrode-electrolyte interface. They concluded that the frequency shift of
"" 3O/cm -1 N was not the result of eithcr coupling of the vibrating CO dipoles nor decreased
back donation of metal electrons into the antibonding 2P* orbital of adsorbed CO. Instead,
they attributed the potential dependent frequency shift to the first -order Stark effect as was
found by Lambert for CO absorbed on Ni (110) in UHV.(48,49)

3.6. In Situ Investigations ofAdsorption on Metal Electrodes in Aqueous Solutions

EMIRS was used by Bewick and colleagues to study .in ID.tu the adsorption of
hydrogen on platinum in aqueous acidic solutions.(50,51) These results supported the
existence of two basic states of adsorbed hydrogen: one strongly bound and the other weakly
bound. The weakly bound hydrogen appeared to exist in the inner Helmholtz plane bondcd to
water molecules through the lone pair molecular orbitals. The strongly bound hydrogen
behaved metallic. Its electron appeared to be associated with the conduction band of the metal.
Experiments by other workers(52,55) suggest that the strongly bonded hydrogen does not exist
on (111) surfaces of platinum and it was concluded by Bewick et al.(50) that strongly bonded
hydrogen occurs at high coordination sites, such as holes in the surface.

A study by Pons and Bewick(52) further illustrates the type of infonnation that is
obtainable by IR spectroscopy. Figure 6 depicts the SNIFTIRS difference spectrum obtained
from p-difluorobenzene (P-dfb) adsorbed on a platinum electrode in 1.0M HCI0 4 . The data
was generated by subtracting spectra obtained at <'\0 =+O.4V (NHE) from those obtaincd at <'\0
=+O.2V(NHE). The spectra in figure 6 can result from two effects. First, as the potential
changes, corresponding changes may have occurred in the amount of p-dfb adsorbed. Second,
there may have occurred changes in the bonding within chemisorbed p-dfb and between
chemisorbed p-dfb and the platinum electrode with potential variation. Figure 7, which was
taken from Bewick et al.,(58) is included to help in interpreting the difference spectra in
 475

LlR > 0
figure 6. The peaks in figure 6(a-c), for which R indicate that as Ll¢ was increased the
amount of adsorption was decreased. The features in figure 6d, in which a peak with
LlR > 0 LlR < 0
R is coupled with a peak for which R suggest that as Ll¢ increased the
absorption peak either shifted to lower frequencies or else the peak broadened towards lower
frequencies.

The shifting of an absorption peak to lower frequencies suggests that the force
constant associated with the vibrational mode has been weakened. Thus, IR spectra (such as
that shown in figure 6) which indicates the influence of Ll¢ on IR absorption may provide
information on the influence of Ll¢ on bonding within the adsorbate and between the adsorbate
and the metal electrode. In addition, the influence of chemisorption on bonding can be
ascertained by comparing the IR spectra of molecules in the gas phase or in the bulk solution to
those adsorbed molecules. Again, any shifts in band positions may provide information on
bonding within the chemisorbed species and between the adsorbate and substrate.
LlR(O
Features in the IR spectra which consist of a single feature or peak for which R
result from the amount of absorption having decreased as Ll¢ was increased. This could result
from a decrease in the amount of adsorbate at the higher value of Ll¢ or possibly from a change
in orientation of the molecule on the metal surface. The latter would cause the orientation of the
derivative of the dipole moment associated with the vibrational mode to move away from the
surface normal.

Thirteen of the normal vibrational modes of "free" molecules of p-dfb are IR active.
Depending on the orientation of the molecule on the metal surface, not all of these thirteen
modes will be IR active. In particular, the b lu and b 2u modes cause dipole changes in the
plane of the benzene ring. These are not IR active at either +O.2V or to +0.4 V, suggesting that
the molecule lies flat on the metal surface.

The spectra in figure 6b also provides information on the nature of the metal-adsorbate
bonding. The coupled positive and negative peaks are associated with two of the b 3u modes.
These absorption peaks occurred at 833 and 509cm -1 at +0.2V but were shifted to 811 and 495
an -1 at Ll¢ =+O.4V. In addition, as shown in figure 7 when the maximum value of Ll¢ is raised
to more positive values during modulation of Ll¢ the absorption peak is shifted to lower
frequencies. Thus, as the metal surface is made more positive, the bonding within the b 3u is
weakened. Pons and Bewick(57) interpret these mcasurcments in terms of molecular bonding
models of adsorption.

During adsorption charge, transfer is thought to occur from filled d levels in the metal
to empty IT* levels in the molecule. As the metal surface is made more positive this electron
transfer will increase. The result is a weakening of the P bonding within the adsorbate.
Consequently, vibrational modes which stretch IT bonds in the adsorbate will shift to lower
frequencies with increasing Ll¢. In addition to charge transfer from metal to adsorbate,
electrons may transfer in one of two ways from the molecule to empty states in the metal. They
may leave either filled IT* or IT states in the molecule. The former will result in a strengthening
of IT bonding within the molecule and the latter will result in its weakening. Thus, when the
metal surface is made more positive, if electrons are leaving IT* states, the location of the
absorption peak will increase in frequency. If, however, electrons are leaving IT states, then
the location of the absorption peak will decrease in frequency with increasing Ll¢.
Consequently, the present results indicate that IT bonding is weakened as the metal surface is
476

1300

I~.o IO-~
I

1300 1100
WAVENUMBERS. CM"

601

AR/R

800 600
WAVENUMBERS. CM· 1

Figure 6(a,b). SNIFfIRS difference spectrum of p-difluorobenzene at a platin~m electrode in

LOM HCI04. Potential was modulated between +D.2V and +DAV vs. NHES .
 477

O.2''IIH~
f

~,
3050

I~.O IO-~
I

OAf

3000 2900
WAVENUMBERS. eM·1

''''
1431

ARfR

l~.o IO-~
I

1500
WAVENUMBERS. eM· 1

Figure 6(c,d). SNIFrlRS difference spectrum ofp-difluorobenzene at a platinum electrode in

1.OM HCI04. Potential was modulated between +D.2V and +D.4V VS. NHE53.
 478

O.2Y

toR/R

4.0 x 10-4

500 400
WAVENUMBERS. CM- 1

Figure 7. SNIFTIRS difference spectra of the b 3u defOlwation mode of p-difluorobenzene.

Potential was modulated between +O.2V and the 4 different upper limits indicated in the
figure 53.

made more positive. which suggests that bonding between platinum and p-dfb is dominated by
charge transfer from d levels in the metal to empty IT* states in the molecule.

The foregoing discussions illustrate the use of .in-situ IR spectroscopy in providing

information about bonding within an adsorbate and between an adsorbate and metal substrate.
Similar experiments would help elucidate the mechanism of inhibition of corrosion by organic
additives, e.g., benzotriazole adsorbed on copper.
 479

3.7. In Situ Inyestigations of Corrosion Product Films Fonned in Aqyeous Solutions

The fonnation of corrosion product films often result from the oxidation of metals in
aqueous solutions. More in-situ infonnation on the nature of these films and their
mechanism(s) of fonnation is required in order to improve our understanding of the processes
of oxidation, passivation and inhibition. Several studies can be identified as examples of the
use of IR spectroscopy to obtain infonnation about oxidation films.

Knolle and Maxwell(59,60) and Nakamura ~.(61) have perfonned IR absorption

experiments to detennine the amount of oxygen present in silicon oxide films fonned by
chemical vapor deposition from SiH 4 and N2 0 gases. In addition IR measurements were
effective in following changes induced by heat treatment in the oxidation state of silicon in such
films.

External reflectance IR spectroscopy was used by Ottesen (62) to study the fonnation
of thennally grown oxide films on chromium and iron. Both the composition and film
thickness of simple oxide films were detennined by this technique for film thicknesses of
-O.OI-lOmm.
Lezna ~.(63)investigated in situ by EMIRS the structural change of iridium oxide
that occurs during transition from colored to bleached states. Films of iridium oxide were
,
grown to thicknesses of approximately 1500 A by potential cycling iridium electrodes between
o and 1.45V (NHE) at 100mV/s in O.5M H 2SO 4 . The coloration and bleaching times of
iridium oxide films in aqueous solutions are very fast (= 40rns). During the EMIRS
experiments, the potential of the oxide-film covered electrodes of iridium immersed in 0.5M
H2 SO 4 were cycled from the region in which the bleached state exists into the rcgion in which
the colored state exists at a frequency of 11 Hz. The experiments revealed the presence of an
O-H stretching mode when the oxide was bleached. This was attributed to the existence of
lrOOH.

The above discussion is not intended to serve as a review of the literature on IR

spectroscopy of adsorbed species. For interested readers there are several excellent review
articles that may be consulted. Rather, the objectives of the present paper are to illustrate the
analysis of vibrational spectra and to illustrate its ability to supply in-situ infonnation about
metal/electrolyte interfaces that are of interest to corrosion scientists.

Appendix I Fundamentals of OUantum Mechanics

Many challenging areas of research in corrosion do not require a knowledge of
quantum mechanics so this is a subject that may be unfamiliar to many corrosion scientists.
Consequently, a brief summary of those aspects of quantum mechanics that are critical to the
understanding of the phenomena occurring during experiments in infrared absorption and
Raman scattering will be presented here. The discussion requires the use of some mathematics
which will greatly facilitate the analysis. We begin with a definition of eigenfunctions and
eigenvalues. If the result of an operator R acting on a function f is the product of a constant c
with the function f, then f is said to be an eigenfunction of the operator Rand c is an
eigenvalue. For example, if R = d/dx, then exp(cx) is an eigenfunction of Rand c is an
eigenvalue. The function f = sin bx would not be an eigenfunction of the operator but it would
be an eigenfunction of the operator d 2/dx 2 and the eigenvalue would be _b 2 .
480

Next, rather than introduce quantum mechanics by hypothesizing that Schrodinger's

equation specifies the energy of a system, three postulates are made, from which Schrodinger's
equation can be derived. More useful, however, is the fact that this approach utilizes the
concept of operators which allows, among other benefits, a ready identification of
experimentally measurable quantities and an efficient proof of the orthogonality of wave
functions. Both of these are required for a discussion of molecular spectroscopy. The three
postulates are:

(1) The wavefunction Y (r l' r2' ... r n;t) for a system composed of n particles contains
all of the information about the system that can be measured experimentally. rl = position
coordinate of the ith particle.

(2) (a) Position and momentum do not commute. Classically, xp - px = O. However,

Heisenberg postulated that rather than precisely equaling zero, xp - px actually equals a very
small number: xp - px =ih.

(b) Any experimentally measurable quantity satisfies the relationship

xp -px = ih (AI)

Note that (AI) suggests that the form of xp-px must be something other than that used
in classical mechanics where xp - px = xmv -mvx since these classical quantities do commute.
Matrices in general do not commute and conscqucntly (A 1) suggests that x and p arc matrices.
In fact, x and p act as operators.

(i) xf= "x operating on the function f' = product ofx and f
(ii) Px6 = "momentum operating on the function f' = product of P x and f

(3) The expected value of the observable R for the system in state "i" is given by

f\fl:R\fIi dV
where 'I'i is a normalized wave function.

In addition to the above three postulates, much use will be made of Hermctian
operators. An operator R is Hermetian if

f \jIrR\jIjdV = [J \jIjR\jIjdV r
Hermetian operators possess a number of important properties. The first to be
considered is that the eigenvalues of Hermetian operators are real. Secondly, is the
eigenfunctions of a Hermetian operator are orthogonal. The validity of these two statements
can be readily shown,i.e.,

(A2)

where \jim and \jIn are eigenfunctions of the operator Rand r m and r n are the
respective eigenvalues. (A2) may be converted into
 481

Integrating over all space gives

J'l'"nR ,¥ mdV = J'l'"n r m'¥mdV = r mJ '¥'n '¥ mdV and J '¥~ R'¥n dV = J'¥~ r n'¥ ndV =r nJ'¥: '¥n dV
(A4)

Since R is Hermetian,

Therefore,

(A6)

Considering the case in which m =n, (A6) becomes

fm =r~
(A7)

The only way in which (A7) could be true is if r m is real. Thus, the eigenvalues of
Hermetian operators are real. Consequently, any measurable property of a system (e.g., the
momentum of a group of particles) must be represented by a Hcrmetian operator.

A second useful property of Hermetian operators is obtained by returning to (A6).

Since the eigenvalues are real, (A6) may be rewritten as

(AS)

States n and m are different,so that r m :t:. r n. Hence, the only way (AS) can be true is if

(A9)

Therefore, the eigenfunctions of an Hermetian operator are orthogonal.

The sum of the kinetic and potential energies of a system is measurable, so that the
operator representing this sum (referred to as the Hamiltonian, H, of the system) must be
Hermetian. The eigenfunctions of the Hamiltonian are, therefore, orthogonal and its
eigenvalues (the total energies) are real.
482

Rr1erences

1. Kruger, I. (1963) I. E1ectrochem. Soc., 110, 654.

2. Kruger I. and Calvert J.P. (1967), J. E1ectrochem. Soc., 114,43.

3. Draper, P. M. G. (1967) Corros. Sci., 7, 91.

4. Cahan B. D. and Spanier R. F. (1969) Surf. Sci., 16, 166.

5. Caham, B. D. (1976)Surf. Sci., 56,354.

6. Aspnes, D. E. (1980) Surf. Sci., 101, 84.

7. Foley J. K., Korzenisewski c., Daschbach J. L., and Pons S. (1986) 'Infrared
'vlbrational Spectroscopy of the Electrode-Solution Interface', in
Electroanalytical Chemistry, v. 14, p.309.

8. Thibeau R. I., Brown C. w., Goldfarb A. Z., and Heidersbach R. H. (1980) J.

Electrochem. Soc., 127,37.

9. Thibeau R. J., Brown C. w., Goldfarb A. Z., and Heidersbach R. H. (1980) J.

Electrochem.Soc., 127, 1702.

10. Thibeau R. J., Brown C. w., Goldfarb A. Z., and Heidersbach R. H. (1980)
J. Electrochem.Soc., 127, 19l3.

11. Keter J. J., Furtak T. E., and Bevolo A. J. (1982) J. Electrochem. Soc., 129,1716.

12. Dorain P. B., von Raben K. U., and Chang R. K., Surf. Sci., 148, 439 (1984).

13. Gardiner D. J., Gorvin A. C., Gutteridge C., Jackson A. R. w., and Raper E.S.
(1985) Corros.Sci., 25, 1019.

14. Shen, Y. R. (1985) in 'New Laser and Optical Investigations of Chemistry and
Structure at Interfaces', eds. R. B. Holland, B. B. Ellis.

15. O'Grady, (1980) W. E. J. Electrochem. Soc., 127,555.

16. Kruger J., Long G. G., Kuriyama M., and Goldman A. 1. (1983) 'Passivity of
Metals and Semiconductors', M. Froment Ed., 163; Amsterdam, Elsevier..

17. Pauling L. and Wilson E. B. (1935) 'Introduction to Quantum Mechanics'. McGraw-

Hill Book Company, N.Y.

18. Wilson E. B., Decius J. C., and Cross P. C. (1955) Molecular Vibrations, McGraw-
Hill Book Co., NY.

19. Cotton, F. A. (1971) Chemical Applications of Group Theory, 2nd Edition, J. Wiley
& Sons, NY.

20. Atkins, P. W. (1983) Molecular Quantum Mechanics. 2nd Edition, Oxford University
Press,Oxford.

21. Graybeal, J. D. (1988) Molecular Spectroscopy, McGraw-Hill, Inc., NY.

 483

22. Born M. and Oppenheimer J. R (1927), Ann. Physik, 84, 457.

23. Co1thup N. B., Daly L. H., and Wiberky S. E. (1964), Introduction to Infrared and
Raman Spectroscopy, Academic Press, NY.

24. Seki H., Kunirnatsu K., and Golden W. G. (1985) Applied Spectroscopy, 39,437.

25. Bewick A., Kunirnatsu K., and Pons B. S. (1980) Electrochernica Acta, 25, 465.

26. Bewick A., Kunirnatsu K, Pons B. S., and Russell J. W.(1984) J. Electroanal.
Chern., 160,47.

27. Davidson T., Pons B. S., Bewick A., and Schmidt P. P. (1981) J. Electroanal.
Chern., 125,237.

28. Pons, S. (1983) J. E1ectroanal. Chern., 150,495.

29. Christenssen P. A., Hamnett A., and Trevellick P. R (1988) J. Electroanal. Chern.,
242,23.

30. Neugebauer B. N., Nauer G., Brinda-Konopik N., and Gidaly G. (1981) J.
Electroanal. Chern., 122, 381.

31. Aurian-Blajeni B., Habib M. A., Taniguchi I., and O'M. Bockris J. (1983) J.
Electroanal. Chern., 157,399.

32. Bell, A. T. (1980) in Vibrational Spectroscopies for Adsorbed Species, p. 13, A. T.

Bell and M. L. Hair, eds. American Chemical Society, Washington, D.C.

33. Russell J. W., Overend J., Scanlon K., Severson M., and Bewick A. (1982) J. Phys.
Chern., 86, 3066.

34. Russell J. W., Severson M., Scanlon K., Overend J., and Bewick A. (1983) J. Phys.
Chern., 87, 293.

35. Kunirnatsu K., Seki H., Golden W. G., Gordon J. G., and Philpott M. R (1985)
Surf. Sci.,158, 596.

36. Kunirnatsu K., Golden W. G., Seki H., and Philpott M. R(1985) Langmuir, 1,245.

37. Greenleer, R. G. (1966) J. Chern. Phys., 44, 310.

38. Foley J. K .• Korzeniewski C., Daschbach J. L.. and Pons S. (1986) in

Electroanalytical Chemistry, v. 14, A. J. Bard, editor.

39. Sheppard N. and Nguyen T. T. (1978) in Advances in Infrared and Raman

Spectroscopy, v. 15, Heyden and Son.

40. Eischens R P., Francis S. A., and Pliskin W. A. (1956) J. Phys. Chern., 60, 194.

41. Hulse J. E. and Moskovits M. (1976) Surf. Sci., 57, 125.

42. Moskovits M. and Hulse J. E. (1976) Surf. Sci., 61, 302.

43. Moskovits M. and Hulse J. E. (1977) J. Phys. Chern., 81,2004.

PASSIVATION AND LOCALIZED CORROSION

MARIO G.S. FERREIRA and ALDA M.P. SIMOES

Dept. of Chemical Engineering
Instituto superior Tecnico
Av. Rovisco Pais
1096 Lisboa Codex / PORTUGAL

ABSTRACT. A review is made of the main models proposed

for the passive state, and also for the breakdown of
passivity leading to localized forms of corrosion like
pitting and crevice corrosion. Some particular aspects of
crevice corrosion are analysed and the use of
electrochemical techniques for its study is illustrated,
like potential and current monitoring, electrochemical
noise and electrochemical impedance spectroscopy (EIS).

1. Passivation

The passive state, although known since the 18th

century, is still a controversial subject.
The first attempt to explain the phenomenon was due
to Faraday [1), more than one century ago, and since then
numerous studies have been carried out leading to several
theories [2). For many years there was great opposition
485
M. G. S. Ferreira and C. A. Me/endres (eds.).
Electrochemical and Optical Techniquesfor the Study and Monitoring of Metallic Corrosion. 485-520.
~ 1991 Kluwer Academic Publishers.
486

between two models: the adsorption model, which described

the passive state based on the adsorption of an oxygen
film on the metal surface, and the oxide film model, which
sustains the formation of a protective oxide film.
Recently, the advent. of spectroscopic techniques,
like Auger electron spectroscopy and XPS, which give
information on the composition of small thicknesses of
matter, has decided in favour of the oxide film theory,
and consequently the discussions have become centered on
other factors, such as the degree of crystallinity of the
passive film and its composition.

1.1. Models for Passivation

Although at present it is accepted that passivation

is a consequence of the presence of oxides on the surface,
a large variety of theories have been proposed involving
different structures and compositions.
Those theories can be considered in two groups:
A) Model of the Crystalline Oxide
B) Model of the Hydrated Polymeric Oxide

1.1.1. Crystalline Oxide Model. This model explains the

passivation based on the formation on the metallic surface
of a crystalline oxide. In the case of iron the films
are composed of either V-Fe203 or Fe203.nH20 or Fe304,
y- Fe 203·
Thecrystalline oxide is assumed in many cases to
have a spinel structure with permeability to ionic
transport. The presence of water in the films formed in
aqueous medium is interpreted as being due either to H20
adsorption in the film-solution interface [3] or to a
process of stabilization of Y-Fe203 by hydrogen in the
outer layers of the film [4].
Thus, the passive film usually contains hydrogen in
 487

addition to metal ions and oxygen, hydrogen being present

in the form of protons, OH- or water. The presence of Cl
and S04 increases the amount of bound water.
Concerning its composition in depth the film can be
considered as homogeneous [5) or having a stoichiometry
continuously changing in thickness [6) or even a sandwich
type of structure [4,7].

1.1.2. The sandwich structure. This model is based on the

fact that the Fe 3 + oxide constituting the passive film is
thermodynamically unstable when in direct contact with
metallic iron [2). The sandwich oxide structure appeared
by analogy with the oxidation in the air at high
temperatures where the film is composed of an outer layer
of a-Fe 203 and an inner layer of Fe304 and was initially
proposed by Nagayama and Cohen [7) .
Several studies
however indicate that for passive iron, the outer layer
may be of Y-Fe203 [5,8].

1.1.3. Polymeric hydrated oxide model. According to this

model, the water has a key role on passivation, acting as
a binder between polymeric oxide chains, in an essencially
amorphous structure.
According to Okamoto [9] the film shows a gel-type
of structure (Fig. 1.).
In the regions where the film is not present, the
production of the metal ions takes place with the
formation of MOH+ species. This species is then captured
by the involving water and precipitates as a solid film
with a high degree of hydration. The anodic polarization
or the ageing of the film leads to the loss of protons
(Fig. 1a).
488

/,
/
o
,OH
-O~M~OH.
"
-O-M-OH.
/
"
OH,OH.
/

MOH+(H,O)

(a)

'"/
-O-M-OI
o/ '"OH
'"
-O-M-CI
'/
/

01 01
,
'Mbl+(H,O)-+MOH++OI-+W
61, 61/
-O-M-OI
, "
o/ /'OH
-O-M-OI
/ "
( b)

Figure 1. Polymeric Oxide Model (a) Dissolution of

metallic ions leading to film formation; (b) Cl ions
replace water molecules resulting in breakdown of the film
(after Okamoto [9).

Okamoto postulates the existence of three kinds of

bridges between metal ions: H20-M-OH2, -OH-M-OH- and
-O-M-O-. The H20-M-OH2 structure, corresponding to a
higher degree of hydration, is the more reactive.
 489

Okamoto's model was originally proposed for

stainless steel, but has also been applied to
iron [10-12].
Table 1 taken from Szklarska-Smialowska's [2], shows
the composition and structure of passive films on iron as
determined by several authors, the lack of agreement being
evident.

1.2. Stainless Steel

The passive films formed on stainless steel are

thinner than those formed on pure iron, show a more
complex composition, and in addition cannot be reduced
cathodically [2], which makes their study quite difficult.
Although it is universally accepted that the films
formed on stainless steel are enriched in chromium, many
doubts still persist concerning the film composition and
structure, as well as the role of the alloying elements.
For example, the reason for a critical concentration of
13% of chromium, responsible for inoxidability is still an
ambiguous point.
Concerning film structure, there are various studies
pointing towards a "duplex" (or sandwich) type of oxide
with an inner layer enriched in chromium and an outer
layer enriched in iron [35-37]. However, XPS measurements
on passive ferritic stainless steel at the corrosion
potential revealed a maximum chromium content in the outer
layer [38].
There is also indication that the increase of
chromium in the alloy decreases the thickness of the film
and is responsible for a certain amorphous character [2].
The chemical composition of chromium in the film is
not clear, the literature indicating species like
[39,40], cro3 [36], CrO [39], Cr(OH)3 [36],
CrOOH [41] and also the mixed oxides (Fe,Cr)203, FeCr204
[35] and Fe1+xCr2-x04 [2].
490

On the other hand, factors such as the formation

potential are known to influence the composition and
structure [35,42] of the passive film.

TABLE 1. Cauposition and Structure of Passive Films on Iron According to Different Investigators [2]

Experimental Experimental Film

Authors Year Techniques Conditions Composition

Foley et al [5] 1%7 Electron diffraction Borate buffer pH 8.4

NaOH pH 11.4
E = 0.5 V
H~4 pH 1.1
E = 1.4 and 1.6 V

Bockris et al. 1971 Ellipsolll!!try and Borate buffer

[13] electrochemical E above - 0.25 VSCE

Osbe et al. [14] 1971 Pbotoelectricity 1 N Na2S04 Close to Pe~ wi tb a small

to amount of defects Fe~_y
1979

Akimv et al. [15] 1966 Reflection Borate buffer

spectroscopy

O'Grady and 1973 l!O.ssbauer Borate buffer pH 6.8 Polymerized bydrated

Bockris [16] spectroscopy E = 0.3 VNHE iron o)(ide

Revie et al. [17] 1975 Auger electron Borate buffer pH B.1 As above
spectroscopy E = 0.05 VSCE

Revie et al. [18] 1975 Auger electron As above Pol1"2~ized fe~. absence
spectroscopy of Fe

O'Grady [19] 1980 !!Ossbauer Borate buffer pH 8.4 AIOOrpbous polymerized

spectroscopy E = 0.35 VSCE Pe404(OH)4(H;P)8

Bloom and 1965 Result of Passive potentials Fe304 structure in which

Goldenberg [4] considerations protons are substituted for
based on literature sane of
the ferrous ion:
data V-fe;P3 remains when all
ferrous ion bas disappeared

wagner [6] 1973 Tbeorical Acidic solutions Single layer film potential
considerations dependent composition fron
Pe304 to fe2.6JD4

Akimv et al. [20] 1979 Ellipsometry and Borate buffer pH7 .35 Multilayer film fron
Auger electron Fe304 to Pe;P3
spectroscopy

caban and Cben 1982 Ellipsometry and Borate buffer pH B.4 Insulating ferric hydroxide
[21] ispedance with small Don-stoichicmetries
primarily Fe2-t near the inner
interface and Fe4+ near the
outer interface
 491

Table 1 - Cont.

ExperiJlental Experimental film

Authors Year Techniques Conditions Ccmposition
-------------------------------------------------------------------------------------------------------------
Nagayama and 1962 Electrochemical oxi- Borate buffer pH 8.4 Inner layer Fe304 outer layer
Cohen [7,22] 1963 dation and reduction, E from 0 to 0.85 V )J-Fe'fl3;
chemical analysis, at higher potentials cati0nt~ vacancies
electron diffraction and fe in the outer layer

Cohen et al. [8] 1979 RHEED, XRE Borate buffer pH 8.4 Inner layer fe304;
Outer layer FetJ3

Haruyama and 1973 Conductivity measured Borate buffer pH 8.4 Inner layer Fe304' outer
Tsuru [23,24] during cathodic layer )J - fe'fl3 with adsorbed
reduction and oxygen
2+
analysis for Fe
determination

Ord et al. 1%6 Ellipsauetry and Borate buffer pH 8.4 Inner layer Fe304;
[25-27] chronovoltarcmetry Outer layer )J-Fe'fl3

Sea et al. [28] 1974 Auger electron Borate buffer As above

spectroscopy E from -0.5 to 1 V

Nisbimura et al. 1980 Ellipsometry Phosphate solution Inner barrier layer Fe304
[29] pH 6.59 outer layer )J-Fe'fl3; at
Borate buffer pH 8.4 interface of ~th layers}
surplus of Fe in pbosphate
and deficit in borate

Sato et al. 1971 Ellipsometry, Borate buffer pH 8.4 Inner layer anhydrous
[30-32) electrocbemical )J -FetJ3 outer layer
oxidation and hydrated )J -Fe'fl3, whole film
reduction contains surplus of oxygen and
can he pictured as Fe'fl3 . HtJ

Sato et al. [31) 1975 Tbermogravity aod Borate buffer pH 8.4 Inner layer Fe'fl3· O. 2HtJ, outer
ellipsometry layer FetJ3·3.25HtJ; outer layer
gr~ by anodic precipitation
2Fe + (3+n)H?i'; _
Fe'fl3.nHtJ + 6H + 2e

Sato ot al. [33) 1976 Ellipsauetry, Borate buffer Inner barrier layer, precipi ted
cathodic reduction, E frOOI -0.15 to ou~!f layer i in both layers 1
chemi~ analysis 1.36 VSCE Fe occurs i outer layer
on Fe hydrated )J -FetJ3

Sea et al. [34) 1977 Auger electron Borate buffer pH 8.4

spectroscopy °
hellow VSCE' Inner layer Fe304i
above 0 VSCE Fe'fl3 • nHtJ, where
n = from 0.6 to loB
492

In spite of the great confusion existing in the

studies performed with stain~ess steel, some general
trends relative to the films are usually indicated, such
as chromium enrichment in the film and nickel enrichment
in the metal phase close to the oxide [43]; concerning
molybdenum there is contradictory information [43,44]
(enrichment, depletion, constancy of content).
Evidence of decrease in film thickness and increase
in non-crystallinity with increase in alloying elements
also exists.

1.2.1. Bipolar model. Usually the models put forward for

the passivation of stainless steel fall into the models
already mentioned for iron.
However, another important model - the bipolar model
- is worth to be referred at this point. Based on results
of RHEED and XPS obtained on AISI 304 in H2S04 solutions
with and without NaCl and in consonnance with works of
Sakashita and Sato [11] and also Brooks et al. [36J the
model describes the film as a bipolar ionic rectifier
constituted by a Cr rich inner layer and an Fe rich outer
layer. The inner layer, resulting from the direct reaction
of Cr with H20, is formed by Cr203 and Cr(OH)3 and is
anion selective. The outer layer, rich in iron,
incorporates S042- ions, which makes it cation selective.
Immediately below the iron layer, Cr04 can be present,
resulting from the dehydration of Cr(OH)3'
When an anodic potential is imposed, a de-
protonation reaction takes place in the internal layer,
near the interface of the duplex film, resulting in the
loss of protons to the solution through the external layer
(cation selective).
Since the inner layer is anion selective, the exit
of metal cations from the substrate is difficult. As a
consequence, the 0 2 - ions originated in the deprotonation
 493

reaction accumulate at the interface, then migra~ing

towards the metal surface, and forming a dehydrated oxide,
Fig. 2.

I
t
cation selective
layer n- 8 8 8 8
Ne04 8 8 8 ~
--------------------_____ OH - - H+ + 0 2- __ ~~ _____ bi po 1ar ~
anion selective • • . _1 2- rectifier ~
::
layer Cr(OH)3. ~_ • lnner oxide

"""""""""""""7"""7""''''''''''''''''''''')to:: ba rri er g

;;~~.i%:t
Figure 2. The bipolar model (after Brooks et al. [36]).

1.3. Remarks on passivation

Investigations of the composition, structure and

properties of the passive films have been carried out in
various metals and alloys, emphasis being put on iron and
its alloys. However, a considerable amount of disagreement
still exists on the results. For example, the enrichment
of certain alloying elements in the film relative to the
substrate is still quite controversial. This is due to
several factors: different experimental conditions,
different surface preparations and limitations of the
techniques
494

At present there are many promising techniques for

characterization of surfaces (SEM, XPS, SIMS, Raman
spectroscopy, ellipsometry, electrochemical techniques
such as EIS, ... ). However, for example, techniques like
AES and XPS are fairly good for analysis, but their
detection levels are much higher than that of SIMS, which
can detect species in extremely low concentrations, but
which in its turn is an essentially qualitative technique.
Also some techniques are not capable of distinguishing
among oxides; for example, in the case of iron, electron
diffraction can not distinguish Fe304 from Y-Fe203 [2].
Objections have also been put to "ex-situ"
techniques like AES, since the passivating oxides can
undergo changes in an environment different from the one
in which they were formed (e.g. loss of water molecules).
Electrochemical techniques like Electrochemical
Impedance Spectroscopy and Mott-Schottky plots and also
photoelectrochemical measurements are "in-situ" techniques
which have become quite popular during the past decade for
the study of the behaviour of passive films and also for
their characterization from the point of view of their
semiconductive properties.
Thus, in spite of all efforts and developments
achieved, many points still remain to be cleared, the
questions being more in fields like: Are passive
films amorphous or crystalline? What is the role of
bonding water molecules? Are the semiconductive properties
of the films related or not with their stability and if
they are in which way? and What is the relation
composition/behaviour?

2. Breakdown of Passivation

The breakdown of the passive film in a situation of

localized corrosion, such as pitting or crevice corrosion,
is commonly accepted as being developed in two stages:
 495

1) initiation, in which stable corrosion sites are formed

on the surface, and
2) progression, in which those microscopic sites develop
into larger or deeper corrosion areas, often leading to
severe attack.

A third phenomenon, consisting of repassivation of

already activated sites, can also occur under favourable
conditions.
Although there is a general agreement concerning the
mechanisms of the progression stage, doubts still remain
on the processes leading to the nucleation of corrosion
sites on the surface of the passive film. The most
relevant theories concerning the initiation stage will be
presented next.

2.1. Anion penetration/migration models

Under the influence of the electric field which

exists across the film, the agressive anion (in general
Cl-) can penetrate and reach the metal/film interface. The
way by which the anion moves across the oxide varies with
the author, and can be either by diffusion through pre-
existing micropores [45], by ion exchange with the 0 2 -
ions of the lattice [46] or by a process involving the
formation of cation vacancies [47].
Irrespective of the diffusion mechanisms, film
breakdown occurs when the agressive anion reaches the
metal.
Another version of this model has been proposed by
Okamoto [9], in the light of the hydrated polymeric oxide
model for the passive film. In this version, the water
molecules adsorbed on the surface are substituted by
chloride ions, which then migrate into the oxide, where
they meet OH- ions. Because chloride ions are stronger
Lewis bases, they tend to substitute the OH ions, with
496

the formation of iron chlorocomplexes, which afterwards

diffuse into the electrolyte.

2.2. The Chemica-Mechanical Models

According to Hoar [48], agressive anions are

adsorbed on the oxide film, substituting water molecules.
The repulsive forces among adsorbed anions lowers the
interfacial tension, leading to the formation of cracks in
the oxide.
Sato (49] has suggested that the mechanical stress
can become critical as the result of an electrostrictive
pressure (dielectric deformation resulting from the
precipitation of opposite charges by the electric field),
and has developed a mathematical model for the
electrocapillary rupture of the passive film [50].
After the breakdown,hidrolysis of the metal ions can
occur, lowering the pH, and accelerating metal
dissolution.
Migration of Cl occurs towards the interior of the
pit in order to maintain electroneutrality, and the
process may become autocatalytic.
On the contrary,non-agressive electrolytes can lead
to the formation of insoluble products which precipitate
leading to repassivation [51].

2.3. Point Defect Model (PDM)

Assuming that the diffusion of cations through the

film results only in metal dissolution (and not in film
growth), metal vacancies form at the metal/film interface.
Although the vacancies tend to diffuse into the metal and
disappear [52,53], a condensate of vacancies can grow if
the speed of diffusion of cations towards the solution
exceeds that of vacancies towards the metal. When this
 497

condensate reaches a critical size, local collapse of the

film takes place. The role of halide ions consists in
occupying anion vacancies in the outer layers of the film,
increasing the electric field through the film and
therefore accelerating the diffusion of metal ions
(Fig. 3).

Metal Film Solution

j
Cation Ejection
M,. .... MX+ (aq) + V X'

(=1- MX+ } SchottXy,.

Pair
... - Reaction

m +V .... +' . \ (J) - - Anion Absorption

,.:\:' M,. xe Cation, ca V- + X- .... X.

t...
- V-.cv : 0 0
J m , Condensate , ,:
,
....... _---_ ... '

Figure 3. Processes leading to passive film breakdown,

according to PDM; m = metal atom, MM = metal cation in
cation site, Xo halide anion in anion site, VMX'
cation vacancy, = anion vacancy, MX+ dissolved
cation, X = dissolved halide ion (after Macdonald [52]).

2.4. Theory of the Chemical Dissolution

Initially proposed by Hoar and Jacobs [53], it

admits that three or four halide ions are adsorbed on the
film surface, around a cation of the lattice, preferably
near an anion vacancy. The complex thus formed can
498

immediately separate from the oxide, and in this way the

dissolution of cations is accelerated when compared to the
aquo-complexes. The thickness 6f the film is therefore
reduced in that area, which will increase the electric
field, provided the anodic potential remains constant.
Consequently, the migration of metal ions through the film
is enhanced, as well as the local dissolution rate. In the
limit, the film becomes locally destructed, and the
underlying metal becomes exposed to the electrolyte.
According to some authors, pit formation should
preferably occur close to film defects or flaws, where the
film is thinner and therefore the electric field across
the oxide/solution interface is higher [2].

2.5. Theory of the Depassivation-Repassivation

The passive film is envisaged as a dynamic system in

which a competition takes place between metal dissolution
and film formation, i.e. between breakdown and
repassivation [54,551. Repassivation is hindered either by
the presence of agressive anions [56] or by anodic
potentials [57]. Therefore, in the presence of high
chloride concentrations or under sufficiently high anodic
polarizations, some pits will not repassivate, giving
start to the process of metal destruction.
Recently, based upon a number of reaction paths
appearing in the literature, a scheme of competion among
the processes of film formation, anodic dissolution, film
breakdown and repassivation (Fig. 4) has been proposed
[58] .
 499

Figure 4. Scheme for the dissolution/passivation of

austenitic stainless steels [58]
passivating species
non-passivating species.

2.6. Theory of Local Acidification

Essencially developed by Galvele [59,60], it assumes

the existence of "micropits" on the surface, even below
the critical potential for pitting corrosion lE * corr )·
Inside those micropits there· is dissolution and hydrolysis
of metal ions, with the formation of free protons.
Corrosion products diffuse towards the electrolyte. Below
E * corr these micropits tend to repassivate, by formation
of a protective oxide. Above E* corr the anodic reactions
Me - Me z + + ze
and
z+ +
Me + zH20 -Me(OH)z + zH
proceed at a sufficiently high rate to provoke
acidification which prevents repassivation.
500

The degree of acidification depends on the product

x.i, ~ being the pit depth and i the current density.
The local acidification theory has been object of
some criticism, because it does not explain effects such
as the specific action of the agressive anions or the
increase of the induction time with film thickness [2,61].
Initially built to explain the initiation of pitting
corrosion, the theory seems to be more adapted to the
progression stage, since the acidification inside the pits
seems to be decisive to its growth.

3. Crevice Corrosion

Crevice corrosion is a form of localized corrosion

which occurs on areas of the metal surface where mass
transfer is limited by the system geometry.
As in other localized corrosion processes, two
stages are considered: initiation and progression.
In spite of a reasonably generalized idea that
pitting and crevice corrosion have similar mechanisms,
there are some fundamental differences, as pointed out by
Smia10wska [2]:
both forms of corrosion have special relevance in
chloride solutions; however, unlike pitting corrosion
(PC), crevice corrosion (CC) often occurs in the presence
of other corrosive media, such as acid solutions.
- CC is specific for stagnant solutions inside a crevice,
as opposed to PC, which can occur either in stagnant or
agitated solutions.
- for the reasons mentioned, PC can occur on any region of
the metal surface, whereas CC is limited in space.
CC generally requires lower potentials and lower
induction times when compared to pc.

The points of view mentioned in the literature

concerning the correlation between the two forms of attack
 501

are mainly the following [2]:

PC is a case of CC in which the micropores existing on
the oxide act as crevices, where local acidification takes
place
CC initiates by pits that nucleate inside the crevice
the electrochemical processes are similar, but the
diffusion path is longer for CC than for PC
both forms of corrosion have identical progression
stages.

3.1. Mechanism for CC

The theories mentioned for the local breakdown of

passive films, although developed for PC, are not to be
excluded in the case of CC.
However, a number of theories have also been
proposed specifically for CC [62]. Fontana and Greene [63]
have developed a model which accounts for the main aspects
supported by the existing theories. The model assumes that
when a passive metal in open circuit is placed in a
solution containing chloride ions, in a first stage, the
electrochemical reactions on the surface occur, having the
same rate inside and outside the crevice. In the case of a
metal with valence 2, the anodic reaction will be

M _M 2 + + 2e

and the cathodic reaction, in a neutral, aerated medium:

The restrictions on mass transfer imposed by the

crevice geometry - in spite of the slow rate at which
these reactions take place in a passive metal - lead to a
decrease in the oxygen concentration inside the crevice,
502

and a situation is reached in which the system behaves as

a differential aeration cell.
Under these conditions the region inside the crevice
behaves as anode, and the metal in contact with the free
electrolyte, as cathode. As a consequence, an acumulation
of positive (metal) ions occurs in the solution inside the
crevice, disturbing the equilibrium of electroneutrality.
As a result, anions migrate from the bulk of the solution
into the crevice. Hydrolysis of the metal chloride inside
the crevice then leads to acidification of the crevice
solution according to the scheme:

When acidity reaches a critical value, the passive

film suffers breakdown, leaving the metal exposed.
At the sites where the metal is exposed, metal
dissolution is accelerated, which leads to an increase in
the rate of anion migration towards the crevice, and
therefore to further acidification, generating an
autocatalytic process. This is the accepted mechanism for
the progression stage, which leads to an extension of the
areas of attack.
Crevice corrosion can therefore be divided into four
steps [64,65]:

1 - depletion of oxygen inside the crevice

2 - increase of acidity and of Cl concentration inside

the crevice

3 - permanent breakdown of the passive film with local

corrosion activation, which may occur according to the
passivation breakdown theories.

4 - CC progression, with propagation of the areas of

attack.
 503

The first three steps are usually considered to

constitute the initiation stage.
This model has good support from experimental
results obtained by a number of researchers. Karlberg and
Wranglen [65] measured the pH inside active crevices on
stainless steel, and observed that initially the pH raised
from 6 to 9.5, and later decreased to 4. The initial raise
was attributed to the reduction of 02, with OH- formation,
and the following decay, to the metal hydrolysis.
According to those authors, the reactions of metal
dissolution and hydrolysis are, respectively:

Fe Fe 2 + + 2e
Cr cr 3 + + 3e

and
cr 3 + + 3H20 Cr(OH)3 + 3H+
Fe 2 + + 2H20 Fe(OH)2 + 2H+

The different steps leading to corrosion are

explained based on a rise of the Flade potential resulting
from the pH drop inside the crevice.
The dissolution of Cr in the initiation period is
contested by other authors [66]. Based on the
determination of the amounts of Fe, Cr and Ni in the
crevice solution, it was concluded that the amount of Cr
is much lower than the corresponding atomic fraction in
the alloy during the initiation period, although i t
increases after the breakdown.
The existence of a single mechanism for CC or of
different mechanisms depending on the metal has been a
subject of discussion. Siitari and Alkire [67], based on
pH and potential measurements and also on current
distribution in crevices on iron and aluminium, suggested
that the initiation phase might have different mechanisms
on different metals.
504

Another type of interpretation, based on the

establishment of equivalent electric circuits, has found
some success in recent years. The kinetics of initiation
and progression has been studied by a.c. impedance and a
circuit has been proposed which includes components
relative to adsorbed intermediates [68), suggesting that
the processes involved in PC and CC are not significantly
different [68,69].
A number of attempts have also been made to describe
CC by means of mathematical models, and which aim at the
determination of parameters such as the time for pH
decrease down to a critical value [64), the oxygen
concentration in the crevice solution [70,71), the
critical concentration of dissolved metal [72) or the
current distribution in the inner surface of the crevice
[73,74).

3.2. Electrochemical Techniques for Crevice Corrosion

Studies

A number of electrochemical techniques have been

used for the study of localized corrosion. Among them,
potential and current monitoring, electrochemical noise
and electrochemical impedance spectroscopy are possibly
the most popular. Some of their applications to crevice
corrosion on stainless steel will be presented next.

3.2.1. Potential and d.c. coupling current. These are

simple and non-perturbing techniques, which follow the
behaviour of systems under free corrosion conditions.
Potential measurement is one of the most widely used
electrochemical techniques for monitoring on-site, due to
the simplicity and low cost of the equipment required.
Measuring the d.c. galvanic coupling current by
means of a zero resistance ammeter (current to voltage
 505

converter) can also give valuable information, provided

cathodic and anodic areas are separated in space, as in
the case of crevice corrosion. A possible experimental
set-up is shown in Fig. 5.

(e)

(a) (A) (DI

Figure 5. Experimental set-up for monitoring galvanic

couples (side view).
(A) "Sandwich" electrode (anode)
(B) Bare electrode (cathode)
(e) Zero resistance ammeter
(D) SeE, used in the potential measurements

The potential record (Fig. 6) makes it possible to

distinguish several stages in the crevice corrosion
processes, as indicated by other authors [64]:

Growth of an oxide film, detected by a rise in potential

- Depletion of oxygen inside the crevice, accompanied by a
retardation of the potential rise
- Breakdown of the film, corresponding to a potential drop
506

Activation stage with permanent breakdown and corrosion

propagation.

The simultaneous recording of the current confirms

the above information in what concerns the transition from
passive to active, since current rise and potential drop
occur simultaneously.

V (.v/seE)

Time of immersion (hours)

Figure 6. Potential and current evolution in a situation

of crevice corrosion.

For longer exposure times a current drop accompanied

by an increase of the potential can be seen in the figure,
and is related to a decrease in activity inside the
crevice, leading to lower corrosion rates. This fact was
observed in many experiments, even for a wide range of
covered/uncovered area ratios, and becomes more evident
when potential and current monitoring is performed during
long periods of time.
 507

3.2.2. Electrochemical Noise. This is a technique which

analyses the fluctuations of the potential (potential
noise) superimposed on the d.c. potential, i.e., the low
frequency noise of the type l/f. Alternatively,
fluctuations of the current flowing in a galvanic couple
can be recorded (current noise).
The first work concerning noise in electrochemical
systems is due to Hagyard and Williams [75], in 1961. In
1968, Iverson [76] detected for the first time a relation
between electrochemical noise and corrosion processes. A
number of researchers have used this technique, especially
during the past decade. Statistical and spectral analysis
has also been applied to the results in order to find
correlations among noise, type of attack and rates of
reactions [77-81).
As far as localized corrosion - namely pitting and
crevice corrosion - are concerned, several workers have
detected the existence of current and potential transients
in the passive state, preceeding the onset of localized
attack [82-84].
Figure 7 shows transients obtained for potential and
current in a situation of passivity. It can be seen that
there are two transients in potential, characterized by a
quick drop followed by a slow (exponential) recovery.
These transients have been ascribed to the nucleation of
metastable corrosion sites, which do not develop,
repassivating instead. This interpretation is confirmed by
the current transients, which reveal an anodic current on
the covered electrode, as expected. At a later stage,
stable corrosion sites occur, which do not repassivate,
and crevice corrosion proceeds.
Electrochemical noise is also sensitive to the
presence of agressive ions. Figure 8 shows potential and
time records of a galvanic couple in which addition of Cl
leads to a sharp drop in potential, together with an
increase in the current noise, which reveals a greater
instability of the passive film.
508

97.5=-------------------.

96.0 (A)

60 90 120
t (min)

0.04

0.02 (B)

0.00

-0. 02 ~ __ ___'L'___ _ _~_ _ _"'__I_ _ _...........

o E W 00 120
t <min>

Figure 7. Potential (Al and current (B) transients for

AISI 304 stainless steel in NaCl solution.
 509

-so r - - - - - - - - - - - - - - - - - ,
..........

-'.
-60
Q
w (A)
~
>-
.!
UJ
-70

-80~------~---~------~--------L-~
o 10 20 30
t (min)

0.4~--------------------~

0.2

0.06-------*'·
V~~~~
'.
'0

-0.2
00

-0• 40
10 20 30
tCmin)

Figure 8. Potential ( A) and current (B) evolution during

addition of NaCl, 10 min after the start of experiment.
510

3.2.3. Electrochemical Impedance Spectrosocopy

(EIS). EIS has been applied to the study of pitting and
crevice corrosion and several models in terms of
equivalent circuit have been proposed.
The main differences when compared to uniform
corrosion lie in the low frequency region of impedance
spectra, where an inductive behaviour is detected, which
has been attributed to a process involving adsorbed
intermediates [58,85] and/or electrocrystallization
[86,87].
One of such results is shown in Figure 9, together
with the equivalent circuit used for simulation.

_------.~17- c

~-

Rt
c.

R.

0. 3500Hz
( b )

o
(a )

Figure 9. a) Nyquist plot in a situation of crevice

corrosion; b) equivalent circuit: Rg - solution
resistance; Cf, Rf = capacitance and resistance associated
to geometry of crevice or passive film; Ce, Rs - double
layer capacitance and charge transfer resistance; Ca , Ra =
adsorption of intermediates.

A different equivalent circuit has been proposed by

Mansfeld [88] for pitting corrosion of aluminium, in
 511

which two RC networks are considered in paralell,

corresponding to the pitted and passive areas.
As crevice corrosion progresses, the shape of the
Nyquist plot becomes a semi-circle, from which the charge
transfer resistance can be obtained. The plot of the
reciprocal of that resistance - which constitues a measure
of the corrosion rate - as a function of time, shows that
after an increase during the first hours, l/Rct passes
through a maximum, decreasing afterwards to lower values
(figure 10).
This result supports the idea that the rate of
crevice corrosion progression does not increase
permanently, but rather tends to decrease, after an
initial period of rapid propagation.
This fact in conjunction with the results obtained
by zero resistance ammetry, contradict the point of view
according to which crevice corrosion is a permanent
auto-catalytic process .

.0 0 1 5 r - - - - - - - - - - - - - - - ,

'(\J- .001
E
u
c:
U .0005
a:
"- •

20 80 100
t (hours)

Figure 10. l/Rct vs time for crevice corrosion of AISI 304

stainless steel in 3% NaCl solution.
512

4. Conc1usions

From the above exposed, it is seen that a good

number of attempts have been made in order to explain both
passivation and localized corrosion. However, a good deal
of work is still needed before a final model can be
reached.
As far as the models for passivation are concerned,
considerable developments have been registered in the past
few years, especially due to the development of
electrochemical, surface observation and microanalysis
techniques. Due to those techniques, it is at present
known that passive films constitute a phase different both
from the metallic matrix and the environment, which
excludes the hypothesis of the simple oxygen adsorption
being responsible for passivation.
Electrochemical techniques have revealed of valuable
importance not only for fundamental studies in laboratory
but also for corrosion monitoring in industry. More and
more the instrumentation in laboratory is becoming adapted
to a wider use in field tests, and in the future great
developments are to be expected in this area.
 513

REFERENCES

l. FaradayM. (1844), "Experimental Research in

Electricity", vol.II, University of London.

2. Szklarska-Smialowska Z., (1986),"Pitting Corrosion of

Metals", NACE, Houston, Texas.

3. Tjong S.C., Yeager E. (1981), J. Electrochem. Soc.

128, 2251.

4. Bloom M.C., Goldenberg L. (1965), Corros. Sci. 2, 623.

5. Foley C., Kruger J., Bechtoldt C. (1967), J.

Electrochem. Soc. 114, 994.

6. Wagner C.(1973), Ber. Bunsenges. Phys. Chern. 77, 1090.

7. Nagayama M., Cohen M. (1962), J. Electrochem. Soc.

109, 781

8. Cohen M., Mitchell D., Hashimoto K. (1979), J.

Electrochem. Soc. 126, 442.

9. Okamoto G. (1973), Corros. Sci. 13, 471.

10. Pou T., Murphy 0., Young V., Bockris J.O'M. (1984),
J. Electrochem. Soc. 131, 1243.

11. Sakashita M., Sato N. (1977), Corros. Sci. 17, 473.

Sakashita M., Sato N. (1978), in R.P. Frankental and
J. Kruger (eds.), Passivation of Metals, The
Electrochemical Society, Corrosion Monograph Series,
Princeton, NY, p.479.

12. Sakashita M., Sato N. (1979), Corrosion-NACE 12, 351.

514

13. Bockris J.O'M., Genshaw M.A., Bousic V., Wroblowa H.

(1971), Electrochim. Acta 16, 1859.

14. A) Oshe E.K., Rozenfe1d I.L., Pozdeeva Yu. V. (1979),

Boshoku, Gijutsu~, 61.
B) Oshe E.K. (1978), Extended Abstract, ISE 29 th
Meeting, Budapest 1, 452.
C) Oshe E.K., Rozenfeld i.l., Doroshenko V.G. (1971),
Zashchita Metallov 1, 38.
(Quoted in reference 2).

15. Akimov A.G., Rozenfeld I.L., Astafev M.G. (1976),

Zashchita Metallov 12, 162. (Quoted in reference 2).

16. O'Grady W.E., Bokris J.O'M. (1973), Surface Sci. ~,

249.

17. Revie R.W., Bockris J.O'M., Barker B.G. (1975),

Surface Sci. 52, 644.

18. Revie R.W., Barker B.G., Bockris J.O'M. (1975), J.

Electrochem. Soc. 122, 1460.

19. O'Grady W.E. (1980), J. Electrochem. Soc. 127, 555.

20. Akimov A.G., Andreva N.J., Rozenfeld I.L. (1979),

Electrokhimiya 15, 1888. (Quoted in reference 2).

21. Cahan B.D., Chen C.T. (1981), in R. Fankenthal (ed.),

Proc. Int. Symp. Honoring H.H. Uhlig's 75 th Birthday,
The Electrochemical Society, Princeton, New Jersey
81-8, p.123. (Quoted in reference 2).

22. Nagayama M., Cohen M. (1963), J. Electrochem. Soc.

110, 670
 515

23. Haruyama S., Tsuru T. (1973), Corros. Sci. 13, 275.

24. Tsuru T., Haruyama S. (1976), Corros. Sci. 16, 623.

25. Ho F.C., Ord J.L. (1972), J. E1ectrochem. Soc. 119,

139.

26. Ord J .L. , Desmet D.J. (1966), J. E1ectrochem. Soc.

113, 1258.

27. Ord J .L., Desmet D.J. (1976), J. E1ectrochem. Soc.

123, 1876.

28. Seo M., Lumsden J.B. (1976), R.W. Staehle, Surface

Sci. 42, 337.

29. Nishimura R., Kudo K., Sato N. (1980), Surface Sci.

96, 413.

30. Sato N., Kudo K., Noda T. (1975), Corros. Sci. 10,
785.

31. Sato N., Kudo K., Noda T. (1971), Electrochim. Acta

16, 1909.

32. Sato N., Kudo K. (1971), Electrochim. Acta 16, 447.

33. Sato N., Kudo K., Noda T. (1975), Z. Phys. Chern. NF

98, 271. (Quoted in reference 2).

34. Seo M., Sato N., Lumsden J.B., Staehle R.W. (1977),
Corros. Sci. 17, 209.

35. Mitrovic-Scepanovic V., MacDougallB., Graham M.J.

(1984), Corros. Sci. 24, 479.
516

36. Brooks A.R., Clayton C.R., Doss K., Lu Y.C. (1986),

J. Electrochem. Soc. 133, 2459.

37. Lorang G., Cunha Belo M. da, Langeron J.P. (1987),

J. Vac. Sci. Technol. A, 2, 1213.

38. Hultquist C., Leygraf C., Brune D. (1984), J.

Electrochem. Soc. 131, 1773.

39. Harrington D.A., Wieckowski A., Rosasco S.D., Schardt

B.C., Sa1aita G.N., Hubbard A.T., Lumsden J.B. (1985),
Corros. Sci. 25, 849.

40.01efjord I., Brox B. (1983), in M. Froment (ed.),

"Passivity of Metals and Semiconductors",
Elsevier Science Publishers B.V. Amsterdam, p.561.

41. Asami K., Hashimoto K., Masumoto T., Shimodaira S.

(1976), Corros. Sci. 16, 909 (1976). (Quoted in
reference 25).

42. Okamoto G., Tachibana K., Shibata T. (1974), J. Japan

Inst. Met. 38. (Quoted in reference 2).

43. Olefjord I., Elfstrom B. (1982), Corrosion~, 46.

44. Goetz R., Landolt D. (1984), Electrochim. Acta 29,

667.

45. Nishimura R., Kudo K. (1981), Proc. 8 th ICMC,

Mainz, 1, p.6.

46. Rozenfeld I.L., Marshakov I.K. (1964), Corrosion lQ,

115t.

47. Heine M.A., Keir D.S., Pryor M.S. (1965), J.

Electrochem. Soc. 112, 24.
 517

48. Hoar T.P. (1967), Corros. Sci. I, 335.

49. Sato N. (1971), Electrochim. Acta 16, 1683.

50. Sato N. (1982), J. Electrochem. Soc. 129, 255.

51. Hoar T.P. (1949), Trans. Faraday Soc. 45, 683.

52. Chao C,Y" Lin F.L., Macdonald D.D. (1981), J.

Electrochem. Soc. 128, 1187, 1194.
Urquidi M., Macdonald D.D. (1985), J. Electrochem.
Soc. 132, 555.

53. Hoar T.P., Jacob W.R. (1967), Nature (London), 216,

1209.

54. Videm K. (1974), Kjeller Report KR-140, Institutt for

Atomenergi, Kjeller, Noruega. (Quoted in
reference 2).

55. Richardson J.A., Wood G.C. (1970), Corros. Sci. 10,

313.

56. MacDonald B. (1979), J. Electrochem. Soc. 126, 919.

57. Barbosa M., Scully J.C. (1982), Corros. Sci. ll,

1025.

58. Dawson J.L., Ferreira M.G.S. (1986), Corros. Sci.

26, 1009.

59. Galvele J.R. (1976), J. Electrochem. Soc. 123, 464.

60. Galvele J.R., Lumdsden J.B., Staehle R.W. (1978), J.

Electrochem. Soc. 125, 1204.
518

61. Janik-Czachor M. (1979), Werkstoffe Korros. 30, 255.

62. France Jr., W.D. (1972), "Crevice Corrosion of

Metals", Localized Corrosion - Cause of Metal Failure,
ASTM STP 516, American Society for Testing and
Materials, p. 164.

63. Fontana M.G., Greene N.D. (1967), "Corrosion

Engineering", MaGraw-Hill, N. York, p. 39.

64. Oldfield J.W., Sutton W.H. (1978), Br. Corros. J. 13,

104.

65. Karlberg G., Wranglen G. (1971), Corros. Sci. 11,

499.

66. Lott S.E., Alkire R.C. (1988), Corros. Sci. 28,

479.

67. Siitari D.W., Alkire R.C. (1982), J. Electrochem.

Soc. 129, 481.

68. Dawson J.L., Ferreira M.G.S. (1986), Corros. Sci. 26,

1027.

69. Szlarska-Smialowska Z., Mankowski J. (1978), Corros.

Sci. 18, 953.

70. Turnbull A. (1980), Br. Corros. J. 15, 162.

71. Turnbull A., Thomas J.G.N., (1979) (1980), NPL DMA

Report (A) 11, 23.

72. Herbert K., Alkire R. (1983), J. Electrochem. Soc.

130, 1007.
 519

73. Sharland S.M., Tasker P.W. (1988), Corros. Sci. 28,

603.

74. Sharland S.M. (1988), Corros. Sci. 28, 621.

75. Hagyard T., Williams J.R. (1961), Trans. Faraday Soc.

57, 2288.

76. Iverson W.P.(1968), J. Electrochem. Soc. 115, 617.

77. Fleischmann M., Oldfield J.W. (1970), J. Electroanal.

Chern. 27, 207.

78. Williams D.E., Fleischmann M., Stewart J., Brooks T.

(1986), in M. Duprat (ed.), "Electrochemical Methods
in Corros. Research", Trans. Tech. Publ., p. 151.

79. Al-Zanki I.A., Gill J.S., Dawson J.L. (1986), in

M. Duprat (ed.),"Electrochemical Methods in Corrosion
Research" Trans. Tech. Publ., p.463.

80. Blanc G., Epelboin I., Gabrielli C., Keddam M. (1977),

J. Electroanal. Chern. 75, 97.

81. Gabrielli C., Ksouri M., Wiart R. (1978), J.

Electroanal. Chern. 86, 233.

82. Bertocci U., Yang-Xiang Y. (1984), J. Electrochern.

Soc. 131, 1011.

83. Williams D.E., Westcott C., Fleischmann M. (1984), J.

Electrochem. Soc. 180, 549.

84. Cavalcanti E., Thompson G.E., Dawson J.L., Wood G.C.

(1989), Materials Science Forum, 44-45, p.235.
520

85. Dechet B., Epelboin I., Keddam M. (1977), J.

Electroanal. Chern. 79, 129.

86. Bignold G.J., Fleischmann M. (1974), Electrochirn. Acta

19, 363.

87. Cachet C., Epelboin I., Keddam M., Wiart R. (1979), J.

Electroanal. Chern. 100, 745.

88. Mansfeld F., Shih H. (1988), J. Electrochern. Soc. 135,

1171.
DmIBlTORS AND COA'l'INGS

FLORIAN MANSFELD
Corrosion and Environmental Effects Laboratory
Department of Materials Science
University of Southern California
Los Angeles, CA 90089-0241, USA

ABSTRACT

The application of Electrochemical Impedance Spectroscopy (EIS) as a

new tool for the investigation of methods of corrosion protection is
illustrated for corrosion inhibitors, conversion coatings, polymer
coatings and oxide layers as well as for cathodic protection of
stainless steels in seawater. It is pointed out that it is essential
for all these cases to develop the appropriate models for the impedance
which can be used to fit the experimental data and extract the
parameters which characterize the corrosion process.

IN'I'RODOCTION

The development of Electrochemical Impedance Spectroscopy (EIS) as a

new tool for corrosion research has provided increased opportunities
for investigating methods of corrosion protection. Especially for the
cases of corrosion protection by polymer coatings and anodic films, it
had been very difficult if not impossible to obtain mechanistic
inf ormation wi th the traditional dc techniques. Examples for the
application of EIS will be given based on the work carried out by the
author and his co-workers. These examples include the use of corrosion
inhibitors, conversion coatings, polymer coatings, anodic films on Al
and cathodic protection of stainless steel in seawater. For these
cases it is important to develop the appropriate models for the
impedance which can then be used to fit the experimental data and
extract the parameters which characterize the corrosion process.

Applications of EIS

1 .0. COrrosion IDhibi tors in Neutral Aerated Media

The evaluation of corrosion inhibitors for steel in neutral aerated

media has shown that the mechanism of inhibition must be different for

521
M. G. S. Ferreira and C. A. Melendres (eds.).
Electrochemical and Optical Techniques for the Study and Monitoring of Metallic Corrosion. 521-544.
© 1991 Kluwer Academic Publishers.
522

an inhibitor such as NaN02 and a mixture of inhibitors such as

phosphonic acid/fatty amine. Fig. 1 shows data obtained for a rotating
cylinder electrode (RCE) of 4340 steel in 0.5 M Na2S04 (open to air)
with and without inhibitors (1). In the presence of NaN02 the shape of
the impedance curve is very similar to that for the blank solution
except for the larger value of the polarization resistance Rp which is
defined as the dc limit of the impedance (2). In the presence of
phosphonic acid/fatty amine (Inh. A) a new time constant appears with a
maximum of the phase angle at about 2.10 5 rad/sec (Fig. 1). Since this
spectrum is very similar to that found for polymer coated steel (see
3.0), it is assumed that the inhibitor mixture covers the steel surface
like a porous polymer film. A mixture of polyacrylic acid/fatty amine
(Inh. B) did not produce the same spectrum, it was more similar to that
for NaN02'

,
LOGw
4)40STEEL (RC£lfO_~M fh:z$04" INHI8ITOR

Figure 1. Bode-plots for 4340 steel rotating cylinder electrode

(RCE) exposed to 0.5 M Na2So4 (open to air) with and without
inhibitors

Fig. 2 shows the time dependence of the polarization resistance

obtained from dc (RpdC) and ac (Rpac) measurements. Very good
agreement is observed between these two independent measurements with
corrosion rates decreasing continuously over a 50 h period. An
inhibi tor efficiency £ can be calculated based on the Rp-value at
t ~ O. As the rust film covers the steel surface, corrosion rates
decrease. This case is an example for the concept of interface vs
interphase inhibition discussed by Lorenz and Mansfeld (3) and the
inhomogeneous surface model (4,5). The decrease of the corrosion
potential Ecorr with time suggests that the total current for the
reduction of oxygen decreases with time as the corrosion product layer
covers the steel surface.
 523

1o,--r--,----,---r--~500 100
a)

50 ~

o
tlhl

Figure 2. Time dependence of Rpdc , Rpac, Ecorr and inhibition

efficiency e for 4340 steel RCE exposed to 0.5 M Na2S04

Fig. 3 shows similar results for Rp ac and Rpdc in the presence of

phosphonic acid/fatty amine. The pore resistance Rpo determined as the
value of the impedance at the minimum of the phase angle in the
high-frequence range has been used to calculate the porosity l/Rpo of
the inhibitor film as a function of expsoure time. This porosity
reaches a constant value after about 20 h.

100 10 500
c)

'7 ';
N ",-

e ,
E m
" E "~
E 50
!
!.. 3
.s
81
~

~a:
8- ~

0 0 700
0 50

Figure 3. Time dependence of ~dc,Rpac, Ecorr and Rpo for 4340

steel RCE exposed to 0.5 M Na2S04 containing phosphonic acid/fatty
amine
524

Results for an iron rotating disc electrode (RDE) in the presence of

O.OlM hexan (1,6)-biphosphonic acid are shown in Fig. 4 as a function
of exposure time (Fig. 4a) and rotation speed (Fig. 4b) (5). The shape
of the impedance curves and the dependence on rotation speed has
suggested that the spectra can only be interpreted by a model which
considers both the effects of charge transfer and of mass transport.
Such a model (5) is shown in Fig.5. An important aspect of this model
is the effect of the porous corrosion product layer which is involved
in the mechanism of corrosion inhibition.

21.00
a)
01
QSHz
E 1600

-
u
C 28h

I t:!. 800
E
I 10 h

800 1600 2l,00 3200 l,000

Re(Z)/Q·cm 2

3000
01

-
b)
E
u
C 2000
IN 0.05 HI
oS·'
E 1000
I
0.005 HI

1000 2000 3000 1.000 5000

Re (Z)ISl-cm 2

Figure 4. Complex plane plots for iron rotating disc

electrode (RDE) in 0.5 M Na2S04 ' pH = 8, containing
0.01 M hexan (1,6)-biphosphonic acid; a) effect of
exposure time, b) effect of rotation speed
 525

Porous
MetaUMe) Oxide Layer Diffusion Layer

~~~:~_/-
~~~~~~~~~,,~
I I I I X
I--lpor-O--lCi--l I
~-lN----1

__-----III c
II
-1-7
(j (1 + 0) Lpor (j La

ZMe

Figure 5. Inhomogeneous surface model and equivalent

circuit (from Ref. 5)
526

2.0. Conversion coatings

2.1. Chromate Conversion Coatings

Chromate conversion coatings have been a standard method o~ corrosion

protection for Al alloys for a long time. For military applications
these coatings have to pass a quality control test such a salt spray
test. Since this test takes a long time and the evaluation of the
result depends on visual observation, a quality control test has been
developed based on EIS. Fig. 6 shows impedance spectra for Al 7075-T73
with a commercial conversion coating (Alodine 600) during exposure to
0.5 N NaCI (open to air) for 7 days. The spectra are dominated by the
capacitance of the coating which remained constant after the first day.
This coating was prepared under the optimum conditions of the coating
process. When these steps were changed, a coating with much poorer
corrosion resistance resulted. For such a coating large changes of the
impedance were observed during the exposure time (Fig. 7). A new time
constant which can be seen as a shoulder in the phase angle occurs at
high frequencies. The capacitance increases continuously and the dc
limit decrease with exposure time. Such an unstable impedance spectrum
is associated with a coating of poor quality.

6 ,-----,-----,-----,------,-----,-----,-----,

en
.2
D
4

C
.~

~ 3
...-N
...-
g' 2
..J

-1 a 2
Logf (f in Hz)

Figure 6. Bode-plots for Al 7076-T73 with a chromate

conversion coating (Alodine 600) as a function of exposure
time to 0.5 N NaCI
 527

90

75

one .
a;
'-
J: • so ~

.
0
.E
.s
!:! 3 <5 01
c
.,-
N
.,-
'"::
01
0 30 ~
-' 0..

15

Logf (f ;n Hz)

Figure 7. Same as for Fig. 6, but with incorrect application

procedure

A quality control test based on EIS has been developed by preparing a

set of six samples using the correct procedure for coating application
which had been used to prepare the sample in Fig. 6. The EIS spectra
were recorded after exposure for 2h in 0.5 N NaCI and a scatter band
was calculated by determining the mean and the standard deviation of
the impedance at each of the frequencies used in the measurement. The
resulting scatter band appears in Fig. 8. It will be noted that the
band is quite narrow except at the lowest frequencies where problems
sometimes occurred as a result of time instabilities of the test
electrodes.

~2~-----~1----~~----~----~2~----~----~----~­

Log t (t In Hz)

Figure 8. Scatter band of Bode-plots for Al 7075-T73

with a chromate conversion coating (Alodine 600)
528

2.2. Conversion Coatings prepared by Immersion in Rare Earth Metal

Chlorides

Hinton et al (6) have shown that rare earth metal (REM) chlorides are
very effective as corrosion inhibitors for Al and its alloys in NaCI
and other neutral media. In this laboratory immersion in REM chlorides
has been used to produce conversion coatings on Al alloys and AI-based
metal matrix composites (7 - 9). Fig. 9 gives a comparison of the
EIS-data for Al 7075-T6 immersed in 0.5 N NaCI (open to air) in the
as-received condition (curves 1 and 2) and after immersion in 1000 ppm
CeCl 3 for 7 days (curves 3 and 4). For the untreated sample large
decreases of the impedance in the capacitive region are observed which
are characteristic for the pitting process on AI. In addition, a
transmission line type impedance occurs at the lowest measured
frequencies. On the other hand, the CeCl3 treated sample shows a very
stable impedance behavior which changes indicating the initiation of
pitting occurring only after 23 days. Similar results have been
observed for Al 6061 and for the metal matrix composites AI/SiC and
AI/graphite (7-9).

~ = :: I ~~¥O'71:~T8
90

:=2:: ;I';~~~:';!'d
1 7~

In N.C)
.
Cii
60
.
E
'-
01

,5
.
~

01
C
./ <t.
N

~
-'
30 ::;"
.c
a.

!5

~2~--L-L-----~----~!----~2------'----~~~--J 0
5
Log f (f in Hz)

Figure 9. Bode-plots for Al 7075-T6 as a function of

exposure time to 0.5 N NaCI; curves 1 and 2: as-received,
curves 3 and 4: after treatment in 1000 ppm CeCl 3 for 7 days

2.2.1. A Model for Localized Corrosion of Al Alloys. The impedance

behavior of Al alloys undergoing localized corrosion can be explained
by the model shown in Fig. 10 (7-10), where Rp and Cp are the
properties of the passive surface and Rpit and Cpit are related to the
 529

pitted area. F is the area fraction on which pitting occurs.

W = (K/F) (jw)n describes the transmission line element. Fig. 11 gives
theoretical spectra for a passive surface and a case where 0.5% of the
total surface is covered with pits. The initiation of this rather
small amount of pits produces large changes in the spectra which are
very similar to those shown for the experimental data in Fig. 9. Among
these characteristic features is the frequency dependence of the phase
angle at the lowest frequencies where a minimum af the phase angle is
observed. It is pointed out here as has been elsewhere (11) that the
details of the frequency dependence of the impedance can be observed
best in Bode-plots such as those shown in Fig. 9 and 11.

A characteristic feature of the impedance upon initiation of pitting is

the large increase of the measured capacitance Ct = Cp (l-F) + FCpit as
shown in Fig. 12 for Al 7075-T6. For the as-received sample Ct
increased by a factor of 4 in 4 days, while for the CeCl 3 -treated
surface Ct started to increase after about three weeks shortly before
pits could be observed visually. Very similar results were obtained
for CeCl3-treated AI/SiC and AI/graphite (Fig. 13). It will be noted
that for AI/SiC a first increase of Ct occurred after about 7 days. No
indication of localized corrosion could be observed visually at this
time. Pits could be detected only after 81 days. When the test cell
was disassembled, crevice corrosion was observed under the O-ring used
for sealing the cell. Since crevice corrosion was never observed on
any other samples, it was concluded that it was due to the SiC
particles. In addition, the first rise of Ct after 7 days must have
been due to the initiation of crevice corrosion. These results show
that EIS is a very sensitive tool for the detection of localized
corrosion on Al alloys and AI-based metal matrix composites.

Rp /11-FI

Rs

O.;F'" 1. W=(K/F)(jwjn

Figure 10. pitting model for Al and Al alloys

530

90

75

'iii 'iii
e
c
0 • 60 ~
go
..
!'; E
.!:! 3 45

-"
N
-"
...c:'"
'"
0
...J
30
"
In

C'"
D-
15

~3~---_~2----_~1----~0----L---~----~~~~~

Logf (f in HZ)

Figure 11. Theoretical Bode-plots for model in Fig. 10 as a function

of pitted area fraction F (curve 1: F = O~ curve 2: F = 0.005),
Rs = 3 ohm, Rp = 5.10 4 ohm, C2 = 2.10- 4 F, Rpit = 2.5 ohm, Cpit =
0.08 F, K = 2.5 ohm. (rad/s) n, n = - 0.25

1. CeCI3 treated AI 7075-T6

2. As received AI 7075-T8

FCpit
····F····················
100- Pitting for 1
~ Pitting for 2
o 10 20 30 40
TIme (days)

Figure 12. Time dependence of Ct for Al 7075-T6 during

exposure to 0.5 N NaCI~ curve 1: after treatment in 1000
ppm CeCl3 for 7 days, curve 2: as-received~ A = 20 cm 2
 531

1. AI I Gr I CeCI3 ( 1 week )
2. 1\1 I SIC I CeCI3 ( 1 week )
In 0.6 N NaCI

Figure 13. Time dependence of Ct for AI/graphite

(passivated in CeCI3) (curve 1) and AI/SiC (passivated
in CeCI3) exposed to 0.5 N NaCI

A software program PITFIT (12) has been written for the fitting of data
such as those shown in Fig. 9 to the model in Fig. 10 . A problem
occurs in determining pit growth rates from the impedance data because
it is not possible at present to determine the values of Rpit' Cpit and
F seperately. Therefore the following approach was used to estimate
these values. At the end of a test, the value of F was determined
visually and Cpit was calculated from the final value of FCpit (Fig.
12). Assuming that Cpit (in F/cm 2 ) was constant, F was calculated as a
function of exposure time. From the fitted values of Rpit/F it was
then also possible to calculate Rpit (in ohm cm2 ).

Fig. 14 shows that pit growth rates expressed qualitatively as 1/Rpit

decrease with exposure time for both AI 6061 and Al 7075-T6. The value
of K is constant for Al 7075-T6 and different from that for Al 6061. F
increases from about 1% of the total surface to about 4% for both
materials during the exposure period. The exponent n has values very
close to 0.5 independent of exposure time (Fig. 15).
532

101~--L-~--~--~~
20 40 60
TIME(days)

Figure 14. Time dependence of 1/Rpit and K for Al 6061 (deoxidized and
treated in CeCI3) and Al 7075-T6 (treated in CeCI3)

AI 6061 <deoxidized) /CeCI 3 <1m). A

AI 7075-T6 / CeCI 3 , 0

..,c
. I

"'

!r
o<t
S:!
.:.'"
N

0
10 30 50 70
TIME (days)

Figure 15. Time dependence of F and n for the cases of Fig. 14

3.0. Polymer Coatings

Perhaps the most successful application of EIS has been in the

evaluation of the corrosion behavior of polymer coated metals. The
simple model shown in Fig. 16 describes· most of the impedance data
obtained so far for fairly thin coatings. In this model Cc is the
capacitance of the coating which can be calculated from its dielectric
constant E and its thickness d:
 533

(1 )

The water uptake of the coating during exposure to an electrolyte can

be calculated from the increase of Cc • A is the exposed area of the
coated metal.

The resistance Rpo has been called "pore resistance" by Kendig and
Mansfeld (13-15) who suggested that the observed decrease of Rpo with
exposure time is due to damage to the coating and the formation of
conductive paths. Haruyama et al (16) have proposed that the decrease
of R 0 is due to an increase of the delaminated area under the coating.
Mans!eld et al (17, 18) have recently discussed these two concepts and
have applied them to coated steel and Mg.

The corrosion reaction at the metal/coating interface is characterized

by Rp and the double layer capacitance Cdl. Some authors have assumed
a constant value of Cdl (in F/cm2 ) and have calculated the corroding
area based on the observed value of Cdl (in F) (19).

The software program COATFIT has been used to fit the experimental data
to the model shown in Fig. 16 and to determine the characteristic
parameters as a function of exposure time.

GENERAL MOOEL

Rp
MOOEL I

Figure 16. Model for the impedance behavior of polymer coated

metals

For very protective coatings the impedance determined in the usual

frequency range of 10 5 Hz to 10- 3 Hz consists mainly of the capacitive
contribution from Cc and a resistive element at the lowest frequencies
which could be due to Rpo + Rp, but could also be a result of the
534

limited input impedance of the potentiostat. Fig. 17 gives results for

AI/SiC with an epoxy layer of about 27 ~m. Spectra as that observed for
curve 1 in 0.5 N NaCI are typical for this system and do not change for
periods exceeding 2 months. In order to study the delamination
tendency of the epoxy coating, a small hole with a diameter of 0.7 rom
was drilled trough the coating into the metal. The impedance spectrum
changed dramatically and was now due to the reactions in the hole. The
impedance contribution at the very highest frequencies can be
calculated based on the geometry of the hole and the conductivity of
the electrolyte. With time this resistance increased due to the
plugging of the hole with corrosion products. The impedance at lower
frequencies measured over a time period of three months showed the
shape expected from the pitting model for Al discussed above (Fig. 10).
No delamination occurred as was concluded from the simple shape of the
impedance spectra. This was confirmed by the negative results in a
pull test with adhesive tape. When this test was repeated with AI-Li
2091-T6, much larger changes in the EIS data occurred in a much shorter
time (Fig. 18). A new time constant was observed at low frequencies
after a week of exposure to NaCI which must be due to delamination.
The coating lifted off and severe corrosion was observed around the
hole. Apparently the dis sol ved Li produced a very alkaline solution
which caused coating delamination.

9
1. wiDa hole
2-S with a hole(Jl=.7mm
8 2. 3h
3. 2d
@
e 4. sod
u S. 93 d
7
e
.r::.
a
....c: 6
t::!
,..-
N
,..- 5
Cl
a
-J

3
-2 -1 o 2 3 4 5
Logf (f in Hz)

Fig. 17. Bode-plots for epoxy coated (27 ~m) AI/SiC

during exposure to 0.5 N NaCI; curve 1: intact coating,
curves 2 - 5: after drilling a hole with a diameter of 0.7 rom
 535

AI - Ll 2091 - T6 Epoxy 0.5 N NaCI

10r-----r----,,----,,----,~--~----_,----_,----_,

1. without a hole; 2h
1
2. with a hole (0= 0.75 mm )
1d
ruE e 3. 9d
u
E
.c
o

"
.....
N
~ 6
/'
N
./
g' 5
...J

3 ~ ____L -__ ~L_ __ ~ ____ ~ ____ ~ ____ ~ ____ ~ ____ ~

-2 -1 0 2 3
Logf (f in Hz)

Figure 18. Bode-plots for AI-Li 2091-T6 during exposure to

0.5 N NaCI; curve 1: intact coating; curves 2 and 3: after
drilling a hole with a diameter of 0.75 rom

Magnesium and its alloys find little application at present because of

their low corrosion resistance. However, recently commercial coating
systems have been developed which allow exposure of coated Mg to
chloride containing media for long time periods (20). Fig. 19 shows
impedance spectra for Mg AZ31 which was pretreated with a Dow *17
(thin) anodic film followed by three layers of a PL 163 polyimide
(curves A) or 985 epoxy (curves B) sealant. The EIS data were
determined during exposure to 0.5 N NaCI at 35 0 C. No deviations from
capaci ti ve behavior were observed in 49 days which suggests that the
coating which was applied by MEL, Ltd. (England) was very protective.
When the pretreatment was Dow *17 (thick) with only one layer of
Araldite 985, the impedance spectra showed the changes which are
expected from the model in Fig. 16 after about 7 days (Fig. 20). The
minimum of the phase angle ~min and its frequency fmin as well as the
breakpoint frequency fb (phase angle equals 45°) can be used to follow
the increasing damage to the coating as shown in Fig. 21 for an
exposure period of about 5 weeks (17,18). With increasing damage to
the coating fmin and fb move to higher frequencies, while ~ min becomes
lower. These results suggest that a coating monitor could be
constructed which measures the impedance starting at very high
frequencies until the phase angle reaches a minimum (fmin) or 45
degrees (fb).
536

75

u; a;
E
L: 6 60 ~
o 0>

....
c
5 E "
t:! 45~
.;-
N
.;-
.'"
c

'"o 30 "

'""
Curve1: 6hlA)
-' 2: 49dlAI ~
3: 5hlB a.
4: 49dlB) 15

oL~_-'-- _ _--L-_ _-L_ _-'-_ _--'L-_ _-'----_----10

-2 -1 2 5
Logf (f In Hz)

Figure 19. Bode-plots for coated Mg AZ31 during exposure

to 0.5 N NaCl, T = 35 0 C; pretreatment: DOW #17 (thin), sealant:
PL163 polyimide (A) or 985 epoxy (B), 3 layers.

90

75

u;
E
L: 6 60
a;
.
L
0
'"
~

....c 5 E
t:! '5

.'"
~
~

./
N
c:
,-
'"
0
...J
30

L:
a.
.'"
~

15

o L-___-'----____-'----____-'----____-'----____-'----____-'--__ ~

-2 -1 1 2
Logf (f in Hz)

Figure 20. Bode-plots for coated Mg AZ31 during exposure

to 0.5 N NaCl, RT; pretreatment: DOW #17 (thick), sealant:
available 985, 1 layer.
 537

4.0. Anodic Films

EIS has also been a very useful tool in the analysis of the corrosion
behavior of anodized Al and its alloys. Mansfeld and Kendig (21,22)
and Lorenz and co-workers (23,24) helve presented recent results. Fig.
22 gives a model for the anodized coating an Al which consists of an
outer porous layer and an inner barrier layer. Various equivalent
circuits have been proposed to explain the EIS data for anodized Al as
shown in Fig. 22 a - c . The simple model in Fig. 22c can be used to
fit the experimental data determined in the usual frequency range.
Mansfeld and Kendig (21,22) have studied the effects of alloy
chemistry, type of anodizing ba1th and sealing treatment on the
corrosion resistance during exposure to NaCI. Fig. 23 shows results
for chromic acid anodized (CAA) Al 7075-T6 which was unsealed, hot
water sealed (HWS) or dichromate sealed (DS). While the HWS sample
showed the expected changes in the high-frequency part of the spectrum
due to sealing, the DS sample was v'=ry similar to the unsealed sample.
Since the DS sample is very corrosion resistant, one can conclude that
the mechanism of sealing in dichromate is not that of plugging the
pores, but of filling the pores with inhibitor. The software program
ANODAL is used to fit the experimen"tal data to one of the models shown
in Fig. 22 a - c .

5 90

4
~ 2
75

60 -;
~
:C
c
..,'"
~

45
"
C>

"-----
0

'"
0
...J
:; 30 ~ "
~
.<:
Q.

Curve 1: Phase Angle ·1:11" 15

2: Log f.. in
3: Log f.
2 0
0 10 20 30 40
I mmersion Days

Figure 21. Time dependence of ¢ min' fmin and fb for the data
in Fig. 20
538

FILM SURFACE

'LUMINUM 0"0'. " "

PORE: IRll

BARRIER LAYf:R IR2 C2 )

ALUMINlJM

,.,

fbi

c,

,,' ~~f-
R,

Figure 22. Model for anodized surface layer on Al and

proposed equivalent circuits

2 x 10 7

2 x 106

~
:I:
0
;;:
4 <31 2 x 10 5 1
x
i2l ~
;;; 3 2 x 104
§.
t:i <!J ;;;
2
'g"
2 x 103
<!J UNSeALED.3h
~ HOT WATER SEAL 20h
<31 DICHROMATE SEAL, 4h

LOG w (w IN RAO/S)

Figure 23. Bode-plots determined in 0.5 N NaCI for chromic

acid anodized Al 7075-T6 which was unsealed (curve 1), hot
water sealed (curve 2) or dichromate sealed (curve 3)
 539

Problems have been observed with anodizing of the metal matrix

composite AI/SiC. The spectra for a hard anodized surface (sulfuric
acid) of AI/SiC do not agree with those for Al 6061 (Fig. 24) and the
corrosion resistance is poor. Apparently, the presence of the SiC
particles interferes with the formation of the oxide film, especially
the barrier layer.

B r-----,------.------~----_r------r_----._----_,

0.5 N NaCI SAA (hard l/HWS

7 1. 2h
2. 8d) Al 6061
rue 3. :55 d
4. 2 h)
u 5. 9 d SiC; Al 6061
. 6
e 6. 51 d
J:=
o
C
... 5
!::!
./
:;: 4
Cl
o
...J

2 ~ ____...J__ _ _ _ ~ ______ ~ _____ ~ _ _ _ _ _ _L __ _ _ _...J__ _ _ __...1

-2 -1 o 2 3 5
Logf (f in Hz)

Figure 24. Bode-plots for sulfuric acid hard anodized

Al 6061 (curves 1 - 3) or AI/SiC (curves 4 - 6) during
exposure to 0.5 N NaCI

Fig. 25 gives an example from tests carried out during outdoor exposure
in California in which the effects of acidic fog on materials were
determined. Curve 1 is for the unexposed sample, curve 2 is for the
side of the exposed sample which was facing the sky at a 30 0 angle and
curve 3 is for the side facing the ground. These results can be
interpreted by assuming that additional sealing had occurred during the
15 months exposure period with the effect being more pronounced for the
side facing the ground. Similar results have been obtained for shorter
exposure periods at this site and also at two additional sites in
California (25). Evidently long periods of acidic fog did not occur
during the exposure time in 1987 and 1988. The results in Fig. 25 show
that EIS can be used to obtain information concerning the structure of
protective films and their changes with exposure to corrosive
atmospheres which are not avai lable with other relatively simple
teChniques.
540

80

75

...
~

E
i
60

r
I:'
•E
.&;
0
.s w
.E ..J
t::!. (!)
z
~ 30 ~
(!) w
0 Ul
..J ~
x
15 0..

0
-2 -1 0 1 2 3 4 5
LOG , (' In Hz)

Figure 25. Bode-plots determined in 0.5 N NaCl for sulfuric

acid anodized AI; curve 1: unexposed blank; curves 2 and 3:
after exposure at Visalia, CA for 15 months; curve 2: side
exposed facing the sky, curve 3: side exposed facing the ground

5.0. cathodic Protection of Stainless Steels in seawater

EIS has been used recently to monitor the formation of calcareous

deposits on stainless steels and titanium during cathodic protection.
Samples of stainless steel (SS) type 304, 316, AL6X and Ti grade 2 were
exposed at the Naval Civil Engineering Laboratory in Port Hueneme,
California with application of a potential of - 850 mV vs SCE in
flowing seawater (26). Seperate sets of samples were exposed for
periods of one to four weeks. The impedance spectra were determined at
Ecorr about 18 h after the potential had been interrupted. Fig. 26
shows spectra obtained for the four different materials after 13 days,
while Fig. 27 shows the spectra for SS 304 as a function of exposure
time. After 5 days the impedance is dominated by the electrode
capacitance (Fig. 27a) and the phase angle shows that only one time
constant occurs (Fig. 27b). A large change in the spectra occurs
between 5 and 13 days at which time the spectra are very similar to
those predicted by the model for polymer coated metals (Fig. 16). The
spectra for the four different materials are very similar at this time
(Fig. 26). With increasing exposure time at -850 mV the calcareous
deposit seems to undergo changes in its structure as can be seen by the
changes of the impedance in the high-frequency region.
 541

(a)

6r------r----~------~----_r------~----~----_,

1. SS304
5 2.SS316
@ 3. AL6X
E
u
4. Ti
. 4
E
~
o
c
." 3
t!
/'
N
/'
01
o
...J

0
-2 -I 0 2 3 5
Logf (f in Hz)

(b)

90
1. 55304
2. SS316
75 3.AL6X
-;:; 4. Ti
Q)
(..
01
Q) 60
~

....01
Q)

c 45
«
Q)
If)
10
~
30
a.

15

0
-2 -I 0 1 2 3 4 5
Logf (f in Hz)

Figure 26. Bode-plots cathodically polarized for SS 304,

316 and Al6X and for Ti grade 2 after exposure to seawater
for 13 days.
542

(a)

6~----~-----r----~------r-----.------r-----'

1. 5 days
2.13
5
3 21
@ 4.42
E
u
• 4
E
1:
o
c:
.... 3
!::!
..-'
~ 2
OJ
o
..J

0
-2 -1 1 2 3 4 5
Logf (f in Hz)

(b)

90

1. 5 days
75
2.13
a;- 3.21
'<-" 4.42
OJ

'"
~
60

'"
rl
OJ 45
c:
""
'"
Ul
III
1: 30
0..

15

-2 -1 0 2 3 5
Logf (f in Hz)

Figure 27. Bode-plots for cathodically polarized 55 304

as a function of exposure time to seawater.
 543

6.0. SUDIDary and Conclusions

The examples given here for the application of EIS in the evaluation of
methods of corrosion protection have shown that much useful information
can be obtained. While the results obtained so far are often of a more
quali tati ve nature, they demonstrate the advantages of EIS over
traditional dc techniques and point the way for a more quantitative
analysis of these complex systems in the future. In order to advance
to this stage, improvements in the models for the different corrosion
systems and in the software for the analyis of the experimental data
are needed.

References

1. F. Mansfeld, M.W. Kendig and W. J. Lorenz, J. Electrochem. Soc.

lB., 290 (1985)
2. W.J. Lorenz and F. Mansfeld, Corr. Sci. ~, 647 (1981)

3. W.J. Lorenz and F. Mansfeld, Electrochim. Acta 11, 467 (1986)

4. K. Juettner, First International Symposium on Electrochemical

Impedence Spectroscopy, Bombannes, France, May 1989, Plenary
Lecture No. 3

5. K. Juettner, W. J. Lorenz, M.W. Kendig and F. Mansfeld, J.

Electrochem. Soc. ill, 332 (198:8)

6. B.R.W. Hinton, D.R. Arnott and N.E. Ryan, Mater. Forum ~, 162
(1986)

7. F. Mansfeld, S. Lin, S. Kim and H. Shih, Corr. Sci. 27, 997 (1987)

8. F. Mansfeld, S. Lin, S. Kim and H. Shih, "Surface Modification of

Al Alloys and AI-Based Metal Matrix Composites", Electrochim. Acta
34, 1123 (1989)

9. F. Mansfeld, S. Lin, S. Kim and H. Shih, "Electrochemical

Impedance Spectroscopy as a Monitoring Tool for Passivation and
Localized Corrosion of Aluminum Alloys", EMCR' 88, Zurich,
Switzerland, July 1988, Matericlls Science Forum, 44, 45, 83 (1989)

10. F. Mansfeld and H. Shih, J. Electrochem. Soc. ill, 1171 (1988)

11. F. Mansfeld, Corrosion 44, 558 (1988)

12. H. Shih and F. Mansfeld, Corrosion 45, 615 (1989)

544

13. F. Mansfeld, M. W. Kendig and S. Tsai, Corrosion 38, 556 (1982)

14. M.W. Kendig, F. Mansfeld and S. Tsai, Corr. Sci. 23, 317 (1983)

15. F. Mansfeld and M. W. Kendig, ASTM STP 866, 122 (1985)

16. S. Haruyama, M. Asari and T. Tsuru, in "Corrosion protection by

Organic Coatings", The Electrochem. Soc., Proc. Vol. 87-2, p. 197
(1987 )

17. F. Mansfeld and C.H. Tsai, "Determination of Coating Delamination

with EIS", submitted to Corrosion

18. F. Mansfeld, C.H. Tsai and H. Shih, "Determination of Coating

Delamination and Corrosion Damage with EIS", Proc. Int. Symp.
"Advances in Corrosion Protection by Organic Coatings", Cambridge,
England, April 1989

19. J. Titz, G. H. Wagner, H. Spaehn, M. Ebert, K. Juettner and W.J.

Lorenz, ibid, Corrosion (in press)

20. F. Mansfeld, S. Kim and S. Lin, "Evaluation of Corrosion

Protection Methods for Magnesium", J. Coating Techn. 61 (774),
33 (1989)

21. F. Mansfeld and M. W. Kendig, Corrosion il, 490 (1985)

22. F. Mansfeld and M. W. Kendig, J. Electrochem. Soc. ~, 828 (1988)

23. J. Hitzig, K. Juettner, W.J. Lorenz and W. paatsch, Corr. Sci. 24,
945 (1984)

24. J. Hitzig,K. Juettner, W.J. Lorenz and W. Paatsch, J. Electrochem.

Soc. lll, 887 (1986)

25. F. Mansfeld, R. Henry and R. Vijayakumar, "The Effects of Acidic

Fog and Dew on Materials", Final Report to California Air
Resources Board, June 1989

26. F. Mansfeld, C.H. Tsai, H. Shih and B.J. Little, "An EIS
Evaluation of Stainless Steels Exposed to Seawater", The
Electrochemical Society, Los Angeles, CA, May 1989, paper No. 57
METHODS FOR STUDYING CORROSION INHIBITION

z. Szklarska-Smialowska
Fontana Corrosion Center
Department of Materials Science and Engineering
The Ohio State University
Columbus, Ohio 43210

ABSTRACT
The important role of adsorption and adsorption
isotherms in the action of inhibitors is emphasized.
The most useful techniques to study adsorption and
inhibition efficiency of various inhibitors in
electrolytic solutions are reviewed and examples of
the results obtained by their use are given.

1. INTRODUCTION

In about the last thirty years, the subject of corrosion

inhibition, which has long been traated empirically, has developed into
a branch of knowledge where theoretical consideration is beginning to
provide guiding light to experimentation. Numerous new experimental
techniques have been introduced to study inhibition and inhibitors
quantitatively. Especially valuable are techniques permitting the
adsorption characteristics of different inhibitors to be defined,
because adsorption is almost always the first and most important step
in the protective action of many organic and inorganic substances. In
this article, a review of the most useful techniques is presented and
examples of the results obtained by their use to study inhibition in
electrolytic solutions are given.

545
M. G. S. Ferreira and C. A. Melendres (eds.).
Electrochemical and Optical Techniquesjor the Study and Monitoring of Metallic Corrosion. 545-570.
© 1991 Kluwer Academic Publishers.
546

2. INHIBITOR EFFICIENCY DETERMINATION

From a practical point of view, it is important to know how
effective protection from corrosion of a metal can be obtained by
adding an inhibitor to the aggressive environment. It is customary to
indicate the degree of inhibition by means of "percentage inhibition",
defined as
( 1)

where Vo is the corrosion rate measured in the absence of inhibitor,

while V is the rate in its presence. The corrosion rate can be
measured either gravimetrically, i.e. by comparing the loss in weight
of a metal specimen subjected to corrosion in the presence and absence
of inhibitor, or by any other quantitative technique, for example,
measurement of the rate of hydrogen gas evolution, or analytical
determination of the concentration of dissolved metal ions in the
corrosive medium with and without inhibitors. These methods can only
be used when uniform (general) corrosion occurs. For other types of
corrosion, e.g., pitting or stress corrosion cracking, special
techniques must be used, such as measurement of the pit nucleation
potential or mechanical tests on tensile specimens. Because inhibitors
are mostly used to protect metals from uniform corrosion, these special
techniques are not reviewed in this article.
The gravimetric method is most commonly employed for accurate
measurements, for example, to determine the effect of inhibitor
concentration on inhibition efficiency with the aim of deriving
adsorption isotherms. However, this method is time consuming. More
convenient, although less accurate, are electrochemical methods based
either on extrapolation of anodic and cathodic polarization curves
(Tafel curves) to their intersection point, or on linear polarization
measurements. Both these electrochemical techniques are so widely used
that there is no necessity to describe them here in detail.

3. ADSORPTION ISOTHERMS; THEIR SIGNIFICANCE. TYPES. AND METHODS OF

DETERMINATION
In most cases, protection of metals from corrosion, by inhibitors,
is brought about by covering the metal surface with an adsorbed layer
which restricts the access of aggressive species from the environment
to the metal surface. One of the most convenient ways of expressing
adsorption quantitatively is by deriving the adsorption isotherm which
characterizes the given metal/inhibitor/environment system. For this
purpose, measurements of inhibitor efficiency as a function of
inhibitor concentration are made at constant temperature, and the
results are expressed by a proper equation.
Determination of the type of adsorption isotherm, taking into
account the degree of surface coverage (8) on the metal/environment
 547

boundary by inhibitor molecules (or ions) as a function of inhibitor

concentration (more precisely activ1ty), provides valuable information
about the adsorption process. In this way important data can be
defined such as the standard free energy of adsorption (6Goads), its
dependence on the degree of surface coverage (9), adsorption
equilibrium constant (K), magnitude and character of the mutual
interaction between molecules of the adsorbed substances and also
between these molecules and the surface metal atoms.
The surface coverage is calculated from the formula:
9 .. 1 - V/V o (2)

This equation can be used when the inhibitor mechanically screens

the electrode surface and does not affect the mechanism of metal
dissolution.
Several adsorption isotherms have been derived for different
situations at the solid/liquid and solid/gas interfaces. The most
commonly used adsorption isotherms are those proposed by Langmuir,
Frumkin, and Temkin.
Frumkin's adsorption isotherm is defined as follows:

_9_ exp (± f6) = Ka (3)

1 - 9
When f>O, Temkin's isotherm results:

_9_ exp (f9) .. Ka (4)

1 - 9
When f=O, Langmuir's isotherm results:

9 = Ka (5)
r8
Other more complicated adsorption isotherms sometimes fit the
experimental results better, e.g., that of Hill de Boer;

_9_ exp _9_ (-f6) .. Ka (6)

1-6 (I-9)

or Parson's isotherm of adsorption:

9 exp 2-9 exp (-f9) = Ka (7)

H (I-9)2
The constant K in the above equations is the equilibrium
constant for the adsorption process which is related to the standard
free energy of adsorption, 6Goads' according to the following equation:
548

flGoads 1
K = exp (- - - ) (8)
RT 55
The constant f in equations 3-8 depends upon the intermolecular
interaction in the adsorption layer and on the heterogeneity of the
surface. Each of the above isotherms has a characteristic maximum
value of f. If f is positive, mutual attraction of molecules occurs; if
f is negative, repulsion occurs. The constant a is the activity of the
inhibitor in the electrolyte.
Fig.l shows adsorption isotherms for aliphatic primary amines
containing 6 to 12 carbon atoms. These isotherms were obtained in a
study of inhibition of the corrosion rate of mild steel in 1 N sulfuric
acid solution at room temperature (1). In this case, adsorption of
primary amines was found to follow Frumkin's isotherm. The f
coefficient was more negative (f<O), the greater the content of C
atoms. An increase in the aliphatic chain length by one CH 2 group
increased the f value by 0.55. Moreover, f changed its sign from
negative to positive which indicated that the interaction between
molecules was chain length dependent and caused the adsorption energy
to increase with surface coverage. The standard free energy of
adsorption also changed regularly with increasing chain length; Table
1. A similar behavior was observed for aliphatic acids and alcohols
(1) •
Frenier, et al. (2) measured the adsorption of
2-benzoyl-3-methoxy-1-propene (BMP) on low carbon steel in strong HCl
solutions at two different temperatures. They found adsorption to occur
according to the following equation:

--:_0_ = K [BMP]n (9)

1-0
Results of these measurements are shown in Fig. 2. The authors (2)
calculated the standard free energy of adsorption (flGoads), average
enthalpy (flHO), and average change of entropy (flSO) (Table 2). The
increase of n with temperature and the positive value of flS o suggested
an oligomerization effect. The increase in entropy was considered to
reflect the increase of inhibitor efficiency with increasing
temperature, which led to an increase in K and a decrease in flGo ads.
In a study by Kaminski and Szklarska-Smialowska (3), adsorption of
thiophene and its derivatives on low carbon steel in sulfuric acid was
found to obey isotherms proposed by Parsons and Hill de Boer (Fig.3).
These isotherms were derived theoretically by assuming the existence of
a mobile adsorption layer, suggesting physical adsorption. The standard
free energy of adsorption was found to increase from 4.95 Kcal/mol for
 549

'Y. 8
90
60
·10

bO

50
40

20

10-'
c, mol/l

Figure 1. Adsorption isotherms for aliphatic amines on mild steel in

1 N H2S04 ; Temp. = 25°C. (i) n-hexylamine; (2) n-octylamine;
(3) n-decylamine; (4) n-dodecylamine (I).

2.0

1.5 ,," •
, .,,"
o 30°C ,
• 65°C
, y"
1.0
CD
.........
I

~ 0.5 . . ... 6"

01
Sl 0.0
... ,,'" •
-0.5

-1.0
-4.0 -3.5 -3.0 -2.5 -2.0
log (BMP(moI/L)]

Figure 2. Fits of inhibition data to linearized power law isotherms

for J 55 rotating disc electrode in 15% HCI at 30 and 65 C
inhibited with BMP (2).
550

Table I: Ther.odynamic Data for Aliphatic Amines Obtained

Fro. Experimental Adsorption Isotherms

Number of C in Aliphatic f ll.GO ads K

Chain cal/mol

6 +1.6 -4.9 74

8 +0.5 -5.7 296

10 -0.6 -6.5 1000

12 -1. 7 -7.4 2500

Table II: Thermodynamic Values for J55 Steel in 15% Hel

Inhi bitor Temperature K n ll.Go ads ll.Ho ll.$O

(C) Kcal/mol cal/mol/K

BMP 30 1.5 x 10 2 0.68 - 2.00

20.7 78.0
BMP 65 5.1 x 10 3 1.19 -5.76
 551

thiophene (weak inhibitor) to 8.5 Kcal/mol for 2-acetylothiophene which

was the most effective inhibitor among the compounds studied in (3).

4. EFFECTS OF INHIBITORS ON THE SHAPE OF POLARIZATION CURVES

Anodic and cathodic polarization curves measured in the presence
of an inhibitor differ from those measured in pure electrolyte
solution. An inhibitor can affect only one branch of the polarization
curve, i.e., either anodic or cathodic, or both branches.
Inhibitors influencing the anodic, cathodic or both branches of
polarization curves are called anodic, cathodic or mixed inhibitors,
respectively.
Schematic plots representing the effects of inhibitors on the
course of Tafel curves without changing their slopes are shown in Fig.
4. However, there are also inhibitors which affect the slopes of Tafel
curves. This occurs when an inhibitor participates in the electrode
reaction by producing some intermedhry substance which controls the
summary electrode reaction.
Fig. 5 from reference (4) and Fig.6 from reference (5) give
examples of polarization curves measured for iron in the presence or
absence of different inhibitors in sodium sulfate solution of pH 2 and
in 1 N hydrochloric acid solution, respectively. In the case shown in
Fig. 5, all the three inhibitors studied increase the hydrogen
overpotential and decrease the corrosion current. However, two of them
do not affect the anodic branch, whilst n-decylamine increases the
anodic overpotential. Hence, homopiperazine and urotropine are
cathodic inhibitors, whilst n-decylamine is a mixed inhibitor.
In Fig. 6, the effects of several organic compounds (containing
phosphorus and sulfur or selenium) on polarization curves measured for
iron on exposure to Hel solutions are shown. All these compounds
influence both branches of polarization curves, which indicates that
they are adsorbed on the metal surface in molecular form. At anodic
potentials, where the corrosion current begins to increase drastically,
desorption of inhibitor molecules takes place.
The schematic plots shown in Fig. 4 apply to many organic
inhibitors. On the other hand, there is a group of inorganic inhibitors
called passivators (e.g., chromates, nitrites) which exert another kind
of influence on the anodic polarization curve. If used in appropriate
concentration, they shift the electrode potential to high positive
values where a protective passive film forms on the metal surface.
However, at too low a passivator concentration, the corrosion rate
(current) can increase further, as shown in Fig. 7. Passivators of
this type behave in the same manner in aerated and deaerated solutions.
Another type of inorganic and organic compounds, (for example
molybdates, tungstates, benzoates) act as passivators only in the
presence of oxygen dissolved in aqueous solution. As with true
passivators, spontaneous passivation occurs when the electrode
potential reaches a high positive value where a passive film is formed.
552

2-Br
80

70
2-Cl
60
3-CH~

40

20

10

10-- 10-' 10-'

109 C. mole Il.

Figure 3. Dependence of the degree of coverage of steel surface in 1 N

H2S04 at 25 C on the concentration of adsorbing substances;
.-(3), x - 2~bromothiophene, 0 - 2- chlorothiophene,
3-methylothiophene. The full line represents the isotherm
giving the best description of all experimental points.

0) + b)+ c) +
,/ ,/

-IT(
/'

~
/'
,/
,/,/
EcOl"; -
Ecor Ecor ECOfi ,Ecor (-
Ecori - I I,
II' I I,
I I,
I '"
I "- I I "-
I I " I I
"-
'cori loor iCOfj ieor leorj ieor
Lg i Lg i Lg i

Figure 4. Schematic plot of anodic and cathodic polarization curves

for dissolution of metal in the presence of inhibitor: (a)
anodic, (b) cathodic, (c) mixed. Ecor , Ecor and i cor , i cor 1.
i
are corrosion potentials and currents with and without
inhibitor, respectively.
 553

03.0"'OI.~
1\3.30"0 ."UIIO~
02.M • 10 • II .. [)(C~

'-='0"""'.•~--:-:'0L_......---:C'0~·74---;';;:0'-.·';---;;;'"'0-.- - . 100
~1Afa0021

Figure 5: Polarization curves for pure Fe in Na 2S0 4 solution at pH 2

(4).

Because passive films playa very important part in the protection

of metals from corrosion. they are intensively studied using different
techniques with the aim of determining their chemical composition,
structure. and thickness.

5. EFFECTS OF INHIBITORS ON THE DOUBLE LAYER CAPACITY

Compared with polarization curve measurements, a much more
accurate determination of the potential range within which inhibitors
are adsorbed on the metal surface can be obtained from measurements of
the double layer capacity (d.l.c.). Adsorption of an inhibitor causes
the structure of the double layer to be modified, and the capacity to
decrease. The changes are related to the coverage of the metal surface
with inhibitor molecules, and hence to the inhibitor efficiency (6,7).
Chemical compounds that are strongly adsorbed on the metal surface
increase the potential range within which a decrease in d.l.c. is
observed. Depending upon the type of inhibitor, the region of decreased
capacity can concern only one branch of the polarization curve (either
cathodic or anodic) or expand to both sides of the potential of zero
charge (p.z.ch.). Cathodic inhibitors are adsorbed at potentials more
negative than p.z.ch .• while anodic inhibitors are absorbed at
potentials more positive than p.z.ch. Compounds that adsorb in
molecular form decrease the d.l.c.oll both sides of p.z.ch.
554

:0

:0

~o

Figure 6: Polarization curves for iron in 1 N HCl and 1 N HCl with

addition of (C2HS)3PSe, (C 4HgO)3 PS , (C6H13)2PSOH,
(C aH17 0) PSOH (S)
 555

+
E
Figure 7: Influence of passivator concentration on the corrosion
potential for metals which can exist in active and pasive
states. I ~ 7 increased concentrations of passivator.

As the effectivness of an inhibitor increases, the larger is the

potential region with decreased d.l.c. Fig.8 gives an example of
d.l.c. measurements on iron in IN Hel solution with and without the
addition of organic compounds containing phosphorus and either sulfur
or selenium (7). The potential of zero charge for iron is -370 mV H (8)
and -350 mV H (9) respectively. At potentials less negative than that
of zero charge, the iron surface is charged positively and adsorbs
anions, whereas at more negative potentials the iron surface is
negatively charged and adsorbs cations. The inhibitors shown in Fig. 8
have been found to decrease d.l.c. uniformly on both sides of p.z.ch.;
they belong to the class of mixed inhibitors in agreement with what has
been found on the basis of polarization curve measurements (Fig. 6).

6. AC IMPEDANCE MEASUREMENTS
For the last twenty years, electrochemical impedance measurements
have been done with the aim of characterizing different electrode
processes. Impedance data provide valuable information and allow
a better understanding of the mechanism of corrosion inhibition and
metal protection by polymer coatings and anodic films. Several
presentations in this Institute deal with impedance, hence only two
examples concerning application of this technique to study inhibition
are given here; both are taken from a paper by Duprat and Dabosi (to).
556

c
btJ o IN Ha +8'1. f.H,pH
0/11 HCl
"0.02 mil (CMPS
2f)() • 0·lXJ25 mil (C,J/I1~PSe
• OfJ0.5 mIL (C,H,IJ),PSe
• OIU mJl (C,Ji,O}IP(S/OII
.. 00004 mII (C,H,p),P(J)OH
• 0 IXXJ4 ",/1 (C,H,iJhP(S)OH
150 o 00004 .,/1 (C,H4P(Sf)H

/00

50

o~~~~~~~ 50 -130 -210 -290 ·370 -450 5.lO

-
610
£.[mVl -

Figure 8: Effect of potential on the double layer capacity in IN HCl

with and without the addition 'of different inhibitors (7).

In their studies, the effect of diamino-I-2, ethane on the corrosion of

mild steel in aerated 3% NaCl solution was examined. This inhibitor
causes strong alkalization; AC impedance was also measured in NaCl
solution to which NaOH was added to obtain the same pH as in the
inhibitor-containing solution. Measurements were made on a rotating
electrode. Results are shown in Fig. 9. In pure 3% NaCl (pH 6.2), the
Nyquist plot exhibits two loops. The right one, which is attributed by
the authors (10) to molecular diffusion of dissolved oxygen through a
porous layer of corrosion products, disappears in the remaining two
alkaline solutions. As seen, in the solution made alkaline with NaOH,
the polarization resistance is much lower than in that containing the
inhibitor. Therefore, it is concluded that the interaction of the
 557

-IG(Il.,m·)
O.4h.
(.)
.00
J • •
• •
100 • •
~4h. .~O.t" •

•• 25hz
•
•
10", .. _--.: . .

100 '00 '00 400 500

R (O.em')
-IG(a.em') ;'"
• • (b)
• •
-'-'01\& •
•
50
•
•• •
10

'0
100 ••

SO 100 ,. I I ~
'00
.~O

•
R(0.,,,,2)
.• ".

(e)

..
G..311t.

• ..3hz

~
•
• ••
I
...-r'.....
" u~.. ~
••
.0

Figure 9: Electrochemical impedance diagrams in galvanostatic mode at

°
i - and w = 1000 rpm; (IO), a) 3% NaCl + H2N-{CH2)2-NH2
(la- 2M), pH = 10.85, b) 3% NaCl, pH 10.85 adjusted by NaOH,
c) 3% NaCl, pH 6.2.

inhibitor with the metal surface causes the polarization resistance to

increase. The shape of the impedance diagram taken in solution of
equal pH is the same; the authors concluded that the inhibitor reduces
the rate of metal dissolution without modification of the anodic
mechanism, and that the tested inhibitor strengthens the "prepassive"
layer of Fe{OH)2 on the steel surface due to a chelating mechanism.
Another impedance diagram taken from the same paper (10) is shown
in Fig.IO for iron in 3% NaCl solution with and without the addition of
zinc fluorophosphate (ZnP0 3F.2.5H 20). The polarization resistance
increases considerably in the presence of this inhibitor. In the low
frequency range (Fig. lOb), a capacitive loop is obtained (capacity
558

_ jG (O.em')
o.63hz.

500

• \ • • •
(a)

••
63hz • .,/63mhZ

•
~ ~6.3hz
• I
~6.3mhz

0 500 1000 1500

R(Il.em2 )

_ jG(O.cm 2 )

.....--S.Jhz

•
(b)
100 •
8.3khz •

Figure 10: Electrochemical impedance diagrams in galvanostatic mode at

i = 0 and w =1000 rpm; 3% NaCl + InP0 3F - 2.5 H20 ( 10 -2
M). Detail of HF range in (b) (10)

5~F/cm2) suggesting the presence of a thick film on the metal surface.

7. THE USE OF OPTICAL AND ELECTRON SPECTROSCOPY TECHNIQUES TO STUDY

PASSIVE FILMS ON METALS IN THE PRESENCE OF INORGANIC INHIBITORS
An extended review of the application of surface analytical
techniques (AES, XPS, SIMS, MOSSBAUER, IR etc.) for studying corrosion
inhibition has been published by Graham (11). Modern electron and
optical spectroscopy techniques such as Auger electron spectroscopy
(AES), secondary ion mass spectroscopy (SIMS), X-ray photoemission
spectroscopy (XPS) (ex-situ techniques) and infrared spectroscopy
(IRS), ellipsometry, Raman spectroscopy (in-situ techniques) are now
 559

widely employed to study thin films on solids. As far as ex-situ

techniques are concerned, it is understandable that the removal of a
metal specimen with a protective film on it from the solution may bring
about significant changes in the chemical composition and structure of
the film. Nevertheless, from these measurements, many important
conclusions can be drawn, especially by comparing results obtained
using different techniques.

7.1 ELLIPSOMETRY AND AES

Ellipsometry is one of the methods enabling examination of the
metal surface without taking the specimen out of the solution; however,
many precautions have to be taken against possible errors in the
interpretation of results. Therefore, ellipsometry is mostly used in
combination with other techniques.
For example, ellipsometry combined with AES and electrochemical
measurements was used to study the passivation of iron in aerated water
solutions containing different concentrations of one of the following
anions: phosphate, chromate, nitrite, molybdate, tungstate, and
arsenate (12). On the basis of the results obtained, it was concluded
that the protective properties of the film could best be qualified by
the potential increase related to the corresponding increase of film
thickness, i.e. by the electric field strength build up in the film
during its formation. Some examples of data taken from (12) are given
o
in Table III. The highest increase as expressed in mV per A was
obtained in Na 3P0 4 (pH 12) and in NaN0 2 (pH 7). The films formed in
these two solutions were composed of iron oxide without inclusions of
other compounds and exhibited the best protective ability, namely, the
lowest ion flux through the film. These films were characterized by
o 0
high electric fields (32 to 35 mY/A) and a small growth rate (3 to 10 A
per decade time).
Films formed at open circuit potentials in chromate (pH 8 to 12),
molybdate (pH 9.6), and tungstate (pH 8.6) solutions were less
protective. All of these films contained metal ions taken from the
o
solution. A lower field existed in these films (2 to 20 mY/A) and the
growth rates were 15 to 130 A/decade time. These films blocked the
iron surface quite effectively, but longer times and thicker films were
necessary to attain the potential range within which oxides exhibiting
better protective abilities were produced.
The less protective films formed in Na 2HP03 (pH 9.1) thickened
rapidly and grew to thicknesses many times greater than those formed in
o
the other solutions. The electric field was about 0.02 mV/A and the
growth rate was several thousand A per decade time. The film formed in
Na 2HP0 4 under open circuit conditions was composed of iron phosphate
o
which is a good conductor. A film several thousand A thick was
obtained before the potential of oxide formation was reached. Thus,
560

Table II I: Properties of Film Formed on Iron While Exposed Under

Open Circuit Conditions to Inorganic Inhibitors

Solution pH Field Growth Rate, Auger

Strength, A/decade Analysis
mV/A time

0.1 N Na 3P04 12 35 5 0, Fe, P

Only on the
Surface

0.01 N NaN0 2 7 32 10 0, Fe, Small

Amount of N

0.01 N Na 2Cr0 4 8-12 3-28 20-33 0, Fe,Cr

near the
substrate
1ack of Cr

01 N Na 2Mo0 4 9.6 13.5 15 0, Fe, Small

amount of Mo

0.05 N Na 2W0 4 8.5 13.5 15 0, Fe

substantial
amount of W

0.01 N Na 2HP0 4 9.1 0.02 Several 0, Fe, P

Thousand
 561

7
Mo 3d5/z
6
232.2 tV I

~
Mo3dl/Z
5 235.4 tV

I&J
4
.::::
I&J

z 3

BINDING ENE RGY. tV

Figure 11: Mo XPS spectrum of the passive film formed on steel in 1000
mg!1 Mo0 4 - Solution, pH 9.0 (13).

the Na 2HP0 4 solution inhibits corrosion of iron as a result of its

buffering ability and capacity to block defect sites in the passive
film by precipitating insoluble ferric phosphate.

7.2 X-RAY PHOTOELECTRON SPECTROSCOPY

While AES is mostly used to determine the content of elements in
the oxide film, XPS provides binding energy data permitting
identification of elements or groups contained in the film being
studied. Unfortunately, there is often uncertainty as to the exact form
in which a given inorganic inhibitor exists in the passive film and it
is impossible to answer decidedly questiops concerning the element or
group responsible for passivity. For example, it is not clear how
molybdate anions act as corrosion inhibitors.
Stranick (13) used AES and XPS techniques to analyze films on 1010
steel in alkaline and neutral solutions containing molybdate ions. He
found the passive film formed in alkaline molybdate solution to be 50
angstrom thick and contain up to 15 at.% molybdenum in the form of
M00 42- In acid solution, the film was black, unprotective, and
primarily composed of Mo0 2. Fig. 11 shows XPS molybdenum spectra
obtained for a film formed in a solution of pH 9. The peaks occur at
binding energies of 232.2 eV and 235.4 eV which characterize M00 42-
562

BLACK FILM

(j) Mo 3d 5/2
f-
z 232.1 eV
::J (Mo+6)
r
cr
<t
t:
Mo3d 3/2
235.3 eV
l'\ Mo 3d 5/2

1:""
(I)
X 1\ 229.1 eV

f\
a:
II
~
>- tI
I
ill I I
t:
Cf)
1\
, I
z I I
w
I I I
j \
l- Mo 3d 3/2
~ 232.2eV I{; I I I
I 1
Vi I I I I
/ ',\,
I I I
I
I
I \ I I
I I I I
I \ I I
I I I \
I ...' \
.- '" \
I I I I I
240 236 232 228 224
BINDING ENERGY, eV

Figure 12: Mo XPS spectrum of the black film formed on steel in 1000
mg/l Mo0 4- 2 solution of pH 4.5 (13).

In Fig. 12, the XPS spectrum obtained after exposure to a solution of

pH 5 is given, showing a peak for a binding energy of 229.1 eV which is
characteristic of Mo0 2. These results indicate that Mo0 42- renders the
passive film on steel to be protective, whilst the presence of Mo0 2
does not do so.
Sastri, et al.(14} studied the protection from corrosion of 1010
steel in coal water slurries (pH 8.5) by the use of chromate, molybdate
and tungstate anions as inhibitors. At the same inhibitor
concentration, the percent corrosion efficiency measured by
electrochemical techniques and from weight loss decreased in the order
cr0 22- > Mo0 42- > W0 42-. Differences between the efficiency of these
three anions were rather small; for example, 96, 89, and 87% for
chromate molybdate, and tungstate, respectively. Electron microprobe,
AES, and XPS analyses were also made showing the presence of hexavalent
molybdenum and tungsten oxides and trivalent chromium oxide species in
 563

c ••

t 000 goo 800 700 600 SOO 400 300 200 100

BINDING ENERGY. eV

Figure 13: XPS spectra from mild steel before and after exposure in
2.45 g/1 NaCl + 0.45 g/1 Na 2S0 4 solution (pH 8.5)
containing 0.01 Mchromate (14).

films that formed on the surface of steel specimens on exposure to

solution containing the respective inhibitors. At the same inhibitor
concentration in the solution, the surface concentration decreased in
the order Cr > W> Mo. Thicknesses of the passive film varied from
about 10 to 40 nm, depending on the amount of Mo, W, and Cr species
adsorbed on the steel surface. The Cr and Wconcentrations and film
thicknesses were significantly larger than those of molybdenum films,
which suggested that there was no simple relation between inhibition
efficiency and thickness or composition of the surface film.
Fig.13 gives examples of XPS spectra obtained in the presence and
absence of cr0 42- in 2.45g/1 NaCl + 0.45 g/1 Na 2S0 4 solution. The Cr
concentration in the inhibitor-containing solution was 0.01 M.
Zanzucchi and Thomas (15) studied by XPS and IRS the inhibitor
nitrilotris (methylene) triphosphonic acid (NTMP), which is employed in
humid environment, for aluminum films used as conductors for integrated
circuits. Fig.14 shows a survey spectrum of sheet aluminum treated with
564

Ols

~Li Ii Cis P2p

\il~AI2P
1\ I
, . ,
, '~'LLd_~~'~
1000 800 600 400 200 0
BINDING ENERGY (eV)

Figure 14: XPS survey spectrum of the surface of sheet aluminum

treated with NTMP (15).

NTMP and rinsed in distilled water. Peaks due to oxygen, carbon,

aluminum and phosphorus are present. The structure of these peaks under
high resolution is shown in Fig. 15 a-c. By analysis of these peaks it
was suggested that the aluminum peak is due to oxidized aluminum as
Al(OH)3 or A100H. The phosphorous peak corresponds well to an A1PO x
species although the presence of carbon suggests that some organic
compound is present in the bonded species. The binding energy of the
oxygen peak indicates the presence of A1P0 4 . The results obtained by
XPS are consistent with those of infrared data. The data suggest that
in the presence of NTMP, a surface aluminophosphonate is formed. The
total oxide NTMP layer is estimated to be about 40-65 A thick.

7.3 INFRARED SPECTROSCOPY

Two examples showing the application of infrared spectroscopy
to study inhibition are given here.
Frenier, et al. (2) studied the behavior of four phenylketones
listed in Table IV, which have been found to be effective inhibitors
protecting oil steel (J 55) from corrosion in strong HCl solutions.
Weight loss tests were conducted at 65 C, using each of the
phenyl ketones tested at concentrations of 0.2 gl100 ml in 15% HC1, and
0.4 glml in 30% HC1. After an exposure of 24 hrs, identification of
the reaction products was made by IRS, both in the bulk solution and
 565

III)

~.,

, , ,
80 75 70 145 135 125
B.E. e.V) B.E. e.V)
.,
ee) ed)

--,,' \ .... -t
, , , ~", ,"'-
sse 533 52. 2M 280 285 280
.... e.V) .... ,.V)

Figure IS: XPS high resolution spectra of an NTMP treated aluminum

surface, (a) the aluminum (2p) spectrum; (b) the phosphorus
(2p) spectrum; (c) the oxygen (Is) spectrum; and (d) the
(Is) carbon spectrum (15).

on the steel specimen surface. Organic reaction products were

extracted from the specimens and from the acid solutions using
chloroform.
It was found that irrespective of the phenyl ketone used as an
addition to HC1, after all tests BBA was the main component of the acid
solution. Tests in 15% HCl in absence of a steel specimen, showed that
conversion of these compounds to BAA was independent of the iron
surface. However, in contact with steel, polymerization of BAA into
polyphenylvinylketone (PPVK) occurred on the metal surface, according
to the following reaction series:
566

The IR spectra obtained for the solution extracted from the

surface of steel specimens after corrosion in the presence of
phenyketones fitted well with the spectrum of PPVK (Fig.16). The high
efficiency of this type of inhibitors could therefore be attributed to
the formation of a polymer film on the steel surface.
In a study by Incorvia and Haltmar (16), internal reflection
infrared spectroscopy (IRIS) was used to examine in situ the adsorption
process for various inhibitor blends and their components as a function
of time. A layer of pure iron was sputtered on two reflective faces of
a 45 0 angle Zn/Se ATR microcrystal, and the metal-coated crystal faces
were swabbed with the substance to be investigated. From the changes in
spectra with time, it was inferred that in the case of filming
inhibitors (e.g.,those containing fatty amide) the surface active
components adsorb on the metal/inhibitor interface, whilst non-active
components diffuse to the outer layer of the film. With the progress of
time, the surface active components bound to the metal surface order
their molecular structure into a microcrystalline state.

7.4 SURFACE-ENHANCED RAMAN SPECTROSCOPY.

This in situ technique was employed by Leygraf and Thierry (17)
together with ellipsometry and electrochemical methods to study the
adsorption and inhibition efficiency of various benzimidazole and
benzotriazole derivatives on copper in 0.1 MNaCl solution.
The surface Raman and bulk Raman spectra of these substances
were very similar, hence it was concluded that they adsorb non-
dissociatively (as molecules) with the imidazole or triazole ring
oriented toward the Cu surface, and are involved in the bonding with
Cu. Replacement of one functional group by another in the imidazole
ring results in detectable changes in orientation and in a different
number of bonding atoms or functional groups. Fig.I7 presents schematic
indications (by arrows) of atoms or functional groups being involved in
the adsorption process of benzimidazole and its derivatives on copper.
The highest inhibitive efficiency was obtained in the case of mercapto
benzimidazole with two functional groups being involved in the
adsorption process. On the other hand, replacement of one group by
another in the benzene ring of benzotriazole did not result in any
changes in orientation of the adsorbed molecules. However,
substitutional groups were found to influence the inhibitive
properties. For example, the best inhibitor in the series of
benzotriazole derivatives was the methyl benzotriazole which contains
 567

Table IV: Phenyl Ketones

Compound Structure Abbreviation

2-Benzolylallyl Alcohol
°
,p-C-C=CH 2 BAA

CH 20H

2-Benzoyl-3-methoxy-l-
°
,p-C-C=CH 2 BMP
propane
CH 20CH 3

2-Benzoyl-l,3-dimethoxy-
°
,p-C-CH-CH 2-OCH 3 BDP
propane
CH 20CH 3

5-Benzoyl-l,3-dioxane ,p-C-C-H BDO

568

WAVENUMBERS

7 8 e 10 '1 12 13 14 15 18
WAvE LENGTH Imlcrons)
88

83

!w 76
U

~ 13

~ 68
~
~ 63

58

4000 3600 3200 2800 2400 2.000 '.800 1 SOO 1 400 1.200 1.000 800
b WAVENUM8ERS

Figure 16. Comparison of (a) reference PPVK IR spectrum with (b) the
spectrum of film extracted from coupon inhibited with BOO
(2).

9
."'= " 9
C-M
9
.N,= .M~=( _N,=
C-N C-N
I I:t
C-N
I I:t

-
I I:t CH, OH
5

Figure 17. Schematic indications (by arrows) of atoms or functional

groups being involved in the adsorption process of
benzimidazole and its derivatives on Cu (17).
 569

the hydrophobic methyl group.

8. OTHER TECHNIQUES
Occasionally, other methods are also employed to study the
interaction of inhibitors with metallic surfaces e.g., Ion Scattering
Spectroscopy (ISS)(I8), Secondary Ion Mass Spectroscopy (SIMS) (19),
Electron Mossbauer spectroscopy (20). Extended X-ray adsorption fine
structure (EXAFS) has also been used by Kruger et aI.(2I) to study the
structure of films formed on iron in chromate and nitrate solution.
They found that the structure of chromate-formed films was more
disordered than that of films formed in nitrate. The difference was
explained by the incorporation of chromium into the film and apparently
promoting the formation of a more glass-like structure.
9. CONCLUDING REMARKS
From this review article it is evident that there is a large
variety of techniques that can be used successfully to study the
inhibition of metallic corrosion. For many practical purposes, simple
weight loss tests combined with electrochemical measurements are
satisfactory for the purpose of determining the inhibitor efficiency
and selecting a proper inhibitor to protect a metal from corrosion in a
given environment. However, for a better understanding of the
inhibition mechanism, development of new and more efficient inhibitors,
and predicting what kind of inhibitor or mixture of different
substances might be the best for a given application, it is necessary
to employ sophisticated surface analytical techniques such as AES,
ellipsometry, XPS, IR, etc. The use of these techniques can help
identify the nature and composition of the films, determine their
thickness, and characterize the type of bonding of the inhibitor
molecules on the metal surface. A good knowledge of the inhibition
mechanism allows a proper choice of inhibitor to be made on the basis
of a greatly reduced number of empirical experiments.
10. REFERENCES
1. Z. Szklarska-Smialowska, G. Wieczorek; Corr. Sci., Vol. 11, p. 843
(1971)
2. W.W. Frenier, F.B. Growcock, V.R. Lopp; Corrosion, Vol. 44, p. 590
(1988)
3. M. Kaminski, Z. Szklarska-Smialowska; Corr.Sci., Vol. 13, p. 557
(1973)
4. S.L.Granese; Corrosion, Vol. 44, p. 322 (1988)
5. Z. Szklarska-Smialowska, B. Dus; Corrosion, Vol. 23, p. 130 (1967)
6. T. Murakawa, S. Nagaura, N. Hackerman; Corr.Sci. Vol. 7, p. 79
(1967)
570

7. B. Ous, Z.Szklarska-Smialowska; Corrosion, Vol. 25, p. 69 (1969)

8. E.O. Ajazjan; Ookl. Akad. Nauk SSSR, Vol. 100, p. 473 (1955)
9. S.Trasatti; J. Electroanal. Chem. and Interface Electroch., Vol.
33, p. 361 (1971)
10. M. Ouprat and F. Oabosi; International Congress on Metallic
Corrosion. Vol. 4, p. 104 (1984) Ed. National Research Council of
Canada.
11. M.J. Graham; Corrosion/89 paper 136
12. J.B. Lumsden, Z. Szklarska-Smialowska; Corrosion, Vol. 34, p. 169
(1978)
13. M.A. Stranick, Corrosion, Vol. 40, p. 296 (1984)
14. V.S. Sastri, R.H. Packwood, J.R. Brown, J.S. Bednar, L.E.
Galbraith, V.E. Moore; Brit. Corr. J., Vol. 24, p. 30 (1989).
15. P.J. Zanzucchi, J.H. Thomas III; J. Electroch. Soc., Vol. 135, p.
1370 (1988)
16. M. J. Incorvia, W. C. Haltmar in "Syrfaces. Inhibition and
Passivation n , E. Mc Cafferty, R.J. Brodd (editors), The
Electrochem. Soc., Vol. 86-7, p. 41.
17. C. Leygraf, O. Tierry; Proceeding of the 10th Scandinavian
Corrosion Congress, Stockholm 1986, p. 249.
18. H. Puderbach; Microchimica Acta, Suppl., Vol. 10, p. 103 (1983).
19. M.F. Abd Rabbo, J.A. Richardson, G.C. Wood, C.K. Jackson; Corr.
Sci., Vol. 16, p. 677 (1976).
20. W. Meisel, E. Mohs, H.J. Guttman, P. Gutlich; Corr. Sci., Vol. 23,
p. 465 (1983)
21. J. Kruger, International Congress on Metallic Corrosion, Vol.l, p.
79 (1984) Ed. National Research Council of Canada.
ATMOSPHERIC CORROSION

Florian Mansfeld
Corrosion and Environmental Effects Laboratory
Department of Materials Science and Engineering
University of Southern California
Los Angeles, CA 90089-0241, USA

ABSTRACT. A brief discussion of the parameters which affect the

atmospheric corrosion process is followed by an explantion of recent
concepts leading to so-called damag,~ functions and classifications of
environments in terms of their corros:lvity. The principles of the use of
electrochemical sensors in outdoor exposure studies are presented next.
The remainder of this paper deals with the results of a recent exposure
study in Southern California and the approach taken in the analysis of
these results. Damage functions for qalvanized steel, nickel and a latex
house paint are presented.

1 .0. Introduction

With the realization of the adverse economic and social effects of acid
rain and acid fog, the subject of atmospheric corrosion has received
increased attention in recent years. Earlier studies of this phenomenon
had consisted mainly of field studies without monitoring of atmospheric
conditions and not much was learned concerning the fundamental mechanisms
of atmospheric corrosion and the role of pollutants. More recently
advanced research techniques have been applied in laboratory tests and
electrochemical probes have been used in field studies. Since at the
test sites which have been chosen for the more recent tests atmospheric
parameters are also being collected, it will be possible to establish
correlations between atmospheric corrosion rates and environmental
parameters. The ultimate goal of many of these studies will be the
establishment of damage functions whieh correlate the corrosion rate of a
metal with the atmospheric parameters which have caused corrosion such as
the concentration of pollutants and the relative humidity (RH).

2.0. Parameters Affecting the Atmospheric Corrosion Behavior

Atmospheric corrosion is a very complicated process which involves the

interactions of the metal, its corrosion products, the surface
electrolyte and the constituents of the atmosphere. The subject of
atmospheric corrosion and the use of electrochemical techniques have been
treated in several recent reviews (1-5). It is quite clear that
atmospheric corrosion is generally an electrochemical process which can
only proceed in the presence of an electrolyte. Atmospheric corrosion
has therefore to be considered as a discontinuous process, where the
total corrosion loss COR in a given time period is determined by the
tim~-of-wetness tWin .and the average corrosion rate rcorr,n during each
per10d n of condensat10n:

571
M. G. S. Ferreira and C. A. Melendres (eds.),
Electrochemical and Optical Techniques/or the Study and Monitoring 0/ Metallic Corrosion, 571-584.
© 1991 Kluwer Academic Publishers.
572

COR = E t •r ( 1)
n w,n corr,n
In order to measure the total corrosion loss of a material for a given
exposure period such as a year it is therefore necessary to determine the
individual values of t w , rcorr and n for a given test site. The factors
which determine these values and some of the experimental techniques for
their measurement are the following:

2.1. Relative Humidity (RH) and Time-of-Wetness two

The latter is the time during which the metal surface is covered by an
electrolyte film making corrosion possible. It depends on RH, duration
and frequency of rain, fog and dew, temperature of the air and the metal,
chemistry of corrosion products, wind speed, hours of sunshine, and other
factors depending on exposure conditions. (1).

2.1.1. Determination of tw:

a) calculated from meteorological data: RH ~ 80% and T :!: OoC. This

might not be the same as the "true" tw during which significant corrosion
occurs and which depends also on the metal and environmental properties.
However, it is quite useful for the analysis of many field data, where
actual values of tw have not been determined.

b) measured with electrochemical devices such as atmospheric corrosion

monitors (ACM) (3-5). These are electrochemical cells which produce a
signal only when sufficient electrolyte covers the electrodes to produce
a corrosion current. The time during these events is taken as two

Typical values of tw are 1000 - 2700 h/year (1).

2.2. Pollutants and Composition of Surface Electrolytes

The surface electrolyte contains species deposited from the atmosphere as

well as species from the metal as a result of corrosion. The chemical
composition of the surface electrolyte is of great importance in
determining the rate of the corrosion process. The major species of
importance in atmospheric corrosion are:

2.2.1. Oxygen: the thin surface electrolyte layer covering the metal is
saturated with 02 resulting in a high rate of the cathodic reaction
leading to corrosion. It has been found that corrosion rates reach a
maximum when this surface layer becomes very thin and dries (6).

2.2.2. S02: it is usually oxidized to H2S04 on moist particles or water

droplets. Corrosion caused by S02 is usually restricted to areas close
to a source such as power plants. S02 has been shown to be the major
cause of corrosion due to acid rain.

2.2.3. NO x : this species can play a role in atmospheric corrosion

processes as N02 which is converted to HN03 by photochemical oxidation.
Recent studies in Southern California reported by Mansfeld and co-workers
(7) have shown that photochemically produced HN03 had a major effect on
atmospheric corrosion in the Greater Los Angeles Basin, while the
 573

concentration of S02 was so low due to strict emission controls that S02
played only a minor role (see 5.0).

2.2.4. Chlorides: In marine environments cl- deposition strongly

decreases with increasing distance from the shore. In urban and
industrial areas the deposition rates of Cl- are usually much less than
those of S02. Only very few studies of atmospheric corrosion in these
areas have reported a significant effect of chlorides.

2.2.5. Temperature: Temperature may influence corrosion rates greatly

under certain conditions, but for long-term exposure in the temperate
climatic zone it is not of decisive importance.

3.0 Damage Functions and Classification of Environments

The correlation between atmospheric corrosion rates and environmental

parameters has been studied in many field exposure tests. More recently
statistical analyses have been performed for these data in order to
obtain mathematical expressions which give the dependence of the
corrosion rate of a material as a function of environmental parameters
such as RH, concentration of pollutants and two These so-called damage
functions allow an engineering approach to atmospheric corrosion with the
possibility to use these data for design and maintenance purposes. It
will also be possible to use these damage functions to estimate the
monetary losses to the society due to pollution and to decide how much
reduction of pollution needs to be achieved in order to reduce these
losses.

Kucera and Mattson (1) have summarized the damage functions developed so
far for carbon steel and zinc. All equations show a detrimental effect
of S02. Dif ferences in the metereological parameters at different
locations in the temperate climatic zone are usually small. Therefore
for technical purposes, simplified expressions can be used which take
into account only the S02-concentration. Atteraas et al (1) and Knotkova
et al (1) have used equations in the form of:

(2 )

where [SoLI is the S02 concentration (e.g. ug/m3) or deposition rate

(e.g. mg/mL • day) and K is the corrosion loss (e.g. g/m2) in a given time
such as one year. The values of a and b are determined by statistical
analysis of weight loss data for a given test site.

Damage functions for zinc are of a very similar form. The equations
compiled by Kucera and Mattson (1) correlate the corrosion rate with the
S02 content of the air and in one case with the chloride concentration.
In most cases the corrosion loss is a linear function of the S02 content.
In some equations tw has been included as an explicit parameter. Often
the time for which RH C! 80% and T ~ OOC is used for two As will be
discussed below, recently developed damage functions for Southern
California do not contain the S02 concentration.
574

Dean (9) has described the approach taken by the International Standards
Organization (ISO) for the classification of atmospheres which makes use
of some of the concepts described here. Only S02 and CI- are considered
as the species producing atmospheric corrosion. The time-of-wetness is
calculated as shown above or measured with electrochemical sensors as for
example described in ASTM G 84. The atmospheric corrosivity of a test
site is determined by a classification system based on the S02 and CI-
concentrations and the value of two

An alternative approach considered by ISO for estimating the atmospheric

corrosivity at a given test site is based on the mass losses of four
different metals - steel, zinc, aluminum and copper - which have been
exposed for one year. Based on the corrosivity class for a given test
area long-term ("guiding") values of corrosion rates for many similar
engineering alloys have been compiled in the ISO document DP 9224. These
long-term corrosion rates can be used to determine the need for
protective coatings such as zinc or polymer coatings and determine
corrosion allowances for bare metals.

4.0 Electrochemical Sensors for the Determination of the Time-of-Wetness

and Atmospheric corrosion Rates

Recent reviews (4,10) describe the use of electrochemical corrosion

probes for the study of atmospheric corrosion phenomena with emphasis on
the determination of the time-of-wetness and corrosion rates in field
exposure. The atmospheric corrosion rate monitor (ACRM) developed by
Mansfeld and co-workers (11,12) uses a zinc or nickel sensor. The
polarization resistance Rp is determined from the current flow as a
result of the application of a potential difference of ± 30 mV between
the "fingers" of the sensor. The data are collected on magnetic tape by
a data logger (ACRMDL) as described in more detail elsewhere (10 - 12).
One of the advantages of the ACRM/ACRMDL system is the simultaneous
determination of Rp-values from up to six sensors at intervals of 10 min.
which makes it possible to follow fast changes in corrosion rates for
sensors of different materials, orientation, etc. during episodes of acid
rain or other rapid changes in environmental conditions. In addition to
the ACRM-data the data for a RH-temperature probe are also collected by
the ACRMDL. Examples for results obtained with this system will be given
in the following discussion of the recent exposure program in Southern
California.

5.0 Atmospheric COrrosion Behavior in Southern california

A two-year exposure program has been conducted in the Greater Los Angeles
basin with the aim of determining damage functions for materials of
economic significance (7,8). ~he three test sites - Burbank, Long Beach
and Upland were selected based on their known pollutant
characteristics. A relatively clean site in Northern California -
Salinas - served as background site giving data which were considered to
be due to natural weathering. The materials studied were galvanized
steel, nickel, aluminum, two flat latex house paints and textile (nylon
fiber) .
 575

5.1 Corrosion Damage Data

Fig. 1-4 show weight loss data for galvanized steel, nickel and aluminum
and the reduction in breaking strength for nylon fiber for exposure
periods between 400 and 500 days at the four test sites. For all
materials it was found that corrosion da~ge is less in the winter than
in the summer which is contrary to results obtained in Europe and the
Eastern part of the United states. At the three Southern California
sites corrosion virtually stopped during the winter months not only for
the metals, but also for the nylon fiber and the latex paint containing
some carbonate filler. Another surprising finding is that corrosion
rates were very low even in the summHr. For galvanized steel corrosion
rates were less than 0.5 11m/year which - based on the ASTM classification
(1) - would characterize these sites as clean, rural areas.

r'------------------1986 -------------------r.------------1987------------~,
MAM ASONDJ FMA M A
"' Du ank

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

V
'DO --~ . .

./
90 -
V 5 !ina.
Upt."d

7
'0

~
L ~8 Dcac

/' /
V
./
V
70

0;
E- 60 ,- v'
R- L -rr
./
/"
/
/'
7
V
~ - - .Lk:::: /"
CI)
CI)

9 50 1 - -- -- : - - - -- V
/'

f--
:r:
(!I
w
I~ /' l---w
./

h / V
.0
;;
30 •
V
20 1- ~~ .'
--..
/~ ~
0.5, licrom lers/ye lr
~/
10
~p
i.....e:
30 60 90 120 150 180 :210 2~O 210 300 330 360 390 0420 0450 .. SO SID

EXPOSURE PERIOD (Days I Months)

FIGURE 1. WEIGHT LOSS DA'rA FOR GALVANIZED STEEL AT THE

FOUR TEST SITES.
 576

rl----------------1986----------------"rl----------1987----------~1
140 M A MAS 0 N D J F M A M A

b".
120 urban.

100 .&
k, Beach

V
-----
Oi
S 80 V kJpland
P
Cfl
Cfl i--s-- V
0
..J
I- 60
V V ~ I---
L-----
/'
",/ /" V
J: I-"

---
(!)
~

r::
ill
;:

,
40
/"
V ..
-
Salina
L-- i--
,~'

VV :-- f-- .........

20
:;- ...
~ ---- t.--I - -f---
~
.'
~ .... 0.5 mj romel rs/year
o
o 30 60 90 120 150 IBO 210 240 270 300 330 360 390 .420 450 480 510

EXPOSURE PERIOD (Days I Months)

FIGURE 2. WEIGHT LOSS DATA FOR NICKEL AT THE FOUR TEST SITES.

1 - - - - - - 19B7 - - - - - - - \ - - - - - - - - 19BB -------------1

J A S o N D J F M A M J A
100
LON >
1
BEACH
.,---L-
90 - - - ----
1 -----

bL
BO ---- --- --- ---

70 ----- - - - - - - - _.---- - - - -----

0;
- - - ----- - - - - - - -----
/
g 60 -----
UJ
UJ SA INAS
0 50 --- --- ---

~ ~
.J
El
l-
I
l'.l 40 - - -t--- ---- BURBANK

ill
lL=
- ---- -/
3:
V"
30

~
~
I-
_... ~ UPLAND

£
20 -----
iF ~
1--
~
'::,---
10

0 ~
o 40 BO 120 160 200 2~0 260 320 360 400

EXPOSURE PERIOD (Days I Monlhs)

FIGURE 3. WEIGHT LOSS DATA FOR ALUMINUM AT THE FOUR TEST SITES.
 577

19B7 ---------r-----------19BB
J A S 0 N o J F M A M J J A
100
95

;.? 90
~
85
~ ,ong Beach
CJ 80 - - - - .__._---- ----- ---_. ------ ---- ...._-
Z
Z
W 75
a: Burbank
I- 70 - - - - - - - - - '---- ----- - - - - - - - - ~
V /
U)
65 Upland
CJ
Z 60
S2
V
«
~
55
W
a:
CD
50 - - ----. ---- ~ ..---
-d' ----
V~ ~ ~
45
~
40

~
~--
U)
U) 35
Salina!
0
---- --- ---- - - - - - - ~- -----
~
...J 30

V
I- 25
Z
W 20

/
--- ....-----
U
a: 15
W
n. 10

0 l /~------
~ ~ 00 00 100 IN I~ 100 100 = ~ m ~ ~ _ ~ ~ ~ ~ _
EXPOSURE PERIOD (Days I Months)

FIGURE 4. REDUCTION OF BREAKING STRENGTH FOR NYLON FABRIC

AT THE FOUR TEST SITES.

5.2 ACRM Results

The unusual time dependence of corrosion rates has been confirmed by

measurements with zinc and nickel ACRMs which had been exposed facing the
sun or facing the sky. Fig. 5 shows the cumulative corrosion loss
obtained for a nickel ACRM facing the sky at the Long Beach test site.
The daily corrosion loss was determined by integration of the 1/Rp - time
curve. The resulting integrated corrosion loss INT (in s/ohm) was then
summed up for the exposure period between March 1986 and September 1987.
Fig. 5 shows with much better resolution than the weight loss data in
Fig. 1-3 that the corrosion rate - the slope of the curve in Fig. 5 - is
lower between September and April than in the remainder of the test
period.

The ACRM data have been analyzed with software developed for this project
(8) • The following parameters have been used to characterize the
atmospheric conditions and the ~orrosion behavior at each test site:

t 60 , t70 and t80 - the time per day for Which RH exceeds 60%, 70% or 80%

tcorr - the time for which Rp is less than 10 7 ohm.cm 2

lNT - the daily integrated corrosion loss.

578

The value of teo corresponds to the ISO defintion of the time-of-wetness

since RH ~ eO% and in most cases T?! OOC. The value of tcorr can be
considered the electrochemical time-of-wetness since it corresponds to
the time for which corrosion rates exceed a very small threshold value.
INT is a qualitative measure of the daily corrosion loss. Due to the
effects of the ohmic drop, current distribution, an unknown area at which
corrosion occurs and uncertainities in converting the measured Rp-value
into a corrosion rate, INT is used here only as an indication of daily or
seasonal trends in corrosion behavior.

100

..
~
~
"
~ 10

'"9'" ,.
z
0
ii'j
0 5.
a:
a:
0
u
w
>
••
;::
:5 ,.
OJ
:.
OJ
U z.

I.

-~
".r-M Jul-1I6 O~t-M 11'0..,·87

EXPOSURE PERIOD (MONTH - YR)

FIGURE 5. CUMULATIVE CORROSION LOSS FOR NICKEL ACRM FACING THE SKY
AT LONG BEACH.

Fig. 6 shows as an example for the data obtained for the ACRMs at the
Southern California sites the monthly average values of t70' t80' tcorr
and INT for the zinc and nickel ACRMs facing the sky and facing the
ground at the test site in Burbank. No significant seasonal effects were
observed for t70' teo or t corr ' nor did there seem to be any effect due
to the position of the ACRM (up or down) (Fig. 6 a-c). However, a very
strong seasonal effect occurred for the corrosion loss INT especially for
the nickel ACRM with high corrosion rates in the summer and low corrosion
rates in the winter and spring of 1986/7 (Fig. 6 d and e). Corrosion
rates seemed to be unaffected by the position of the ACRM (facing the sky
or the ground).
 579

24
a
20
.,60
D 170

16

12
19S5 1981

Month· Yr

24

b
a Zn up
20
• In down

16

~ 12

12
1886 1967
Month Yr

.4

c
20 D Nlup
• Nldown

1986 1987
Month. Yr

FIGURE 6. MONTHLY AVERAGE VALUES OF t70 AND tao (FIG. 6a),

tcorr (FIG. 6b,c) AND INT (FIG. 6d,e) AT BURBANK
580

Burbank, 4/86-9/87
120
d
Il Zn up
100 • Zn down

eo
E
<:
0
U., 60
S'"
I-
~ 40

20

0
e 12 e
1986 1987
Monlh - Yr

Burbanl<, 4/86-9/87
120

100 Il HI up
• NI down

eo
E
~ 60
ill
.§.
~ 40

20

0
4 e 12 8
1986 1987
MonU,- Yr

FIGURE 6. MONTHLY AVERAGE VALUES OF t 70 AND tao (FIG. 6a),

tcorr (FIG. 6b,c) AND INT (FIG. 6d,e) AT BURBANK
 581

5.3 Pollutant Data

An example for the seasonal variations of the pollutant concentrations is

given in Fig. 7a for the monthly average values of the concentrations of
N02 and 03 at Burbank which were found to be important factors in the
damage functions for galvanized steel, nickel and the house paint
containing carbonate extender (see below). A significant seasonal effect
was observed for 03' but not for N02' At Long Beach a strong seasonal
effect was observed for both pollutants with maxima for 03 coinciding
with minima for N02 (Fig. 7b). These seasonal variations paralell those
found for the weight loss data for galvanized steel and nickel (Fig. 1
and 2 ) and the corresponding ACRM-data (Fig. 6 d and e). At Upland both
03 and N02 showed the same time dependence, while at the background site
in Salinas no significant seasonal variations were detected (8).

5.4 Damage Functions

Damage functions were determined by regression analysis of weight loss

rates (weight loss divided by the number of exposure days) against the
explanatory variables which had been determined by principal component
analysis (PCA) ( 03' N02' t60, 03xt60' N02xt60' 03xN02) (8). It is clear
that 03 and N02 do not themselves cause corrosion damage, but are
surrogates for the components of photochemical smog such as HN03 vapor,
organic acids and acidic particles that presumably are the cause of the
observed corrosion behavior. It will be noted that S02 does not appear
in these damage function. The S02 concentrations at the three sites in
the Greater Los Angeles area were less than 10 ppb and therefore did not
have a significant effect on atmospheric corrosion rates.

The damage functions derived for the three sites in Southern California
during exposure between March 1986 and March 1988 are of the form:

CORR = Ei a·V·
1 1,
(3)

where the explanatory variables Vi are either the average gas

concentrations or the humidity related variable t60' The coefficients
and associated errors for the damage functions for galvanized steel,
nickel and latex paint containing carbonate extender are listed in Table
1. It has to be noted that these functions are only valid for estimating
seasonal or yearly average corrosion damage and that the input gas
concentrations are daily averages for a season or a year. Damage
functions for aluminum and the nylon fabric could not be established
during this project because of insufficient corrosion and aerometric
data. The California Air Resources Board which has funded this study (8)
plans to extend this project to obtain such data over a sufficiently long
test period with sufficient amounts of reliable data.
582
Burbank, 3/86-3/88

80
a
70
[] Ozone

60 • N02

50
:0-
.s
Q.
40

I
'a
0-
30

20

10

0
3 7 11 3 7 11 3
1986 1987 1988

Month· Yr

Long Beach, 3/86-3/88

eo

70 [] Ozone

• N02
60

50
:0-
Q.
.s 40
I!I

I
0-
30

20

10

0
3 7 11 3 7 11 3
1986 1987 1988

Month - Yr

FIGURE 7. MONTHLY AVERAGE VALUES OF TilE 03 AND N02 CONCENTRATIONS

AT BURBANK (7a) AND LONG BEACH (7b) FOR THE TIME BETWEEN
MARCil 1986 AND MARCH 1988.
 583

TABLE 1. COEFFICIENTS FOR DAMAGE FUNCTIONS FOR CURRENT

POLLUTION LEVELS AT SOUTHERN CALIFORNIA SITES (1986-1988)

Zinc Nickel Paint

Constant 10.41 8.79 -5.716

error 2.98 2.24 2.465

03 Coeff. -0.3408 -0.3305

error 0.1025 0.0770

N0 2 Coeff. -0.1536 -0.1244 0.1494

error 0.0519 0.0389 0.0486

T60 Coeff. 19.14

error 4.89

03 x T60 Coeff. 0.1874 0.1184

error 0.0496 0.0372

N0 2 x T60 Coeff. -0.3284

error 0.1034

03 x N0 2 Coeff. 0.004641 0.00507

error 0.001728 0.00130

Standard Error
of Estimate 0.768 0.576 0.579

R-squared 0.560 0.518 0.572

584

References

1. V. Kucera and E. Mattson, "Atmospheric Corrosion", in "Corrosion

Mechanisms", F. Mansfeld, editor, M. Dekker, 1986, p.211

2. E. Mattson, Mater. Perf. 21(7), 9 (1982)

3. F. Mansfeld," New Approaches to Atmospheric Corrosion Research using

Electrochemical Techniques", in "Corrosion Processes", R.N. Parkins,
editor, Appl. Publ., London, p.l, 1982

4. F. Mansfeld, "Electrochemical Methods for Atmospheric Corrosion

Studies", Proc. Symp. "Atmospheric Corrosion", The Electrochemical
Society, p. 139 (1980)

5. F. Mansfeld, ASTM STP 727, 215 (1981)

6. F. Mansfeld and S. Tsai, Corr. Sci. 20, 853 (1980)

7. F. Mansfeld and R. Vijayakumar, Corr. Sci. 28, 939 (1988)

8. R. Vijayakumar, F. Mansfeld and R. Henry, "Investigation of the

Effects of Acid Deposition on Materials", Final Report, Contract
Nos. A4-110-32 and A5-137-32, June 1989

9. S.W. Dean, ASTM Stand. News, May 1987, p.36

10. F. Mansfeld, "Monitoring of Atmospheric Corrosion Phenomena with

Electrochemical Techniques", Proc. "Chemical Sensors", The
Electrochemical Society, Oct. 1987

11. F. Mansfeld, J. Electrochem. Soc. ~, 1354 (1988)

12. F. Mansfeld, S.L. Jeanjaquet, M.W. Kendig and D.K. Roe, Atmosph.
Env. 20, 1179 (1986)
STRESS-CORROSION CRACKING AND CORROSION FATIGUE:
APPLICATION OF ELECTROCHEMICAL TECHNIQUES

R.C. NEWMAN
Corrosion and Protection Centre
UMIST
P.O. Box 88
Manchester M60 IQD
United Kingdom

ABSTRACT: The applications of electrochemical concepts and methods in

stress-corrosion cracking and corrosion fatigue are reviewed. Steady-state and
transient electrochemical techniques are shown to give useful correlations with
cracking behaviour, but their predictive ability remains limited. Measurement and
analysis of spontaneous fluctuations in current or potential is a developing area of
research.

Introduction

Environmental cracking of metals in aqueous solutions has an inherently

electrochemical character, and there is no stress-corrosion cracking (SCC) or
corrosion fatigue (CF) process whose rate cannot be changed by electrochemical
means. In some cases it can even be shown that an anodic reaction is
quantitatively responsible for crack growth, i.e.

v = iA (1 )
zFp

where v is the crack velocity, i the average anodic current density at the crack
tip, A the mean atomic weight of the metal or alloy, and p its density [1]. In
others the anodic process is correlated with crack growth rate, but microscopically
the crack advances by a series of brittle events, each nucleated by a thin anodic
surface film such as a de-alloyed layer [2-6]. Finally, many systems show
cracking due to hydrogen at the crack tip, even though the net current in the
crack may be anodic [7]. In all these cases the crack growth is related to the
transient electrochemical behaviour of the bare metal surface produced by
microscopic plastic strain at the crack tip, a subject which has been
comprehensively reviewed elsewhere [8].

Our discussion will begin with a survey of the steady-state electrochemical

conditions that are known to promote SCC (CF is more ubiquitous). We shall
then consider the application of transient electrochemical techniques, including some
of the pitfalls therein. Finally the measurement of current or potential 'noise'
will be reviewed.

One major technique will not be considered in detail, namely the electrochemical
measurement of hydrogen permeation through metal membranes, and the correlation
of the results with hydrogen-related environmental cracking. The reader is

585
M. G. S. Ferreira and C. A. Melendres (eds.),
Electrochemical and Optical Techniques/or the Study and Monitoring 0/ Metallic Corrosion. 585-610.
© 1991 Kluwer Academic Publishers.
586

referred to recent reviews for information on this topic [e.g. 9].

Steady-state Electrochemical Behaviour and SCC

While steady-state anodic polarization curves cannot be used to predict SCC,

especially as the chemistry within a crack is normally different from that outside
the crack, it is observed that SCC often occurs in particular regions of these
curves. The most obvious correlation, e.g. for aluminium alloys or stainless steels
in chloride solutions, is between SCC and actual or incipient pitting corrosion as
shown in Fig. 1a. For the aluminium alloys, closer examination often shows that
the pitting is intergranular, e.g. because of Cu depletion in precipitate-free zones
(A1-Cu systems [10]) or because of reactive phase precipitation (A1-Li-Cu systems
[11 ]). Other systems, such as low-alloy steels, show correlation of SCC with
active/passive or (rarely in practice) passive/transpassive transitions [13] as shown in
Figs. 1 b-c. In all these cases the walls of the stress-corrosion crack are passive,
and the occurrence of SCC in borderline conditions reflects a relative reduction in
passivation rate in these conditions, compared with potentials well within the
passive region of the polarization curve. This retarded passivation, following film
rupture at the crack tip, permits more dissolution, acidification, hydrogen formation,
etc. to occur than in a rapidly repassivating system, and is not specific to anyone
microscopic cracking mechanism.

The electrochemistry of non-passive SCC systems is complex and confusing,

especially as the microscopic mechanisms of cracking are often under fierce debate.
The only straightforward correlation is between de-alloying and SCC in noble metal
alloys such as Au-Cu (Fig. 1d). In copper alloys, such as Q-brass in ammonia
solutions, poorly protective surface films form as a result of copper
dissolution/reprecipitation, and the alloy surface is close to equilibrium between
copper and cuprous ammonia complexes. Both anodic and cathodic (Cu(II) ~
Cu(I» reactions can be diffusion-limited [14]. In these circumstances the products
of the cathodic reaction affect the anodic reaction, and one cannot gain a simple
kinetic interpretation by studying anodic behaviour in a solution free of cathodic
reactants (0 2 or Cu 2+). Within a crack, the conditions approach the Cu/Cu+
equilibrium more and more closely as the crack length increases (Fig. 2), and it is
the de-alloying of zinc, aluminium, etc. in this local environment that causes the
crack growth [3-6,15]. This is a good example of the need to understand the
local conditions within the crack before one can begin to design an appropriate
transient electrochemical experiment to explore the crack-tip kinetics.

The above comments about the specific role of the cathodic reaction also apply to
SCC of steels, where chemical oxidation of Fe 2+ ions by dissolved oxygen leads to
basic ferric salt films of varying protectiveness, e.g. in phosphate solutions. This
is only relevant to crack nucleation; the crack tip normally grows in an essentially
deoxygenated state.

Transient Electrochemical Behaviour and SCC or CF

This topic has been reviewed elsewhere [8]. Most experimental methods, whether
based on potential stepping or rapid scanning, straining, scratching, fracturing or
laser illumination, have the same aim of exposing fresh metal surface to the
electrolyte at a controlled potential, and measuring the consequent current transient.
This transient, provided it is recorded under conditions relevant to the crack tip,
 587

pitting

passive

E
passive
E
~2 ::;flCC
~1
log log

trans passive

passive E
E

slow dealloyi ng

log log

Figure Typical anodic behaviour and its correlation with SCC:

(a) passive metals or alloys in neutral chloride solution, e.g. AI
alloys or austenitic stainless steels; (b) iron or low-alloy steel in
caustic, nitrate carbonate-bicarbonate, etc., solutions [see Fig. 3
for explanation of numbers]; (c) stainless steels or nickel alloys
in hot caustic solutions; (d) noble-metal alloys (note that a or c
may conceal behaviour of this type).
 588

Free surf ace Crack tip

(og i

Figure 2
Electrochemistry of see for a-brass in ammonia solution, showing
establishment of de-alloying conditions (low net eu dissolution) at
crack tip. Again there is a concealed Similarity to Fig. 1 (d).

Figure 3 Schematic results of creating a bare surface at potentials 1, 2 and

3 in Fig. 1(b). see correlates with an intermediate rate of
repassivation.
 589

may be used to test various models of the cracking process. Intermediate rates
of repassivation are normally correlated with SCC in passive systems, as shown in
Fig. 3. Until recently, it was widely thought that many, or even most,
stress-corrosion cracks grew by an essentially anodic mechanism, slip-dissolution.
That is, the crack velocity v is related to the average anodic current density at the
crack tip, i, via Faraday's laws:

v = iA (1 )
zFp

where A, z and p were defined earlier.

For a repetitive film-rupture process at the crack tip, whereby a film of ductility
(strain to fracture) ff is fractured by a crack-tip strain rate fc'

v = q(T)A with T = ff/Ec (2)

zFpT

where T is the interval between film rupture events and q( T) is the charge density
under the current transient from time 0 to T. The crack-tip strain rate fC is a
function of v, K (the stress intensity factor) and (as v ~ 0) time, thus providing a
potential mechanical-electrochemical model of the entire SCC process (Fig. 4).

Complications in the above approach include the choice of electrolyte for the
measurement (bulk solution or simulated crack-tip), hydrodynamics, IR potential
drops especially in low-conductivity media, choice of alloy where the crack path
may contain segregants or be depleted in major solute elements, and the
occurrence of simultaneous anodic and cathodic reactions on the bare surface.
These will be considered in turn.

CHOICE OF ELECTROLYTE

Historically, most transient electrochemical investigations related to SCC used the

bulk electrolyte. Only a few authors, such as Diegle and Vermilyea [17,18],
studied behaviour in simulated crack environments, in their case a caustic solution
saturated with ferrous (HFeO 2 -) ions (Fig. 5). It is particularly regrettable that
the many investigations of chloride SCC in austenitic stainless steel, even as
recently as 1986 [19], have not tried to incorporate the strong local acidification
(by Cr 3+ hydrolysis) measured as early as 1970 [20,21].

There are basically four types of complication related to crack chemistry:

1. Diffusional effects in passive systems without a significant pH change in the

crack - e.g . in alkaline or buffered solutions with iron alloys, where ferrous
ions are sparingly soluble in equilibrium with oxide.

2. Crack acidification by cation hydrolysis - e.g. for aluminium alloys or stainless

steels in chloride solutions.

3. Active-state systems where the crack-tip potential approaches the metal/metal

ion equilibrium - e.g. copper alloys in ammonia solutions [14]
(Cu/Cu(NH 3) 2+) or steels in NaCl solutions [7] (Fe/Fe 2+).
 590

-1:-
-'max
i ,,
( Crack -'av = q /T.
tip)

Figure 4 Current density-time behaviour at a crack-tip. according to the

slip-dissolution model.

10-1 10
Time from creation of bared surface, S

Figure 5 Effect on transient anodic behaviour of iron of saturating a hot

NaOH solution with magnetite [17]
 591

4. Salt precipitation.

Effect (1) was addressed by Diegle and Vermilyea [17,18], and later Ford [22], by
pre-saturating the NaOH solution with magnetite; this virtually eliminates
dissolution at the bare metal surface, except for supersaturation effects, and gives a
conservatively low anodic reaction rate (Fig. 5). Effect (2) was considered by
Newman et al [23,24] in their work on sensitized stainless steel in thiosulphate
solutions, where it was unusually easy to estimate the crack-tip pH, since
thiosulphate decomposes and buffers the pH at around 3; the results are shown in
Fig. 6. Those wishing to study aluminium alloys using this approach are
referred to Edwards [25], who gives a particularly clear explanation of the crack
chemistry problem. Effect (3) has been modelled by Bertocci [14], and the
cuprous-rich solutions predicted for the crack tip were used by Cole et al [5] in
their dealloying studies using the scratching electrode technique (Fig. 7). The
Fe/Fe 2+ equilibrium effect on crack chemistry of steels (SCC and CF of
high-strength; CF of low-strength) has been discussed by Turnbull [7]; this is a
particularly complex problem as the cracking is due to hydrogen absorption, but
the net reaction at the crack tip is anodic; also the system is markedly
diffusion-controlled (Fig. 8). In our own laboratory we are attempting to bypass
some of these problems by using an abraded hydrogen permeation electrode.
Finally, effect (4) (salt precipitation limiting the achievable current density) has
been given by Galvele [26] as his reason for abandoning the slip-dissolution model
of SCC, but recent data on Al-Cu alloys [10] suggest that his defection may be
premature as anodic current densities of 0.5 Alcm 2 (equivalent to > 0.1 Jlm/s)
were demonstrated in small cracks (Fig. 9). Models of dissolution and transport
in cracks have difficulty when the crack tip is the dominant source of dissolving
cations, as the width of the dissolving zone is usually unknown; Galvele's simple
calculations may have been pessimistic.

The most satisfactory way to resolve many of the above problems would be to
carry out the bare-surface electrochemical measurement inside a real or simulated
crack. One approach using real cracks would be to apply a step increase, or
pulse, in load and measure the current response of a cracked specimen; this has
been tried for the sensitized stainless steellthiosulphate system [24], but led to large
and unknown brittle increments in crack length on application of the load pulses,
making the results difficult to interpret. In any case, this method is only likely
to be useful if the crack walls are completely passive, so that their depolarization
by the solution motion (due to opening of the crack) is minimal. Another
problem is the calculation of the bare surface area responsible for the current
transient, so that a current density can be derived. Similar problems of data
interpretation arise from attempts to measure the coupling between fatigue cycling
and anodic current [27], where an 'admittance' transfer function can be defined
but usually reflects crack wall rather than crack tip behaviour.

So long as one is dealing with a metal that has a reducible oxide, e.g. iron,
successful transient measurements can be made by incorporating electrically isolated
microelectrodes into crevices, or artificial cracks, made with large flat metal
specimens. Zheng [28] studied the anodic behaviour of flush-mounted, 50 JIm or
125 JIm iron and platinum electrodes inside 1 cm 2 iron/resin crevices in anhydrous
methanollammonia solutions. The large iron crevice-forming electrode efficiently
removed oxygen and water contamination from the crevice solution, enabling the
anodic behaviour of the microelectrodes to be defined much more exactly than in
 ~
x
o
..J -9
-400-200 0
POTENTIAL (mV SeE)

Figure 6 Repassivation transients of scratched Fe-9Cr-10Ni alloy (0),

simulating a sensitized grain boundary, in pH 3 sodium
thiosulphate solution, compared with SCC velocities eX) [23].

N
E
u
-.
<
E 2
....>-
'VI
c
c:.l
"0
....C
~
:J
u
..c 0
....IIIu
L-
U
1I)
0 10 20 40
Time, S
Figure 7 Dealloying transients of scratched Cu-Zn alloys in (NH3 + Cu+)
solution at the equilibrium potential of Cu, showing onset of de-
alloying at 20% Zn which correlated with the onset of SCC [5].
 593

Cathodic
excess H production
due to bare surface

Figure 8 The SCC/CF problem for steels in the active state. Even if the
crack-tip environment is simulated successfully, the hydrogen which
causes the cracking is produced by a partial cathodic transient
under a net anodic transient.

-6

--
t il
E
IGSCC
'-"

0:: I G cor rosion

I>
en
..Q
-1

0 10 20
KI (MNm-3/2 )
Figure 9 Potentiostatic v-K curve for underaged AI-Cu alloy in NaCI
solution, showing the effect of small stresses in enhancing the
intergranular corrosion rate [10].
594

I
CREVICE
I 114-,\. C02
I 20mV/s
I
\ -250~mFe
-02 \
\ --- 250~m PI
I
,,'
~ -0·4 ;", "
III

~ -06
w
,
I
I
,J

-0 1-=~:;'''';;-=-~=-
-l ...~~:.:::_

0·1 lO
i (mA/cm 2 )

Figure 10 Anodic behaviour of Fe and Pt microelectrodes within a 180

f.l m - wide steel/resin crevice in methanol/ammonia/CO 2 solution
[28]. Shows the correlation between passivation of Fe and
oxidation of ammonia on Pt.

CE

.RE

bare surface coup'led to

passive surface (measured)
passive 'I
surface

WE passive surface contri bution

when coupled to bare surf ace

Figure 11 Schematic result of the coupling of passive film capacitance to a

bare surface reaction where there is some IR drop: (a) geometry;
(b) effect of passive surface on experimental result.
 595

the bulk solution; in addition the local pH, etc. were allowed to reach realistic
crack values. The IR problems were minimal, despite the modest electrolyte
conductivity, as the radial current flow from a microelectrode incurs only a small
IR drop. The results (Fig. 10) showed a correlation between ammonia oxidation
on the platinum and an abrupt passivation of the iron microelectrode, and
supported a film-induced cleavage model where a thin nitrogen-containing layer
repeatedly nucleates a cleavage crack of ~ pm in length. No fast transient
measurements were reported, but these can readily be done. This approach can
be extended to many other systems, particularly as it is no longer believed that the
first ms of extremely fast anodic reaction are relevant to SCC at normal strain
rates [1,8,16,19]. We are currently working on steel in CO/CO 2 solutions,
studying similarities to the anhydrous ammonia system.

Straining electrode measurements inside artificial crevices with fully-developed

crevice (crack) chemistry are a natural development of Zheng's method, but have
not been successful as yet. There are many problems related to IR drop and
current distribution; also there are some systems (though probably not many)
where the deforming crack tip is the main source of cations.

HYDRODYNAMICS

Some of the methods used to study 'bare surface' electrochemistry use a relatively
quiescent electrolyte, while others (e.g. the scratched rotating disc electrode [8])
may have very rapid mass transfer to or from the electrode. Flow was one of
the main variables used by Hoar and West [29] in their pioneering investigation of
bare-surface electrochemistry in the stainless steellMgCl 2 system; they found that
current densities equivalent to the crack propagation rate could only be sustained at
the cracking potential by rapidly flowing the solution. They should have
concluded from this that there was a problem with what came to be known as the
slip-dissolution model, but in fact proposed that the crack-tip hydrodynamics were
similar to a flOwing solution, a view now rejected by Turnbull [30] even for
low-frequency fatigue cycling.

There is no doubt that stagnant conditions are most appropriate for studying
bare-surface kinetics in the context of crack growth in SCC or CF; however, in
many cases one is also interested in the fundamental processes occurring at very
high current densities. The best advice is to know your system, i.e. to explore a
number of variables including flow rate and (if appropriate) simulation of the crack
conditions. It is also important to realize that the earliest stages of crack (or pit)
nucleation occur with very short effective diffusion lengths, and hence very high
limiting current densities, since diffusion and current flow in the bulk solution are
roughly hemispherical. For such conditions, a flowing system or a microelectrode
may be necessary to reproduce the limiting current densities found in practice.

IR DROPS

The IR-drop problem in bare-surface electrochemistry has been discussed elsewhere

[8]. The basic result is that the IR drop, for a given current density, scales with
the electrode radius (for a disc) or width (for a scratch or other long, narrow
rectangle); thus any technique which generates a very small patch of bare surface
will suffer relatively less from IR errors (see Table 1).
596

Table 1

Solution Resistances and IR drops for Common Electrode Types Used in Transient
Electrochemical Investigations Related to Environmental Cracking (p = resistivity)

R I.R.

Disc with ~1riap

remote counter
and reference
electrodes

Scratch or In(4b/a) 2pa/1r In(4b/a)

other rectangular
electrode

Straining wire plhL In(rla) iap In(r/a)

with radial
current flow L r .RE
2a
 597

A serious consequence of the IR drop has recently been studied [31,32], though
earlier users of the scratching technique were aware of it and knew how to detect
when it was affecting their results [8,33-35]. This is the coupling of the bare
surface to any filmed surface forming part of the working electrode in a
potentiostatic circuit. As one decreases the electrolyte conductivity or increases
the bare surface current density, the resistance for current flow to the bare surface
increases, and the passive surface (especially near the bare surface) experiences a
drop in potential, Le. some of the current from the bare surface flows locally to
the passive surface, discharging the film capacitance, and not to the
counterelectrode. Instantaneously, the effect of this is to reduce the apparent
bare-surface current, but later on, as the bare surface repassivates, some of the
charge lost to the passive surface is won back as the film capacitance is recharged
(Fig. 11). According to Wei et al [32], the effect of increasing the amount of
passive surface (keeping the bare area constant) is to increase the apparent
bare-surface current: assuming that there were no other experimental artifacts in
their experiment, this must be because they were sampling the current so
infrequently that they missed the initial current peak and detected only the
film-recharging phase. They also stated erroneously that the ratio of filmed to
bare surface area would determine the severity of this effect: in fact, since the
solution resistance scales with the bare surface radius, the problem should disappear
at sufficiently small absolute bare surface areas, e.g. in scratch tests. This is
evident in pitting studies on stainless steel, where there is no undershoot of the
current when pits abruptly repassivate (unless this comes from the electronics as in
the work of Frankel et al [36]).

IR drops within cracks are especially troublesome if one wishes to understand the
technologically important SCC problems in high-temperature water. After various
earlier attempts to fit a slip-dissolution model to the SCC of sensitized stainless
steels, Andresen [37] has concluded that the crack-tip potential hardly changes with
external potential (Le. O 2 concentration) and that the effect of oxygen is to
increase the concentration of impurities in the crack - Le. external polarization
changes the crack chemistry, not (appreciably) the crack potential. Large IR
drops in the low alloy steel/high-temperature water system have been measured
experimentally [38]. The same conclusion was reached by Edwards [25], who
applied it to all SCC systems showing crack acidification (aluminium alloys, stainless
steels). In buffered passive systems, such as iron in caustic or
carbonate/bicarbonate solutions, the external potential influences the crack-tip
potential much more strongly, as shown by Parkins and others [39,40]. Note,
however, that small changes in potential in aluminium or stainless steel could cause
large changes in dissolution rate if Tafel's law applies (the anodic Tafel slope for
aluminium has been variously quoted as 20 or even 10 mY).

The latest position of Ford and Andresen amounts to a recommendation that one
should do rapid-straining (or other bare-surface) electrochemistry in various
simulated crack environments at a single potential, in order to interpret the effect
of varying the free-surface potential. This obviously requires testing.

CHOICE OF ALLOY

The use of simulated crack-tip alloys is an obvious approach if one is dealing with
sensitized stainless steels, and the greater activity of Cr-depleted materials is
especially obvious in thiosulphate systems [23,24] see Fig. 12. In
598

N
I
e
0
60
u
e
• S0 (;)

'"
(;)

IX 40 '94
w
I-
u.. .
a: 30 (;)

>-
....I-
(J)
20
z e 184
w e e e X..I
Q 10
X X (;) e e
w X X X
'-'
IX
e ~
a: -400-200 0 200 400
I
u POTENTIAL (mV SeE)

Figure 12 Effect of Cr content on repassivation charge density of Fe-Cr-Ni

alloys in O.SM Na 2 S 2 0 S solution [23].

-
N
E
u
~

>- 0·8
.....
·iii
c
~ .... 0·6

tip 5 content. ppm

Figure 13 Predicted effect of sensitization, crack chemistry and alloy

composition on SCC in oxygenated high-temperature water, using
rapid-straining transient experiments on normal and Cr-depleted
alloys [41].
 599

high-temperature water systems the differences are more subtle (Fig. 13), but still
large enoj.lgh to account for the localization of cracking at the grain boundaries
[41]. With modern analytical transmission electron microscopes, the minimum Cr
concentration at the boundaries can readily be measured; this may range from 7
to 15% depending on carbon content and thermal history. In thiosulphate
solutions, not much impairment of passivity was observed above 11% Cr [23,24].

A much more complex subject is the incorporation of grain boundary segregation

into kinetic models of SCC. It is not obvious that· a 2-dimensional array of
solute or impurity atoms can be simulated by any bulk material, especially as the
oxide film thickness at the crack tip is normally much greater than the width of
the susceptible zone. However, there is plenty of evidence for segregation effects,
ranging from carbon in iron to phosphorus in low-alloy steels [42-44], and one
must start somewhere. Bandyopadhyhay et al [44] showed that an amorphous
Fe-10 wt% P alloy repassivated about five times more slowly than pure iron or
Fe-3% Ni in 9M NaOH at 98C (Fig. 14); this correlated with the high crack
growth rate in a Fe-3Ni-o.06P alloy, but did not explain why a high-purity
Fe-3Ni alloy showed no cracking at all, or why small additions of Mo (0.6%) also
induced cracking. The most likely explanation is that no SCC can occur unless
there is a chemical active path at the grain boundary (and, consequently, that the
slip-dissolution model cannot apply to transgranular SCC where there is no active
path).

The availability of RF magnetron sputtering should enable this problem to be

addressed in more detail in the future, using metastable crystalline solid solutions,
especially as the normal levels of segregants are more like 1-10% than the > 20
at. % used by Bandyopadhyay et al.

SELF-CORROSION OF THE BARE SURFACE

In most passivating systems, such as steel in NaOH solution, bare surfaces behave
as essentially pure anodes at the potentials relevant to SCC. In aluminium alloys,
or in stainless steels, there will often be a minor hydrogen evolution component
which has not been measured (this is not to say that this is an unimportant
problem, especially in aluminium alloys).

A more subtle situation, which has generated many misconceptions, is the cracking
of low-alloy steels (SCC or CF in high-strength; CF in low-strength) in the
active state, e.g. in natural waters. If one simulates the crack chemistry using
the measurements and modelling of Turnbull and others [7,30], then carries out a
potentiostatic scratching or rapid fracture experiment at the free-corrosion potential,
in deaerated solution, there is invariably a net anodic current transient on the bare
surface. This conceals the partial cathodic current, which mayor may not show
a transient increase on exposure of the bare surface (Fig. 8). Since the cracking
is caused by hydrogen, i.e. the product of this cathodic reaction, one must try to
extract the time variation of the partial cathodic current, or preferably the
hydrogen absorption rate (since this actually causes the cracking). We are
currently trying to do this using a mechanically polished hydrogen permeation
electrode. In view of these complexities, it is surprising to find Alavi et al. [45]
measuring the net anodic current transient, then equating it to a cathodic current
transient and arguing that this is the amount of hydrogen available for crack
advance. This appears to be incorrect, as pointed out several years ago [8].
600

C'I <
E ;J
u
3
VI

....
--
0"

....

Timq, S

Figure 14 Repassivation transients of Fe-3Ni alloy in hot NaOH solution,

compared with Fe-3Ni-2Mo alloy and amorphous Fe-lOP alloy
[44).

o C steel .n NO)
A C sted .n OW
o Fernt.c N.-st~el .n M9CI2
• C steel .n COf IHCO)
ImI Type 304 .n M9CI2
• C steel .n CO/C02' H20
.& AI-7 i00i9 in NaCI
E
)C Brass.n N H4
,5.1(f4

Current Oens.ty on Bare Surface

Figure 15 Faraday's law plot of maximum see velocity versus maximum

bare surface current density measured with various techniques [1].
 601

Correlation of Anodic Kinetics with sec Velocity: the Current Position

Having reviewed the use of bare-surface electrochemistry in some detail, we can

now review the status of the slip-dissolution model, i.e. the ability of transient
anodic currents to explain crack growth rates quantitatively.

In the 1970's a diagram emerged, and has often been reproduced, showing a good
correlation between maximum sec velocity for a range of steel, stainless steel,
aluminium alloy and copper alloy systems (Fig. 15) [1]. Hidden in this figure
was a very wide range of repassivation rates: some of the current densities were
sustained for a few ms, while others were sustained for hundreds of ms, or even
seconds. Since the strain rate at the tip of a stress-corrosion crack is not
extraordinarily high, the very short-lived transients must be rejected as sources of
sustained crack growth: they would require crack-tip strain rates on the order of
100%/s. The longer-lived transients, especially in ferritic steels exposed to
nitrate, carbonate/bicarbonate or caustic environments, remain convincingly
correlated with crack velocity. Once again we draw attention to the recent work
of Rota and Bohni [10] showing that crack opening under stress reduces the ohmic
polarization and readily permits dissolution rates on the order of 1 Alcm 2,
disposing of the difficulties which led Galvele [26] to abandon the slip-dissolution
model (Fig. 9).

A detailed study of the relationship between anodic kinetics and sec velocity has
been done for the system a-brass/NaNO 2 solution [46]. Among other things, it
was shown that the rapid-straining method of Vermilyea [17] leads to an
underestimate of the area of bare surface, owing to film spalling (as shown by
Bubar and Vermilyea in 1966 for thick alumina films [47]); thus scratching was
the principal technique used. The results (admittedly neglecting possible changes
in crack chemistry) show that the velocity predicted from equation 2 parallels the
sec velocity, but is consistently lower by 1-2 orders of magnitude, depending on
the value of T. If the results are processed so as to test a film-induced cleavage
model with a critical anodic charge density (= film thickness?; 0' film thickness?)
and a 1.5 pm crack jump distance, then the log (velocity)- potential line from
equation 2 is parallel to but 1.5 orders of magnitude lower than the actual log
(velocity)-potential line (Fig. 16). This shows that the film-induced cleavage
model is not kinetically distinct from the slip-dissolution model, but also relies on
a fast (but less fast) anodic reaction to produce the film which nucleates cleavage.

At this point it is appropriate to mention that the film-induced cleavage model has
been (virtually) proved correct for the brass/ammonia system, using thin foils which
were fractured completely by a single cleavage event [6].

Spontaneous Electrochemical Transients and 'Noise' during sec

It was found in the earliest electrochemical studies of sec, by Hines and Hoar
[48], that considerable current flow accompanied cracking (in their case this was
deduced indirectly from a depression in the corrosion potential, since they had no
potentiostat). The significance of rapid current or potential transients, in that
they might indicate brittle crack advance, was realized by Hoar and later by
Baumel and Engell, who unfortunately mistook discontinuous yielding (Luders band
formation) for discontinuous cracking in the mild steel/nitrate system [see ref. 49
602

Act~t sec velocities

0 Allllil.rcz 2 1 .I 10-4 ,,,

••
Vu 6-6/( lO"'"'Js

! 1 Vu '.8.1 10-6'5 0

.
0
E
• 0

••
:l.

...
>'
u ~. :
9UJ 0 0
> • 8
"
u
«
01
8 []

••
cz: 0
u
-'
«
...u •
::J

«
• 0
0
cz:
00·01 [] 8
0
...u
UJ 0

Ci
UJ 0
g:
B a Predicted crack velocity
USI"9 .... t =(Cl 1 ''t)(A/IF"p)

0001 •• with vaflAb6c 1:

-100 0 100 200

POTENTIAL, mV (SeE)

Figure 16 Relationship between bare surface reaction and sec velocity for
a-brass in 1M NaN0 2 solution [46). Assuming that the crack
grows by film-induced cleavage, with a 1.5JLm crack jump distance
(0), equation 2 has been used with a variable T (= Mv). The
parallel lines are consistent with a critical charge density required
for cleavage nucleation.

i~11 IIIII~
<[

13'7~
II1IIII111111

,.:3.5
~ 3,3
~ 3.1 L_--:."...._-;!-~--f;o;--+.;---.;;:,:---tn
u

Figure 17 Correlation of electrochemical and acoustic events for a growing

transgranular stress-corrosion crack in large-grained a-brass in
1M NaN0 2 solution [2,55).
 603

for discussion of this subject]. The earliest reliable reports of transients due to
crack advance were in underaged A1-Zn-Mg alloys, where brittle crack increments
of several tens of /tm cause simultaneous mechanical and potential transients [50].
The mechanical transients have been reproduced by Baumgartner and Kaesche [51],
who do not discuss any electrochemical transients. More recently this system has
ben studied potentiostatically, and the current transients correlated with acoustic
events [52].

Following the successful use of acoustic emission to monitor discontinuous SCC in

a-brass monocrystals [53,54], it was apparent that a correlation of acoustic and
electrochemical events would confirm and extend these findings. This could not
be done (then) in ammonia solutions, as the expected current transients were small
(~ nA) and hard to detect above a fluctuating active-state (low impedance)
background. However, sodium nitrite solution proved to be an ideal medium for
this experiment, and successful correlations were obtained [2,55] as shown in Figs.
17 -18. The advantage of the nitrite solution was the higher anodic overpotential
(Le. transient current density on the new surface) and the more passive baseline
condition of the alloy.

Single-crack studies on monocrystals or large-grained polycrystals do not require

sophisticated analysis to extract information on the microscopic crack growth
behaviour. Different problems arise in commercial materials where the crack
front is highly fragmented by the grain structure, and more than one crack may be
present. Here the best approach is to carry out a statistical or spectral analysis
of the current or potential noise, with appropriate unstressed control specimens to
provide a baseline. The resulting noise spectra may show features due to the
presence of cracks, although a careful recent study has shown no change for a
high-strength steel in H 2S solution [56]. Often a simple plot of standard
deviation versus time is more interpretable than the full noise spectra. Noise
enhancement by cracks could comprise lowish-frequency features due to reinitiation
and deceleration of segments along a complex front, or distinct higher-frequency
features due to discontinuous brittle crack advance [57]. Gas bubble evolution
may also produce noise. A co-ordinated study needs to be done, e.g. on brass
in nitrite solution, using specimens of varying grain size.

The most ambitious 'noise' study of SCC is a recent program at Harwell laboratory
[58,59], in which sensitized stainless steels were tested in a dilute thiosulphate
solution, at room temperature, and in oxygenated high-temperature water of similar
conductivity. Current fluctuations were measured with a 2-electrode ZRA
technique where the second electrode was a large piece of annealed stainless steel
placed very close to the tensile specimen. Crack initiation in the thiosulphate
solution occurred at many sites just after yielding, in a slow strain rate test, and
was indicated by numerous current transients whose rise time was ~ seconds.
These corresponded to small cracks compriSing " 1 grain boundary facet. The
initiation rate and noise level then fell dramatically, but increased again as stable
cracking began at very large strains by coalescence of small cracks (Fig. 20). In
the high-temperature water, cracking was accompanied by transients with a much
shorter rise time, possibly due to fracture of whole grain boundary facets (Fig. 21).
In this case it is not clear whether the noise is due to the elementary crack
advance mechanism, or to an incidental, rare process accompanying 'quiet' crack
growth.
604

~ -I~mv'

~ -100m'" '

~
lOLA T:
1 -50m\l

~W~\ ~
20 40 60
OmIJ •

TIME hI

Figure 18 Variation of anodic current transients, due to discontinuous crack

advance, with potential applied to a-brass in 1M NaND 2 solution
[2,55].

AlZnMg, 3·5~NaCl

1·4

> 1-2
E
i0
I'
\/429MPa
~06
~ ~1I./
U1 0.4f\
_./ '. .-.
No stress
x....->t"_.-:....... ..........-,
02 -¥......
.....:::
/e ___ •
........

o 40 ~ 120 160 200 240 280

Time (min)

Figure 19 Current noise, and the effect of stress, for a high-strength

aluminium alloy [57].
 605

«
c

Time from yield, hours

Figure 20 Current-time ZRA record for a slow strain rate test of sensitized
304 stainless steel coupled to annealed stainless steel in aerated,
dilute thiosulphate solution [58].

5
(a)
4
«
::L
_,3
c
......
(01

~
u

675 690 705 720 735 7SO 765

Time from yield, mins

Figure 21 As Figure 20, for oxygenated 290 0 C water [59].

606

Future Priorities and Possibilities

Some of these have been mentioned already, but are gathered here for
convenience.

1. The mechanically polished hydrogen permeation electrode technique should be

developed for application to steels and aluminium alloys.

2. Electrochemical and acoustic noise should be correlated in a controlled series

of tests in well-behaved systems known to crack by various mechanisms - e.g.
a-brass in nitrite; low-strength ferritic steel in caustic; martensitic stainless
steel in chloride. Grain sizes should be varied from ~ mm to commercial
values.

3. The technique of embedding isolated, independently polarizable microelectrodes

in crevices should be used for ferritic steels, copper alloys and othLr
cathodically reducible systems, so as to carry out potential stepping
experiments in realistic crack/crevice environments.

4. The limitations of bare surface electrochemical techniques claimed by Wei et

al. [32] should be explored further as a function of absolute electrode size,
solution conductivity, etc. This should show that small working (bare)
electrodes minimize these problems.

5. Magnetron sputtering can now be used to prepare substantial electrodes of

metastable solid solutions to simulate grain boundaries with segregation.

References

1. Parkins, R.N. (1980), 'Predictive approaches to stress corrosion cracking

failure', Corros. Sci. 20, 147-166.

2. Sieradzki, K. and Newman, R.C. (1985), 'Brittle behaviour of ductile metals

during stress-corrosion cracking', Phil. Mag. A 51, 95-132.

3. Newman, R.C., Kim, J.S. -and Sieradzki, K (1986), 'Stress-corrosion crack

growth by film-induced cleavage: some kinetic considerations', in R.B. Jones
and W.W. Gerberich (eds.), Modeling Environmental Effects on Crack Growth
Processes, TMS-AlME, Warrendale, PA, pp. 199-208.

4. Sieradzki, K and Newman, R.C. (1987), 'Stress-corrosion cracking', J. Phys.

Chern. Sol. [special issue on fracture] 48, 1101-1113.

5. Sieradzki, K, Kim, J.S., Cole, A.T. and Newman, R.C. (1987), 'The
relationship between de-alloying and transgranular stress-corrosion cracking of
Cu-Zn and Cu-AI alloys', J. Electrochem. Soc. 134, 1635-1639.

6. Newman, R.C., Shahrabi, T. and Sieradzki, K. (1989), 'Film-induced cleavage

of alpha-brass', Scr. Metall. 23, 71-74.
 607

7. Turnbull, A, Saenz de Santa Maria, M. and Thomas, N.D. (1988),

'Steady-state electrochemical kinetics of structural steel in simulated fatigue
crack-tip environments', Corros. Sci. 28, 1029-1038.

8. Newman, R.C. (1987), 'Measurement and interpretation of electrochemical

kinetics on bare metal surfaces', in A Turnbull (ed.), Corrosion Chemistry
within Pits, Crevices and Cracks, HMSO, London, pp. 317-356.

9. Zakroczymski, T. (1985), 'Entry of hydrogen into iron alloys from the liquid
phase', in R.A Oriani, J.P. Hirth and M. Smialowski (eds.), Hydrogen
Degradation of Ferrous Alloys, Noyes Publications, Park Ridge, NJ, pp.
215-250.

10. Rota, A and Boehni, H. (1989), 'Kinetics of the intergranular stress corrosion
of AI-Cu alloys', in B. Elsener (ed.), Electrochemical Methods in Corrosion
Research III, Trans Tech Publications, Zurich, pp. 177-190.

11. Craig, J.G., Newman, R.C., Jarrett, M.R. and Holroyd, N.J.H. (1987), 'Local
chemistry of stress-corrosion cracking in AI-Li-Cu-Mg alloys', J. de Physique
48, Colloque C3, supplement au No.9, pp. 825-833.

12. Pessall, N. (1980), 'Prediction of stress corrosion cracking in 10% caustic soda
solutions at 3150 C', Corros. Sci. 20, 225-242.

13. Lee, K.H., Cragnolino, G. and Macdonald, D.D. (1985), 'Effect of heat
treatment and applied potential on the caustic stress corrosion cracking of
Inconel 600', Corrosion 41, 540-553.

14. Bertocci, U. (1984), 'Modeling of crack chemistry in the alpha brass-ammonia

system', in R.P. Gangloff (ed.), Embrittlement by the Localized Crack
Environment, TMS-AIME, Philadelphia, pp. 49-58.

15. Newman, R.C. and Shahrabi, T. (1989), 'De-alloying and stress-corrosion

cracking of copper alloys in Cu(I) solutions', ref. 10, pp. 169-176.

16. Ford, F.P. and Emigh, P.W. (1985), 'The prediction of the maximum
corrosion fatigue crack propagation rate in the low alloy steel-deoxygenated
water system at 288 oC', Corros. Sci. 25, 673-692.

17. Diegle, R.B. and Vermilyea, D.A (1975), 'A study of transient corrosion in
the iron-caustic system by means of a drop-weight apparatus', J. Electrochem.
Soc. 122, 180-188.

18. Diegle, R.B. and Vermilyea, D.A (1976), 'Concerning strain-enhanced

corrosion mechanisms of stress-corrosion cracking', Corrosion 32, 26-29.

19. Nakayama, T. and Takano, M. (1986), 'Application of a slip

dissolution-repassivation model for stress corrosion cracking of AISI 304
stainless steel in a boiling 42% MgCl2 solution', Corrosion 42, 10-15.

20. Marek, M. and Hochman, R.F. (1970), 'Stress corrosion cracking of austenitic
stainless steel', Corrosion 26, 5-6.
608

21. Baker, H.R., Bloom, M.C., Bolster, R.N. and Singleterry, C.R. (1970),
Corrosion 26, 420-426.

22. Ford, F.P. (1982), 'Mechanisms of environmental cracking in systems peculiar

to the power generation industry', EPRI Report NP-2589, Electric Power
Research Institute, Palo Alto, California.

23. Newman, RC., Sieradzki, K. and Isaacs, H.S. (1982), 'Stress-corrosion

cracking of sensitized type 304 stainless steel in thiosulfate solutions', Metall.
Trans. A 13A, 2015-2026.

24. Newman, R.C. and Sieradzki, K. (1983), 'Electrochemical aspects of

stress-corrosion cracking of sensitized stainless steels', Corros. Sci. 23,
363-378.

25. Edwards, R.AH. (1989), 'A simple mass transport analysis of localized
corrosion', in H.S. Isaacs (ed.), Advances in Localized Corrosion, NACE,
Houston, in press.

26. Galvele, 1.R (1987), 'A stress corrosion cracking mechanism based on surface
mobility', Corros. Sci. 27, 1-34.

27. Kendig, M.W., Lumsden, 1.B., leanjaquet, S. and Allen, AT. (1989),
'Fatigue admittance of aluminium-lithium alloys in 0.5 molar sodium chloride',
Corrosion/89, paper no. 40, NACE, Houston.

28. Newman, R.C., Zheng, W., Tilley, C.R and Procter, R.P.M. (1989),
'Exploration of a nitrogen-induced cleavage model for the anhydrous ammonia
cracking of steel', Corrosion/89, paper no. 568, NACE, Houston.

29. Hoar, T.P. and West, 1.M. (1962), 'Mechano-chemical anodic dissolution of
austenitic stainless steel in hot chloride solution', Proc. R. Soc. (Lond.) A268,
304-315.

30. Turnbull, A and Ferriss, D.H. (1987), 'Mathematical modelling of the

electrochemistry in corrosion fatigue cracks in steel corroding in marine
environments', Corros. Sci. 27, 1323-1350.

31. Oltra, R. and Indrianjafy, G .M. (1989), 'Effect of the ohmic coupling on
AC and transient measurements in localized corrosion', in B. Elsener (ed.),
Electrochemical Methods in Corrosion Research III, Trans Tech Pubs., Zurich,
in press.

32. Wei, R.P., Gao, M. and Xu, P.Y. (1989), 'Peak bare-surface current densities
over-estimated in straining and scratching electrode experiments' , 1.
Electrochem. Soc., 136, 1835 -1836.

33. Burstein, G.T. and Newman, R.C. (1980), 'Anodic behaviour of scratched
silver electrodes in alkaline solution', Electrochim. Acta 25, 1009-1013.
 609

34. Newman, R.C. and Burstein, G.T. (1981), 'Reactions of scratched copper
electrodes in aqueous solutions', J. Electrochem. Soc. 128, 2270-2276.

35. Pearson, H.J. Burstein, G.T. and Newman, R.C. (1981), 'Resistance to flow
of current to scratched electrodes', J. Electrochem. Soc. 128, 2297-2303.

36. Frankel, G.S., Stockert, L., Hunkeler, F. and Boehni, H. (1987), 'Metastable
pitting of stainless steel', Corrosion 43, 429-436.

37. Andresen, P.L. (1988), 'Modeling of water and material chemistry effects on
crack-tip chemistry and resulting crack growth kinetics', in G.J. Theus and
J.R. Weeks (eds.), 3rd International Symposium on Environmental Degradation
of Materials in Nuclear Power Systems - Water Reactors, TMS-AIME,
Warrendale, Pennsylvania, pp. 301-312.

38. Gabetta, G. and Caretta, E. (1987), 'Corrosion-potential measurements inside

and outside a growing crack during environmental fatigue tests at 288 0 C, with
different oxygen contents', ref. 8, pp. 287-300.

39. Landles, K., Congleton, J. and Parkins, R.N. (1987), 'Potential measurements
along actual and simulated cracks', ref. 8, pp. 453-470.

40. Parkins, R.N. Liu, Y. and Congleton, J., (1988), 'Effects of dynamic straining
in an artificial crack on the stress corrosion of a NCMV steel in 10M
NaOH', Corros. Sci. 28, 259-278.

41. Andresen, P.L. (1988), 'Environmentally assisted crack growth rate response of
nonsensitized AISI 316 grade stainless steels in high temperature water',
Corrosion 44, 450-460.

42. Krautschick, H.J., Grabke, H.J. and Diekmann, W. (1988), 'The effect of
phosphorus on the mechanism of intergranular stress corrosion cracking of mild
steels in nitrate solutions', Corros. Sci. 28, 251-258.

43. Bandyopadhyay, N. and Briant, C.L. (1982), Corrosion 38, 125.

44. Bandyopadhyay, N., Newman, R.C. and Sieradzki, K. (1984), 'Caustic cracking
of a Fe-3Ni alloy: effects of alloying additions or impurities', in Proceedings
of the 9th International Congress on Metallic Corrosion (Toronto, 1984),
National Research Council, Ottawa, vol. 2, pp. 210-215.

45. Alavi, A, Miller, C.D. and Wei, R.P. (1987), 'A technique for measuring the
kinetics of electrochemical reactions with bare metal surfaces', Corrosion 43,
204-207.

46. Cole, AT., Newman, R.C. and Sieradzki, K. (1988), 'A comparison and
evaluation of 'bare surface' electrochemical techniques for the investigation of
stress corrosion cracking in alpha brass', Corros. Sci. 28, 109-118.

47. Bubar, S.F. and Vermilyea, D.A (1966), 'Deformation of anodic oxide
films', J. Electrochem. Soc. 113, 892-895.
6\0

48. Hoar, T.P. and Hines, J.G. (1956), 'The stress-corrosion cracking of
austenitic stainless steels. Part 1. Mechanism of the processs in hot
magnesium chloride solutions', J. Iron Steel Inst. 182, 124-143.

49. Kaesche, H. (1985), Metallic Corrosion, NACE, Houston, pp. 388-469.

50. van Rooyen, D. (1960), Corrosion 16, 421.

51. Baumgartner, M. and Kaesche, H. (1988), 'Intercrystalline corrosion and stress

corrosion cracking of AlZnMg alloys', Corrosion 44, 231-239.

52. Sadeli, Y. (1987), M.Sc thesis, University of Manchester.

53. Beggs, D.V., Hahn, M.T. and Pugh, E.N. (1984), 'Recent observations on the
propagation of stress corrosion cracks and their relevance to proposed
mechanisms of stress corrosion cracking', in R. Gibala and R.F. Hehemann
(eds.), Proceedings of the AR. Troiano Honorary Symposium on Hydrogen
Embrittlement and Stress-Corrosion Cracking, ASM, Metals Park, Ohio,
pp. 181-205.

54. Pugh, E.N. (1985), 'Progress towards understanding the stress corrosion
problem', Corrosion 41, 517-526.

55. Newman, R.C. and Sieradzki, K. (1983), 'Correlation of acoustic and

electrochemical noise in the SCC of alpha brass', Scr. Metall. 17, 621-624.

56. Cottis, R.A and Loto, C.A (1989), 'Electrochemical noise generation during
stress-corrosion cracking of a high strength carbon steel', Corrosion, in press.

57. Loto, C.A and Cottis, R.A. (1989), 'Electrochemical noise generation during
stress corrosion cracking of the high-strength aluminium AA 7075-T6 alloy',
Corrosion 45, 136-141.

58. Stewart, J., Scott, P.M., Williams, D.E. and Wells, D.B. (1987), 'The
development of electrochemical techniques to study the initiation of
intergranular stress corrosion cracking in sensitized stainless steels', Harwell
Laboratory Report no. AERE R12889, United Kingdom Atomic Energy
Authority, Harwell Lab., UK. (also Corrosion, in press).

59. Stewart, J., to be published.

MATERIAL CORROSION IN HIGH TEMPERATURE AQUEOUS
SOLUTIONS IN POWER GENERATING SYSTEMS

D. D. MACOONALD
SRI International
333 Ravenswood Avenue
Menlo Park, CA 94025

ABSTRACT. The corrosion of structural materials in high temperature aqueous

environments is reviewed with reference to problems encountered in the nuclear and
geothermal power industries. Emphasis is placed on identifying the causes and remedial
measures for: (1) Pitting of stainless steels in geothermal brines, (2) Intergranular stress
corrosion cracking of Type 304 SS under simulated Boiling Water Reactor (BWR) heat
transport circuit conditions, and (3) Irradiation assisted stress corrosion cracking of in-core
components in water cooled nuclear reactors. Experimental methods for investigating these
phenomena are described and various experimental results are reviewed to indicate the
mechanisms of the corrosion processes. Particular attention is paid to the use of
electrochemical techniques for studying corrosion processes in high temperature aqueous
systems because they have contributed greatly to the development of remedial measures.
Finally, we comment on the institutional problems facing corrosion science and
engineering, particularly in the nuclear industry where our failure to properly consider the
impact of corrosion phenomena has extracted an enormous price from reactor operators and
consumers and has contributed significantly to the public's mistrust of nuclear power.

1. Introduction
Many electrical power generating systems, including light water nuclear reactors (BWRs,
PWRs, CANDUs, SGHWRs), fossil plants, and geothermal systems employ high
temperature aqueous fluids as heat transport media. The temperature at which many of
these systems operate may exceed 300°C and in the case of some fossil plants the maximum
operating temperature may be well above the critical temperature of pure water (374°C).
Other systems (e.g., direct cycle geothermal plants) employ aqueous geoenergy fluids
which may contain high concentrations of dissolved solids (e.g., NaCl, MgC12, ... ) and
acid gases (CCh, H2S). These fluids, including pure water, are inherently corrosive
towards common structural materials (steels, stainless steels, etc.) so that corrosion-
induced failures are a major cause of system down time and unavailability. Accordingly,
considerable economic losses can be avoided through minimizing the effects of corrosion in
power generating facilities by maintaining tight control over the chemistry of the heat
transport fluid.
In this paper, we describe some of the corrosion problems encountered in geothermal
and nuclear power generating facilities. Our objective is not to exhaustively review the
subject (that would require far more space than is available here) but rather our goal is to
highlight some of the more interesting and controversial issues associated with the
611
M. G. S. Ferreira and C. A. Melendres (eds.),
Electrochemical and Optical Techniquesjor the Study and Monitoring oj Metallic Corrosion. 611-650.
© 1991 Kluwer Academic Publishers.
612

mechanisms of attack and with the remedial measures that have been developed to counter
these costly phenomena.

2• Geothermal Systems
Geothermal brines are well recognized [1] as being corrosive fluids by virtue of the high
temperatures (200-350°C) and high chloride contents (frequently> 105 ppm Cl-), and
because they commonly contain appreciable amounts of acid gases (H2S and C02).
Additional factors that lead to enhanced corrosivity include the presence of scale-forming
components (e.g., silica), which upon precipitation may lead to "under-scale attack" and
other localized forms of corrosion, and the high flow velocities at which geothermal fluids
are frequently produced. These characteristics lead to a variety offorms of corrosion,
including general corrosion (e.g., acid attack), pitting attack, erosion-corrosion, stress
corrosion cracking, and corrosion fatigue.
Because geothermal brines are formed under conditions of high temperature and high
pressure, their properties and compositions in the reservoir differ substantially from those
that exist under ambient conditions. Unfortunately, the fluids are frequently so remote
(e.g., hundreds or even thousands of meters below the surface) that direct chemical
analysis, including the measurement of pH, has proven to be virtually impossible.
Accordingly, a great deal of effort has been expended in recent years to develop chemical
and thermodynamic models for the down-hole environment, so that corrosion and scaling
phenomena may be properly interpreted.
In this section, we discuss the chemical and thermodynamic bases for interpreting
corrosion phenomena in down-hole geothermal environments. Emphasis is placed on
deriving basic thermodynamic data for high temperature/high pressure geothermal systems,
because these can be used to predict the conditions under which various electrochemical
and corrosion phenomena might occur, and for predicting related phenomena such as scale
formation and dissolution. We also discuss the pitting corrosion of Type 304 SS and Alloy
600 under flowing conditions, because this is a major cause of corrosion damage in
geothermal systems. These studies illustrate the various chemical and electrochemical
parameters that determine the susceptibility of engineering alloys to degradation in
geochemical systems. In particular, knowledge of the electrochemical potential and pH,
both measured in the high temperature/high pressure fluid, is essential for the proper
interpretation of corrosion phenomena in geothermal systems.

2.1. THERMODYNAMICS

The thermodynamic properties of chemical systems are of great importance, since they
determine whether or not a given reaction is spontaneous under the prevailing conditions.
The parameter of greatest significance in this regard is the Gibbs energy, calculated for the
conditions of temperature and pressure of interest. Related parameters, such as equilibrium
potentials for charge transfer reactions, partition coefficients for two phase gas/liquid
systems, mineral solubility, etc., are also of great relevance.
Because geothermal systems normally exist under conditions of high temperature and
high pressure, the relevant equation relating the change in Gibbs energy, dG, to changes in
total pressure (dP) and temperature edT) can be written as [2]

dG = VdP - SdT (1)

where V is the volume and S is the entropy, both of which may be functions of T and P.
Integrating equation (1) yields [3]
 613

where Cp is the heat capacity, Po is the reference pressure (usually the vapor pressure of
pure water), and ll.tG~ is the Gibbs energy of fonnation at the reference temperature of To
= 298.15 K. In this eqBation we have chosen to carry out the integration of the SdT tenn at
the reference pressure (Po) and that of the YdP tenn at the system temperature (T). We
could just as easily have chosen to integrate SdT at the system pressure and YdP at the
reference temperature. However, we have chosen the fonner because the YdP tenn is
nonnally less significant than the entropy tenn, and entropy and heat capacities are
commonly reported for the reference pressures.
An important feature of equation (2) is that the quantity G~ is not the normal Gibbs
energy of formation at temperature T, which is based on the convention that the Gibbs
energy of fonnation of an element is zero at all temperatures, but rather corresponds to the
free energy of formation of the species at temperature T from the elements at the reference
temperature (To =298.15 K) and at the reference pressure (Po = 1 atm). Thus, we have
chosen a single standard state at (To, Po), which is consistent with the finite values for S,
Cp, and Y (see equation (2)). We have used this convention extensively for calculating
thennodynarnic parameters for high temperature aqueous solutions in water-cooled nuclear
reactors [3-11], and we have found it to be particularly efficient for computer application.
A number of methods have been developed to integrate the SdT tenn contained in
equation (1), including the "correspondence principle" of Criss and Cobble [12], the
approximate integrand of Macdonald and Butler [3], and the "linear ionic heat capacity
approximation" of Naumov et al. [16], Tremaine [15], and Taylor [14,17]. This latter
approach, which follows directly from Criss and Cobble's fonnalism, is the most useful
for computations, since the partial molal heat capacity of an ion is approximated by

CO=bT (3)
p

where b is a constant. For solids, liquids (excluding solutions), and gases, more precise
heat capacity functions of the fonn

(4)

are commonly available [16, 18-21], which may be inserted directly into equation (2).
Methods for integrating the YdP tenn are equally well-developed through the work of
Ellis and co-workers [22-24] and Millero et al [25-27]. Two problem exist; the first is to
calculate the molal volume of the spe.fies of interest at temperature T, and the second is to
estimate the pressure dependence of yo at this temperature. The first problem has been
addressed by Ellis [22-24] in tenns of a "correspondence principle" for partial molal
volume

- -
YO(T) = aY°(25°C) - b (5)

where a and b are temperature-dependent constants. On the other hand, Millero et al [25-
27] expand the partial molal volume as a Taylor series in temperature
614
- -
Yo(T) = Vo(To) + EO(T - To) (6)
in which only the ftrst order term is retained and where

(7)

Millero [25-27] lists extensive data for partial molal expansibilities, most of which are valid
only at temperatures less than 100°C, in contrast with Ellis' correlations, which extend to
200°C.
With respect to the second problem, the partial molal volume may be expanded as a
Taylor series in pressure

- - -
YO(P) = YO(Po} - KO(P - Po) (8)

where

(9)

The quality j(o is referred to as the isothermal partial molal compressibility, and extensive
tabulations of this parameter for a wide variety of ions are available in the literature [25-27].
Higher deri"yatives of the partial molal volume with respect to temperature and pressure
(dEo/dT, dKo/dP, and cross derivatives) can also be included, but generally reliable values
for these higher derivatives are not available for most ions, and the precision required in the
calculations frequently does not warrant their use.

2.1.1. Potential-pH Diagrams. The initial step in analyzing any chemical or

electrochemical system is to inquire as to which reactions are spontaneous in the
thermodynamic sense. In classical chemical thermodynamics this amounts to determining
whether or not the change in chemical potential or Gibbs energy is negative (spontaneous)
or positive (not spontaneous). In electrochemical thermodynamics, this same requirement
of the Second Law of Thermodynamics is expressed in terms of the potential (E) measured
with respect to some arbitrary reference [28]

(E - Ee}1 > 0 (10)

where Be is the equilibrium potential and I is the current. Thus, for E > Be, I must be
positive and the reaction is spontaneous in the oxidation direction but for E < Be the
reduction reaction is spontaneous. It is evident that equation (10) represents a powerful
basis for determining which reactions in a complex chemical (and corrosion) system might
occur. However, Be is also a function of a number of variables, including pH,
temperature, ion activity, etc. More than thirty years ago, Pourbaix [28] simplifted the
presentation of the vast amount of information contained in the thermodynamics of complex
chemical systems by plotting the equilibrium potential for each reaction as a function of the
pH, whilst holding the activities of all other components constant. This formalism gave
rise to the immensely useful potential-pH or "Pourbaix" diagrams that are now so
extensively used in analyzing the corrosion behavior of metals and alloys in condensed
media.
The computational methods used for deriving potential-pH relationships for metaVwater
systems in geothermal brines at elevated temperatures closely follow those described
 615

previously [3-11, 29-31]. Briefly, the equilibria used to describe the chemistry of a system
in which sulfur does not undergo a change in oxidation state are written in the form [3-10]

(NR)R + (NHzS)HzS + (NH)H+ + (NHz)Hz =

(NP)P + (NCl)CI- + (NHzO)HzO (11)

where the quantities in parentheses refer to the stoichiometric coefficients, R is a general

rea2tant, and P is a general product. 1 In cases that involve the oxidized sulfur species,
4,
S04- and HS0 the equilibria are written in the closely related forms [10]

(NR)R + (NS04)SO~- + (NH)H+ + (NHz)Hz = (NP)P + (NHzO)HzO (12)

(13)

The decomposition of Reaction (11) into the appropriate half-cell reaction gives

= (NP)P + (NCl)CI- + (NH20)HzO (14)

and

2(NHZ)H+ + 2(NHz)e- = (NHz)Hz (15)

Comparable decompositionsf-Te easily performed for Reactions (12) and (13), which
4.
involve the sulfur oxyanions SO4- and HS0 Both half-cell reactions are written in the
reduction sense so as to conform with the Stockholm convention for the analysis of
electrochemical cells. Because the potential of Reaction (14) is referred to the standard
hydrogen scale, the potential-pH relationship is written as

E = C(1) + C(2) log (aR) + C(3) log (aHzS) + C(4) log (ap)

+ C(5) log (aCI-) + C(6) log (aHzO) + C(7)pH (16)

where
C(1) = -l1G~zF;
C(2) = 2.303 RT(NR)/zF;

C(3) = 2.303RT(NHzS)/zF;
C(4) = -2.303RT(NP)/zF;

1 The formalism is easily extended to handle more complex reactions - for example, those that involve
multiple reactants and products.
616

C(5) = -2.303RT(NCI)/zF;

C(6) = -2.303RT(NH20)/zF;

C(7) = -2.303RTx/zF;
x = (NH) + 2(NH2); and

z = 2(NH2).

Likewise, the potential-pH relationships for equilibria involving sulfur oxyanions are
written as

E = C(l) + C '(2) log (aR) + C '(3) 10g(aHSO- or aS0 2-)

4 4

+ C '(4) log (ap) + C '(6) log (aH20) + C '(7) pH (17)

where the coefficients C '(1) to C(7) are defined in a manner analogous to those for Eq
(16).
From Eqs (16) and (17) it is clear that derivation of the potential-pH relationships
requires a knowledge of the change in standard Gibbs energy (~G~) for the cell Reactions
(11) to (13i (see Eq 9) and of the activities of water (H20), Cl-, hydrogen sulfide (H2S),
4,
HS0 SO4-' R, and P. In the work discussed in this paper, the activity of water is
assumed to equal 1 (Raoult's law standard state); the activity of H2S is derived from the
constant total sulfide constraint (see Activitylunctions); aCl- is calculated from the system
4,
composition; and the activities of HS0 SO 4-' R, and P are assigned arbitrary values.

2.1.2. Activity Functions. Because the total sulfide content of a geothermal brine is fixed
by geochemical processes, we considered [10] the conventional method of fixing the
activities of individual sulfide species (H2S, HS-, S2-), irrespective of the pH of the
medium, to be inappropriate. Instead, we believe that a more realistic description of a
geothermal system involves calculation of the activity of dissolved H2S in the system as a
function of pH, and then substituti.vn of this quantity into Eq (16) to derive the potential-pH
relationship for the reaction of interest.
The equilibria used to describe the chemistry of the H2S/H20 system are written as

H2S = H+ + HS- (KI) (18)

HS- :: H+ + S2- (K2) (19)

where Kl and K2 are, respectively, the first and second dissociation constants [9] for H2S.
From the definitions ofK} and K2, we obtain the activity ofH2S in terms of the total
sulfide concentration ([H2S] + [HS-] + [S2-] :: [S]) as follows

where Yl and Y2 are activity coefficients for univalent and divalent ions, respectively.
 617

The substitution ofEq (2) into Eq (16) therefore gives

E = C(1) + C(2) log (aR) + C(4) log (ap) + C(5) log (ac1-)
+ C(6) log (aH20) + C(3) log [S] + [C(7) - 2C(3)] pH

According to Eq (21), the potential varies nonlinearly with the pH, in contrast with the
linear relationship predicted by Eq (16) when constant activities are assumed for all species
involved in the half-cell reactions.
Other activity functions can be defmed for the derivation of potential-pH diagrams for
complex systems. For example, Macdonald and Hyne [9] used a constant H2S pressure
constraint in their thermodynamic analyses of the iron/H2SIH20 system at elevated
temperatures.
The problem of estimating activity coefficients for ions in high salinity brines requires
special attention, but the details of how they are calculated is beyond the scope of this
review. Briefly, activity coefficients are normally estimated using various extended forms
of the Debye-Htikel theory, principally by including higher terms in ionic strength to take
into account and ion-solvent and ion-ion interaction effects [16, 32]. However, these
equations generally are valid only for pure electrolytes (e.g., NaCl), and are not applicable
to the mixed electrolytes (e.g., NaCl + CaCl2 + NaHC03 + Na2C03) found in nature.
Activity coefficients for these latter systems may be estimated to reasonable accuracy using
a number of methods ranging from Harned's rule [32] to the more sophisticated (and
cumbersome) relations derived by Pitzer and coworkers [32].

2.3. POTENTIAL-pH DIAGRAMS FOR GEOTHERMAL SYSTEMS

Potential-pH diagrams for iron and nickel in h1-h-salinity geothermal brine containing a
total of 10 ppm dissolved sulfide (H2S + HS- + S -) are shown in Figures 1 to 4. Also
plotted are the corresponding diagrams for the sulfur/water system, whose equilibrium
relationships are designated by the light broken lines and by numbers ending with the
letters "s" (Is, 2s, etc.). Equilibria between dissolved metal-containing species are
represented by heavy broken lines, whereas those reactions that involve at least one metal-
containing solid phase are represented by solid lines. Lines A and B represent the
thermodynamic equilibrium limits for stability of liquid water. For example, at voltages
above line A, oxygen spontaneously evolves, whereas at voltages more negative than those
given by line B, hydrogen evolutio~yends to occur. All four diagrams were derived for
4
constant activities for HS0 and SO4 - equal to 10-6 (molal scale).2 In the case of the
iron/brine system activities of dissolved metal-containing species were arbitrarily set equal
to lO-4(m), whereas for the nickel, brine system dissolved species activities of 1O-6(m)
were assumed. The higher activities for the iron/brine system were required so as to
exceed the calculated minimum solubilities of the oxide phases at elevated temperatures.
Had the lower activity been assumed, the oxides would not have appeared as stable phases
on the diagrams. Potential-pH diagrams for the iron/sulfur/water system at elevated
temperatures in the absence of chloride ion have been reported elsewhere [9, 33,34], and,
where comparison is possible, the diagrams reported here are in good agreement with those
previously published.

2 Activities are dimensionless quantities, but the concentration scale is included here for reference purposes.
 618

$
FeIOHI'
1.0
t.O
i

"'"
CIl
CIl
~
'"
"'"
'"CIl
:g
~

a: o
'"> w
>
o i"
~
g
..J
o
>
",' w'
..J ..J
< <
;:::
~ ~ -1.0
'"
5 F. I-
... -1.0 ~

FeIOHI·

-2.0 '--_-'---_-'-_-'--'--'-_..LJ.._--'.....l.--.J -2.0 L[_ _~~--..l._.L.:...._-'-----'--"-----'

o 4 6 B 10 12 14 o
RA-320S22-7 RA.-320522--8

FIGURE 1 POTENTIAL-pH DIAGRAM FOR IRON IN HIGH-SALINITY BRINE FIGURE 2 POTENTIAL-pH DIAGRAM FOR IRON IN HIGH-SALINITY BRINE
AT 25"'C IN THE PRESENCE OF 10 ppm TOTAL DISSOLVED SULFIDE AT 250°C IN THE PRESENCE OF 10 ppm TOTAL DISSOl VED SULFIDE
(H 2 S + HS· + 5 2 -) (H 2 S + H5- + 5 2 -)
ActIvities of HSO~ AND SO~-,. 10~ molal. Activities of di::solved iron Activities of HSO; and 50;-" 10~ molal. Activities of dissolved Iron
species" 10-' molal. IS) Indicates 8 soluble molecular species [10J. spec;es" 10-4 molal (5) Indicates a soluble molecular species [10]

II HNiOi!
2.0 SO~-
I),NiIOHI·
Ni 2• II 1.0
I I HNiOi
w
II w
"'"3 II "'"3
'"a:~ $1 -! '"'"a:
LO I Ni0 2 >
~
o
~ 0
o
> >
",' w'
..J ..J
<
;:::
«
;:::
Z 0 Z
'" '"
...5 ~ -1,0
Ni
"-

;-'>-
~

I
I
H5- I 5 2-
LI_--'--_-'-__-'-~-'-_.l....L_--'-__'__'
Ni
-1.0 -2.0 '--_.L--'--'-_--'-L-LL_-'----'---'-_-.!
Q 4 6 B 10 12 14 o 4 6 B 10 12 14

FlA-320522-10

FIGURE 3 POTENTIAL-pH DIAGRAM FOR NICKEL IN HIGH-SALINITY BRINE FIGURE 4 POTENTIAL-pH OIAGRAM FOR NICKEL IN HIGH-SALINITY BRINE
AT 2S"C IN THE PRESENCE OF 10 ppm TOTAL DISSOLVED SULFIDE AT 2SOoC IN THE PRESENCE OF 10 ppm TOTAL DISSOLVED SULFIDE
(H 2S'" HS· ... 5 2 -1 !H 2 S + HS- + 57 -)
Activities of HSO~. SO~-, and dissolved nickel specie~"' 10~ molal Activities of HSO~, SO!-, and dissolved nickel specles:- 10-6 molal
(5) indicates a soluble molecular species [10J. (5) indicates a soluble molecular species [10]
 619

For both iron and nickel at 25°C, the most stable solid corrosion product under highly
reducing conditions is the M(ll) sulfide. Only troilite (stoichiometric FeS) was considered
in the derivation of the diagram for iron, although it is recognized that other sulfide phases,
such as mackinawite (Fel+xS) and pyrrhotite (FeSl+x) exist, and indeed mackinawite may
form at lower potentials than troilite under certain conditions [9]. As the potential is
increased, oxidation of MS to MS2 is predicted to occur over almost the entire pH range.
Note that iron sulfide (FeS2) and nickel disulfide (NiS2) are Fe(ll) and Ni(II) phases,
respectively; the change in oxidation state is associated with the sulfur anions in the lattices.
Thus, FeS2 and NiS2 are best described as iron and nickel disulfides, in which the average
oxidation state of sulfur is -1. Note also that the conversion of MS to MS2 involves
reaction with H2S (or its anions HS- and S2-), as shown by the location of the equilibrium
lines 11 (Figure 1) and 52 (Figures 3 and 4) in the stability region for sulfide (H2S and HS-
) in the potential-pH diagram for the sulfur/water system. On the other hand, equilibria
between MS2 and the oxides Fe304, Fe203, and NiO involve sulfur oxyanions as
demonstrated ~Y the potentials for these processes lying within the stability regions for
HS04, and SO4 .
Iricreasing the temperatures from 25 to 250°C proves to have several important
consequences for the thermodynamic equilibrium behavior ~f iron, nickel, and sulfur in
geothermal brine. Thus, the stable iron (TIl) complex (FeCI) at 25°C is no longer the
predominant species at 250°C; the decrease in the dielectric constant of the medium favors
the formation of the neutral complex FeCI3. Similarly, pKa for the dissociation of bisulfate
ion (HS04,) increases from approximately 1.98 at 25°C to 4.39 at 250°C. Increasing the
temperature also has a marked effect on the stability regions for the dissolved metal-
containing ions in geothermal brine. For example, the stability regions for cations at low
pH values become more restricted as the temperature increases, whereas those for the
anions HFeO; and HNiO; [or, equivalently, Fe(OH); and Ni(OH);] increase. These
predicted changes in the stability domains for dissolved species have been noted previously
[3-10], and, in the case of the anions, the shifts in the lines for equilibrium between the
ions and the oxides (for example, line 32, Figure 3) appear to be greater than can be
accounted for by the change in the dissociation constant of water alone. This observation
suggests that the anions become stabilized in relation to the cations at elevated temperatures,
probably because of their smaller charge-to-radius ratios.
Probably the most imponant change in the diagrams due to increasing temperature
occurs in the relative stabilities of the sulfides and oxides. Thus, in both cases the
disulfides FeS2 and NiS2 are predicted to exhibit domains of stability at 25°C that extend to
very low pH values in the cation stability regions. The lower boundaries of these regions
are determined by equilibrium between MS2 and dissolved M2+ and H~S, whereas the
upper boundaries represent eqUilibrium between MS2 and dissolved M +, at some point the
formation of solid FeS2 (line 37, Figure 1) or NiS2 (line 48, Figure 3) becomes
spontaneous by oxidative deposition. However, if the potential is increased still further,
oxidative dissolution of the disulfides can occur. These relationships are of considerable
importance in the mining and geothermal industries, because they provide the
thermodynamic boundaries for the oxidative and reductive dissolution of pyrite ores, for
example, and pyrite-rich scales in geothermal systems. At elevated temperatures, however,
the stability domains for the sulfides, and panicularly for the disulfides FeS2 and NiS2, are
reduced sharply in size, thereby indicating much more restrictive conditions for the
formation of these phases in high-temperature geothermal systems. Nevenheless, Seward
[34] has argued that the down hold redox potential and pH in the Broadlands geothermal
field south of Reporoa, New Zealand, lie within the stability region for FeS2, and, indeed,
pyrite is observed in core samples from the steam-producing formation.
620

2.3. PITIING CORROSION

Localized corrosion processes, such as pitting attack, stress corrosion cracking, and
corrosion fatigue are most detrimental since component failure frequently occurs with little
loss in weight or little direct evidence of attack. Furthermore, these processes are most
prevalent on passive alloys, which are used for some components in geothermal
environments because of their high general corrosion resistances. Because of the difficulty
in examining corrosion phenomena in situ in industrial high temperature aqueous
environments, considerable emphasis has been placed on laboratory studies to characterize
the mechanisms involved.
In any laboratory corrosion study, it is important to reproduce the in-service conditions
as accurately as possible. In the case of geothermal systems, the conditions that need to be
simulated are temperature, pressure, ionic strength, acid gas (C02, H2S) content, and
hydrodynamic state (laminar vs turbulent, flow velocity). This latter condition seldom is
simulated in the laboratory, because of the difficulty in generating well-defined flow
conditions at elevated temperatures and pressures. However, we have recently reported
[35, 36] pitting corrosion studies on Alloy 600 and Type 304 stainless steel in brines at
temperatures to 275°C and at flow velocities up to 1.5 mls. These flow velocities are
sufficiently high to yield highly turbulent conditions in the annular flow channel employed
in the study. Pitting or "breakdown" potentials were determined potentiodynamically, at a
sufficiently low voltage scan rate that the observed breakdown potential was negligibly
different from the stationary value.
Typical plots of breakdown potential as a function of flow velocity and temperature for
the pitting of Alloy 600 in buffered sodium chloride solution (0.1 M NaCI + 0.001 M
B(OHh + 0.001 M LiOH) are shown in Figure 5. Clearly, flow velocity has little effect on
the breakdown voltage, even at very high flow velocities where the flow is highly
turbulent. While no effect of flow rate is evident, temperature has a very marked influence
on the pitting potential, as shown in Figure 6. Regardless of the flow velocity, increasing
the temperature from 25°C to 250°C shifts the breakdown potential by almost 600 m V in the
negative direction. This finding suggests that Alloy 600 may not be resistant to pitting in
geothermal brines at very high temperatures (> 150°C), particularly if air in-leakage occurs
and the corrosion potential is displaced in the positive direction. Finally, microscopic
examination of the Alloy 600 surface indicates that flow velocity has only a minor influence
on the morphology of attack; possibly because flow modifies only the kinetics of mass
transport of reactants (e.g., Cn or products into and out of the pit.
In contrast to the above, both temperature (Figure 7) and flow velocity (Figure 8) were
found to influence the pitting potential (albeit slightly) and morphology of pits on Type 304
SS in 0.01 M NaCI at temperatures ranging from 25°C to 275°C. As shown in Figure 9,
the pits tend to nucleate in the direction of flow at well-defined spacings. This morphology
probably results from the establishment of a standing pressure wave emanating from an
initial breakdown point, presumably with the down stream pits nucleating at the points of
maximum hydrostatic pressure. If this interpretation is correct then, to our knowledge, this
is the first reported instance of pressure-induced pitting corrosion.
The competing effects of environmental factors (flow velocity) and substrate
microstructure and chemistry on the morphology of pitting attack on Type 304 stainless
steel in high temperature 0.01 M NaCI solution are shown in Figures 10 and 11. In this
case, the steel was sensitized to form Cr23C6 precipitates along the grain boundaries, which
may (apparently) act as nucleation sites for pitting attack. The observed morphology is
therefore expected to display features of the competing effects of nucleation along flow
lines and nucleation around grain boundaries, with the latter dominating at low velocities
and the former at higher velocities. The pitting morphologies shown in Figures 10 and 11
confinn this hypothesis.
As an example of how potential-pH diagrams can be used to interpret localized
corrosion phenomena in geothermal environments, we refer the reader to the work of
 621

700

600

W
:r 500
til

~
2:
500 - ...J
..:
400

.•
0 25"C
;:

..
l00"C zw
400 200"C - t- 300
0
o 250"C
-
Q.

300 z
w ~
:r 0 200
!!l 0
~
200
'"
..:
w
~ a: 100
100 II)

'"
0 0

o 20 40 60 80 100 120 140 160 o 50 100 150 200 250 300

Urn. em/sec TEMPERATURE I"CI
RA.-320S22-11
FlA-320522-12

FIGURE 5 PLOT OF BREAKDOWN POTENTIAL VERSUS FLOW

VELOCITY FOR INCDNEL 600 IN BUFFERED 0.1 M FIGURE 6 PLOT OF BREAKDOWN POTENTIAL VERSUS
N.CI SOLUTION AT 250. 200. 100. AND 2S"C 135J TEMPERATURE FOR INCONEl600 BUFFERED
0.1 M NaCI SOLUTION
Flow velocity· a cm/s {35].

10-1

Type 304 Stainless Steel Solution

Annealed 1 Hr at 105Q°C
Scanning Rate"" 25 mV/min
0.01 M NaCI "V 0 RPM
10- 2
"'E "V 27S"C
.!:! "'E
< A 22S·C
)-' ~
r- IJ 120·C )0'
iii 0 BO·C r-
zw 10'3 ~ 10- 3
C w
t- c
Z r-
w z
a: w
a: II:
...,::J a:
10-4 a 10--4 Type 304 Stainless Steel Solution
Annelaed 1 Hr at 1050°C
Scanning Rate = 25 mV/min
0.01 M NaCI

10-S L-.l-_..ILL_--'_ _ -'-_-L_--'_ _.L...J 10-S L-;!-::-_!,:-_...,J-::_.J_--.JL-_.L._..L.J

-0.6 -0.4 -0.2 o 0.2 0.4 0.6 -0.6 -0.4 -0.2 0 0.2 0.4 0.6
POTENTIAL, VH POTENTIAL. VH
RA-320522-13

FIGURE 8 POLARIZATION CURVES FOR SOLUTION ANNEALED

FIGURE 7 POLARIZATION CURVES FOR SOLUTION ANNEALED
TYPE 304 STAINLESS STEEL IN 0.01 M NaCI SOLUTION
TYPE 304 STAINLESS STEel IN 0.01 M N,CI SOLUTION
AT 27S"C. AS A FUNCTION OF ROTATION VELOCITY
AT TEMPERATURES RANGING FROM ao"c TO 27S"C 136J
OF THE IMPELLER [36J
622

Syrett, Macdonald, and Shih [37] on the pitting resistance of various engineering alloys in
low salinity geothermal brines. For example, comparison of the breakdown (Eb) and
protection (EproJ potentials for Type 316L in this environment with the potential-pH
diagrams for chromium at 25°C (Figure 12) and at 200°C (Figure 13) shows that E!2~d
~jJCot are more negative than the equilibrium lines for the oxidation of Cr20:3 to no4. •
However, the Seco~ Law of Thermodynamics requires that Eb should be more posItive
than the Cr2D3lCJ04- equilibrium potential, so that the oxidative dissolution of chromic
oxide particles in the passive film is unlikely to be the mechanism of pit nucleation. On the
other hand, this condition is satisfied for the oxidation of manganese sulfide, so that the
experimental and thermodynamic data are consistent with the oxidative dissolution of MnS
inclusions as being the electrochemical event that initiates pits on this low carbon stainless
steel. Note that this alloy has sufficiently low carbon that extensive carbide formation
along grain boundaries, as evident in the case of Type 304 SS (Figure 10), does not occur,
even upon severe sensitization.
Considerable experimental work has shown that pitting corrosion (and other forms of
localized attack) in geothermal systems is most effectively avoided by alloying. Thus, the
addition of 2-3 weight percent of molybdenum or tungsten to stainless steels significantly
increases the resistance to localized attack even under acidic conditions at elevated
temperatures. In those instances where stainless steels are prohibitively expensive, (e.g.,
in tubulars) the use oflow alloy steels (such as 2-1/4 Cr-l Mo) affords some protection
against localized corrosion compared with carbon steels (e.g., N-80, J-55), but in the most
aggressive environments severe corrosion (including erosion/corrosion) may render these
materials impractical. In these instances, the economic development of geothermal energy
may not be currently feasible.

3. Water Cooled Nuclear Reactors

The heat transport circuits of typical Boiling Water Reactors (BWRs) and Pressurized
Water Reactors (PWRs) are shown schematically in Figures 14 and 15, respectively.
Briefly, the fluid (water) within the core of a BWR is allowed to boil to produce steam,
which then drives high pressure and low pressure turbines to generate electricity. After
vacuum stripping of non-condensable gases, the condensed steam (water) is returned to the
core. In the case of a PWR, the primary circuit is pressurized to maintain the heat transport
fluid in the liquid state. The heat is exchanged across a steam generator to produce high
quality steam in a secondary circuit, which then drives high pressure and low pressure
turbines, as before.
An important difference between BWRs and PWRs relates to radiolysis in core. In the
case of a boiling water reactor, stripping of noncondensible gases in-core by boiling,
together with radiolysis, gives rise to a stoichiometric excess of oxygen in the recirculation
lines. Under normal operating conditions, the equivalent oxygen level is about 200 ppb,
which has the effect of displacing the redox potential and the corrosion potentials of
structural materials to quite positive values. We show later that this phenomenon is
responsible for one of the major corrosion problems which has plagued BWRs in recent
years. On the other hand, the primary circuit of a PWR contains a high level of hydrogen
(up to 25 cc (STP)lkg H20), which is added to "suppress radiolysis," as shown in Figure
16. However, hydrogen also shifts the redox potential and the corrosion potentials of
structural materials in the negative direction, so that many of the corrosion problems
encountered in BWRs might not be expected to occur in the primary circuit of a PWR.
However, this generalization must be tempered by the fact that in-core temperatures in
PWRs are considerably higher (-340°C) than those in BWRs (typically 290°C at the outlet).
A second important difference between PWRs and BWRs is the steam generator in the
former that is not present in the latter. It is evident that PWR steam generators have been a
major source of trouble in the operation of pressurized water reactors, with some designs
 623

..
FIGURE 9 EFFECT OF FLOW VELOCITY ON PIT MORPHOLOGY FIGURE 10 PIT DISTRIBUTION AND MORPHOLOGY
FOR SOLUTION-ANNEALED TYPE 304 STAINLESS FOR SENSITIZED TYPE 304 STAINLESS
STEEL AT SOoC AS A FUNCTION OF ROTATIONAL STEEL AT 275°C [36 ]
VELOCITY OF THE IMPELLER [36 )
Rotational yelocity .. 0 RPM for both (a) and (b)
(a) 0 RPM ; (b) 606 RPM; (ele63 RPM ; (d) 1494 RPM

FIGURE 11 EFFECT OF FLOW VELOCITY ON PIT MORPHOLOGY

FOR SENSITIZED TYPE 304 STAINLESS STEEL AT
27S"C 1361
Rotationa l velocity: (a) 606 RPM ; (bl 863 RPM;
Ie) 1494 RPM; (d) 1494 RPM.
 624

1.4 1.4

@
I
I
I
1.0 crO~- 1.0
®
CrO~-
0.6 0.6
w
J:
VI
VI
0.2 :J
VI
0.2
a:
w
>
VI -0.2
~
0
>
w' -0.6
...J
«
~
-1.0 zw -1.0
I-
2
-1.4
Cr

1
Cr
-1.8 -l.B

,
-2.2 ' -2.2
Q 2 4 6 8 10 12 14 0 4 6 10 12 14
pH T
RA-320522~18 RA.-J20S22-19

FIGURE t2 POTENTIAL-pH OIAGRAM FOR STABLE Cr SPECIES IN LOW FIGURE 13 POTENTIAL-pH DIAGRAM FOR STABLE Cr SPECIES IN lOW
SALINITY BRINE AT 2SoC SALINITY BRINE AT 2OO'C
Activities of dissol..,ed species'" 10-6 m. Eb and Epro\ values for Activities of dissolved species'" 10-6 m. Ranges of Eb v,lues and EPf'01
Type 316L stain leu steel are shown 85 an open circle and an open values for high chromium ferrous alloys are shown, respectively. as an
tri.ngle, respectively. at pH - 5.4 and II dosed Circle and a clO!ed open circle and an open triangle (with error bars) al pH 2S ..... 5.3, and as
triangle, respecti..,.ely. a1 pH -- 7.2. line 8 is the equilibrium line
for the retc1ion, Mn2.... SO~- + SH+ .. ae- '"'
MnS ... 4H:2 D, with
aMn2.... aSO~- .. 10-6 m; E '" 0.161 - 0.05916 pH [37J
~~~::d i~i;~~e :Q~i\~b~li::dli~~a;::~~:~::::i;~, :~SJ/: ~~~5_ ~H+ + :
ae- =MnS + 4H:zO, With 'Mn 2+ = 'SO;- ~ 10--6 m: E '" 0 116 - 0.0939 pH
1371
 625

RA-320543-28

FIGURE 14 BOILING WATER REACTOR (BWR)

After Roberts (38).

«: .. «: )
primary water loop secondary steam/water
loop
RA-320543-71

FIGURE 15 PRESSURIZED WATER REACTOR (PWR)

After Roberts (38).
626

10-2 . . . - - - - - - - - - - - - - - - - - - - - - - - - .

~
g 10-5
Z
o
~
a:
I-
~ 10- 6
(j '0

~
Z
o
(j
a:
10-7 I-

10-8 r

10-9 L..-_ _ _ _0..L"1_pp_m-'--_ _ _ _ _1L...P_pm_.1..-_ _L-_1..J.O_p_p_m-l

10-5 10-4 10-3 mollkg
INITIAL H2-CONCENTRATION

RA-320543-29

FIGURE 16 STEADY STATE CONCENTRATIONS OF SEVERAL RADIOLYSIS PRODUCTS

IN PWRs VERSUS THE INITIAL HYDROGEN CONCENTRATION
 627

being considerably more troublesome than others. Indeed, recent analyses [39] indicate
that steam generator corrosion phenomena are a major cause of unscheduled down time.
Finally,low pressure turbines have proven to be a major source of concern ever since
the catastrophic failure at Hinkley Point "A" in the United Kingdom in 1969 [40]. The
exact cause of failure is still being debated, and it is clear that no simple solution exists to
this problem. Rather, the solutions(s) will very likely involve simultaneous changes in
design, metallurgy, and system chemistry, all of which require continued in-depth scientific
and engineering study before full implementation.

3.1. IGSCC OF BWR PIPING

Intergranular stress corrosion cracking of recirculation piping in boiling water reactors

has proven to be a major operating problem in BWRs world wide [41-45]. Cracking
occurs in the heat affected zones adjacent to welds in Type 304 SS pipes (Figures 17-19).
Furthennore, the incidence of cracking has not only increased since it was fIrst reported in
the early 1970s, but it has begun to appear in large diameter pipes (Figure 20). This trend
is of concern because of the diffIculty in repairing or replacing large diameter pipes in the
fIeld. However, it is important to emphasize that IGSCC of BWR recirculation piping is a
"leak-before-break" phenomenon that apparently presents little risk of a catastrophic pipe
failure, and hence of a "loss of coolant" accident. Nevertheless, leaks force shutdowns,
which represent signifIcant economic losses for the operators (in excess of $2.4 million per
day for a 1000 MWe plant at a unit cost of 1O¢ per kW. hr).

3.1.1. IGSCC o/Type 304 SS. Extensive work [46-51] on the intergranular stress
corrosion cracking (lGSCC) of Type 304 SS and related alloys in high temperature
aqueous solutions has established that the cause is sensitization due to the precipitation of
chromium carbides along the grain boundaries. The adjacent matrix is depleted of
chromium to the extent that corrosion occurs preferentially along the boundaries, resulting
in intergranular penetration. These same studies have shown that the fracture process is
greatly affected by electrochemical, environmental, and metallurgical parameters, and some
of these effects fonn the bases for countenneasures. In this section, we review various
fundamental studies of IGSCC in sensitized Type 304 SS and show how this work has led
to a phenomenological understanding of the fracture process. We also briefly review some
of the countenneasures that are currently being developed to prevent cracking of
recirculation pipes in operating BWRs.
Figure 21 shows the effect of potential on the SCC susceptibility of sensitized Type
304 SS (in this case, heat-treated at 650°C for 1 hour after solution-annealing at 1050°C) in
deaerated 0.01 m Na2S04 solution at 250°C [49]. A drastic decrease in the elongation-to-
failure ratio (the elongation to failure in the solution compared with that in argon) is
observed at potentials higher than -0.35 VSHE, defining a critical potential for IGSCC.
Good agreement was obtained between the test results at controlled potentials and those
carried out under open circuits conditions at oxygen concentrations that yield equivalent
electrode potentials. This observation demonstrates that it is the potential and not the
oxygen concentration per se that is the important parameter in controlling the susceptibility
of the steel to IGSCC.
If the sensitization time is increased, the critical potential for IGSCC decreases
(becomes more negative), but a threshold value (-0.4 VSHE) is reached at this temperature
(250°C) for prolonged sensitization heat-treatment times. Also, the elongation-to-failure
ratio decreases at any given potential above the critical value with increasing sensitization
time, indicating an enhanced susceptibility to IGSCC because of more severe chromium
depletion along the grain boundaries.
The cracking behavior of Type 304 SS, sensitized for 12 hours at 650°C, in 0.01 m
Na2S04 solution is summarized in Figure 22 [52] in the fonn of a potential-temperature
plot, in which domains of IGSCC and ductile failure are delineated by a temperature
628

FIGURE 17 INTERGRANULAR STRESS CORROSION CRACK

IN HEAT-AFFECTED ZONE NEXT TO WELD IN
TYPE 304 STAINLESS STEEL (J. Y_Park, ANL)
 629

Inlet Safe End

Valve
Dutchman
Core Spray

Recirculation
Inlet

Recirculation
Outlet
Manifold

Recirculation
Pump
Discharge
Valve

Bypass
Valve

RA-320543-31

FIGURE 18 SCHEMATIC OF A BOILING WATER REACTOR RECIRCULATION PIPING SYSTEM

After Danko (44).
630

Reactor
Vessel

Recirculating Inlet
Nozzle (typ)

Cross Tie Valve 305 mm (12 in.) Upper

(not installed on Discharge Riser
some plants)
End Cap to Manifold
Discharge

508 mm (20 in.) 559 mm (22 in.)

RHR Outlet Discharge Manifold
(one loop only)
559 mm (22 in.) x
710 mm (28 in.) 711 mm (28 in.) Cross
Suction Riser 305 mm (12 in.)
RHR Inlet
Motor 711 mm (28 in.)
Lower Discharge Riser
102 mm (4 in.)
Bypass Line
Pump.../ / (not installed on
some plants)
Suction / '
Valve Discharge
Blank Flange Valve
(for decontamination)
RA-320543-33

FIGURE 19 A TYPICAl RECIRCULATION LOOP ARRANGEMENT OF A BWR MODEL 4

After Danko (44).
 631

FREQUENCY OF IGSCC INCIDENTS

AS A FUNCTION OF PIPE DIAMETER

100 r---------------------------------------------------------------------------------~
Total incidents
9Or- worldwide through
m
!zw 80 r- 132 (:::,:,';·,·16-1-78
432c:=J~1~
oQ
z
70 r- -
- 60 r-
8
~ 50 r-
~ 40 r-
t ;

::
ffi
OJ
30 r- ~~.
~;.~
~ 20 r- ~.: "!'!
z
10 r- ;. f' .~~ ,.
.,
,.
o ::: ~:~. :1.;
~.~.
::.:.
o 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 (in)
o 50 102 150 200 250 310 360 410 460 510 560 610 660 710 (mm)
TYPE 304 AND 316 55 PIPE DIAMETER
RA-J20543-J4

AGURE 20 WORLDWIDE TOTAL OF PIPE CRACKING INCIDENTS AS A FUNCTION OF PIPE DIAMETER

After Danko (44).
632

Sen. Type 304 Sla.nl... Stell

0.01 M No.SO.
T. 250"C

'.0

0.'
0 .•
c_

...-
Ol
a.. '

16'
&I

• Au
o
::-:~-:tso..)
Se ••
0.0111 Na.SO.
"t.,
o...rottCl K,:O
J At ..... " _ _ Ial

At • • • CIt'CIIII , .• •tIO.
0.2

0
te • A., S,fut.". "10
7
~O.. -0.6 -0.4 -0.2 0 0.2 0.'
Potential, VH
RA-350543-4

FIGURE 21 STRESS CORROSION CRACKING SUSCEPTIBI LlTY

AS A FUNCTION OF POTENTIAL FOR SENSI-
TIZED TYPE 304 SS IN 0.01 MNazS04AT250"C

0.'
Type ~ Stainless
S8nsitired at 650-c for 12 Hrs.
Sf.'
aOlm~So. Solution
0.6 i=lxlO Y.s.rc

o. • ••
0
• •
>1 02
I IGSCC
j
j 0

002
DJctile
00.

00.6
::::..,{:=
:,.{:=..
110

RA-350543-5

FIGURE 22 POTENTIAL-TEMPERATURE DIAGRAM SHOWING

DOMAINS OF IGSCC AND DUCTILE FAILURE
FOR SENSITIZED TYPE 304 SS IN 0.01 M
Na2S04
 633

dependent critical potential for IGSCC. It is apparent that the critical potential decreases
linearly with temperature. Similar results were obtained in 0.0l m B(OH)] + 0.001 m
LiOH solution [53]. However, the critical potential for IGSCC was more positive by about
100-150 mV in this latter medium. Furthermore, since an upper potential limit for IGSCC
was observed in borate solution at temperatures equal to or lower than 200°C, the IGSCC
domain is narrower than that for sulfate solution, and at 150°C intergranular cracking was
detected only at 0.35 V SHE.
In 0.01 m NaCl solution, the cracking behavior is more complex [49, 51]. Although
IGSCC occurs above a critical potential, which is linearly related to temperature, different
fracture morphologies are observed, depending upon the potential and temperature, as
illustrated in Figure 23. Pure IGSCC occurs over a wide potential range at high
temperatures, but this range becomes narrower with decreasing temperature, as noted
before for sulfate and borate solutions. However, at and above 150°C, simultaneous
intergranular and transgranular cracking occurs in chloride solutions at potentials equal to
or higher than 0.0 VSHE. This is precisely the potential at which breakdown of passivity is
observed on both unstressed and strained samples under potentiostatic conditions.
indicating that pitting may be a fundamental cause of the nucleation of transgranular cracks.
On the other hand, no TGSCC is observed at 100 and 125°C. At both temperatures.
specimens fail by IGSCC accompanied by pitting corrosion.
In all of the environments discussed above, the critical potential for IGSCC lies well
above the corrosion potential measured in de aerated solutions « 0.01 ppm Oz), so that
under normal deareated conditions cracking is not expected to occur. However, if the pH
of the medium is lowered sufficiently, sensitized Type 304 SS apparently becomes
susceptible to fracture. even at very negative potentials. This is illustrated in Figure 24.
For 0.01 m Na2S04 solution (x = 0), where open, closed. and semiclosed circles denote
experiments carried out using internal or external reference electrodes in either recirculating
loops or in static autoclave systems. a well-defined critical potential is observed. However,
if the pH of the solution is lowered by the addition of H2S04 at a constant total sulfate
concentration ( x = 0.005). intergranular cracking occurs at much more negative potential;
even at cathodic potentials, as indicated by the low values of the elongation-to-failure ratio.
These results clearly indicate that, in the presence of acidic contaminants, IGSCC may
occur even in fully deaerated solutions.
A specific inhibitory effect of the solute on the susceptibility of Type 304 SS towards
IGSCC is observed in phosphate solution [54]. For example, the pH of 0.002 m
Na2HP04 + 0.008 m NaH2P04 solution at 250°C is approximately 6.5 [55]. However,
the critical potential for IGSCC. as shown in Figure 25 [54], is significantly higher than
that expected from the pH alone [53]. Furthermore, no cracking is observed in phosphate
solution at 200°C (Figure 25), although severe IGSCC was obtained in all the other
environments investigated. Additional evidence for the inhibitory effect of phosphate
anions on the development of intergranular cracks was obtained by comparing the
elongation to failure ratios measured under open circuit conditions at 275°C in phosphate
solutions and in high purity water, both containing an oxygen concentration of -8 ppm.
The values were 0.81 and 0.50, respectively [49].

3.1.2 Crack Propagation Studies. Considerable interest exists in measuring crack growth
rates using fracture mechanics techniques, in which the stress state at the crack tip is well
defined. As shown in Figure 26, the stress intensity for Mode I loading (KI) has a strong
influence on the crack velocity [56]. Crack velocity is plotted in terms of the maximum
crack growth rate, because it was found that the curvature of the front of the intergranularly
cracked area is very dependent on applied potential. At low potentials, the intergranular
crack extends further at the edges of the specimen than in the center, whereas the opposite
is true at high potentials. By plotting the maximum crack growth rate as a function of KI,
the effects of potential and heat-treatment time become apparent. It apJ>ears that a value of
KI exists below which the crack growth rate becomes smaller than 10- 11 mls. This value,
634

0.3
A Deep --L Pitting _ Shallow D
I
0.2 0 Pitting I TGSCC D
+ I +

•
0 IGSCC I IGSCC
0.1 0 D
I
I
:I:
> 0 D
16 ...c:
J!! -0.1 0
0
ll. 0 I
0 I+P
-0.2 A P
D I+T

-0.3
•• T
D
Dubie
-0.4

0 100 150 200 250

Temperature, 0 C
RA-350543-6

FIGURE 23 POTENTIAL-TEMPERATURE DIAGRAM FOR

SENSITIZED TYPE 304 SS IN 0.01 M NaCI
SOLUTION SHOWING REGIONS OF DIFFERENT
FAILURE MODES

Sensitized TYpe 304

Stainless Stel
1.2
0.01 m Na2 504
T~25O°C

1.0

.S!
.tj 0.8

-6 x=o
.[ 0.6
w

D
i

~
0.4
'x = 0.005
0.2

0
-0.4
Potential, VH
RA-350543-7

FIGURE 24 EFFECT OF POTENTIAL ON IGSCC SUSCEPTI-

BI LlTY FOR SENSITIZED TYPE 304 SS IN
Na 2SO 4/H2S0 4 AT 250 0 C
 635

____ ~C

1.01)-_......__-~

! A
D

o.S
_ o.S "7"~_'--o-"'~
e:
... 0.4 Type 304 stainless steel
sensitized at 6500 C for
0.2 12 hrs 0.008 m NaH2P04

o
+ 0.002 m Na2HP04 e= 1 x 10"6/s
-400 -300 -200 -100 0 100 200 300
Potential, mlH
RA-350543-S

FIGURE 25 EFFECT OF POTENTIAL ON IGSCC SUSCEPTI-

BILITY FOR TYPE 304 SS IN PHOSPHATE
SOLUTION AS A FUNCTION OF TEMPERA-
TURE

Sensitized TYpe 304

Stainless Stel
0.01 m Na2 $04
T =25QDC

II!O'C r;;;;;;;
VjHl

£I>
• -0.2
-0.1

•••
-o.os
0 0.0
0 0.2

10 15 20 25 30 35 40 45
KJ,MPIIIfm
RA-350543-9

FIGURE 26 MAXIMUM CRACK GROWTH RATES AS A

FUNCTION OF STRESS INTENSITY FOR
TYPE 304 SS IN SULFATE SOLUTION
AT 250°C
636

KISCC, apparently decrease with increasing applied potential. The crack growth rate is
strongly dependent on KI for KJ < 20-25 MPa. mla; but for KJ values above this level a
plateau is observed. Under these conditions the crack growth rate tends towards 1 x 10-9
mls for an applied potential of 0.0 VSHE. It should be noted that significant
microbranching occurred during the propagation of intergranular cracks for initial KJ values
greater than 35 MPa.m l12, particularly at high anodic potentials. Accordingly, the
calculated stress intensity values are not very reliable, but the trends are significant.
The effect of potential on crack velocity is more easily seen in Figure 27 [56], where
the maximum crack velocity is plotted as a function of potential for specimens loaded to
different KJ values within the plateau region. From these data it is apparent that a critical
potential exists (- -0.30 VSHE) below which intergranular crack growth does not occur or
is extremely slow. Note that this critical potential is about the same as that determined
using constant extension rate tests.
Recent studies [57,58] on cyclic crack growth in sensitized Type 304 SS in deaerated
0.01 m Na2S04 solution at 250°C, have also revealed the existence of a critical "cracking"
potential (-0.4 to -0.2 V SHE), as shown in Figure 28. The corrosion fatigue tests were
conducted at various applied potentials using a sinusoidal loading wave form, with R =
KminlKmax = 0.5 and ill{ varying from 15 to 30 MPa.m l12 . At high frequencies, the
fatigue crack growth rates are least, and are almost independent of potential. The fracture
morphology was predominantly transgranular, whereas at a lower frequencies « 0.08 Hz),
the enhancement of the fatigue crack growth rate at potentials above -0.4 V SHE is
accompanied by a change in the fracture morphology from transgranular to the typical
intergranular mode observed in stress corrosion cracking tests. These results indicate that,
at low frequencies, the corrosion fatigue behavior of sensitized Type 304 SS becomes
fractographically indistinguishable from that observed at potentials above EIGSCC on
precracked specimens loaded under constant displacement conditions or for smooth
specimens subject to slow strain rate testing.
The corrosion fatigue behavior is further illustrated in Figure 29, [57] where the crack
growth rate on a time basis [da/dt) = f[da/dN)] is plotted as a function of frequency (f,
cycles/s). The dashed line represents the extrapolation of the fatigue crack growth rate to
low frequencies assuming that (da/dN), in the absence of environmental effects, is a
constant. In essence, this is the "cycle-dependent" fatigue crack growth rate that is
expected if fracture occurs by a purely mechanical mechanism. At the open circuit potential
(-0.67 VSHE) there is a slight environmental effect on the fatigue behavior, since the log
(da/dt) versus log f curve deviates upwards from the straight dashed line at low
frequencies. Nevertheless, the environmental enhancement is so small that crack growth is
predominantly cycle-dependent. Furthermore, transgranular fracture morphologies are
found, similar to those observed in air, including the presence of fatigue striations.
However, at 0.035 VSHE, which is above the critical potential for IGSCC in the sulfate
solution, da/dt becomes independent of frequency as the frequency decreases. The
environmental effect is therefore dominant, and the fracture morphology becomes
completely intergranular. It should be noted, however, that the value of daldt is
approximately two orders of magnitude greater than that measured under constant
displacement conditions, as shown in Figure 27. This apparent discrepancy can be
explained qualitatively on the basis of the arguments development by Ford [59, 60]. By
increasing the strain rate at the crack tip, the crack growth rate will increase due to the
increase in the rate of film rupture. Under constant displacement conditions, the rate of
film rupture is extremely low and it may be essentially determined by the creep rate of the
material. Under cyclic loading, the rate of film rupture will increase, as is the case in slow
strain rate tests. This may explain why crack propagation rates measured in slow strain rate
tests (0.01 m Na2S04, 250°C, E = 0.1 VSHE), - 1 x 10-7 mis, are very close to those
measured in the corrosion fatigue tests (f < 8 x 10-2 Hz).
An important finding of these fatigue studies is that intergranular crack propagation
occurs at potentials above the critical value only at low frequencies [f < 0.08 Hz]. Thus,
 637

~~.-~-r----r----r-----r----r----r---~
s.ttsilired ryp. 304
S/rJinl.ss SIHI
aOlmND6 So.
T~250-C

e"2. z
,(/ .all
.6
.23.5 . .

0'....
0".' 024....

~·~7~~~~3--~~~2--~-Q~'----O~--~Q~'--~O~2--~~
i>ow1"aJ '''_
RA-350543-10

• FIGURE 27 EFFECT OF POTENTIAL ON THE MAXIMUM

CRACK GROWTH RATE IN SENSITIZED
TYPE 304 55 IN 0.01 M Na2S04 AT 250°C
(Numbers denote KI values.)

TyIM 304 s.s.

(s.nsifiz6d 1Ofus. at 6!X? C)
aOlm i\q.So.
2!X?C
RzQ5
N=20 MPrJ.",m

~~~~~~--~~~~~-iW~

RA-350543-11

FIGURE 28 FATIGUE CRACK GROWTH RATE (da/dN) AS

A FUNCTION OF POTENTIAL FOR SENSITIZED
TYPE 304 5S IN 0.01 M Na2S04 AT 250°C
638

~.t[Hz)

RA-350543-12

FIGURE 29 EFFECT OF FREQUENCY ON THE TIME-BASIS

FATIGUE CRACK GROWTH RATE (da/dt) IN
SENSITIZED TYPE 304 SS IN 0.Q1 M Na2S04
AT 250°C

TIME Ilwl
I~~__~I~O~~~~__~~~__~~~__~DO~~

400

DO

~TIONI"1 II)

RA-350543-13

FIGURE 30 CRACK DEPTH vs ELONGATION AND NOMINAL

STRESS vs ELONGATION CURVES FOR SENSI-
TIZED TYPE 304 SS IN 0.Q1 M NaCI SOLUTION
AT 250°C
 639

we conclude that IGSCC is not due to high R-ratio, high frequency, loading by rotating
machinery (pumps, for example), but it appears to be due to low frequency loading
phenomena, perhaps associated with changes in the operating conditions of the plant.
Presumably, transient loading due to sudden changes in the operating conditions provides
the crack tip strain rates that appear to be necessary to induce rapid IGSCC crack growth.
Accordingly, IGSCC is most likely an intermittent process, in which cracks may exist in a
near dormant state until appropriate (low frequency) loading causes them to propagate
through the sensitized microstructure.

3.1.3. Significance of the Critical Potential. As shown in Figures 22 and 23, sensitized
Type 304 SS is susceptible to IGSCC in dilute aqueous solutions containing various
anionic species over broad ranges of potential and temperature. For a given solution and
temperature, EIGSCC is also dependent on the degree of sensitization, but reaches a
threshold value after prolonged sensitization times (-16 to 24 hrs).
It has now been verified experimentally, by using scanning transmission electron
microscopy (STEM), that a chromium-depleted region is formed along grain boundaries
upon sensitization of Type 304 SS [61]. Different compositions of the chromium-depleted
zone are expected as a function of heat treatment time. Chromium concentrations as low as
-10 wlo have been measured in specimens sensitized for 24 hrs at 650°C [61]. Also, the
width of the chromium-depleted zone increases and the chromium content decreases with
increasing sensitization time [62], which indicates that a greater volume of low chromium
alloy is available for dissolution at the crack tip for those alloys that have been heat-treated
for longer times. Therefore, we expect that EIGSCC will decrease with increasing
sensitization time, because a lower anodic overpotential will be required to enhance the
dissolution rate of the chromium depleted grain boundaries with respect to that of the
matrix. Evidence for this behavior can be gleaned from the anodic polarization curves for
Fe-XCr-lONi (where X = 8, 10, 18) measured by Indig and McIlree [63] in high
temperature sulfate solutions.
Although the phenomena discussed above appear to be sufficient to account for the
effect of sensitization time on the susceptibility of Type 304 SS to IGSCC, the application
of stress or straining is required to produce mechanical rupture of the passive fllms as a
precondition for this form of localized attack, when the bulk environment is neutral or
slightly acidic. The necessity for straining (or applied stress) is indicated by the fact that no
intergranular attack occurs on unstressed samples. It should be noted, however, that
Vermilyea et al [64-66] have observed the formation of thicker oxide fllms on emergent
grain boundaries than on grains, when sensitized Type 304 SS is exposed to dilute sulfuric
acid (pH 3 to 4) and neutral solutions at 290°C. Thus, it is possible that, in the absence of
straining, the susceptible emergent grain boundaries are, in fact, preferentially protected.
However, upon straining these thicker, and presumably more crystalline (brittle) films
rupture occurs to mark a nucleation event. After rupture, the film on the matrix rapidly
reforms due to the higher repassivation rate, while at the grain boundaries the greater
dissolution rate, coupled with a lower rate of repassivation, tends to confine the attack to
these areas. No data exist for the repassivation kinetics of Fe-Cr-Ni alloys at high
temperatures (> 150°C). Results obtained by Ford and Silverman [59] in 0.01 m Na2S04
solution at 100°C using very fast straining, and those by Newman et al [31] in 0.5 M
Na2S203 solutions through scratching techniques, suggest that repassivation is
significantly delayed for alloys simulating the grain boundary composition within specific
potential ranges. This hypothesis is somewhat countered by the observation [51] that at
moderate strain rates (on CERT specimens), cracks nucleate transgranularly and propagate
in this manner until they intersect a sensitized grain boundary, at which point crack growth
occurs intergranularly. However, this is not expected to be the case at the much lower in-
service strain rates, where differences in the properties of the passive film over the bulk
grain and chromium-depleted emergent grain boundary regions are expected to dominate
the nucleation process.
640

Although plastic defonnation produced by continuous straining is responsible for the

rupture of the passive film, results such as those shown in Figure 30 [51] clearly indicate
that a certain amount of elongation (or time beyond the onset of yielding) is required to
initiate the growth of intergranular cracks. In these experiments, straining was interrupted
at various elongation values, and the specimens were removed from the autoclave. The
maximum crack depth was then measured on the longitudinal cross section of the tensile
specimens by optical microscopy. It is seen that cracks start to grow rapidly at elongations
corresponding to a stress value close to the maximum nominal stress. Shallow
transgranular cracks may fonn on the metal surface at short times (or strains), but
additional conditions must be met to induce the nucleation and propagation of deep
intergranular cracks. The principal additional factor seems to be the electrode potential. If
the potential is lower than EIGSCC, a sufficiently aggressive environment cannot build up at
the crack tip, so that the shallow cracks will remain inactive due to the fonnation of a
protective film, and the specimen will fail in a ductile mode. However, if the potential is
high enough, the rate of metal dissolution at the crack tip will be enhanced, and hydrogen
ions will be produced as a consequence of the hydrolysis of metal cations. This
acidification will further enhance the dissolution rate, and hence, in the presence of such an
aggressive local environment, only a poorly protective film can fonn at the crack tip, thus
causing the crack to propagate. The high aspect ratio of the crack is preserved by the faster
repassivation of the crack walls, which are composed principally of the non-depleted alloy.
In order to maintain ~ectroneutrality, anions must migrate to the crack tip, but only non-
OH- anions (X = SO -, Cl-, B(OH)4) are able to maintain a localized acidic environment.
However, as the pH ~ecreases the rate of hydrogen evolution at the crack tip will increase
and ultimately the drop in pH will be limited by the reduction of hydrogen ions.
Considering the equilibrium potential for the hydrogen evolution reaction at 250°C, the
minimum pH that could be reached at the EIGSCC determined in chloride and sulfate
solutions (- 0.35 VSHE), assuming minimal potential drop down the length of the crack,
should be around 4.0.
This interpretation provides a partial explanation of the results presented in Figure 22
through 25. Similar values for EIGSCC were obtained for chloride and sulfate solutions at
250°C, probably reflecting comparable acidities for the localized environment inside the
crack. In the presence of B(OH) - , a higher potential is necessary £ecause of the
neutralizing effect of this anion. 'On the other hand, H2POJIIPO - has a specific inhibitory
effect, due to the buffering properties of this system and the possiile fonnation of insoluble
ferrous phosphates. However, if the bulk pH is too acidic (HS04)' localized dissolution of
the chromium-depleted grain boundaries might be sufficiently rapId at lower anodic
overvoltages to lead to rapid initiation of intergranular cracks.
The preexistence of microcracks, per se, is not the controlling factor in the development
of macroscopic intergranular cracks, at least when the bulk environment is neutral. The
establishment of appropriate electrochemical conditions in the system will determine
whether or not any given crack will arrest or propagate. This is clearly illustrated in Figure
31 [51]. In these tests, specimens were strained at 250°C in deaerated 0.01 m NaCI
solution at a strain rate of I x 10-6/s and at an applied potential of - 0.2 V SHE (above
EIGSCc) until the load started to relax after reacting a maximum, indicating the initiation of
rapid intergranular crack growth. The potential was then maintained at - 0.2 VSHE or was
(e
switched to a more negative value, while straining was simultaneously interrupted = 0).
Figure 31 shows that the load response, reflecting the extension of a crack under constant
strain conditions, is clearly dependent on potential. Curve (a) shows the behavior observed
when the potential was maintained at - 0.2 VSHE until fracture occurred. A monotonic
decrease of load is observed. A similar monotonic decrease was observed when the
potential was switched to - 0.3 VSHE [curve (b)], but at a lower rate of load relaxation. By
switching the potential back to - 0.2 VSHE [curve (c)], a change in the rate of load
relaxation becomes noticeable, indicating an acceleration in the crack growth rate.
However, the most significant result is that crack arrest can be induced by switching the
 641

1500r---r----....,..---~--__r---~--__,

r,,,.
J04 51aifl'-SS 51H1
5M/silir«J 12Hrsaf650·C
1200 OOIM NaCI
r""".ratu,.. 6250 -C
------- -o.ZV.
\ - -o.3V.
900 \ zoo
,,
_·04V.
z. \
\
! \
\
\
\
\
600
\ \
\ \
\ \ 100

300 \ \
\
\0, \
Ie
\
\
0~~~-----+------~~~~--~---~~---~2f
Time ,Hour.
RA-350543-14

FIGURE 31 EFFECT OF POTENTIAL ON LOAD VERSUS TIME PLOTS DURING PROPAGATION

OF INTERGRANULAR CRACKS IN SENSITIZED TYPE 304 SS IN 0.Q1 M NaCI AT 250°C

100r--__r--r-~-~~~h'I----rI---~Ir--~I~I'~~r--~---~---~---~

Type ~ Sfainl.ss SlNI

StNtsifinld Of 65Cr C lor 12 Htus 500
aOlm M:7,sa, Soluh'on
T~tur. -275- C
-L_
Solution Flow
----5001 s.s.] M,.
Raf~-I20cc/mi".
~.............. .......
. . . .....:
,~
400

..••••..... Plot_ PlMrtttatI •.•.••• , , , B

t ,. t " ~\ 300 ~
" ... oS
:
.
·500 200

-600
100
·700

~~~~~~~---~I~---~~I---~I·,w---~~--~--~~~'O
26 28 42 44 46 48 50 52
Time, Hours
RA-350543-15

FIGURE 32 EFFECT OF ENVIRONMENTAL CHANGES (H2 AND 02 ADDITIONS) AND THE ASSOCIATED
VARIATIONS IN Ec ON THE PROPAGATION OF INTERGRANULAR CRACKS IN SENSITIZED
TYPE 304 SS IN 0.Q1 M Na2S04 AT 275°C
642

potential to - 0.4 VSHE (50 mV below EIGSCc), as indicated by the constant value of the
load [curve (d)]. When the potential was switched back to - 0.3 VSHE (50 mV more
positive than EIGSCC, point X) crack growth reinitiated, and finally a further potential
change to - 0.2 VSHE (point Y) induced even faster crack growth, which ultimately resulted
in failure. It is evident that crack growth can be arrested and restarted by cycling the
potential below and above EIGSCC.
A propagating crack can also be arrested by decreasing the corrosion potential <Be) of
Type 304 SS to a value lower (more negative) than EIGSCC as a consequence of a change
in the composition ofthe environment [53]. After a crack was initiated and allowed to
propagate to a depth of at least 200 J..Un under slow straining conditions (£ = 1 x lO-6/s) in
0.01 m Na2S04 solution containing 0.2 ppm D2 at 275°C <Be > EIGSCc), straining was
interrupted (Figure 32). Almost simultaneously, a mixture ofH2 and N2 was injected into
the system. The potential of a platinum electrode exposed to the environment (Ered)
decreased rapidly, indicating a change in the "redox" potential of the environment from
oxidizing to reducing conditions. On the other hand, Be decreased slowly, probably due to
slow changes in the surface oxide film, while the crack was still growing as indicated by
the decrease of the load. However, when Be decreased to a value lower than - 0.45 VSHE,
which is the EIGSCC for this environmental condition in terms of solution composition,
pH, and temperature, the propagating crack was arrested, as seen by the constant value
attained by the load. From this result it is clear that it is not the D2 concentration per se, as
indirectly measured by the Pt electrode, that controls the initiation and propagation of
intergranular cracks, but it is the corrosion potential of the stainless steel relative to the
critical potential that is the important parameter. When the concentration of D2 was
increased to 0.2 ppm after 22 hrs, Ered and Be, began to increase, and the propagation of
the crack was reinitiated after a prolonged induction time. An additional increase in the 0:2
concentration accelerated the crack growth due to the increase in Be, but a further decrease
in the D2 concentration, although it was able to produce a sudden drop in £roo, did not shift
the Be below EIGSCC rapidly enough to avoid the fmal failure of the specimen.
The two experiments described above, one under potentiostatic control and the other
one under open circuit conditions, indicate that the establishment of appropriate
electrochemical conditions in the system, in terms of potential and possibly pH, will
determine whether or not a crack will arrest or propagate. Indeed, recent work by the
author, which will be published shortly, indicates that the rate of crack growth in sensitized
Type 304 SS in BWR environments is detennined by the properties of the environment
external to the crack because of charge conservation - i.e. the current exiting the crack must
be consumed by some cathodic reaction occurring on the external surfaces and the high
resistivity of the external environment severely limits the projection of the current from the
crack mouth.
3.1.4. Potential Control Techniques. With the discovery of a critical potential for IGSCC
of sensitized Type 304 SS in high temperature aqueous systems, as discussed above, it is
not surprising that potential control techniques are being actively developed as a means of
preventing pipe failures in operating BWRs. The most important of these techniques is the
"hydrogen water chemistry" (HWC) program that is now being evaluated in BWRs in both
the U.S. and Sweden [45]. Since the goal of this program is to displace the electrode
potential of the Type 304 SS pipe surfaces to a value that is more negative than the critical
potential for IGSCC, we briefly outline the origin of the electrode potential and discuss the
mechanisms by which the potential can be modified by the addition of hydrogen to an
operating BWR. The fundamental aspects of this problem were addressed some years ago
by Macdonald, Scott, and Wentrcek [68] and more recently by Niedrach and Stoddard
[69].
In developing a physically realistic model for BWR systems, it is important to
recognize that the aqueous phase is not an equilibrium, as evidenced by the simultaneous
presence of oxidizing and reducing species (e.g., 02, H2D2, and H2,) due to the radiolysis
 643

of water. Accordingly, the electrochemical properties of the system (including the electrode
potential) cannot be discussed in terms of equilibrium thermodynamics. The most
important consequence is that the electrode potential cannot be accounted for by the
concentration of a single component (e.g., oxygen). If the metal substrate is completely
inert, or is nearly so (e.g., platinum), the electrode potential, which is a mixed potential,
depends only upon the relative concentrations and kinetic parameters for the various
oxidizing and reducing species in the environment. In this case, we refer to the mixed
potential as a "redox potential". On the other hand, if the substrate is not inert (as in the
case of steel), oxidation reactions involving the metal (corrosion) also contribute to the
observed electrode potential, which we refer to as the "corrosion potential". Clearly, the
two converge as the corrosion activity of the substrate decreases.
Macdonald, Scott, and Wentrcek [68] developed a mixed potential model (MPM) for
non-equilibrium high temperature aqueous systems of the type that exist in BWRs. The
basis of this model is that the oxidation of hydrogen and the reduction of oxygen in a non-
equilibrium system occur simultaneously on a metal surface, and that these reactions may
be subject to mass transport control depending upon the potential and the kinetic parameters
for the two reactions. The redox potential (note that corrosion processes involving the
substrate are not taken into account) is simply that potential at which the rates of the two
reactions are equal. The MPM has successfully accounted for the "S-shaped" electrode
potential vs oxygen concentration curves that have been reported by various workers
[70,71]. Some typical results of this model, together with redox potentials for non-
equilibrium H2IOz systems measured in the laboratory [68], are shown in Figure 33. In
this figure, the redox potential is plotted as a function of the ratio of hydrogen to oxygen in
the fluid and the "S-shaped" form of the redox potential curve is clearly apparent. Two
important findings of this work are that: (i) The redox potential is determined essentially by
the ratio of the concentrations of oxidizing to reducing species and (ii) the redox potential
(and by implication the corrosion potential) may be displaced to sufficiently negative values
to prevent IGSCC (e.g., to -0.6 VSHE at 275°C) by the addition oflow levels (e.g., 1-1.5
ppm) of hydrogen without a concomitant reduction in the concentration of oxygen
(although this does occur). This latter point is particularly important, since the ability of
HWC to provide protection against IGSCC in operating BWRs need not depend upon the
"suppression of radiolysis" by the relatively small amounts of hydrogen added to the
feedwater (c.f. PWR primary systems that employ much higher concentrations).
Interestingly, recent calculations by Ishigure, et al [72], for BWR systems indicate that as
little as 300 ppb of hydrogen added to the feedwater is effective at reducing the
concentration of oxygen to the 25 ppb level, in agreement with observations in a Swedish
BWR [73]. It is also important to note, however, that the ability of hydrogen to modify the
[0z]![H2] ratio, and hence to depress the corrosion potential, depends in a complex manner
on system configuration (e.g., whether or not jet pumps are used). This arises because the
extent to which the corrosion potential and oxygen level can be suppressed depends upon
the rates of various homogeneous reactions involving a large number of redox species
(e.g., 02, H202, H2, OH, H, HOz, 0;, ... ), and hence on the residence time of these
species in the n and 'Y field in-core. One encouraging observation, particularly from
laboratory studies, is that the redox potential (Figure 34) and corrosion potential of Type
304 SS (Figure 32) may be cycled by changing the relative concentrations of hydrogen and
oxygen in the water. This finding suggests that the addition of small amounts of hydrogen
may be effective in arresting IGSCC after periodic exposures to highly oxidizing
conditions.
It is fitting to end this discussion by emphasizing that laboratory studies (Figure 24)
indicate that pH may be a critical parameter in determining the effectiveness of HWC for
preventing IGSCC in operating BWRs. Although the critical potential for IGSCC
apparently is not reduced substantially by lowering the pH of the fluid, slight acidification
results in the elongation-ta-failure ratio remaining substantially below one and in
intergranular fracture even at very cathodic potentials. Such a reduction in pH might occur
644

W 0.6
:J: BWR
'"::>'" Shutdown
0.4
'"a:w
>
0.2
~
..J
0
~ 0
..J
«
~ -0.2
w
~
0
11. -0.4
X
0
<>
::! -0.6

-3 -2 -1 o 2 3

SA-7307-66

FIGURE 33 SIMULATED REDOX POTENTIAL Ern (0)

VERSUS COMPOSITION CURVES FOR SETS
B, C, D LISTED IN REF. [68J
• - equilibrium potentials Ee. 1, for the H2/H+ reaction;
I!. - equilibrium potentials Ee.2,for the 02/H20 reaction;
• - experimental data.

nHOURr--.----------------r---------------,----------------r-----,
100"(. H2

too

-too

-200

-Joo

' " REDOX PROBE

"00

-500

...,.,
RIGHT S C A U - - - _

-700
t80 t90 200 210
TIME/"ou~

SA-7307-44A

FIGURE 34 RESPONSE OF HYDROGEN MONITOR (UPPER CURVE)

AND REDOX PROBE (LOWER CURVE) TO HYDROGEN
AND OXYGEN IN O.lm BORIC ACID AT 275°C
 645

because of the thennal or radiolytic decomposition of cationic resin fines to fonn sulfuric
acid, or because of local acidification within a pre-existing crack. Clearly, total control of
the water chemistry, and not just the specification of the corrosion potential of the steel
pipes, will be necessary for effective prevention of rosee in operating BWRs.

3.2. IASCC OF IN-CORE MATERIALS

Over the past few years [74], intergranular fracture has been observed in various
stainless steel and high alloy components in the cores of water cooled nuclear reactors
(BWRs and PWRs). The phenomenon apparently is caused by a combination of two or
more of the following factors: high neutron fluence (> 5 x lOLO n/cm2, E > 1 MeV), high
temperature (290-300 0 e for BWRs, but -340°C for PWRs), segregation of impurities (P,
Si, S) to grain boundaries, grain boundary chromium depletion, hydrogen and/or helium
uptake, aggressive radiolytic species fonned in the water, and the existence of grain
boundary stresses. The mechanistic factors involved are sufficiently vague that it is not
certain whether or not IASee is substantially different from the IOSee of sensitized
austenitic steels, which occurs in BWR environments at lower temperatures (see above).
However, IAsee is also observed in PWR primary systems, which contain high
concentrations of dissolved hydrogen and are characterized by very negative redox and
corrosion potentials.
The possibility that IASee is an irradiation-accelerated fonn oflOSee is supported by
extrapolating the critical potential data shown in Figure 22 to the in-core temperatures for
BWR (300°C) and PWR (340°C) systems. For BWR conditions, the extrapolated potential
is of the order of -0.6 V (vs. SHET), which is probably slightly more negative than the
potential that can be achieved by adding hydrogen (1-1.5 ppm) to the feedwater (see
previous section) and is considerably more negative than the measured corrosion potentials
for Type 304 SS in nonnally operating boiling water reactors. In the case of PWRs,
extrapolation of the potential data shown in Figure 22 to 340°C yields a critical potential for
IOSee of - -0.8 V(SHET). Noting that 1.5 ppm of dissolved hydrogen corresponds to
about 17 cc(STP)/kg, and that this amount of hydrogen yields a redox potential of - -600
mV(SHET) at 275°C, we estimate that the redox potential and structural materials corrosion
potentials in-core in a PWR will be no more negative than about -650 mV(SHET). This is
so, because the equilibrium pressure for the hydrogen electrode reaction (H+ + eH 112
H2) is shifted in the negative direction by 2.303 RT/2F (- 60 mY) for every decade
increase in the dissolved hydrogen concentration. The shifts in the mixed (redox and
corrosion) potentials are likely to be of similar magnitudes, because the structural materials
of interest all tend to act as "hydrogen electrodes" in hydrogen-rich environments. Thus,
the shift in potential at 340°C on increasing the hydrogen concentration from 17 to 25 cc
(STP)/kg amounts to approximately 0.010 V. On the basis of these calculations, the in-
core environments of both BWRs and PWRs are likely to be such that rosee of sensitized
austenitic stainless steels will occur spontaneously; in BWRs by virtue of the relatively high
redox and corrosion potentials, even upon the addition of hydrogen, and in PWRs because
the high in-core temperature possibly displaces the critical potential for cracking to an
excessively negative value.
The hypothesis presented above argues that the cause of IASee of austenitic alloys in
the cores of nuclear reactors is basically the same as that for rosee in sensitized stainless
steels in the lower temperature out-of-core environments. However, considerable evidence
exists that neutron irradiation renders the alloys more susceptible to this type of attack,
although the effect has yet to be quantified. Neutron radiation may increase the diffusivities
of impurities such as P, S, and Si to the grain boundaries by the fonnation of point defects,
and it may also increase the diffusivities of chromium and carbon, which can then react at
the grain boundaries to fonn chromium carbides and simultaneously deplete the
surrounding matrix of chromium. Because of the enhanced diffusivities, sensitization
would be expected to occur at lower temperatures than in the unirradiated case, and in this
646

sense IASCC might be viewed as an irradiation-assisted low temperature sensitization

phenomenon.

4. The Institutional Problem

There is a little question that corrosion has been responsible for many of the operating
problems experienced in nuclear power plants, and that materials degradation phenomena
will continue to impose economic penalties on the nuclear industry in the future. In
reviewing the origin and impact of corrosion problems in power reactors, it has been
possible to pinpoint a number of root causes for our current situation:

(i) Lack of corrosion expertize in the design and materials specification

stages of developing nuclear power reactors.

(ii) Simple and sometimes naive transfer of materials technology from

fossil power to nuclear systems, without proper appreciation of the
differences that exist in the operating conditions or materials properties
between them, at least with regard to corrosion.

(iii) The desire to minimize the capital cost of constructing nuclear reactors
without careful analysis of the impact of corrosion on the operating
costs.

(iv) Inadequate prior research on the corrosion of materials under realistic

environmental conditions.

In view of the rapidly escalating cost of building nuclear power plants and the high cost
of replacement power, it is clear that the trends of the past cannot be allowed to continue if
we are to maintain an economically-viable nuclear power industry. Furthermore, the
reliability of nuclear systems is an issue that, in the public's eye, transcends all other issues
connected with this industry, and the perception (real or imagined) that nuclear reactors are
"unsafe" will surely kill nuclear power if reasonably economic alternatives are available.
Because corrosion phenomena are closely related to "safety issues" (e.g., fracture of
coolant pipes, cracking of pressure vessels), the need to continue corrosion research is
clear. However, materials and corrosion research in the nuclear power industry has
become "reactionary," in that the resources necessary to deal with a problem are made
available only when the problem occurs. This is hardly "preventive medicine," and I fear
that the price which the nuclear industry will have to pay in the future for this attitude may
well contribute significantly to its demise.
Accepting for the moment that the nuclear power industry will survive the present
crises, and that we will move on to more advanced reactor systems (e.g., liquid metal fast
breeders, AGRs, advanced light water reactors), including fusion reactors, we would do
well to heed the mistakes of the past. Accordingly, we should begin to carry out the
required research systematically, so that the necessary corrosion data base will be available
in the design phase to ensure that the most appropriate materials are selected. This should
be done without economics being the most important factor in deciding whether or not a
particular material could be used. Clearly, highly skilled materials scientists and corrosion
engineers will have to be involved at all points of the design and specification stages, and I
would go so far as to suggest that designs for reactor systems should be reviewed by
independent corrosion and materials science bodies before approval for construction is
given. The continued and future reliability of reactor systems is that important.
 647

5. Acknowledgments
The author gratefully acknowledges the financial assistance of SRI International in the
preparation of this paper.

6. References
1. Casper, L. A. and Pinchback T. R. Eds., (1980) 'Geothermal Scaling and
Corrosion', ASTM STP 717, Philadelphia, PA.

2. Lewis, G. N. and Randall, M. (1961) Thermodynamics, Revis. K. A. Pitzer and L.

Brewer, 2nd edit, McGraw-Hill, NY.

3. Macdonald, D. D. and Butler, P. (1973) Corros. Sci., 13, 259.

4. Macdonald, D. D. (1975) Modem Aspects of Electrochemistry, J. O'M. Bockris and

B. E. Conway (eds.), Plenum Press, NY, 11, p. 14l.

5. Macdonald, D. D., Shierman, G. and Butler, P. (1972) 'The Thermodynamics of

Metal-Water Systems at Elevated Temperatures I. The Water and Copper-Water
Systems', Atomic Energy of Canada, Ltd., report, AECL-4136.

6. Macdonald, D. D., Shierman, G. and Butler, P. (1972) 'The Thermodynamics of

Metal-Water Systems at Elevated Temperatures II. The Iron-Water System', Atomic
Energy of Canada, Ltd. report, AECL-4137.

7. Macdonald, D. D., Shierman, G. and Butler, P. (1972) 'The Thermodynamics of

Metal-Water Systems at Elevated Temperatures. III. The Cobalt-Water System',
Atomic Energy of Canada Ltd. report, AECL-4138.

8. Macdonald, D.D. (1972) 'The Thermodynamics of Metal-Water Systems at Elevated

Temperatures. IV. The Nickel-Water System', Atomic Energy of Canada Ltd.
AECL-4139.

9. Macdonald, D. D. and Hyne, 1. B. (1976) 'The Thermodynamics of the

Iron/Sulfur/Water System', Final Report to Atomic Energy of Canada Ltd., Pinawa,
Manitoba.

10. Macdonald, D. D. (1980) Thermodynamics of Corrosion for Geothermal Systems,

Spec. Tech. Pub!. 717, ASTM, Philadelphia, PA.

11. Macdonald, D. D., Syrett, B. C. and Wing, S. S. (1979) Corrosion 35, p. l.

12. Criss, C. M. and Cobble, 1. W. (1964) Journal of the American Chemical Society,
~, pp. 5385-5390.

13. Murray, R. and Cobble, 1. W. (1977) Personal communication.

14. Taylor, D. F. (1978) Journal of the Electrochemical Society, 125, p. 808.

15. Tremaine, P. and Goldman, S. (1978) J. Phys. Chern., 82, p. 2317.

648

16. Naumov, G. B., Ryzhenko, B. N. and Khodakovsky, 1. L. (1974) Handbook of

Thennodynamic Data, Transl. U. S. Geological Survey USGS-WRD-74-001.

17. Taylor, D. F. (1977) 'Thennodynamic Properties of Metal-Water Systems at

Elevated Temperatures', Report No. 77CRD213, General Electric Co., Schenectady,
NY.

18. National Bureau of Standards, Technical Note, No. 270-1,1965.

19. National Bureau of Standards, Technical Note, No. 270-4,1969.

20. Kelley, K. K. (1960) Bulletiin of the United States Bureau of Mines. Vol. 605.

21. Wicks, C. E. and Bock, F. E. (1968) Bulletin of the United States Bureau of Mines,
Vol. 605.

22. Ellis, A. J. (1966) J. Chern. Soc. (A), p. 1579.

23. Ellis, A. 1. (1968) 1. Chern. Soc. (A), p. 660.

24. Ellis, A. J. (1968) J. Chern. Soc. (A), p. 1138.

25. Millero, F. 1. (1969) Activity Coefficients in electrolyte Solutions, Vol. II, CRC
Press, Boca Raton, Fl.

26. Millero, F. J. (1971) Chern. Rev., p. 147.

27. Millero, F. 1. (1982) Geochim. Cosmochim. Acta, 46, p. 11.

28. Pourbaix, M. (1974) Atlas of Electrochemical Equilibria in Aqueous Solutions,

NACE, Houston, TX.

29. Townsend, H. E. (1970) Corros. Sci., 10, p. 343.

30. Cowan, R. Al and Staehle, R. W. (1971) 1. Electrochem. Soc., 118, p. 557.

31. Lee, J. B. (1981) Corrosion, 37.

32. Pytkowicz, R. M. (1979) 'Activity Coefficients in Electrolyte Solutions', Vol. I and

II, CRC Press, Boca Raton, FL.

33. Biernat, R. J. and Robins, R. G. (1972) Electrochimica Acta, 17, 1261.

34. Seward, T. M. (1974) Am. J. Sci. 274, p. 190.

35. Karaminezhaad-Ranjbar, M., Mankowski, J. and Macdonald, D. D. (1985)

Corrosion, 41, p. 197.

36. Choi, H. J., Hu, Y-H., Karaminezhaad-Ranjbar, M., Mankowski, J., Beck, F. H.,
Szklarska-Smialowska, Z. and Macdonald, D. D. (1983) Proc. Int. Symp. Env.
Deg. Mat. Nuclear Power Syst.-Water Reactors, EPRI/NACE, Myrtle Beach, Sc.

37. Syrett, B. C., Macdonald, D. D. and Shih, H. (1980) Corrosion, 36, p. 180.
 649

38. Roberts, J.T.A. (1981) Structural Materials in Nuclear Power Systems, Plenum
Press, NY.

39. Green, S. J. and Paine, 1.P.N. (1983) Proc. Int. Symp. Env. Degrad. Mat. Nucl.
Power Syst.-Water Reactors, NACE/AIME/ANS, Myrtle Beach, SC, p. 53.

40. Karlderson, D. (1972) 'Steam Turbine Failure at Hinkley Point A', Proc. Inst.
Mech. Eng., 186, p. 341.

41. Danko, 1. C. (1983) Proc. Int. Symp. Env. Degrad. Mat. Nucl. Power Syst-Water
Reactors, NACE/AIME/ANS, Myrtle Beach, SC, p. 209.

42. Cowan, R. L. (1978) Stress Corrosion Cracking and Hydrogen Embrittlement of

Iron Base Alloys, 1. Hockmann and R. W. Staehle (Eds) NACE, Houston, TX, p.
1023.

43. Akashi, M. (1980) Boshoku Gijutsu, 22, p. 142.

44. Danko, J. C. and Smith, R. E. (eds), (1980) Proceedings: Seminar on

Countermeasures for Pipe Cracking in BWRs, EPRI WS-79-174, Electric Power
Research Institute, Palo Alto, CA.

45. Danko, J. E. (ed), (1984) Proceedings: 2nd Seminar on Countermeasures for Pipe
Cracking in BWRS, EPRI NP-3684-SR, Electric Power Research Institute, Palo
Alto, CA.

46. Szklarska-Smialowska, Z. and Cragnolino, G. (1980) Corrosion, 36, p. 653.

47. Andresen, P. L. (1983) Corrosion/83, Paper Nos. 11 and 123, NACE, Anaheim,
CA.

48. Ruther, W. E., Soppet, W. K., Ayrault, T. and Kassner, T. F. (1983)

Corrosion/83, Paper No. 125, NACE, Anaheim, CA.

49. Cragnolino, G., Lin, L. F. and Scklarska-Smialowska, Z. (1981) Corrosion, 31, p.

312.

50. Park, H. C., Cragnolino, G. and Macdonald, D. D. (1984) Proc. Int. Symp. on
Env. Degrad. Mat. Nucl. Power Syst. Water Reactors, NACE/AIME/ANS, Myrtle
Beach, SC, NACE, Houston, TX, p. 604.

51. Lin, L. F., Cragnolino, G., Szklarska-Smialowska, Z. and Macdonald, D. D. (1981)

Corrosion, 37, p. 616.

52. Park, H. C, Cragnolino, G. and Macdonald, D. D., unpublished results.

53. Cragnolino, G. and Macdonald, D. D. (1984) Proc. 9th Int. Congress Metallic
Corrosion, National Research Council, Otawa, Canada, Vol. 2, p. 185.

54. Soma Sekhar, N. (1984) MS Thesis, The Ohio State University, Columbus, OH.
650

55. Kinoshita, K., Scott, A C., Leach, S. c., Macdonald, D. D. and McKubre,
M.C.H. (1981) 'Monitoring Techniques for pH, Hydrogen and Redox Potential',
SRI International Annual Report, EPRI Project RP1168-1.

56. Chung, P., Yoshitake, A, Cragnolino, G. and Macdonald, D. D. (1985) Corrosion,

41, p. 159.

57. Tsai, Wen-ta, Moccari, A, Szklarska-Smialowska, Z. and Macdonald, D. D. (1984)

Corrosion, 40, p. 573.

58. Tsai, Wen-ta (1983) Ph.D. Dissertation, The Ohio State University, Columbus, OH.

59. Ford, F. P. and Silverman, M. J. (1980) Corrosion, 36, p. 597.

60. Ford, F. P. (1982) 'Mechanisms of Environmental Cracking in Systems Peculiar to

the Power Generation Industry', EPRI NP-2589, Electric Power Research Institute,
Palo Alto, CA

61. Pande, C. S., Suenaga, M., Vyas, B. and Isaacs, H. S. (1977) Scr. Metall, 11, p.
681.

62. Chung, P. and Szklarska-Smialowska, Z. (1981) Corrosion, 37, p. 39.

63. Indig, M. E. and McIlree, A. R. (1979) Corrosion, 35, p. 288.

64. Vermilyea, D. A (1975) Corrosion, 31., p.421.

65. Indig, M. E. and Vermilyea, D. A (1975) Corrosion, 31, p. 51.

66. Vermilyea, D. A and Indig, M. E. (1972) J. Electrochem. Soc., 119, p. 39.

67. Newman, R., Sieradzki, K. and Isaacs, H. S. (1982) Trans. A., 13A, p. 2015.

68. Macdonald, D. D., Scott, A. C. and Wentrcek, P. (1981) 1. Electrochem. Soc.,

128, p. 250.

69. Niedrach, L. W. and Stoddard, W. H. (1985) Corrosion/85, NACE Boston, MA,

Paper No. 104.

70. Indig, M. E. and McIlree, A. R. (1979) Corrosion,.15., p. 288.

71. Lee, J. B. (1978) Ph. D. Dissertation. The Ohio State University, Columbus, OH.

72. Kshigure, K., Takagi, J., Kato, I., Shigeto, T. and Fugita, N. (1983) Proc.
Symposium Water Chemistry Corrosion Problems in Nuclear Power Plants, IAEA-
SM-264/35.

73. Feyes, P., Ref. 39, Paper No. 54.

74. Jones, R. H. (1985) Proc. Sec. Int. Symp. Env. Degrad. Mat. Nucl. Power Syst.-
Water Reactors, ANS/AlME/NACE, Monterey, CA, Sept. 9-12.
DEVELOPMENTS IN THE STUDY OF BIOMATERIALS
DEGRADA nON

M.A. BARBOSA and R.A. SILVA

Department of Metallurgy,
Faculty of Engineering, University of Oporto,
Rua dos Bragas,
4099 PORTO CODEX,
PORTUGAL

ABSTRACT. Degradation of metallic materials in the body is a

process with important clinical consequences and therefore
research is being carried out to minimize the release of toxic
species. This paper focus on some methods presently available to
reduce the release of toxic metal ions and to improve fixation of
implants to bone. The need to characterize the in vitro behaviour
of new materials and devices has led to the application of
electrochemical corrosion testing. Some examples are presented,
which show the usefulness of these techniques in providing basic
data enabling a reduction in the number of animal tests.
Polarization resistance and corrosion potential measurements are
useful in predicting the biological tolerance of tissues to metallic
implants. Zero-resistance am metry can be used to study the
behaviour of galvanic couples between carbon composites and
various common metallic materials. Capacitance measurements
account for differences in corrosion behaviour of stainless steel in
normal and simulated infected media.

1. Introduction

Traumatic injuries caused by road accidents and increased life

expectancy have created greater scope for the development of
biomaterials. Medical care has a social importance which is
difficult to quantify in economic terms but the evolution of the
biomaterials market indicates what society expects from this area
651
M. G. S. Ferreira and C. A. Melendres (eds.J,
Electrochemical and Optical Techniquesfor the Study and Monitoring of Metallic Corrosion, 651--663.
© 1991 Kluwer Academic Publishers.
652

of materials engineering. World growth is estimated at 15 % per

annum. By the turn of the century biomaterials may have reached
an economic importance equivalent to that of pharmaceutical
products.
Hip and knee prostheses, fixation devices for fractured bones,
dental amalgams and prostheses, catheters, heart valves,
mammarian and penis prostheses are only a few examples of the
application of biomaterial research, each de manding special
mechanical and biological properties.
For many years man has been attempting to use materials
developed for other purposes in the body. When we hear that
com mon steel( 1) and Plaster of Paris(2) were used in orthopaedic
surgery perhaps the psychological impact this causes is not too
different from the physiological impact on the patients. However,
most of the materials used nowadays, like stainless steel, titanium
alloys or carbon fibre composites, simply reflect a more advanced
technology developed in other engineering fields, but which is not
adapted to the specific needs of a living organism. The human
body is a dynamic system where millions of cells are created and
destroyed every second. Most materials are designed to resist all
sorts of aggressive environments, thus creating a conflict with the
behaviour of living tissues.
A new wave of thought has emerged from the need to design
materials that can be accepted by the body. Several concepts have
been adopted, but perhaps those that best describe what are still
essentially long-term goals are biointegration and biocooperation.
By this we mean that the material should be a very special type
of scaffold, one that enables the reconstruction of tissues,
becoming part of them or being eliminated when no longer
required. Biodegradable polymers of the polylactic or the
polyglycolic type, and hydroxyapatite, the mineral constituent of
bone, are some examples of materials that are being increasingly
used because of their active relationship with the body. In
accordance with this thinking, using Plaster of Paris makes a lot of
sense because it degrades in the body, although a bit too quickly
for most uses. However, the incorporation of hydroxyapatite in
varying proportions produces a wide range of degradation
rates(3).
 653

2. Where do Metals Fit?

Metals are too inert for the body and when they react, i.e. when
they corrode, ions of varying degree of toxicity are released. They
have another big disadvantage: they are mechanically too stiff.
The modulus of elasticity of stainless steels is ten times larger
than that of bone, which is ca. 20 GPa(4). For titanium it is only
five times but this is still too much. When a bone with a fixation
plate is loaded, the stresses on the bone are minimal. When this
happens, the bone, which has to be subjected to stresses to remain
strong, becomes brittle and a second fracture may often occur.
In spite of these drawbacks metals have a market that still
exceeds that of any other material. For many applications
requiring high mechanical strength metals are still unreplaceable.
One of the present trends is to coat them with inert or active
ceramics. In the first case, of which the coating of titanium with
titanium nitride is a typical example, the objective is to avoid
excessive wear. In the second case, plasma spraying of
hydroxyapatite on to the stem of femural hip prostheses is
becoming com mon practice to facilitate bone adherence to the
implant.
Porous metallic coatings can also be used to enable tissue
ingrowth. Implants coated with beads or with a metallic mesh do
not have to be embedded in acrylic cement, contrary to what is
required for smooth implants. The rise in temperature during
polymerization, with the occurrence of tissue necrosis and the
toxicity problems associated with residues of the monomer are
two of the reasons that have led to the progressive replacement of
cemented prostheses by cementless porous prostheses. The
possibility of bead loosening, due simply to weak bonding
between the particles or to corrosion at the interface, has caused
some concern among doctors. Another problem is the higher
amount of metallic elements that can be released into the tissues
due to the large area of exposure to the physiological environment
and the increased risk of crevice corrosion and intergranular
attack. With the development of new polymers and preparation
techniques, smooth prostheses are creeping back into orthopaedic
surgery, as a result of the insatisfactory performance of some of
the porous coated prostheses.
654

3. Corrosion as a Major Biological Proble m

The body constitutes a particularly aggressive environment. If we

try to simplify the characteristics of this very complex corrosion
medium we can select the following key variables: chloride and
oxygen concentration, temperature, and type and magnitude of
mechanical stresses.
Extracellular fluids have a chloride concentration of 0.9% NaCl, i.e.
it is not much less aggressive than sea water. In fact the pitting
potential of stainless steel in a 3.5% NaCl solution is only ca. 35
m V lower than in a 0.9 % NaCl solution. It has been assumed that
Ep = a-b log ICI -I with b = 60 m V.
The partial pressure of oxygen in arterial and venous blood is(S),
respectively, 94.2 and 38.0 mm Hg. Here it is not the absolute
values that matter, since commonly employed metallic
biomaterials passivate with very low oxygen concentrations. The
difference in oxygen concentration between parts of the same
implant may be responsible for the passage of galvanic currents.
Their importance is not so much related to what they may do to
the material but to the way in which they may affect tissues,
since a potential difference of ca. 6 m V may develop between
regions of the same metal.
The difference in corrosion resistance at room and body
temperatures may be irrelevant in most cases. Temperature
variations in the mouth cavity can, however, be very significant,
extremes of SQC and 60 QC being easily reached.
In the body, mechanical stresses are dynamic. That is why
fatigue and corrosion fatigue have always been recognized as two
of the main possible causes of failure of implants. Fretting and
fretting corrosion are frequently responsible for the large
accumulation of metal particles and ions found near some
implants. The poor wear behaviour of titanium has prevented its
wider use in fem ural prostheses, with some surgeons still
preferring the harder Co-Cr alloys. A very large proportion of
stainless steel plates and screws show signs of corrosion. 70%(6)
and 90%(7)are two figures taken from the literature. Crevice
corrosion is a possible cause, but fretting corrosion seams more in
accordance with experimental evidence. Fig. 1 indicates
schematically the regions in a plate that can be subjected to this
 655

form of degradation. The interfaces between bone and screw,

screw and plate, and plate and bone are possible areas of attack.
However, it is the metal to metal fretting at the holes that is
responsible for the vast majority of cases of attack.

PLATE

BONE

Fig. 1 Places where crevice corrosion may occur on fixation plates

and screws

Engineers tend to look at corrosion from a materials point of view,

often worrying about a decrease in mechanical strength. Plates,
screws, hip replacements, like many other devices, also fail
mechanically. But what is of greater concern are the consequences
of their failure as biologically tolerated devices. Rejection,
loosening, tissue necrosis, and infection are problems that in one
way or another have been associated with corrosion of metallic
materials. Nickel and chromium in stainless steels, and vanadium
in titanium alloys, are particularly important due to their toxicity.
Local reactions are the more tangible evidence of their lack of
biocompatibility. Even if no discomfort to the patient results from
their use, the implants fail to be covered by bone and become
isolated by a capsule of fibrous tissue whose thickness tends to
increase as the material's toxicity increases. Metals can also be
carried in the blood stream and cause systemic effects in various
organs, like the kidneys and the liver.
656

4. Materials Development. In vivo vs. in vitro Testing

Progressive replace ment of conventional biomaterials by new,

mechanically and biologically more compatible ones, has placed a
great stress on animal testing. Pressures from the community
have imposed drastic restrictions on the use of animals. In vivo
testing is also very expensive, time consu ming and not easy to
replicate. All these circumstances are turning the tide of testing to
in vitro methods, leaving animal experimentation to the final
stages of materials development.
Since corrosion is an electrochemical process it is not surprising
that electrochemical techniques playa major role in materials
selection, and a few examples will be given later. But their
usefulness goes beyond this aspect. They have also been used as a
means of studying biocompatibility. Thrombus deposition occurs
in those metals that have a corrosion potential, Ecorr ' in blood
lower than the equilibrium potential. In general, a metal with a
tendency to be reduced causes clotting(8). Two typical examples
are nickel and gold. On the contrary, metals that are oxidised
behave in nonthrombogenic manner, like magnesium for example
Measurements of the polarization resistance, Rp ' have been
correlated with the tendency of tissues to grow on metallic
surfaces(9). Human fibroblast growth is inhibited on stainless
steel and gold, two materials with a lower Rp than titanium.
Vanadium, nickel, copper and cobalt have higher toxicity and
lower Rp than stainless steel, Co-Cr-Mo and titanium alloys,
showing that a correlation exists between biological tolerance and
electrochemical behaviour.
The electrical charge distribution on metal surfaces is an essential
element to the understanding of their interaction with cells and
proteins, and the examples below were selected because they
demonstrate how important this aspect is.
The blood compatibility experiments focused above are clearly
related to the electrical properties of the interface, and the
streaming (or zeta) potential provides an interpretation for the
thrombosis caused by some biomaterials( 1 0,11). Hydrophobic
interaction is involved in the adherence of streptococcus m utans
to several metallic and non-metallic materials, a process that is
considered an essential step in any infection(1Z).
Protein adsorption depends on surface energy and electrostatic
forces. When a surface has a positive charge, negatively charged
 657

proteins have a tendency to become adsorbed. At a pH lower than

their isoelectric point proteins are positively charged and hence
adsorption by coulombic forces is not favoured. The isoelectric
point of albumin is 4.5. The corrosion rate of stainless steel in
albumin solutions is lower at pH 8 than at pH 3(13). At pH 8
albumin adsorbs on a stainless steel surface because of its
negative charge and inhibits the corrosion process. At pH 3
adsorption does not occur and, as a result, no inhibition is
observed. The inhibition of serum solutions at physiological pH
(7.4) can be accounted for by this mechanism. Albumin adsorption
is stronger on Ag, Au and Cu than on stainless steel, Ti6AI4V, the
pure metals Ti, AI, Mo, Co, Ni and Ta and AI203(4). If we consider
that the former group has more noble electrode potentials the
logical consequence is that electrostatic adsorption of albumin
should be stronger.

S. Zero Resistance Am metry: A Techniq ue Against

Prejudice

The use of different materials subject to environmental

degradation must be carefully considered in any engineering
work because of the risks of galvanic corrosion. In orthopaedic
and dental applications surgeons are normally aware of this
elementary precaution, but sometimes an overcautions attitude
is taken and materials with an identical specification but from
different suppliers are not used on the same patient. The fear of
galvanic corrosion is justified by some disastrous failures that
occurred in earlier times. when stainless steels of different types
were used together( 15). The resistance to the use of dissimilar
materials is, however, not always justifiable. We have applied the
technique of zero resistance ammetry to the study of galvanic
couples between carbon and com mon metallic biomaterials
(stainless steel, Co-Cr-Mo and Ti-6AI-4V al1oys)(I6). The
experiments aimed to identify safe and unsafe couples,
particularly with a view to developing carbon-fibre reinforced
carbon (CFRC) plates fixed to bone by metallic screws. A cell,
specially designed to reproduce crevice corrosion conditions,
was used. All the couples gave currents that decreased with
time, but those for the titanium alloy were much lower and
varied much less than those for the other two alloys. Table 1
gives the maximum and final currents recorded in these
658

tests. It was considered unsafe to use stainless steel screws for

the fixation of carbon plates and therefore in subsequent animal
studies titanium screws were employed.
Another method consists of measuring the anodic polarization
curve of the more anodic material and the cathodic polarization
curve of the more cathodic material, and determining the
corrosion rate from their intersection. In spite of the agree ment
between this method and the zero-resistance ammetry technique,
the latter is more reliable since it does not depend on the
assumption that only one process occurs on each surface. Long-
term (6 months) exposure to a physiological environment gave
results that were totally in accordance with the predictions made
on the basis of the zero-resistance ammetry tests. To the best of
our knowledge, this is the first time it has been applied to the
study of biomaterials but it can be a very powerful tool for future
materials development.

TABLE 1. Galvanic Currents Measured by Zero-Resistance

Ammetry for Metal/Glassy Carbon Couples in the Presence of
Crevices.

Max. Current Final Current

Metal (nA) (nA)

Stainless steel 1300 500

Co-Cr-Mo 120 30

Ti-6Al-4V 5 3

The total area of each specimen was 0.785 cm 2 . The area under
the crevice was 0.392 cm 2 . The duration of the tests was 6 days.

6. Application of Electrochemical Impedance to Infected

Implants

Infection has sometimes been associated with accelerated

corrosion of metallic implants. It may occur some time after the
 659

operation and it produces a decrease in pH, which may be

responsible for an enhanced rate of dissolution. Acidification is
caused by the production of lactic acid, as a result of the activity
of pathogenic bacteria, like staphylococus a ureus. In our
laboratory an infected medium was simulated( 17) by adding 53.2
mM of lactic acid to a physiological solution containing (in mM):
137 Na+, 4 K+, 6.6 Ca 2+, 5.0 Mg, 110 C1- and 36.8 acetate. The final
pH was 4.0. The addition of lactic acid was responsible for the
enoblement of the open-circuit corrosion potential of stainless
steel, as observed in Fig. 2. Potential oscillations were less
frequent and with lower magnitude than in the normal solution,
indicating a higher film stability to breakdown. Polarization
resistance measurements produced lower currents in the lactic
acid solution, but practically no difference in the pitting potential
was found.
The above data seem to contradict clinical experience, since no
acceleration of corrosion is anticipated as a result of lactic acid
additions. The problem may be more complex than initially
envisaged and protein denaturation will have to be taken into
account in the next series of tests.
Mott-Schottky C-2 vs. E curves were determined in both
solutions. Fig. 3 shows that lactic acid suppresses the maximum
observed at 300 mV in the normal solution. Although total
capacitance includes that of the electrical double layer and that of

400r-------------------------~

2. ·...... ··. . . ·· ....·

w . ......
u l{········· ······v·········
(/)

>"
.5
w 1
-200

-400~~--~--~~--~~~~--~~
o 20 40 60 80 100 120 140 160
t (hour)
Fig. 2 - Corrosion potential as a function of time in normal (1) and
simulated infected solution (2).
660

60r-----------------~

N
LL
;;-.
E
N
U
-...

O~~~~~~_L_L~~~
-0.5 0 0.5
POTEN TIAUV. SCE)
40r-------------------~

N
LL
...."
E
N
U
-...

o 0.5
POTENTIALlV,SCE)

Fig. 3 - Mott-Schottky impedance data in normal (1) and

simulated infected (2) physiological solutions, showing that
lactic acid affects the capacitance of the electrode surface.
 661

the passive film we have considered, as a first approximation,

only the latter. This enabled the calculation of some semi-
conducting parameters for the oxide film. Lactic acid was
responsible for an increase of ca. 150 mV in the flat band
potential and a ten-fold decrease in the concentration of electron
donors. These changes seem to imply that the passive film
becomes more insulating when lactic acid is present in solution. It
is conceived that lactate adsorption is responsible for this effect,
since no changes in film composition were detected by X-ray
photo-electron spectroscopy. One possible explanation for the
effect on the corrosion rate and open-circuit behaviour is that
lactate increases film stability. No effect on pitting potential was
found. This can be justified if we consider that in this type of
attack the properties of the passive film are not as im portant as
those of non-metallic inclusions US ), particularly when sulphides
are the most susceptible sites for attack.

7. Conclusion

This paper has reviewed some topical subjects of biomaterials

development and focused on critical aspects of their
electrochemistry. Biocompatibility prediction, the role of proteins,
the use of dissimilar materials and passive film properties are all
important areas where electrochemistry has a fundamental role to
play. Metallic corrosion is one of the biggest problems that
patients, doctors, biologists and engineers have to face. A new
approach is emerging, one that tries to understand better the
interaction between life and man-made materials and that
attempts to find new solutions for a long-term goal: bioanalogy.

8. References

1. JT. Scales, G.D. Winter, and H.T. Shirley,

Corrosion of Orthopaedic Implants,
J of Bone and Joint Surgery, 4tB, 810-820 (1959)

2. H. Kawahara,
Bioceramics for Hard Tissue Replacements,
Clinical Materials, 2, 181-206 (1987)
662

3. ].S. Hanker, M.R. Tucker, B.C. Terry, R.A. Carnevale, and B.L.
Giammara,
Composite Plaster/Hydroxylapatite Implants for Jaw Bone
Reconstruction,
Biomedical Materials, ].M. Williams, M.F. Nichols, and W.
Zingg, Eds., Materials Research Society, Pittsb urgh,
Pennsylvania, 1986, pp. 77 -96

4. D.T. Reilley, A.H. Burstein, and V.H. Frankel,

The Elastic Modulus of Bone,
]. Bone Jt. Surg. Am. Vol., 56,1001, 1974

5. B.C. Syrett,
The Application of Electrochemical Techniques to the Study
of Corrosion of Metallic Implant Materials,
Electrochemical Techniques for Corrosion, R. Baboian, Ed.,
NACE, 1977, pp. 93-100

6. R.A. Silva, M.A. Barbosa, A.S. Costa, and j.A.C. Ribeiro,

Estudo Analitico de Pr6teses e Implantes Ortopedicos,
Materiais 85, Vol.5, Sociedade Portuguesa de Materiais, 1985

7. S.D. Cook, K.A. Thomas, A.F. Harding, C.L. Collins, R.]. Haddad
Jr, M. Milicic and W.L. Fischer,
The in vivo Performance of 250 Internal Fixation Devices: a
Follow-up Study,
Biomaterials, 8,177-184 (1987)

8. P.N. Sawyer and S. Srinivasan,

Essential Criteria for the Selection of Metallic and
Non - metallic Vascular Prostheses,
Adv. Biomed. Eng. Med. Phys., 3, 273-294 (1970)

9. S.G. Steinemann,
Corrosion of Titanium and Titanium Alloys for Surgical
Implants,
Titaniu m Science and Technology, Proc. 5th Intern. Conf. on
Titanium, Vol. 2, DGF, 1985, pp. 1373-1379

10. P.N. Sawyer,

The Relationship Between Surface Charge (Potential
Characteristics) of the vascular Interface and Thrombosis,
Ann. N.Y. Acad. Sci., 416, 561-583 (1983)
 663

11. S. Srinivasan and B.R. Weiss,

Electrokinetic Approach to the Selection of Blood Compatible
Materials and Anticoagulant Drugs,
Colloid Dispersions and Micellar Behaviour, ACS Symposium
Series, Vo1.9, K.L. Mittal, Ed., American Chemical Society,
Washington, D.C., 1975, pp. 322-341

12. Y.F.-Hirai, A. Akagawa, S. Minagi and H. Tsuru,

Adherence of Streptococcus m utans to I mplant Materials,
]. Biomed. Mater. Res., 21,913-920 (1987)

13. K. Merritt and S. Brown,

Effect of Proteins and pH on Fretting Corrosion and Metal Ion
Release;
]. Biomed. Mater. Res., 22. 111-120 (1988)

14. R.L. Williams and D.F. Williams,

Albumin Adsorption on Metal Surfaces,
Biomateria1s, 9.206-212 (1988)

15. O.E.M. Pohler,

Degradation of Metallic Orthopaedic Implants,
Biomaterials in Reconstructive Surgery, L.R. Rubin, Ed., Chap.
15, The C.V. Mosby Company, 1983, pp. 158-228

16. R.A. Silva, M.A. Barbosa. G.M. Jenkins and I. Sutherland,

Electrochemistry of Galvanic Couples Between Carbon and
Com mon Biomaterials in the Presence of Crevices,
Biomateria1s (in publication)

17. R.A. Silva, M.A. Barbosa, A.M. Costa, M. da Cunha Be10 and I.
Sutherland
Electrochemical Behaviour of A lSI 316L Stainless Steel
Physiological Solution Containing Lactic Acid,
8th European Conference on Biomaterials, Heidelberg, RFA,
September 1989

18. M.A. Barbosa, A. Garrido, A. Campilho, and I. Sutherland,

Surface Composition and Corrosion Behaviour of AISI 304
Stainless Steel After Immersion in 20% HN03
Sub mitted for publication
ADVANCED MATERIALS: CORROSION AND ELECTROCHEMISTRY

R.G. KELLY, R.C. NEWMAN

Corrosion and Protection Centre
UMIST
P.O. Box 88, Sackville St.
Manchester M60 lQD
United Kingdom

ABSlRACT. The term "advanced material" encompasses a wide variety of materials which are produced
by unconventional means and/or have interesting properties. The use of electrochemical techniques for the
evaluation of the corrosion behaviour of such materials is widespread. However, the interpretation of such
tests must take into account a number of complications which arise from the nature of these materials.
The importance of true control experiments has often been overlooked, and has often led to conclusions
that do not withstand scrutiny. A critical review of the use of electrochemical techniques in corrosion
studies of amorphous and rapidly solidified alloys, metal-matrix composites, aluminium-lithium alloys,
sputtered aluminium alloy films and oxide superconductors is presented.

1. Introduction

The development of advanced materials is usually driven by the need for improved mechanical
(or optical, chemical, magnetic) properties. The surface reactions of the material, including
corrosion, are usually not considered at the initial design stage, where it is assumed that
corrosion problems in service can be dealt with by surface treatment. There is nothing wrong
with this philosophy, so long as corrosion, and especially stress-corrosion cracking, are
considered at an appropriate stage of material development. Such corrosion evaluations rely
heavily on electrochemical techniques, and often involve problems of testing or interpretation
which are different from those encountered with conventional materials.
A few "advanced" materials have been selected for discussion in this paper:

Metallic glasses (amorphous alloys)

Rapidly solidified materials, including microcrystalline alloys
Metal-matrix composites based on aluminium alloys (MMCs)
Aluminium-lithium alloys
Sputtered aluminium alloy films
Oxide superconductors

We would have included nickel and titanium aluminides, but were unable to locate sufficient
literature on these materials.

665
M. G. S. Ferreira and C. A. Melendres (eds.).
Electrochemical and Optical Techniques for the Study and Monitoring of Metallic Corrosion. 665-696.
© 1991 Kluwer Academic Publishers.
666

Electrochemical studies on advanced materials may be done for a variety of reasons, not just
to evaluate potential engineering performance. Insight into fundamental questions, especially
about passivity, may be achieved by exploiting the metastable structures and phase compositions
found in many advanced materials (or produced by surface treatment, e.g. ion-implantation). We
shall review this aspect in detail.
A number of themes related to scientific method recur in the literature on advanced materials.
In particular, the use of control experiments to support novel findings has often been poorly
thought out. Examples of inadequate control experiments are :

1. Use of multi-phase recrystallized alloys to assess the role of the amorphous state
in passivity.
2. Use of conventionally processed AI alloys to determine the role of the reinforcing
phase in metal-matrix composites prepared by spraying or other unconventional
processes.
3. Use of abraded surfaces of conventional stainless steel to compare with
as-processed laser-melted surfaces.

There is also a lack of flexible thinking in some of the literature on corrosion of advanced
materials, amorphous alloys excepted. Standard engineering tests such as potentiodynamic
pitting tests are often the only evaluation reported, yet with modem electrometers it is easy to
measure the current fluctuations preceding the so-called pitting potential : detailed microscopic
models based on passive film breakdown should be supported by experimental measurements
which make some attempt to measure breakdown, not just stable pit growth. Sometimes, the
evaluation of standard tests shows that they are totally unsuitable for a particular advanced
material, e.g. AI-Li-Cu-Mg alloys are fairly resistant to stress-corrosion cracking in the standard
3.5 % NaCI medium, but crack in a variety of other solutions including artificial seawater and
Manchester tapwater [1].
Two major problems in corrosion investigations of advanced materials are chemical
heterogeneity and porosity, both of which lead to irreproducibility of results and the inability to
apply microscopic models satisfactorily. A particular problem is the necessity (or perceived
necessity) of using the as-processed surface for electrochemical investigations where one is
accustomed to mechanical polishing as the standard preparation technique. Such considerations
make it particularly desirable to measure localized corrosion behaviour in detail with high
current sensitivity, not just measure "pitting potentials", and to pay attention to experimental
statistics.

2. Use of Metallic Glasses to Gain Insight into Fundamental Questions

2.1 ROLE OF THE AMORPHOUS ST ATE VS. COMPOSITION IN CORROSION RESIS-

TANCE

A major emphasis in the research that has been conducted on metallic glasses has been the
determination of the relative contributions of the amorphous state and the composition of the
materials to their corrosion and electrochemical behaviour. Despite many years of research, it is
a question that is still under debate. The main difficulty in determining this partition is the
scarcity of compositions that can exist both as a glass and as a homogeneous single phase
material. Since most metallic glasses are of a composition at or near a deep eutectic in the alloy
 667

system phase diagram, crystallization nonnally results in the fonnation of a multi phase system,
with all its complications (e.g. galvanic effects, phase interfaces acting as sites for pit initiation,
etc.). At that point, the question of the relative importance of homogeneity becomes an issue.
For example, Latanision, Turn and Compeau [2] studied the changes in polarization behaviour
for 57.9Cu-Zr and 54.7Cu-Zr metallic glasses upon crystallization. The high Cu alloy crystal-
lized into "largely" a single phase while the lower Cu alloy composition is near a eutectic and
thus decomposes into a multiphase material upon crystallization. All four alloys showed
spontaneous passivity in the 1 N H2SO. as shown in Figures 1 and 2. Comparisons showed little
difference in behaviour between the glass and the crystalline material, though both glasses did
show slightly lower open circuit corrosion rates. Hence, these and other workers have used these
data to propose that: (a) structure is of little importance, and (b) glassy alloys corrode more
slowly than their crystalline equivalents.
Moffat and Lichter [3] also stated the amorphous structure had no effect on the active
dissolution of Fe-B and Fe-P-C in acidic sulphate solution. However, it should be pointed out
that they compared the results with crystalline iron, as neither of the two alloys crystallize into a
homogeneous phase, so conclusions about the lack of an effect of the amorphous structure must
be viewed with some care. In fact, these workers point out that if the Fe-B glass were to dissolve
according to the same mechanism as Fe, the current density measured should be higher since
boron is known to dissolve as B3+ . However, the data presented show that the Fe-B glass
exhibits a lower current density than crystalline iron, indicating that something is amiss.
Others [4-9] have found that the amorphous alloys perfonn better than the crystalline material.
However, in all of the cases cited, a true control experiment has been lacking. The amorphous
alloys were either compared with crystalline alloys that were heterogeneous [4-7] or were merely
conventional alloys of "similar" composition [8,9]. As Archer et ai. [10] have pointed out in a
recent excellent review of the electrochemistry of metallic glasses, there are very few alloys that
are amenable to lit test of the effect of the amorphous state per se. In the vast majority of cases in
the literature, where the glass was found to perfonn better than the crystalline material,
complications due to heterogeneities can be found which make data interpretation difficult.
Two metallic glasses that do crystallize into a homogeneous phase are Ni2B and C0 3B.
Heusler and Huerta [11,12] studied the active dissolution of these materials in acidic sulphate
solutions. These workers also developed a theory based upon the idea that the atoms in a glassy
metal should be less tightly bound and hence have higher Gibbs energy. They predicted that a
glassy metal should dissolve at a higher rate than a homogeneous crystal phase of the same
composition, whose atoms would have a lowered Gibbs energy due to their higher coordination.
Heterogeneous alloys were expected to corrode much faster, potentially even faster than the
glasses, due to a galvanic effect between the different phases. In Ni2B and C03B, this was indeed
found as shown in Figures 3 and 4.
Figure 3 shows the effect of heating the Ni2B alloy for 5 hours at different temperatures. The
drastic decrease in dissolution rate after heating at 670°C corresponds to the point at which the
majority of the crystallization occurred. The C03B alloy also crystallizes into a homogeneous
phase, but then decomposes into C02B and Co upon heating at higher temperatures. The effect
of these changes on the dissolution rate can be seen in Figure 4. The same work also showed
that as the crystal size increases, the dissolution rate should decrease due to a further lowering of
the Gibbs energy as predicted by the Gibbs-Thomson equation.
The analysis of Heusler and Huerta [11] gives a transparent explanation of the results of so
many workers who have found that an amorphous structure imparts a slightly higher corrosion
resistance to an alloy: the control experiments were perfonned with heterogeneous materials.
These would be expected to corrode faster than the glasses for the reasons detailed above.
However, the difficulties in producing homogeneous alloys with the same compositions as the
glasses does lead into a discussion of the role of heterogeneity in the corrosion resistance of such
 668

o Glass alloy A
01
1 ,
,
579 '4 Cu 6
W 6 Crystalline alloy A 0 6
U
<fl
579'1. Cu
0
\ ,
U
I

,) 0 ,6
>
>
0 b\ 6
I

Ii;.
06
iii
c
-E -0·1
a..
I
o ,A
C>I
/0 I6
"t; -0·2
0
01
o 6
AI
C>I
0/'
w ___ o/-c.~

-0·3
o~ ~~,
.I 6 ,l
0.01 0·1 1·0 10 100
log Curr~ Dlil:ns'ty. IJA/cm2

Figure 1. Polarization diagrams for glassy and recrystallized 57.9 % Cu-Zr alloy in 1 N
~SO., from [2].

01
o GI.ss IiIloy B
51.7'" Cu

Figure 2. Polarization diagrams for glassy and recrystallized 54.7 % Cu-Zr alloy in 1 N
~SO., from [2].
 669

-4f.- -
0
"/ 0
E
u
«
J1I 0

-5
m
I
400 500 600 700 800 900
T/k

Figure 3. Dissolution rates of Ni)J at 298 K and E = -0.1 V(SCE) in 0.5 M Na.SO.
(acidified to pH 1.8) after 5 hr heat treatment at temperature T, from [11].
.
-2~---------r----~~---~.-------r-------'-------,
r-
. 0-

0
0
-
N 0 -
'E o 0
0
u
«
-. 0
0

-3 I I

300 500 700 900

TI K
Figure 4. Dissolution rates of C03B at 298 K and E = -0.3 V(SCE) in 0.5 M Na.SO.
(acidified to pH 1.8) after 5 hr heat treatment at temperature T, from [12].
670

materials. Many workers [13-16] have found that rapid solidification can improve the localized
con-osion resistance of conventional materials without making them amorphous. This is
accomplished by either the removal (by evaporation) or redistribution (into smaller dispersals) of
inclusions and second phase particles. This treatment has been found to increase the pit density
of microcrystalline Fe-Cr alloys [15], but to lower the severity of each pit as compared to the
conventional alloy. In rapidly solidified Al alloys, a reduction in the number of the initiation
sites for pits (the interfaces between the Al matrix and the cubic silicon and intermetallic phases)
occurs which was found to increase the pitting potential [16].
Laser surface-melting, though strictly a surface modification strategy rather than a material
fabrication method, can be used to gain useful insights into localized corrosion mechanisms of
stainless steels as well as possibly improving performance in service. McCafferty and Moore
[14] found that Epi' was increased by at least 200 mV on 304 stainless steel in 0.1 M NaCI by the
removal of large scale sulphide inclusions during the laser processing. On the other hand, Chiba
et al. [13] found that too many laser passes on ferritic and austenitic surface alloys caused a
decrease in both the pitting potential and the passivating ability due to a decrease in the Cr
content of the layer. This decrease was thought to occur by Cr evaporation.
The population and size of sulphide inclusions are dramatically reduced by laser surface-melt-
ing, and a number of workers have reported that this improves the pitting resistance of the
material. One problem with most of this work is the lack of a suitable control experiment: if one
compares as-melted surfaces with normal mechanically polished stainless steel, the differences
could be due to mechanical polishing, not the laser surface-melting. This has been resolved by
Stewart [17] who melted high- and low-sulphur stainless steels to a sufficient depth (with a high
power laser) that he could give them a normal 240 grit abrasion treatment without removing the
melted layer. Subsequent comparison with 240 grit treated base alloys showed that the melted
zone was dramatically more resistant to pit nucleation in NaCl despite its duplex microstructure
as can be seen in Figure 5. This work provides evidence that nearly all metastable pitting events,
at least at low potentials, occur at inclusions rather than in re-entrant surface roughness from
mechanical abrasion (though the mechanical surface finish has a large effect which remains to be
explained). This is a satisfying piece of work on an "advanced material," since it is at the same
time a penetrating evaluation of the material's performance and a test of an important
fundamental question with important implications for pitting in general.

2.2 ROLE OF ALLOYING ELEMENTS IN PASSIVITY

Due to the wider range of composition available in glassy metals, theories concerning the role of
different alloying elements on corrosion processes can, in principle, be more completely tested.
One alloying element that has been the focus of a great deal of research is phosphorus, which is
added to glassy alloys to help achieve the amorphous state upon rapid quenching from the melt.
While much work has been done in Fe- and Ni-based alloys aimed at understanding the role of
phosphorus in both active dissolution and passivation, there exists little general agreement in the
literature.
Moffat, Flanagan and Lichter [18] found that P decreased the dissolution rate of their
Fe-40Ni-based alloy in the active region, which facilitated the passivation process. They showed
EDS evidence to support their argument that a P-rich prepassive film forms. Hashimoto et al.
[19] stated that P enhances dissolution in Fe-Cr alloys before passivation. This was said to lead
to the enrichment of Cr on the alloy surface which can then form the uniform, hydrated
chromium oxyhydroxide film which contains little phosphorus as found by XPS [20]. This view
that P enhances the dissolution rate has been supported by the impedance measurements of
Virtanen, Elsener, and Boehni [21] on a variety of Fe-Cr alloys containing different metalloids.
However, after having made impedance measurements to determine the polarization resistance at
 671

a
Potential 115 -161 mV SCE

<l
C
....c
CioI
L..
L..
:J
U

4000Orb--------------------------------------~

Potential 142 -188 mV SCE

<l
c
.....
c
CioI
L..
L..
:J
U

Figure 5. Electrochemical current noise from commercially pure 304L stainless steel
mechanically polished to 240 grit during slow potential scan in NaCl solution: (a)
conventional alloy, (b) laser surface melted material, from [17].
672

both the corrosion potential and in the passive region, these workers stated that a thick, porous
iron phosphate film fonns during the passivation process which accelerates the hydrogen
evolution reaction and decreases the anodic area. This leads to easier passivation. This work
provides a link of sorts between the ideas of Moffat et al. [18] and Hashimoto et al. [19].
Diegle and Lineman used a scratching technique on Metglass 2826A (Fe-36Ni-14Cr-12P-6B),
an Fe-36Ni-14Cr-6Si-12B glassy alloy and crystalline 316 stainless steel in 1 N HzSO. with and
without 1 N HCl [9]. These workers found that the Si-containing glass repassivated faster than
the P-containing alloy as shown in Figure 6, in contrast to what was found by Hashimoto et al.
[22J.
It has been found by Moffat et at. [3] that Mo alone does not improve the passivation
characteristics of Fe-based glasses, but when present along with P, dramatic improvements in the
passivation ability are abruptly achieved [23,24]. This is even more apparent if Cr is also present
[25,26]. For example, Fe-1OCr-lOMo-13P-7C spontaneously passivates in 12 N HCI [27].
Studies of the behaviour of amorphous alloys of different compositions have resulted in more
support for the applicability of percolation theory to corrosion phenomena [28-30]. One
example of this is the work of lanik-Czachor [7] on the passivation of Fe-B-Si glasses in borate
buffer. Pitting in the absence of aggressive ions was found for alloys containing less than 12
atomic % Si. Alloys containing more than this critical concentration fonned stable passive films.
Ruf and Tusei [5] found that for amorphous Cr-B alloys, complete protection was achieved for a
four day immersion test in 12 N HCI when the Cr concentration was greater than 50 at.%.
Others [23,27,31] have also observed an abrupt dependence of the corrosion process on Cr
concentration for more complex amorphous alloys.
Amorphous alloys such as Fe-P have been used to simulate the grain boundary chemistry in
steels where stress-corrosion cracking (SCC) is associated with grain boundary segregation. In
one study by Bandyopadhyay et al. [32] it was shown that the response of Fe-lOP to an anodic
potential step in the cracking environment (9 M NaOH, 98 C) showed a much-impaired
passivation rate compared to iron or Fe-3Ni, as shown in Figure 7. However, the overall results
(the alloys included Fe-Ni-Mo) showed that a simple correlation between transient electrochem-
istry and SCC did not exist for caustic SCC.

2.3 STUDIES OF LOCALIZED CORROSION USING AMORPHOUS ALLOYS

Surprisingly little has been published on the localized corrosion of amorphous materials. While
it is well known that these materials usually exhibit vastly superior corrosion resistance when
compared to conventional materials in conventional solutions, there are a number of cases where
this immunity to attack disappears in more aggressive solutions. It is unfortunate that so little
work has been done in such solutions, as a great deal of insight into the phenomena of localized
corrosion could be gained. Indeed, what work has been done on the pitting of amorphous alloys
has relied almost completely on the use of standard engineering tests, especially potentiodynam-
ic polarization. Again, this is unfortunate, as there is a wealth of other techniques which are
more suited for such studies. It is not surprising therefore, that discrepancies can arise as to
whether or not a material pits in a given environment. For example, while Naka et al. [33] found
that Fe-Cr-13P-7C did not pit in sulfuric acid upon anodic polarization, Hanham et al. [34] found
that it did if a nitrogen atmosphere and a slower scan rate were used. Hanham also found that
this same alloy would pit at very low potentials (0.37 V(SCE» if held potentiostatica1ly for 1.5
hours. This highlights the need to study the initiation of pitting as well as its propagation for
both a better understanding of the fundamental processes involved and in order to use these
materials in an engineering application.
 673

10- 2 ", 1N H2S04

Passiw potentials

N
E
u

..a
I

Time. ms

Figure 6. Anodic current transients after scratching Type 316L, Fe-Ni-Cr-P-B (Met-
glass), and an Fe-Ni-Cr-Si-B glassy metal in 1 N ~SO., from [9]. The base current
density, ib, was approximately 40, 40, and 200 IlNcm2 for Metglass 2826A,
Fe-Ni-Cr-Si-B, and 316 stainless steel, respectively.
674

10

N
<
E :l
u
-... 3
~ 10 VI

-
E
-...
cr

Time,s

Figure 7. Anodic charge density transients from potentiostatic stepping of various alloys
in 9 M NaOH at 98 C. Crack velocity, v, calculated based upon a film rupture model for
see from charge transient (q/t vs. t), where t is the time interval between film rupture
events, from [32].
______________________________________

-
NE1~ ~

u
D~cay
<l
:l.
_,75 ~
c
....~
:l
U
c
.Q
~25
E
....
01

~ O~~~~~~--~~--~~--~~~
o 2 3 4 5 6 7 8 9 10 11
Time, hours

Figure 8. Normal hydrogen permeation transient for an Fe-Ni-P-B glass for low charging
current density (4.15 mNcm2), from [2].
 675

Problems associated with making electrochemical measurements without inducing crevice

corrosion are always a difficulty when dealing with passive alloys. Experimental difficulties in
working with melt spun ribbons are dealt with in section 3. Moreover, in order to use melt spun
ribbons in any engineering application, the ribbons must be consolidated. A few studies of the
effects of such processing have been performed. Devine [35] studied the effect of cold rolling
on the behaviour of a 55Ni-20Fe-5Cr-14P-6B alloy. The microcracks which developed during
the processing acted as initiation sites for crevice corrosion. Hanham et al. [34] found that a
number of glassy alloys were susceptible to crevice corrosion in 1 M HiiO •. For example,
Fe-Ni-B alloys with a Ni:Fe ratio of 0.6 were susceptible, as were Co-Fe-B-Si, Ni-Cu-Si-B, and
both Ni-Si-B andi Fe-Si-B. However, Fe-1OCr-13-P-7C and Fe-lOCr-10Ni-13-P-7C were not,
indicating that the presence of relatively small amounts of Cr was extremely beneficial. In this
vein, it is interesting to note that while Devine [35] found that Fe-35Ni-15Cr-14P-6B was not
susceptible to pitting upon anodic polarization at any pH or concentration of CI·, it was
susceptible to crevice corrosion even in dilute NaCI solutions.
Diegle [36] found that a sample of Fe-30Ni-16Cr-2V-15P-6B could be made to crevice
corrode at a rapid rate, but only if the potential was held at +1.4 V(SCE). The crevice solution
pH was then found to fall to a value of 2 at which time the corrosion increased. However, higher
potentials and solutions of lower pH did not cause crevice attack. In a later study of the same
allOY, Diegle [37] found that cold rolling did indeed cause crevice initiation, but that given time,
the crevices were repassivated. Searson and Latanision [38] found that while consolidated iron
powder or compacted amorphous chips generally gave improved corrosion resistance, sintered
iron-based alloys generally performed more poorly. However, Saito, Searson and Latanision
[39] found that rapidly solidified titanium-modified nuclear grade 316 stainless steel foils
performed as well as the conventionally processed alloy in terms of general corrosion, though
pitting resistance was lower due to dendritic microsegregation. Consolidation of the foils led to
the degradation of both general corrosion and pining resistance due to the presence of prior foil
boundaries, which acted as corrosion initiation sites.
The environmentally-assisted cracking (EAC) of amorphous alloys has received some
attention, though the limited dimensions of many of these materials do make such studies
difficult and they tend to be qUalitative. Most cases of EAC in these materials have been shown
to be due to hydrogen embrittlement. Much has been made of the role of metalloids in the
susceptibility of these alloys, as these elements are known recombination poisons, and would
therefore be expected to increase the absorption of hydrogen into the alloy. Compeau and
Latanision [2] found that if the hydrogen permeation rate exceeded a critical value, abnormalities
were observed in the transient, as shown in Figures 8 and 9, which they ascribed to the trapping
of hydrogen within the glass at micron sized voids which formed during the preparation of the
material. It was observed that the hydrogen pressure within these voids can build to the point
where the membrane shatters from the internal pressure.
Some cases of stress-corrosion cracking have been reported. For example, Kawashima et al.
[40] showed that the cracking of glassy Fe-7.5Cr-23Ni-13P-7C, while rapid at low cathodic
potentials, essentially ceased over a wide range of potentials (0 to l.5 V(SCE». However, in
that range of potential, the addition of chloride ions caused cracking to occur as shown in Figure
lO.
676

Time, hours

Figure 9. Abnonnal hydrogen penneation transient for an Fe-Ni-P-B glass for high
charging current density (24.7 rnA/cm2), from [2].

.... 08

0 "'
"0 0·6
III
0

04

-05 o 05 10 15
Potential, V vs SeE

Figure 10. Ratio of the fracture stress in solution to that in air, from [40], for
Fe-Cr-Ni-P-C glass in solutions with and without chloride ion.
 677

3. Practical Problems Associated with the Application of Electrochemical Techniques to

Amorphous Alloys

The major experimental difficulties in the application of electrochemical techniques to

amorphous alloys involve the preparation, defmition and reproducibility of test surfaces.
Different material preparation techniques produce different surfaces with respect to chemical
composition and heterogeneities, physical heterogeneities, and sample size. Therefore, the same
alloy composition can give different results when prepared by different means. Thus, the
corrosion behaviour as a material property is more difficult to defme.
Most amorphous alloys are prepared by techniques such as melt spinning, which produce
essentially two dimensional samples. Typical sizes for melt spun ribbons are I to 20 mm wide,
20 to 60 lUll thick and often many centimeters long. Such physical dimensions lead to
difficulties in specimen preparation, as reliable electrical connections must be made on samples
that are small, delicate and sensitive to heating. In addition, the definition of a test surface area
by insulation of the rest of the specimen can be very difficult. A variety of insulating materials
have been used such as lacquers [3), epoxies [34,41), and even polystyrene tape [42). The
prevention of crevices forming during the application of the insulating material can make the
process very tedious and difficult. However, few reports of crevice corrosion of glassy metals
during testing have been made, though Hanham [34) found crevice corrosion was so extreme for
Fe-Ni-B alloys in 2 M H2S04 that the entire thickness of the ribbon was dissolved at the interface
with the lacquer. This prevented the generation of interpretable polarization curves for this
alloy/electrolyte combination. Since most amorphous alloys are used only once for electrochem-
ical testing, many cases of crevice corrosion may go undetected, as removal of the insulating
material without destroying the specimen is virtually impossible. Since many of these materials
show active/passive behaviour and dissolve faster in acidic solutions, they are candidates for
crevice problems. One possible solution for this problem is the use of a loop electrode design as
used extensively by Boehni and coworkers [43). In this technique, the two dimensional nature of
melt spun ribbons can be advantageous, as the length allows the electrical connection to be made
outside the solution, and the great surface area reduces the effect of the waterlines. Such a
design is often of great use in pitting studies where crevice attack at the insulation/sample
interface can invalidate results. Waterlines rarely nucleate pits if evaporation is prevented.
The preparation of a reproducible surface is usually accomplished in conventional materials
by grinding and polishing. This often cannot be done in amorphous metals for a variety of
reasons. First, their two dimensional nature often precludes the removal of any material, as such
abrasion can lead to mechanical perforation. In addition, the mechanical work induced by
polishing can lead to localized crystallization of these metastable materials. However, as-manu-
factured materials are usually not satisfactory either, as the rapid solidification procedure leads to
both chemical heterogeneities [44,45) and physical heterogeneities such as gas entrainment
cavities on the wheel side of melt spun ribbons [18). Such gas entrainment cavities act as pit
nuclei [10), as one would expect. Melt spun ribbons also suffer heavier oxidation on the free
side and at the edges [46). While such oxide films may prove to be very protective, they are not
usually what one wants to study when experimenting on glassy metals.
A number of workers have found differences in the electrochemical behaviour of the two sides
of melt spun ribbon [47-49). This is not surprising, as both melt spinning and sputtering have
been found to cause chemical heterogeneities or compositional gradients [44,45). One example
of this behaviour can be seen in Figure 11, from the work of Moffat, Hanagan and Lichter [18],
who showed that the roller side of an Fe-40Ni-14P-6B alloy exhibited an open circuit potential
that was over 50 mV higher than the free side when immersed in deaerated I N H2S04 • These
678

-0·05r--~--...--.....,....--r---....,
o Rough I roller side

o o
Smooth I air side
A Platinum ellZctrodlZ

~ -~l
; -0·15 \

." -0.

~
~ -0·25 \
0

\
Ill ____

------ -0

~---------------------~
-0·3·U'---'----'---'----'-----'
o 2 468 10
Time, hours

Figure 11. Corrosion potential as a function of time for different sides of a glassy
Fe-Ni-P-B ribbon immersed in 1 N ~S04 , from [18].

10

345678910
Chromium ContlZnt. at. "I.

Figure 12. Comparison of passive current density in 1 M ~S04 of glassy Fe74Cr6P13C7

electrodes (diamond = end of ribbon, triangle = mechanically polished side of ribbon,
square = unpolished wheel side, circle = unpolished free side) from [46] with the data of
Naka et al. [33].
 679

workers stated that at overpotentials greater than 100 mV, no differences in the electrochemical
behaviour could be found. No such effect was seen for an Fe-38Ni-4Mo-18B alloy. The
existence of the differences in corrosion potential for the two sides of the P-containing alloys
was attributed to the formation of a more stable air-formed oxide on the free side of the
P-containing alloy. However, the possibility that compositional differences existed in the near
surface region of the P-containing alloy due to the tendency for phosphorus to segregate was not
investigated.
Corke and Archer [46] found that the gross lateral variations in composition of the native
oxide film of an Fe-Cr-P-C alloy led to variations in the electrochemical behaviour, as shown in
Figure 12. When compared to the results of Naka et al. [33], it can be seen that the passive
current density varies, for different parts of the same ribbon, by an order of magnitude. Devine
[35] found a similar situation for Metglass ribbons in the as-spun state. The top of the ribbon
was Cr-rich and did not show an active/passive transition, while the bottom of the ribbon (which
had been in contact with the wheel) was Fe-rich, and exhibited an active/passive transition as
shown in Figure 13.

4. Metal-Matrix Composites (MMCs)

Metal-Matrix Composites (MMCs) based on aluminium alloys have not produced a large body
of corrosion literature. In the case of Al alloy/silicon carbide composites, this is good news;
these materials can have almost as good a corrosion resistance as their parent alloys with greatly
improved strength and stiffness. When prepared by metal spraying and subsequent consolidation
they may even have better performance owing to their reduced content of damaging phases or
inclusions which can act as pit nucleation sites. Aluminium alloy/graphite composites, on the
other hand, usually have rather poor corrosion resistance.
The galvanic effects of the non-metallic phases in MMCs are the subject of a confusing
literature. Some authors make extravagant claims about the galvanic effects of SiC - e.g. Paciej
and Agarwala [50] state : ".. aluminum alloy matrix materials are very susceptible to corrosion
when the dispersoid is a noble phase, such as SiC." However, most authors agree that SiC has a
rather small (or non-existent) effect, and in alloys containing copper, the galvanic effect of the
alloyed copper will certainly outweigh that of SiC. Graphite is a different matter, and most
authors state that serious galvanic effects exist tending to promote localized corrosion of the
aluminium alloy matrix. Even here, one of the more detailed studies, by Aylor and Moran [51]
found no galvanic effect, but ascribed the detrimental effects of graphite (shown in Table 1) to
interfacial effects on passive film integrity.

Table 1. Corrosion, Protection and Pitting Potentials (m V vs. SCE) for Al MMCs
[51]
Material Avg. Ecorr Range of Ep ,,, Range of E,.,
Gr/6061 Al -840 -870 to -890 -800 to -850
SiC/6061 Al -740 -760 to -850 -200 to -500
SiC/6061-T6 AI -735 -820 to -850 -50 to -350
6061-T6 Al (Control) -750 -750 to -850 o to -700
PM 6061 Al (Control) -720 -900 to -950 -450 to -700
680

2000 o Metglas 2826A top (shiny) surface, as received

• Metglas 2826A bottom (du II) surface, as received
l1J
:r 1600 _a-.-O
II) -<a-c.-o-e
0.0'~:!:e

~
. r.•.--;-..... r::.:::.,,-.
C1\1
'0
Q
o
(-~0-0

Figure 13. Anodic polarization curves for top (free side) and bottom (wheel side)
surfaces of a Metglass ribbon in 0.17 M NaCI, from [35].

:::a....------r-- A (no graphite)

A (graphite)

C (graphite)

C (no graphite)

log

Figure 14. Schematic Evans' diagram for the possible interaction of graphite reinforce-
ment on the pitting behaviour of Al MMCs. The "A" branches are for the anodic
polarization behaviour, while the "C" branches are for the cathodic polarization
behaviour of the alloy with and without graphite reinforcement.
 681

a
IJJ
~
~ 0·5
>
.~

~ 0·0
~

e
--0·5
1\1

u
1\1
IJJ
I
-1·0_L7--'---l_6,....-~_~S-"--_~4--'-Z-:_3~.........-_~2~--::-1
log i, A/em

2 ,
b
r
IJJ
u
I/) l
I

,!
III
>
> 0
J
:;:iiic a I

?0
1\1
I
0
Q. 0
I

e 0>
1\1
0
-1
U
1\1
w
-2 I
-8 -7 -6 -5 -4 -3 -2
log i, A/em 2

Figure 15. Anodic polarization curves for (a) a planar six-ply plate of graphite/6061-T6
Al MMC in aerated 0.5 M N~S04and (b) monolithic 6061-T6 Al in deaerated 0.5 M
N~S04,from [52].
 682

Or--------------------------------

-0·5

-uw
tf)
-l
-
>
w
/
-1·5

-7 -6 -5 -4 -3 -2
log i I [A/cm 2 ]

Figure 16. Cathodic polarization curve of graphite fibers in aerated 0.5 M Na,SO•• from
[52].
 683

It is worth examining some of the papers closely to establish what the authors mean by a
galvanic effect, and how they arrived at their conclusions. Aylor and Moran [51] reported that
the incorporation of graphite fibers in 6061 alloy caused a negative shift in the corrosion
potential (Eear) in aerated artificial seawater, associated with a negative shift in the pitting
potential. They stated that this is, therefore, not a case of "classical" galvanic corrosion, since
this would normally involve an elevation in E"",. In terms of dictionary defInitions, this might be
true, but the galvanic effect of the graphite undoubtedly influences the the rate of corrosion once
pitting has initiated. In this and several other papers on MMC corrosion, one wishes that the
authors had included cathodic polarization curves (starting at just above E"",) in the aerated
electrolyte. An Evans' diagram which shows one possible result of a galvanic effect on the
observed polarization behaviour is displayed in Figure 14.
The very strong galvanic effect of graphite can be displayed in solutions not containing
chloride. Hihara and Latanision [52] used N~SO. solutions, primarily to study the corrosion due
to pre-existing chloride in the composite structure (Figure 15). This arises from the use of
compounds such as TiCl. and BCll in the TiB deposition process used to improve the wettability
of the graphite fibers by liquid aluminium. They found that oxygen reduction on the graphite
fibers was already diffusion-limited at potentials higher than -0.5 V(SCE), as shown in Figure
16. This means that any natural corrosion process occurring at potentials around -700 to -800
mV(SCE) is necessarily under oxygen diffusion control, albeit possibly at a reduced rate owing
to cathodic inhibition effects (deposition of calcareous and/or alumina films on the graphite
cathode). Cathodic inhibition in AI alloy/graphite composites has not been studied to our
knowledge.
Galvanic effects in the corrosion of Mg alloy/graphite composites are more obvious than in AI
alloy systems, as shown by Trzaskoma [53] and others. Here there is a less sharp delineation of
the transition from passivity to pitting corrosion of Mg, so the action of the graphite shows up as
a more "classical" galvanic effect.
Saxena et al. [54] showed that composites of LM 13 (AI-Si casting alloy) and particulate
graphite pitted readily in salt water, but made no proper evaluation of the contributions of anodic
and cathodic processes.
Paciej and Agarwala [55] made detailed electrochemical measurements on AI 7091/SiC
composites and unreinforced AI 7091 alloy, but made these measurements in deoxygenated 3.5
% NaCl (PH 2 and 6) using a polarization resistance technique. It is not clear what one can
deduce from such data at pH 6, but the pH 2 results are interesting in that they display higher
corrosion rates for the composites. It is possible that these may reflect some minor galvanic
effect of the SiC, but there are many possible interfering factors and full experimental details are
not given.
The best way to explore possible galvanic effects of SiC would be to incorporate it into pure
aluminium, or at any rate, to study alloys not containing copper.
Some of the special hazards, or sources of irreproducibility, in work on Al/SiC MMCs are
given by Paciej and Agarwala. It is likely that porosity is an important variable, since nucleation
of pitting will normally be favored in tight, crevice-like cavities. No defmite evidence for a role
of porosity can be found, but most investigators report an increased scattering of data on MMCs
as compared with their base alloys. It would be valuable to make statistical measurements of
unstable ("metastable") pitting, as done by Williams et al. [56] on stainless steel, provided that
the control experiments use a monolithic alloy prepared by the same process (e.g. spraying and
extrusion) as the MMC. Even then there are difficulties in obtaining identical ageing conditions
if one wishes to study the precipitation-hardened material used in practice; Al/SiC MMCs age
more rapidly because (inter alia) diffusion is more rapid in the plastic strain fields around the
SiC particles.
684

In our own laboratory we have seen striking examples of rare, local effects in MMC corrosion.
In a pitting experiment on Al2014/SiC using 50 ppm Cl' and 14 ppm Cu++, Whyte [57] found
that a few spots on the surface grew large dendritic copper structures which elevated the
potential some 50 m V above that of the base alloy, with a corresponding increase in the pitting
rate. When these copper structures were removed with a probe, the potential fell to that of the
base alloy, as shown in Figure 17.
Spectral analysis of electrochemical (two electrode current) noise was used by Kendig et al.
[58] to compare the pitting behaviour of a fine-grained aluminium alloy with that of a 7064/10 %
SiC composite, both presumably processed by the same route. The environment was aerated 0.5
M NaCl solution. Virtually no difference was found, independent of ageing, suggesting that SiC
has little or no effect on the pit nucleation behaviour in materials of otherwise similar
composition and microstructure (Figure 18). However, another comparison of a 2024 alloy
(presumably solution-annealed) with solution-annealed 2124 alloy and 2219/SiC composites
showed a greater amplitude of noise in the composites over a wide range of frequency. This last
result is consistent with earlier reports that pining of AI/SiC MMCs tends to to be more frequent
(but not necessarily more damaging) than in the base material [59,60]. Obviously, more work
needs to be done in this area ; it is plausible that both the results are valid. but to be sure, one
needs a properly controlled set of tests on the 2000-series systems. If SiC really does promote
pining in the 2000 alloys, but not in the 7000 alloys, this could be associated with the higher
corrosion potentials of the former.
Protection of MMCs by anodizing. coating or inhibition has been explored by several authors
and is obviously of major practical importance. Mansfeld and leanjaquet [60] showed that, as
one would expect, significant reductions in the corrosion rate could be achieved using organic
coatings. Changes in low frequency impedance with time could be related to the progress of
corrosion under the coating. Anodizing of AI/SiC composites is a complex subject and it is not
clear whether bridging of the SiC particles can be achieved reliably over a large surface area
(given that there is always clumping of the SiC particles). Anodizing of AI/graphite composites
cannot be done satisfactorily. although it no doubt increases the corrosion resistance to some
extent. Inhibition using chromates or rare earth salts is effective for AI/SiC composites. but
probably not for A1!graphite composites.

S. Aluminium-Lithium Alloys

Aluminium-lithium alloys are regarded as advanced materials. despite their monolithic metallic
nature, conventional processing and long history, because they are not yet in widespread
engineering use. Recent work on corrosion of these materials has revealed some unique effects
of lithium alloying, but also some effects which were enthusiastically ascribed to lithium, but
later found to occur in other aluminium alloys. This reflects a good side of research in advanced
materials - that new types of experiments need to be done - but also a bad side. that one can
overestimate the novelty of what is found (naturally, there is a political-economic side to this as
well).
Aluminium alloys containing lithium have a unique corrosion behaviour in limited volumes of
mildly alkaline solutions where the presence of lithium ions in solution inhibits the Al ~ Al02 •
dissolution reaction. ling and Devine [61] reported passivation of Al-Li (similarly pure Al) in
Li2CO,. but not in Na2CO, or ~CO, solutions. This passivation plays an important role in the
SCC of Al-Li-Cu-Mg alloys, particularly under atmospheric conditions [1]. Craig [62] has
studied the pasSivation phenomenon in detail using rotating disc electrodes, and concludes that
 O~E~cO~r_r~(m_V~V~S~S_C_E~)______________________________________--, 685

-100

-200

-300

-400

-500

-600L-----------L-----------~----------~--------~
o 5 10 15 20
Time (hours)

~ Ecorr(2014/SiC) -+-- Ecorr(2014)

Figure 17. Corrosion potential versus time for Al 2014 and A12014/SiC in 50 ppm
chloride + 14 ppm cupric solution. The higher corrosion potential of the MMC was due
to the growth of copper dendrites, which were knocked off after about 17 hours, causing
the corrosion potential of the MMC to fall to that of the base alloy, from [57].

10- 5
a b

0 0
10-6 r- 0
0
0
.., 00 0
0 000
0
0 0 0-&-00 0
N 0
0 a o 00 0 0 0
0-0-
0
E o 00 0 0 00
u 0 0
10-7 a 0
0
<t.III
~
10
10- 8

10- 9
0 10 20 0 20
Time, h Time, h

Figure 18. Typical rms two electrode current noise versus time data for (a) aged Al 7064,
and (b) aged Al 7064/10 % SiC, in aerated 0.5 M NaCl, from [58].
686

an inner, dense film of lithium aluminate is plugging pores in the film of alumina normally found
in mild alkali. In all this work, one must take care to carry out suitable control experiments on
conventional alloys - otherwise, one might overestimate the uniqueness of Al-Li systems.
Stress-corrosion cracking of Al-Li-Cu-Mg alloys responds to potential (as in other systems)
like a pitting process. Lumsden and Allen [63] carned out slow strain rate tests in deaerated 0.5
M NaCI solution at various potentials, and obtained a "critical" potential which is above the
normal corrosion potential in aerated solutions - i.e. SCC would not be expected in aerated
solution, as shown in Figure 19. Unfortunately, the values of the corrosion potential and critical
cracking potential are reversed in artificial seawater (Figure 20), where craCking occurs under
natural immersion conditions. This is due to two additive effects : the presence of sulphate,
bicarbonate, etc. raises the pitting potential slightly and reduces the number of pits per unit area
produced at a given elevation above the pitting potential. Thus the corrosion potential floats
above the critical cracking potential as shown in Figure 20. This complex situation could not be
resolved by carrying out short-term electrochemical tests, but requires (as, ultimately, do all
corrosion problems with passive metals) an understanding of both free-corrosion behaviour (in
aerated solution) and potentiostatic behaviour (in de aerated solution).
Stress-corrosion cracks of Al-Li-Cu(-Mg) alloys normally initiate at intergranular pits in
which the grain-boundary phases are selectively attacked. Buchheit and Stoner [64] studied the
electrochemistry of the Tl (Al 2CuLi) phase, which is commonly present on the grain boundaries
of the 2090 alloy, and found it to be extremely reactive: e.g. the corrosion current density at its
corrosion potential of -1055 mV(SCE) was 80 )lNcm2• No passivation was observed at higher
potentials, though the behaviour was certainly not conventionally active either (Figure 21).
Probably Al, and especially Li, are dealloyed leaving a Cu-rich residue. The results of Buchheit
and Stoner show that the Tl phase is plausibly responsible for the intergranular corrosion of
Al-Li-Cu alloys at their typical corrosion potential of -750 to -800 mV(SCE), where the Tl phase
corrodes at approximately I mNcm 2 •
Spectral analysis of electrochemical (two electrode current) noise was used by Kendig et al.
[58] to study the pitting and/or intergranular corrosion of solution-annealed 8090 (Al-Li-Cu-Mg)
alloy in deaerated 0.5 M NaCI solution. The authors reported that the amplitude of the noise
dropped to essentially zero at all frequencies over a period of 6 hours for two 1 cm2 electrodes in
a limited volume (200 ml) of stagnant solution, whereas the electrodes remained noisy in
refreshed solution as shown in Figure 22. This was ascribed to a pH increase, from 6.4 to 8-8.5,
in the stagnant solution. We are unable to reproduce these results, possibly due to differences in
alloy composition or microstructure ; our electrodes remain noisy for several days in small
volumes of solution, although there is a reduction in noise amplitude at some frequencies with
time, as unstable pitting sites (Fe-rich inclusions) are used up [65]. More work is needed to
resolve this point.

6. Sputtered Al Alloy Films

An interesting area of alloy corrosion, which has generated a considerable literature in the past
five years, is the behaviour of metastable aluminium alloys prepared by vapour deposition.
especially RF magnetron sputtering. Normally these are studied (and potentially used) in the
form of thin films, but a parallel development is the ambition to prepare macroscopic aircraft
components [66,67]. The latter can be compositionally modulated through-thickness, and recent
attention has been focussed on Al-Cr-Fe compositions. Not much corrosion work has been done
on these materials, but presumably the compositional modulation will lead to unusual anodic
behaviour and pit or crack development - e.g. Meyer [68] found porous structures in Al-Cr-Fe
which were attributed to selective dissolution of aluminium. Such research requires more
 687

, -/---------
--
MI~
w w
c'
o
8190
170 o C/4h

}
UJ

iN
~
-
E
z6 •• ~-L ____L-__-L____L-__~~~~__~~~~~~=-~
-690 -710 -7~ -750 -770 -790 -810 -830 -850 -870 -890
Pot(lntial, mV vs SCE

Figure 19. Effects of potential on the ratio of the strain-to-failure in solution (de aerated
0.5 NaCl) to that in air for Al 8090 aged at 170 C for 4 hr, from [64].

Reduction
in Area
j
Ecorr (art. seawater
oc Nael. Na2SO,)

Artificial seawater
or NaCI. Na2S04 solution

Potential
Figure 20. Reduction in area as a function of potential for an AI-Li-Cu-Mg alloy in
artificial seawater, from [63], showing how the corrosion potential floats above the
critical cracking potential, in contrast to NaCl solution.
 688

-0·5
w
u -0·7
V)
til
>
> -0·9
~

ttJ
....c -1·1 B--II!II;:•
....0
~

Cl. -1-3

-1-5
10- 8 1()6 10-5 10-2
log i, AI cm 2

Figure 21. Polarization scan for synthesized T, (A1 2CuLi) in de aerated 3.5 wt. % NaCl,
from [65].
10-2r------~-----~-----..,

I~ 10- 4
t:;:

-
<t Refreshed

~~ 1O-S.--_ _ _ _ _ _ _ _ _ _ _ _ _ _s_ta....9;..,n_e_n_t_

10-10~-----L_::_-----L_;_-------J
10-3 10-2 10- 1
f, Hz

Figure 22. Power spectral density for Al 8090 in refreshed and stagnant aerated 0.5 M
NaCl after 6 hours exposure, from [58].
 689

thought about specimen preparation, mounting, etc., than similar work with conventional bulk
materials, and time assumes a different significance than in most corrosion experiments, as
described above for metallic glasses with unintentional compositional non-uniformity.
The pitting corrosion resistance of sputtered AI alloy thin films can be spectacular [69-73].
Moshier et al. [70] showed that an AI-6 at. % Mo thin film pitted at -210 m V(SCE) in 0.1 N KCl,
as compared with -750 mV(SCE) for pure AI, as shown in Figure 23. They also showed that the
corrosion potential of AI-6Mo in aerated solution did not approach its pitting potential - an
important precaution since some elements such as copper raise the pitting potential and the
corrosion potential by about the same amount and hence can be harmful, not beneficial. In
similar work by Natishan et al. [72] this interplay between pitting potential and corrosion
potential was not addressed.
Several authors propose that solid-solution alloying elements affect the pitting of AI alloys by
incorporating into the oxide film, e.g. Mo as MoO.--. Such suggestions are vulnerable to the
following question : how does the Mo increase the pitting potential from (say) -750 to -500
mV(SCE) ? This cannot be due to highly oxidized forms of Mo, unless the thermodynamics of
the system are altered drastically by alloying. In order to evaluate models based upon film
composition and integrity, it would be useful to have data for other alloying elements, especially
noble metals which defmitely do not oxidize into the alumina film at the relevant potentials. If
such experiments are done, they will probably show that 6 at.% of (e.g.) Cu will be as beneficial
as 6 at. % Mo ; hence that the latter exerts its influence mostly in the metal phase, not in the film.
As mentioned before, it is also necessary to make more sensitive measurements of pit nucleation,
not just propagation as done by all authors who have studied these alloys to date. If a model is
based upon the maintenance of an intact oxide film by incorporation of some alloying element,
then one must measure pit nucleation at the nA, or preferably pA, level. Such measurements are
not difficult to make using a small specimen with a well-shielded 3- (or better 2-) electrode cell
with a good electrometer to measure the current.
Frankel et al. [73] have recently shown that delicate pit growth measurements can be made on
sputtered thin films of Al and AI-X (X= Mo, Cr, Nb, Ta, or Ti). Single pits grew radially from
initiation sites and could be observed microscopically as well as by monitoring the pit current.
Pits grew under the passive film, leaving the latter to provide the necessary occlusion for the 2-D
pit growth process. Hydrogen evolution tended to destroy this flap of passive film, stopping pit
growth. An alloy with about 40 % Ta did not pit at all up to + 1.5 V(SCE), and the other
elements also gave large increases in pitting potential at the 10 to 30 % level, as can be seen in
Figure 24. In order to resist salt water environments in practice, the 5 % level used by Moshier
et al. seems to be adequate, and here there is general agreement between the pitting potentials
reported by the two groups.

7. Oxide Superconductors

Oxide superconductors have a remarkable aqueous electrochemistry which is relevant to their

interaction with moist air in practice. One can also envisage protection methods based upon
electroless deposition of metals, oxides or salts from aqueous solution.
Various families of oxide superconductors have been developed, but we shall address only the
YBlIzCuP7.• type. When low-porosity, freshly abraded material is immersed in aqueous media
such as HCl, NaCl or NaOH solution, a highly oxidizing potential is established and oxygen is
normally evolved as the anodic reaction, the cathodic reaction being reduction of a peroxide
[Cu(II)-O(I)]+ species [74-76]. In neutral or acid solutions, barium is selectively dissolved from
a reaction zone which thickens almost linearly with time (in moist air, the final product is BaC03
[76]). In sulphate or fluoride solutions, on the other hand, film formation occurs and both the
690

5 10 15
MoIybdRrum Ccnc:rmtration. at '7.

Figure 23. The effect of average Mo content in the alloy on the pitting potential of
sputtered aluminium thin films detennined by potentiodynamic polarization in 0.1 N
NaCI, from [70].

o ,1
,,
AI-Mo
JC AI-Nb
1/1 • AI- Ta
> A AI-Ti
• AI-Cr I
/ •
=e 1 • AI o·-~·-·o~

" ......
,,0'
.-
7.. ,.". .
··•....
~...
LQ
_ ........JC••••,:... -x ~
.'

;' .........
o· ..X" ' .,.'
o~~;;~ .... ~- /
~
.~
,.'
JC •••~•• "" .........

.~
•
o 10 20
Solute Concentration. at ·1.

Figure 24. Pitting potentials of various Al alloy thin films in deaerated pH 10, 0.1 M
NaCI, as a function of concentration, from [73].
 691

COrTso 2-
4
log i

Figure 25. Schematic summary of the effect of anion additions to 0.5 M acetic
acid/sodium acetate buffer on the polarization behaviour of YB~CuP7_x' from [76].
692

anodic and cathodic reactions are significantIy inhibited (Figure 25). There are excellent
possibilities of extending this approach to produce protective films which will permit storage of
these materials in untreated, room temperature air.

8. Summary

Without attempting to produce a complete review of the literature on the corrosion behaviour of
advanced materials, we have tried to point out the need for flexible thinking when testing new
materials. By using solely standard engineering tests, such as potentiodynamic pitting tests, the
opportunity to gain new insights into a wide variety of fundamental questions is missed. In
addition, the use of well·designed control experiments is critical in making conclusions about the
uniqueness of the properties of advanced materials. Finally, we have tried to point out some of
the possibilities for future research directions with the advanced materials included in this
review.

9. Acknowledgements

One of the authors (RGK) gratefully acknowledges the financial support of the the Fulbright
Commission.

10. References

[1] Craig, J.G., Newman, R.C., Jarrett, M.R., Holroyd, N.J.H. (1987) "Local Chemistry of
Stress-Corrosion Cracking in Al-Li-Cu-Mg Alloys," Proc. of the 4th IntI. Lithium Conf., J. de
Physique Colloque C3, supp. to no. 9, 48, 825-833.
[2] Latanision, R.M., Tum, J.C., Jr., Compeau, C.R. (1980) "The Corrosion Resistance of
Metallic Glasses," in Miller, K.J., Smith, R.F. (eds.), 3rd IntI. Conf. on Mech. Behav. of
Materials, Pergamon Press, Oxford, pp. 475-483.
[3] Moffat, T.P., Hanagan, W.F., Lichter, B.D. (1988) "The Active Dissolution of Iron-Metalloid
Metallic Glass" J. Electrochem. Soc., 135,2712-2719.
[4] Tong, H.S. (1985) "The Corrosion Behavior of a Ni Based Amorphous Alloy," Corr., 41,
10-12.
[5] Ruf, R.R., Tsuei, c.c. (1983) "Extremely High Corrosion Resistance in Amorphous Cr-B
Alloy," J. Appl. Phys., 54, 5705-5710.
[6] Janik-Czachor, M. (1985) "Localized Corrosion of Fe-B-Si Glassy Alloy," J. Electrochem.
Soc., 132,306-309.
[7] Janik-Czachor, M., Mazurkiewicz, B. (1987) "Effect of Crystallization on the Anodic
Behavior of FeB and CoB Glassy Alloys," Corr., 43,194-198.
[8] Hashimoto, K., Osada, K., Masumoto, T., Shimodaira, S. (1976) "Characterisitics of
Passivity of Extremely Corrosion Resistant Amorphous Alloys," Carr. Sci., 16,71-76
[9] Diegle, R.B., Lineman, D.M. (1984) "Repassivation Behavior of Two Glassy Alloys in
Sulfuric Acid with and without Chloride Ions," J. Electrochem. Soc., 131, 106-109.
[10] Archer, M.D., Corke, C.C., Harji, B.H. (1987) "The Electrochemical Properties of Metallic
Glasses," Electrochim. Acta, 32, 13-26.
[11] Heusler, K.E., Huerta, D. (1989) "Kinetics and Mechanisms of the Anodic Dissolution of
Metallic Glasses," 1. Electrochem. Soc., 136,65-71.
 693

[12] Huerta, D., Heusler, K.E. (1983) "Influence of Crystallization of Glassy C07sB25 on its
Dissolution Rates in Acid Sulfate Solution," J. Non-Cryst Sol., 56, 261-269.
[13] Chiba, S., Sato, T., Kawashima, A., Asami, K., Hashimoto, K. (1986) "Some Corrosion
Characteristics of Stainless Steel Surface Alloys Laser Processed on a Mild Steel," Corr. Sci.,
26,311-328.
[14] McCafferty, E., Moore, P.G. (1986) "Corrosion Behavior of Laser-Surface Melted and
Laser-Surface Alloyed Steels," J. Electrochem. Soc., 133, 1090-1096.
[15] Kawashima, A., Hashimoto, K. (1986) "The Pitting Corrosion Behavior of Rapidly
Solidified Fe-Cr Alloys in 0.5 M NaG Solution," Corr. Sci., 26, 467-478.
[16] Yoshioka, H., Yoshida, S., Kawashima, A., Asami, K., Hashimoto, K. (1986) "The Pitting
Corrosion Resistance of Rapidly Solidified Aluminum Alloys," Corr. Sci., 26, 795-812.
[17] Stewart, J. (1989) Harwell Laboratory, private communication.
[18] Moffat, T.P., flanagan, W.F., Lichter, B.D. (1987) "Effect of Alloy Chemistry on the
Corrosion Resistance of FeNi Metal-Metalloid Metallic Glasses," Corr., 43, 589-593.
[19] Hashimoto, K., Masumoto, T. (1981) "Corrosion Behavior of Amorphous Alloys," in
Herman, H. (ed.) Treatise on Materials Science and Technology, v. 20, Academic Press, New
York,291-324.
[20] Asami, K., Hashimoto, K., Masumoto, T., Shimodaira, S. (1976) "ESCA Study of the
Passive Film on an Extremely Corrosion Resistant Amorphous Iron Alloy," Corr. Sci., 16,
909-914.
[21] Virtanen, S., Elsener, 8., Boehni, H. (1988) "Stability of Passive Films on Amorphous
Fe-Cr Alloys with Boron and Phosphorus Studied with Impedance Spectroscopy," in Elsener,
B. (ed.) Proc. 3rd Inti. Symp on Electrochem. Methods in Corr. Res., Trans-Tech Publishers,
Zurich.
[22] Hashimoto, K., Naka, M., Noguchi, J., Asami, K., Masumoto, T. (1978), in Frankenthal,
R.P., Kruger, J. (eds.) Proc. 4th IntI. Symp. on Passivity, Electrochemical Soc., Princeton, NJ.
[23] Naka, M., Hashimoto, K., Masumoto, T. (1978) "High Corrosion Resistance of Amorphous
Fe-Mo and Fe-W Alloys in HG," J. Non-Cryst. Sol., 29,61-65.
[24] Hashimoto, K., Naka, M., Asami, K., Masumoto, T. (1979) "An X-ray Photoelectron
Spectroscopy Study of the Passivity of Amorphous Fe-Mo Alloys," Corr. Sci., 19, 165-170.
[25] Hashimoto, K., Kobayashi, K., Asami, K., Masumoto, T. (1981) "Corrosion Resistant
Amorphous Alloys in Hot Concentrated Hydrochloric Acids," in Proc. 8th Inti. Congo Met.
Corr., I, Dechema, Frankfurt, 70-75.
[26] Kawashima, A., Sato, T., Hashimoto, K. (1985) "Amorphous Fe-Cr-Mo-13P-7C Alloys
Unsusceptible to Hydrogen Embrittlement in Deaerated Solutions of Different pH," J.
Non-Cryst. Sol., 70, 55-68.
[27] Naka, M., Hashimoto, K., Inoue, A., Masumoto, T. (1979) "Corrosion Resistant Amorphous
Fe-C Alloys Containing Chromium and/or Molybdenum," J. Non-Cryst. Sol., 31, 347-354.
[28] Sieradzki, K., Corderman, R.R., Shukla, K., Newman, R.C. (1989) "Computer Simulations
of Corrosion: Selective Dissolution of Binary Alloys," Phil. Mag. A, 59, 713-746.
[29] Newman, R.C., Foong, T.M., Sieradzki, K. (1988) "Validation of a Percolation Model for
Passivation of Fe-Cr Alloys: I. Current Efficiency in the Incompletely Passivated State," Corr.
Sci., 28, 523-527.
[30] Qian, S., Newman, R.C., Sieradzki, K. (1989) "Validation of a Percolation Model for
Passivation of Fe-Cr Alloys: 2-Dimensional Computer Simulations," J. Electrochem. Soc., in
press.
694

[31) Naka, M., Hashimoto, K., Masumoto, T. (1979) "Effect of Addition of Chromium and
Molybdenum on the Corrosion Behavior of Amorphous Fe-20B, Co-20B and Ni-20B Alloys,"
J. Non-Cryst. Sol., 34, 257-266.
[32) Bandyopadhyay, N., Newman, RC., Sieradzki, K. (1984) "Caustic Cracking of an Fe-3Ni
Alloy : Effects of Alloying Additions and Impurities," Proc. 9th IntI. Congo Met. Corr.,
Toronto, 2, Nat. Res. Council, Ottawa, 210-215.
[33) Naka, M., Hashimoto, K., Masumoto, T. (1976) Corr., 32, 146.
[34) Hanham, C.M., Waseda, Y., Aust, K. (1980) "Corrosion Behavior of Amorphous Iron-base
Alloys," Mater. Sci. Engng., 45, 71-76.
[35) Devine, T.M. (1977) "Anodic Polarization and Localized Corrosion Behavior of Amor-
phous Ni)5Fe30CrlsPI4B6 in Near Neutral and Acidic Chloride Solutions," J. Electrochem. Soc.,
124,38-44.
[36) Diegle, R.B. (1979) "Localized Corrosion of Amorphous Fe-Ni-Cr-P-B Alloys," Corr., 35,
250-258.
[37) Diegle, R.B. (1980) "Crevice Corrosion of Glassy Fe-Ni-Cr-P-B Alloys," Corr., 36,
362-368.
(38) Searson, P.C., Latanision, R.M. (1985) "The Corrosion and Oxidation Resistance of Iron-
and Aluminum-based Powder Metallurgical Alloys," Corr. Sci., 25, 947-968.
[39) Saito, N., Searson, P.e., Latanision, RM. (1986) "The Corrosion Perfonnance of
Microcrystalline Titanium-modified 316 Stainless Steel," Corr. Sci., 26, 629-645.
[40) Kawashima, A., Hashimoto, K., Masumoto, T. (1976) in Grant, N.J., Giessen, B.e. (eds.),
Proc. 2nd Conf. on Rapidly Quenched Metals, MIT Press, Cambridge, MA, 475.
(41) Thorpe, S.J., Ramaswami, B., Aust, K. (1988) "Corrosion and Auger Studies of a
Nickel-base Metal-Metalloid Glass," I. Electrochem. Soc., 135,2162-2170.
(42) Bhattacharya, RS., Raffoul, C.N., Rai, A.K. (1987) "Corrosion Behavior of Ion Beam
Mixed Amorphous NiNb Coating," I. Electrochem. Soc., 134,400-403.
[43) Stockert, L., Hunkeler, F., Boehni, H. (1985) "A Crevice-Free Measurement Technique to
Detennine Reproducible Pitting Potentials,'· Corr., 41, 676-677.
[44) Walter, I.L., Bacon, F., Luborsky, F.E. (1976) "An Auger Analysis of the Embrittlement of
Amorphous Alloy NiJeJI4B6'" Mater. Sci. Engng., 24, 239-245.
(45) Brunsch, A. (1979) "Magnetic Properties and Corrosion Resistance of (CoFeB)I()()..Cr. Thin
Films," I. Appl. Phys., 50, 7603-7605.
(46) Corke, C.C., Archer, M.D. (1988) "The Electrochemical Corrosion of Fe74Cr6PI3C7 Metallic
Glass Produced by Vacuumschmelze GmbH," Corr. Sci., 28, 353-370.
[47] Massiani, Y., Gas, P., Crousier, I., Bernardini, J. (1984) "Effect of Heat Treatments on the
Corrosion Behavior of 30Fe-35Ni-14P-6B-15Cr Amorphous Alloy - Electrochemical and
Corrosion Studies," Appl. Surf. Sci., 20, 132-144.
(48) Kiss, L., Kovacs, P., Farkas, I. (1982) "Electrochemistry and Corrosion Properties of
Amorphous and Crystalline Alloys," Prot. of Met., 18, 193-194.
[49] Kovacs, P., Farkas, I., Takacs, L., Awad, M.Z., Vertes, A., Kiss, L., Lovas, A. (1982)
"Effect of Composition and the Processing Parameters on the Electrochemical Corrosion of
Iron-Boron Metallic Glasses," I. Electrochem. Soc., 129, 695-700.
[50] Paciej, RC., Agarwala, V.S. (1988) "Influence of Processing Variables on the Corrosion
Susceptibility of Metal-Matrix Composites," Corr., 44, 680-684.
[51) Aylor, D.M., Moran, PJ. (1985) "Effect of Reinforcement on the Pitting Behavior of
Aluminum-Based Metal Matrix Composites," 1. Electrochem. Soc., 132, 1277-1284.
[52] Hihara, L., Latanision, RM. to be published.
 695

[53] Tnaskoma, P.P. (1986) "Corrosion Behavior of a Graphite Fiber/Magnesium Metal Matrix
Composite in Aqueous Chloride Solution," Corr., 42, 609-612.
[54] Saxena, M., Modi, D.P., Yegneswaran, A.H., Rotiatgi, P.K; (1987) "Corrosion Characteris-
tics of Cast Aluminum Alloy-3 wt.% Graphite Particulate Composites in Different Environ-
ments," Corr. Sci., 27, 249-257.
[55] Paciej, R.C., Agarwala, V.S. (1986) "Metallurgical Variables Influencing the Corrosion
Susceptibility of a Powder Metallurgy SiCw/Al Composite," Corr., 42, 718-728.
[56] Williams, D.E., Westcott, C., fleischmann, M. (1985) "Stochastic Models of Pitting
Corrosion of Stainless Steels, II : Measurement and Interpretation of Data at Constant
Potential," 1. Electrochem. Soc., 132, 1804-1811.
[57] Whyte, S.M., Newman, R.C. (1988) to be published.
[58] Kendig, M.W., Jeanjaquet, S., Mahoney, M. (1988) "Electrochemical Noise Analysis of the
Corrosion of Aluminum Alloys and Composites," Paper 383, Corrosion '88, Nat. Assoc. Corr.
Engr., Houston.
[59] Tnaskoma, P., McCafferty, E., Crowe, C. (1983) "Corrosion Behavior of SiC/Al Metal
Matrix Composites," J. Electrochem. Soc., 130, 1804-1809.
[60] Mansfeld, F., Jeanjaquet, S.L. (1986) "The Evaluation of Corrosion Protection Measures for
Metal Matrix Composites," Corr. Sci., 26, 727-734.
[61] Jing, G., Devine, T.M. (1987) "Influence of Lithium on the Corrosion of Aluminum,"
Scripta Metall., 21, 853-857.
[62] Craig, 1.G. (1988) to be published.
[63] Lumsden, J.B., Allen, A.T. (1988) "Stress-Corrosion Cracking Behavior of Al-Li Alloy
8090," Corr., 44, 527-532.
[64] Buchheit, R.G., Stoner, G.E. (1989) "Corrosion Behavior of Tl [Al2CuLi] Intermetallic
Compound in Aqueous Environments," Al-Li V, Williamsburg, Va, 1989, in press.
[65] Ryder, J.C., (1988) M.Sc. Dissertation, Univ. of Manchester.
[66] Bickerdike, RL. et aI. (1984-85) IntI. 1. Rapid Sol., 1, 305.
[67] Bickerdike, RL. et aI. (1986) IntI. J. Rapid Sol., 2, 1.
[68] Meyer, C. (1989) Ph.D. Thesis, Univ. of Manchester.
[69] Moshier, W.e., Davis, G.D., Cote, G.O. (1989) "Surface Chemistry of Sputter-Deposited
Al-Mo and AI-Cr Alloys Polarized in 0.1 N KCl," J. Electrochem. Soc., 136,356-362.
[70] Moshier, W.C., Davis, G.D., Ahearns, J.S., Hough, H.F. (1986) "Influence of Molybdenum
on the Pitting Corrosion of Aluminum Films," J. Electrochem. Soc., 133, 1063-1064.
[71] Moshier, W.C., Davis, G.D., Ahearns, 1.S., Hough, H.F. (1987) "Corrosion Behavior of
Al-Mo Alloys in Chloride Solution," 1. Electrochem. Soc., 134, 2677-2684.
[72] Natishan, P.M., McCafferty, E., Hubler, G.K. (1988) "Surface Charge Considerations on the
Pitting ofIon-Implanted Aluminum,"J. Electrochem. Soc., 135,321-327.
[73] Frankel, G.S., Russak, M.A., Jahnes, C.V., Minamaani, M., Brusic, V.A. (1989) "Pitting of
Sputtered Al Alloy Thin Films," J. Electrochem. Soc., 136, 1243-1244.
[74] BachtIer, M., Lorenz, W.J., Schindler, W., Saemann-Lechenko, G. (1988) "Electrochemical
BehaviorofYB~Cu307'" Mod. Phys. Lett. B, 2, 819-828.
[75] Rosamilia, J.M., Miller, B., Schneemeyer, L.F., Waszczak, J.V., O'Bryan, H.M., Jr. (1987)
"Aqueous Electrochemistry of Cuprate-Based High T. Superconductors," J. Electrochem. Soc.,
134,1863-1864.
[76] Hepburn, B.1., Lau, H.L., Lyon, S.B., Newman, RC., Thompson, G.E. (1989) "Electro-
chemistry of Superconductor YB~CUP7.• in Solutions Containing Various Anions," submitted
for publication to J. Electrochem. Soc.
696

[77] Bacbtler, M., Lorenz, W.J., Schindler, W., Saemann-Lechenko, G. (1988) "Electrochem-
istry of YB~CUP7'" J. Electrochem. Soc., 135,2284-2287.
 INDEX

Absorption 245
Acid Rain 571
Acoustic Emission 603
Active-passive Transition 42
Active Site 418
Adsorption 474, 545, 553
Adsorption Isotherms 38, 545-547
Affinity 2,3
Ageing Effects 99
Aliasing 128
Aluminium 175, 574, 576
Aluminium-litium Alloys 665, 684
Aluminium Oxide 75, 84
Amorphous Alloys 665-667, 672, 675, 677
Angle of Incidence 289
Annealing 267
Anodic Films 537
Anti-Stokes Lines 356
Atmospheric Corrosion 333, 571
Atmospheric Corrosion Monitor 572, 574, 577, 578

Band Structure 192, 111

Beer's Equation 243
Biointegration 652
Biomaterials Applications 652, 653, 657
Biomaterials Market 651, 653
Bleaching 325
Boiling Water Reactors 622
Bone 652, 653, 655, 657
Breakdown of Passivation, Models 494
Anion Penetration Migration 495
Chemical Dissolution 497
Chemico-Mechanical 496
Depassivation - Repassivation 498
Local Acidification 499
Point Defect Model 56, 58, 496

Capacity of Electrodes 104

Cathodic Protection 540
Cobalt 364, 656-658
Coloring 325
Computer Simulation 123
Conversion Coatings 521, 526, 528
Copper 364
Copper Oxidation 256
Copper Oxide 88
Corrosion by Abrasion 177
Corrosion Fatigue 585, 636
Corrosion Products 572
Corrosion Rates 571, 572
697
698

Crack Propagation 633

Crevice Corrosion 500, 653-655, 657
Mechanisms 501
Critical Potential 639
Current and Potential Transients 586, 601, 603
Current Detection 71
Current Transients 139

Damage Functions 571, 573, 581, 583

Damping Constant 242
Delamination 337
Dezincification 278
Dielectric Constant 532
Differential Reflectometry 241, 242
Dipole Moment 392, 451
Double Layer Capacitance 45
Drude Equations 247
Dynamic Ellipsometry 339

Electrochemical Impedance Spectroscopy 31, 123, 510, 521

Electrochemical Noise 135, 507
Electrodissolution of Aluminium 33
Electron Spectroscopy 558
Ellipsometry 247, 285, 286, 559
Equilibrium Conditions of Reactions 4, 6

Faradaic Admittance 45
Fe-Cr 174
Fermi Level 253
Film Thickness 292
Fourier Transform 32

Galvanic Corrosion 654, 657

Galvanized Steel 574, 575
Gaseous Oxidation 343
Geothermal Systems 611, 612
Gold 362, 363
Gold Oxide 99
Group Theory 453
Basis 453, 457
Character 459
Direct Products 466
Matrices 453
Orthogonality 461, 462
Representation 453, 457
Similarity 457
Symmetry Operations 453

Hamiltonian 393, 442, 443

Harmonic Oscillator 440, 445
 699

High-Field Model 75
High Temperature Corrosion 611
High Temperature Raman Spectroelectrochemistry 376
Humidity Relative 571, 572, 581, 583
Hydrogen Permeation Electrode 599
Hydrogen Water Chemistry 642
Hydroxyapatite 652, 653

Index of Refraction 244, 288, 289

Infrared Adsorption 392
Infrared Spectroscopy 439, 564
Inhibition
Interface 522
Interphase 522
Inhibitor Efficiency 546
Inhibitors 328, 545, 551, 558
Inhomogeneous Surface Model 522
Interband Transition
Direct 253
Indirect 253
Interferometry 249
Ion Implantation 234
IR Drop 595
Iron 9, 10, 12, 171, 177, 181, 183, 229, 367
Iron Oxide 110
Irradiation 234

Kramers-Kronig Transforms 31

Laser Raman Spectroscopy 355

Lead 360, 378
Linear Polarization Resistance 123
Localized Corrosion 528, 529

Magnox 319
Mass Transients 85
Mercury 360
Metal-matrix Composites 665, 679
Metal Alloys 372
Metallic Glasses 665-667
Microscopic Ellipsometry 337
Modulation Spectroscopy 254
Mott-Schottky Plots 225, 658

Nickel 269, 367, 380, 574, 576

Noise 130
Nuclear Power System 611
Numerical Models 130
Nylon 574, ~i77
Nyquist Frequency 128, 129
700

Optical Constants 244, 245, 249, 298

OPtical Spectroscopy 558
Optical Techniques 242, 285
Oxide Films 299
Oxide Superconductors 665, 689
Oxygen Reduction 125

Paint 574
Passivation,Models 486
Bipolar Model 492
Crystalline Oxide 486
Polymeric Hydrated Oxide 487
Sandwich Structure 487
Passivators 551
Passive Films 219, 299
Electronic Structure 222
Interfaces 220
Structure 236
Capacity Measurements 222
Passivity 423
Phase Angle 523
Phonons 253
Photocorrosion 112, 213
Photoelectrochemical Measurements 226
Photoelectrochemical Process 191
Photoelectrochemistry 219
Experimental Techniques 220
Theoretical Aspects 208
Pitting Corrosion 14, 15, 158, 331, 528, 620, 654, 659, 661
Place Exchange 322
Plaster of Paris 652
Polarizability 356, 390
Polarization 390, 471
Polarization Curves 4, 5, 7
Polarization of Light 287, 288, 293
Polarization Resistance 656, 659
Polarized Light 246
Pollutants 571, 572, 581, 582
Polymer Coatings 521, 532
Polymers 652, 653
Pore Resistance 523, 533
Porosity 523
Potential-pH Diagrams 7, 614, 617
High Temperatures 21-23
Iron 9, 10, 12
Other Metals And Metalloids 12-14, 16
Water 8
Potential-Temperature Diagram 22, 25
Power Spectra 128
Power Spectral Measurement 155
Pressurized Water Reactors 622
Proteins, Role of 656, 657, 659, 661
Pulse Techniques 69, 87, 104
Quantization 130
 701

Quasi Fermi Levels 211

Raman Activity 391, 392

Raman Scattering 358
Random Phenomena 136, 137
Rapid Solidified Materials 665
Redox Potential 642
Reflection Coefficients 288, 289
Reflectivity 243,245
Ref1ectogram 255
RF Magnetron Sputtering 599

Selection Rules 451, 468

Semiconductors 192
Electron Distribution 194
Quantum Electrode Kinetics 201
Reactions Kinetics 202
Semiconductor Liquide Interface 195
Stability and Decomposition 206
Thermodynamics 199
Silver 362
spectroscopic Ellipsometry 337
Sputtered Aluminium Alloy Films 665, 686, 689
stainless Steel 229, 489, 504, 652-659
Statistical Counting 151
Steel, Carbon 652
Stochastic Process 140
Past-Dependant Point Process 144
Point Process 140, 142
Poisson Point Process 143
Power Spectrum of Stochastic Pulses 145
Stokes Lines 356
Stress Corrosion Cracking 181, 334, 585
Film Induced Cleavage Mode 595, 601
Hydrodynamics 595
Intergranu1ar 627
Mechanical-Electrochemical Model 589
Slip-Dissolution Model 589, 591, 595, 601
Steady State Electrochemical Techniques 586
Transient Electrochemical Techniques 591
Surface-Enhanced Raman Spectroscopy 389, 566
Surface Coverage 547
Surface States 215
Surgical Implants 20, 21

Thermodynamics 1, 2, 612
Time-of-wetness 571, 574
Time Records 139, 165
Titanium 230, 365, 652-658
Titanium Nitride 653
Titanium Oxide 115
Transients Measurements 123
702

Transmission Line 529

Transpassive Dissolution 33, 38

Uniform Corrosion 831

Vacancies 56
Vanadium 655, 656
Vibrational Mode 393, 394, 446
Internal Mode 395
Normal Coordinate 395
Normal Modes 398
Projection Operator 401
Representation 396
Symmetry 396
Symmetry Coordinates 403

Weight Loss 575, 581

X-ray Photoelectron Spectroscopy 561

Zero-Resistance Ammetry 657, 658

Zeta Potential 656
Zinc 365
Zirconium Oxides 107